Dorsal Excitor motor neuron DE-3 in the medicinal leech plays three very different dynamical roles in three different behaviors. Without rewiring its anatomical connectivity, how can a motor neuron dynamically switch roles to play appropriate roles in various behaviors? We previously used voltage-sensitive dye imaging to record from DE-3 and most other neurons in the leech segmental ganglion during (fictive) swimming, crawling, and local-bend escape (Tomina and Wagenaar, 2017). Here, we repeated that experiment, then re-imaged the same ganglion using serial blockface electron microscopy and traced DE-3’s processes. Further, we traced back the processes of DE-3’s presynaptic partners to their respective somata. This allowed us to analyze the relationship between circuit anatomy and the activity patterns it sustains. We found that input synapses important for all the behaviors were widely distributed over DE-3’s branches, yet that functional clusters were different during (fictive) swimming vs. crawling.

The ability to recognize motivationally salient events and adaptively respond to them is critical for survival. Here, we tested whether dopamine (DA) neurons in the dorsal raphe nucleus (DRN) contribute to this process in both male and female mice. Population recordings of DRN(DA) neurons during associative learning tasks showed that their activity dynamically tracks the motivational salience, developing excitation to both reward-paired and shock-paired cues. The DRN(DA) response to reward-predicting cues was diminished after satiety, suggesting modulation by internal states. DRN(DA) activity was also greater for unexpected outcomes than for expected outcomes. Two-photon imaging of DRN(DA) neurons demonstrated that the majority of individual neurons developed activation to reward-predicting cues and reward but not to shock-predicting cues, which was surprising and qualitatively distinct from the population results. Performing the same fear learning procedures in freely-moving and head-fixed groups revealed that head-fixation itself abolished the neural response to aversive cues, indicating its modulation by behavioral context. Overall, these results suggest that DRN(DA) neurons encode motivational salience, dependent on internal and external factors. SIGNIFICANCE STATEMENT Dopamine (DA) contributes to motivational control, composed of at least two functional cell types, one signaling for motivational value and another for motivational salience. Here, we demonstrate that DA neurons in the dorsal raphe nucleus (DRN) encode the motivational salience in associative learning tasks. Neural responses were dynamic and modulated by the animal’s internal state. The majority of single-cells developed responses to reward or paired cues, but not to shock-predicting cues. Additional experiments with freely-moving and head-fixed mice showed that head-fixation abolished the development of cue responses during fear learning. This work provides further characterization on the functional roles of overlooked DRN(DA) populations and an example that neural responses can be altered by head-fixation, which is commonly used in neuroscience.

SIGNIFICANCE: Mid-infrared (IR) imaging based on the vibrational transition of biomolecules provides good chemical-specific contrast in label-free imaging of biology tissues, making it a popular tool in both biomedical studies and clinical applications. However, the current technology typically requires thin and dried or extremely flat samples, whose complicated processing limits this technology’s broader translation. AIM: To address this issue, we report mid-IR photoacoustic microscopy (PAM), which can readily work with fresh and thick tissue samples, even when they have rough surfaces. APPROACH: We developed a transmission-mode mid-IR PAM system employing an optical parametric oscillation laser operating in the wavelength range from 2.5 to 12 μm. Due to its high sensitivity to optical absorption and the low ultrasonic attenuation of tissue, our PAM achieved greater probing depth than Fourier transform IR spectroscopy, thus enabling imaging fresh and thick tissue samples with rough surfaces. RESULTS: In our spectroscopy study, the CH₂ symmetric stretching at 2850 cm¯¹ (3508 nm) was found to be an excellent source of endogenous contrast for lipids. At this wavenumber, we demonstrated label-free imaging of the lipid composition in fresh, manually cut, and unprocessed tissue sections of up to 3-mm thickness. CONCLUSIONS: Our technology requires no time-consuming sample preparation procedure and has great potential in both fast clinical histological analysis and fundamental biological studies.

Persistent neural activity in cortical, hippocampal, and motor networks has been described as mediating working memory for transiently encountered stimuli(1,2). Internal emotional states, such as fear, also persist following exposure to an inciting stimulus(3), but it is unclear whether slow neural dynamics are involved in this process. Neurons in the dorsomedial and central subdivisions of the ventromedial hypothalamus (VMHdm/c) that express the nuclear receptor protein NR5A1 (also known as SF1) are necessary for defensive responses to predators in mice(4-7). Optogenetic activation of these neurons, referred to here as VMHdm(SF1) neurons, elicits defensive behaviours that outlast stimulation(5,8), which suggests the induction of a persistent internal state of fear or anxiety. Here we show that in response to naturalistic threatening stimuli, VMHdm(SF1) neurons in mice exhibit activity that lasts for many tens of seconds. This persistent activity was correlated with, and required for, persistent defensive behaviour in an open-field assay, and depended on neurotransmitter release from VMHdm(SF1) neurons. Stimulation and calcium imaging in acute slices showed that there is local excitatory connectivity between VMHdm(SF1) neurons. Microendoscopic calcium imaging of VMHdm(SF1) neurons revealed that persistent activity at the population level reflects heterogeneous dynamics among individual cells. Unexpectedly, distinct but overlapping VMHdm(SF1) subpopulations were persistently activated by different modalities of threatening stimulus. Computational modelling suggests that neither recurrent excitation nor slow-acting neuromodulators alone can account for persistent activity that maintains stimulus identity. Our results show that stimulus-specific slow neural dynamics in the hypothalamus, on a time scale orders of magnitude longer than that of working memory in the cortex(9,10), contribute to a persistent emotional state.

Among animals with visual processing mechanisms, the leech Hirudo verbana is a rare example in which all neurons can be identified. However, little is known about its visual system, which is composed of several pigmented head eyes and photosensitive non-pigmented sensilla that are distributed across its entire body. Although several interneurons are known to respond to visual stimuli, their response properties are poorly understood. Among these, the S-cell system is especially intriguing: it is multimodal, spans the entire body of the leech and is thought to be involved in sensory integration. To improve our understanding of the role of this system, we tested its spectral sensitivity, spatial integration and adaptation properties. The response of the S-cell system to visual stimuli was found to be strongly dependent on the size of the area stimulated, and adaptation was local. Furthermore, an adaptation experiment demonstrated that at least two color channels contributed to the response, and that their contribution was dependent on the adaptation to the background. The existence of at least two color channels was further supported by transcriptomic evidence, which indicated the existence of at least two distinct groups of putative opsins for leeches. Taken together, our results show that the S-cell system has response properties that could be involved in the processing of spatial and color information of visual stimuli. We propose the leech as a novel system to understand visual processing mechanisms with many practical advantages.

Two-photon microscopy is a key imaging technique in life sciences due to its superior deep-tissue imaging capabilities. Light-weight and compact two-photon microscopes are of great interest because of their applications for in vivo deep brain imaging. Recently, dielectric metasurfaces have enabled a new category of small and lightweight optical elements, including objective lenses. Here we experimentally demonstrate two-photon microscopy using a double-wavelength metasurface lens. It is specifically designed to focus 820 and 605 nm light, corresponding to the excitation and emission wavelengths of the measured fluorophors, to the same focal distance. The captured two-photon images are qualitatively comparable to the ones taken by a conventional objective lens. Our metasurface lens can enable ultracompact two-photon microscopes with similar performance compared to current systems that are usually based on graded-index-lenses. In addition, further development of tunable metasurface lenses will enable fast axial scanning for volumetric imaging.

Sensitivity to water waves is a key modality by which aquatic predators can detect and localize their prey. For one such predator - the medicinal leech, Hirudo verbana - behavioral responses to visual and mechanical cues from water waves are well documented. Here, we quantitatively characterized the response patterns of a multisensory interneuron, the S cell, to mechanically and visually cued water waves. As a function of frequency, the response profile of the S cell replicated key features of the behavioral prey localization profile in both visual and mechanical modalities. In terms of overall firing rate, the S cell response was not direction selective, and although the direction of spike propagation within the S cell system did follow the direction of wave propagation under certain circumstances, it is unlikely that downstream neuronal targets can use this information. Accordingly, we propose a role for the S cell in the detection of waves but not in the localization of their source. We demonstrated that neither the head brain nor the tail brain are required for the S cell to respond to visually cued water waves.

In this protocol, we introduce an effective method for voltage-sensitive dye (VSD) loading and imaging of leech ganglia as used in Tomina and Wagenaar (2017). Dissection and dye loading procedures are the most critical steps toward successful whole-ganglion VSD imaging. The former entails the removal of the sheath that covers neurons in the segmental ganglion of the leech, which is required for successful dye loading. The latter entails gently flowing a new generation VSD, VF2.1(OMe).H, onto both sides of the ganglion simultaneously using a pair of peristaltic pumps. We expect the described techniques to translate broadly to wide-field VSD imaging in other thin and relatively transparent nervous systems.

Studies of neuronal network emergence during sensory processing and motor control are greatly promoted by technologies that allow us to simultaneously record the membrane potential dynamics of a large population of neurons in single cell resolution. To achieve whole-brain recording with the ability to detect both small synaptic potentials and action potentials, we developed a voltage-sensitive dye (VSD) imaging technique based on a double-sided microscope that can image two sides of a nervous system simultaneously. We applied this system to the segmental ganglia of the medicinal leech. Double-sided VSD imaging enabled simultaneous recording of membrane potential events from almost all of the identifiable neurons. Using data obtained from double-sided VSD imaging we analyzed neuronal dynamics in both sensory processing and generation of behavior and constructed functional maps for identification of neurons contributing to these processes.

VScope is a software package for the acquisition and analysis of data from multiple cameras as well as electrophysiology. Its main intended purpose is to record fluorescent traces from neurons loaded with voltage-sensitive dyes along with associated electrophysiology. VScope can record simultaneously from any number of cameras and frame rates can be set independently for each camera. VScope can also record synchronized electrophysiology traces and digital channels. VScope is mainly intended to acquire fixed-duration trials, but it can also acquire electrophysiology continuously. A variety of electrical stimulation protocols can be created through an easy-to-use GUI.

The “local bend response” of the medicinal leech (Hirudo verbana) is a stimulus–response pathway that enables the animal to bend away from a pressure stimulus applied anywhere along its body. The neuronal circuitry that supports this behavior has been well described, and its responses to individual stimuli are understood in quantitative detail. We probed the local bend system with pairs of electrical stimuli to sensory neurons that could not logically be interpreted as a single touch to the body wall and used multiple suction electrodes to record simultaneously the responses in large numbers of motor neurons. In all cases, responses lasted much longer than the stimuli that triggered them, implying the presence of some form of positive feedback loop to sustain the response. When stimuli were delivered simultaneously, the resulting motor neuron output could be described as an evenly weighted linear combination of the responses to the constituent stimuli. However, when stimuli were delivered sequentially, the second stimulus had greater impact on the motor neuron output, implying that the positive feedback in the system is not strong enough to render it immune to further input.

Multielectrode arrays (MEAs) allow for acquisition of multisite electrophysiological activity with submillisecond temporal resolution from neural preparations. The signal to noise ratio from such arrays has recently been improved by substrate perforations that allow negative pressure to be applied to the tissue; although, such arrays are not optically transparent, limiting their potential to be combined with optical-based technologies. We present here multi–suction electrode arrays (MSEAs) in quartz that yield a substantial increase in the detected number of units and unit signal to noise ratio from mouse cortico-hippocampal slices and mouse retina explants. This enables the visualization of stronger cross correlations between the firing rates of the various sources. Additionally, the MSEA’s transparency allows us to record voltage sensitive dye activity from a leech ganglion with single neuron resolution using widefield microscopy simultaneously with the electrode array recordings. The combination of enhanced electrical signals and compatibility with optical-based technologies should make the MSEA a valuable tool for investigating neuronal circuits.

The medicinal leech (genus Hirudo) is a classic model animal in systems neuroscience. The leech has been central to many integrative studies that establish how properties of neurons and their interconnections give rise to the functioning of the animal at the behavioral level. Leeches exhibit several discrete behaviors (such as crawling, swimming and feeding) that are each relatively simple. Importantly, these behaviors can all be studied – at least at a basal level – in the isolated nervous system. The leech nervous system is particularly amenable to such studies because of its distributed nature; sensory processing and generation of behavior occur to a large degree in iterated segmental ganglia that each contain only ∼400 neurons. Furthermore, the neurons are relatively large and are arranged with stereotyped topography on the surface of the ganglion, which greatly facilitates their identification and accessibility. This Commentary provides an overview of recent work on the leech nervous system, with particular focus on circuits that underlie leech behavior. Studies that combine the unique features of the leech with modern optical and genetic techniques are also discussed. Thus, this Commentary aims to explain the continued appeal of the leech as an experimental animal in the 21st century.

While moving through their environment, medicinal leeches stop periodically and wave their head or body back and forth. This activity has been previously described as two separate behaviors: one called ‘head movement’ and another called ‘body waving’. Here, we report that these behaviors exist on a continuum, and provide a detailed description of what we now call ‘scanning’. Scanning-related behavior has been thought to be involved in orientation; its function has never before been assessed. While previous studies suggested an involvement of scanning in social behavior, or sucker placement, our behavioral studies indicate that scanning is involved in orienting the leech towards prey stimuli. When such stimuli are present, scanning behavior is used to re-orient the leech in the direction of a prey-like stimulus. Scanning, however, occurs whether or not prey is present, but in the presence of prey-like stimuli scanning becomes localized to the stimulus origin. Most likely, this behavior helps the leech to gain a more detailed picture of its prey target. The display of scanning, regardless of the presence or absence of prey stimuli, is suggestive of a behavior that is part of an internally driven motor program, which is not released by the presence of sensory stimuli. The data herein include first steps to understanding the neural mechanisms underlying this important behavior.

Many of today’s radiofrequency-emitting devices in telecommunication, telemedicine, transportation safety, and security/military applications use the millimeter-wave (MMW) band (30-300 GHz). To evaluate the biological safety and possible applications of this radiofrequency band for neuroscience and neurology, we have investigated the physiological effects of low-intensity 60 GHz electromagnetic irradiation on individual neurons in the leech midbody ganglia. We applied incident power densities of 1, 2, and 4 mW/cm2 to the whole ganglion for a period of 1 minute, while recording the action potential with a standard sharp-electrode electrophysiology setup. For comparison, the recognized U.S. safe exposure limit is 1 mW/cm2 for 6 minutes. During the exposure to MMWs and gradual bath heating at a rate of 0.04 ºC/sec (2.4 ºC/min), the ganglionic neurons exhibited similar dose-dependent hyperpolarization of the plasma membrane and decrease in the action potential amplitude. However, narrowing of the action potential half-width during MMW irradiation at 4 mW/cm2 was 5 times more pronounced, as compared to equivalent bath heating of 0.6 ºC. Even more dramatic difference in the effects of MMW irradiation and bath heating was on the firing rate, which was suppressed at all applied MMW power densities and was increased in a dose-dependent manner during gradual bath heating. The mechanism of enhanced narrowing of action potentials and suppressed firing by MMW irradiation, as compared to gradual bath heating, is hypothesized to involve specific coupling of MMW energy with the neuronal plasma membrane.

Creating visually pleasing graphs in data visualization programs such as Matlab is surprisingly challenging. One common problem is that the positions and sizes of non-data elements such as textual annotations must typically be specified in either data coordinates or in absolute paper coordinates, whereas it would be more natural to specify them using a combination of these coordinate systems. I propose a framework in which it is easy to express, e.g., "this label should appear 2~mm to the right of the data point at (3, 2)" or "this arrow should point to the datum at (2, 1) and be 5 mm long." I describe an algorithm for the correct layout of graphs of arbitrary complexity with automatic axis scaling within this framework. An implementation is provided in the form of a complete 2D plotting package that can be used to produce publication-quality graphs from within Matlab or Octave.

The medicinal leech, Hirudo verbana, is an aquatic predator that utilizes water waves to locate its prey. However, to reach their prey, the leeches must move within the same water that they are using to sense prey. This requires that they either move ballistically towards a pre-determined prey location or that they account for their self-movement and continually track prey. We found that leeches do not localize prey ballistically. Instead, they require continual sensory information to track their prey. Indeed, in the event that the prey moves, leeches will approach the prey’s new location. While leeches need to continually sense water disturbances to update their percept of prey location, their own behavior is discontinuous—approaching prey involves switching between swimming, crawling and non-locomoting. Each of these behaviors may allow for different sensory capabilities and may require different sensory filters. Here, we examined the sensory capabilities of leeches during each of these behaviors. We found that while one could expect the non-locomoting phases to direct subsequent behaviors, crawling phases were more effective than non-locomotor phases for providing direction. During crawling bouts, leeches adjusted their heading so as to become more directed towards the stimulus. This was not observed during swimming. Furthermore, in the presence of prey-like stimuli, leeches crawled more often and for longer periods of time.

Microfluidic and optical sensing platforms are commonly fabricated in glass and fused silica (quartz) because of their optical transparency and chemical inertness. Hydrofluoric acid (HF) solutions are the etching media of choice for deep etching into silicon dioxide substrates, but processing schemes become complicated and expensive for etching times greater than 1 hour due to the aggressiveness of HF migration through most masking materials. We present here etching into fused silica more than 600 μm deep while keeping the substrate free of pits and maintaining a polished etched surface suitable for biological imaging. We utilize an HF-resistant photosensitive resist (HFPR) which is not attacked in 49% HF solution. Etching characteristics are compared for substrates masked with the HFPR alone and the HFPR patterned on top of Cr/Au and polysilicon masks. We used this etching process to fabricate suspended fused silica membranes, 8–16 μm thick, and show that imaging through the membranes does not negatively affect image quality of fluorescence microscopy of biological tissue. Finally, we realize small through-pore arrays in the suspended membranes. Such devices will have applications in planar electrophysiology platforms, especially where optical imaging is required.

Neuroscience research increasingly relies on optical methods for evoking neuronal activity as well as for measuring it, making bright and stable light sources critical building blocks of modern experimental setups. This paper presents a method to control the brightness of a high-power light emitting diode (LED) light source to an unprecedented level of stability. By continuously monitoring the actual light output of the LED with a photodiode and feeding the result back to the LED’s driver by way of a proportional-integral controller, drift was reduced to as little as 0.007% per hour over a 12-h period, and short-term fluctuations to 0.005% root-mean-square over 10 seconds. The LED can be switched on and off completely within 100 µs, a feature that is crucial when visual stimuli and light for optical recording need to be interleaved to obtain artifact-free recordings. The utility of the system is demonstrated by recording visual responses in the central nervous system of the medicinal leech Hirudo verbana using voltage sensitive dyes.

Medicinal leeches, like many aquatic animals, use water disturbances to localize their prey, so they need to be able to determine if a wave disturbance is created by prey or by another source. Many aquatic predators perform this separation by responding only to those wave frequencies representing their prey. Since leeches’ prey preference changes over the course of development, we examined their responses at three different life stages. We found that juveniles more readily localize wave sources of lower frequencies (2 Hz) than their adult counterparts (8–12 Hz), and that adolescents exhibited elements of both juvenile and adult behavior, readily localizing sources of both frequencies. Leeches are known to be able to localize the source of waves through the use of either mechanical or visual information. We separately characterized their ability to localize various frequencies of stimuli using unimodal cues. Within a single modality, the frequency response curves of adults and juveniles were virtually indistinguishable. However, the differences between the responses for each modality—visual and mechanosensory—were striking. The optimal visual stimulus had a much lower frequency (2 Hz) than the optimal mechanical stimulus (12 Hz), frequencies that matched, respectively, the juvenile and the adult preferred frequency for multimodally sensed waves. This suggests that in the multimodal condition, adult behavior is driven more by mechanosensory information and juvenile behavior more by visual. Indeed, when stimuli of the two modalities were placed in conflict with one another, adult leeches, unlike juveniles, were attracted to the mechanical stimulus much more strongly than to the visual stimulus.

Medicinal leeches (Hirudo spp.) swim using a metachronal, front-to-back undulation. The behavior is generated by central pattern generators (CPGs) distributed along the animal’s midbody ganglia and is coordinated by both central and peripheral mechanisms. Here we report that a component of the venom of Conus imperialis, alpha-conotoxin ImI, known to block nicotinic acetylcholine receptors in other species, disrupts swimming. Leeches injected with the toxin swam in circles with exaggerated dorsoventral bends and reduced forward velocity. Fictive swimming in isolated nerve cords was even more strongly disrupted, indicating that the toxin targets the CPGs and central coordination, while peripheral coordination partially rescues the behavior in intact animals.

BACKGROUND: Medicinal leeches (Hirudo spp.) are simultaneous hermaphrodites. Mating occurs after a stereotyped twisting and oral exploration that result in the alignment of the male and/or female gonopores of one leech with the complementary gonopores of a partner. The neural basis of this behavior is presently unknown and currently impossible to study directly because electrophysiological recording techniques disrupt the behavior. RESULTS: Here we report that (Arg(8))-conopressin G and two other members of the oxytocin/vasopressin family of peptide hormones induce in Hirudo verbana a sequence of behaviors that closely mimic elements of spontaneous reproductive behavior. Through a series of progressively more reduced preparations, we show that one of these behaviors, a stereotyped twisting that is instrumental in aligning gonopores in preparation for copulation, is the product of a central pattern generator that consists of oscillators in ganglia M5 and M6 (the ganglia in the reproductive segments of the leech), and also in ganglion M4, which was not previously known to play a role in reproductive behavior. We find that the behavior is periodic, with a remarkably long cycle period of around five minutes, placing it among the slowest behavioral rhythms (other than diurnal and annual rhythms) yet described. CONCLUSION: These results establish the leech as a new model system for studying aspects of the neuronal basis of reproductive behavior.

To quantify locomotory behavior, tools for determining the location and shape of an animal’s body are a first requirement. Video recording is a convenient technology to store raw movement data, but extracting body coordinates from video recordings is a nontrivial task. The algorithm described in this paper solves this task for videos of leeches or other quasi-linear animals in a manner inspired by the mammalian visual processing system: the video frames are fed through a bank of Gabor filters, which locally detect segments of the animal at a particular orientation. The algorithm assumes that the image location with maximal filter output lies on the animal’s body and traces its shape out in both directions from there. The algorithm successfully extracted location and shape information from video clips of swimming leeches, as well as from still photographs of swimming and crawling snakes. A Matlab implementation with a graphical user interface is available online, and should make this algorithm conveniently usable in many other contexts.

Changing gain in a neuronal system has important functional consequences, but the underlying mechanisms have been elusive. Models have suggested a variety of neuronal and systems properties to accomplish gain control. Here, we show that the gain of the neuronal network underlying local bending behavior in leeches depends on widespread inhibition. Using behavioral analysis, intracellular recordings, and voltage-sensitive dye imaging, we compared the effects of blocking just the known lateral inhibition with blocking all GABAergic inhibition. This revealed an additional source of inhibition, which was widespread and increased in proportion to increasing stimulus intensity. In a model of the input/output functions of the three-layered local bending network, we showed that inhibiting all interneurons in proportion to the stimulus strength produces the experimentally observed change in gain. This relatively simple mechanism for controlling behavioral gain could be prevalent in vertebrate as well as invertebrate nervous systems.

^* These three authors contributed equally to this work.

Light-activated ion channels provide a precise and noninvasive optical means for controlling action potential firing, but the genes encoding these channels must first be delivered and expressed in target cells. Here we describe a method for bestowing light sensitivity onto endogenous ion channels that does not rely on exogenous gene expression. The method uses a synthetic photoisomerizable small molecule, or photoswitchable affinity label (PAL), that specifically targets K(+) channels. PALs contain a reactive electrophile, enabling covalent attachment of the photoswitch to naturally occurring nucleophiles in K(+) channels. Ion flow through PAL-modified channels is turned on or off by photoisomerizing PAL with different wavelengths of light. We showed that PAL treatment confers light sensitivity onto endogenous K(+) channels in isolated rat neurons and in intact neural structures from rat and leech, allowing rapid optical regulation of excitability without genetic modification.

Recurring patterns of neural activity, a potential substrate of both information transfer and transformation in cortical networks, have been observed in the intact brain and in brain slices. Do these patterns require the inherent cortical microcircuitry of such preparations or are they a general property of self-organizing neuronal networks? In networks of dissociated cortical neurons—which lack evidence of the intact brain’s intrinsic cortical architecture—we have observed a robust set of spontaneously repeating spatiotemporal patterns of neural activity, using a template-matching algorithm that has been successful both in vivo and in brain slices. The observed patterns in cultured monolayer networks are stable over minutes of extracellular recording, occur throughout the culture’s development, and are temporally precise within milliseconds. The identification of these patterns in dissociated cultures opens a powerful methodological avenue for the study of such patterns, and their persistence despite the topological and morphological rearrangements of cellular dissociation is further evidence that precisely timed patterns are a universal emergent feature of self-organizing neuronal networks.

Three remarkable features of the nervous system—complex spatiotemporal patterns, oscillations, and persistent activity—are fundamental to such diverse functions as stereotypical motor behavior, working memory, and awareness. Here we report that cultured cortical networks spontaneously generate a hierarchical structure of periodic activity with a strongly stereotyped population-wide spatiotemporal structure demonstrating all three fundamental properties in a recurring pattern. During these “superbursts,” the firing sequence of the culture periodically converges to a dynamic attractor orbit. Precursors of oscillations and persistent activity have previously been reported as intrinsic properties of the neurons. However, complex spatiotemporal patterns that are coordinated in a large population of neurons and persist over several hours—and thus are capable of representing and preserving information—cannot be explained by known oscillatory properties of isolated neurons. Instead, the complexity of the observed spatiotemporal patterns implies large-scale self-organization of neurons interacting in a precise temporal order even in vitro, in cultures usually considered to have random connectivity.

BACKGROUND:

We have collected a comprehensive set of multi-unit data on dissociated cortical cultures. Previous studies of the development of the electrical activity of dissociated cultures of cortical neurons each focused on limited aspects of its dynamics, and were often based on small numbers of observed cultures. We followed 58 cultures of different densities—3000 to 50,000 neurons on areas of 30 to 75 mm²—growing on multi-electrode arrays (MEAs) during the first five weeks of their development.

RESULTS:

Plating density had a profound effect on development. While the aggregate spike detection rate scaled linearly with density, as expected from the number of cells in proximity to electrodes, dense cultures started to exhibit bursting behavior earlier in development than sparser cultures. Analysis of responses to electrical stimulation suggests that axonal outgrowth likewise occurred faster in dense cultures. After two weeks, the network activity was dominated by population bursts in most cultures. In contrast to previous reports, development continued with changing burst patterns throughout the observation period. Burst patterns were extremely varied, with inter-burst intervals between 1 and 300 s, different amounts of temporal clustering of bursts, and different firing rate profiles during bursts. During certain stages of development bursts were organized into tight clusters with highly conserved internal structure.

CONCLUSIONS:

Dissociated cultures of cortical cells exhibited a much richer repertoire of activity patterns than previously reported. Except for the very sparsest cultures, all cultures exhibited globally synchronized bursts, but bursting patterns changed over the course of development, and varied considerably between preparations. This emphasizes the importance of using multiple preparations—not just multiple cultures from one preparation—in any study involving neuronal cultures.

These results are based on 963 half-hour-long recordings. To encourage further investigation of the rich range of behaviors exhibited by cortical cells in vitro, we are making the data available to other researchers, together with Matlab code to facilitate access.

We attempted to induce functional plasticity in dense cultures of cortical cells using stimulation through extracellular electrodes embedded in the culture dish substrate (multi-electrode arrays, or MEAs). We looked for plasticity expressed in changes in spontaneous burst patterns, and in array-wide response patterns to electrical stimuli, following several induction protocols related to those used in the literature, as well as some novel ones. Experiments were performed with spontaneous culture-wide bursting suppressed by either distributed electrical stimulation or by elevated extracellular magnesium concentrations as well as with spontaneous bursting untreated. Changes concomitant with induction were no larger in magnitude than changes that occurred spontaneously, except in one novel protocol in which spontaneous bursts were quieted using distributed electrical stimulation.

We constructed a simulated spiking neural network model to investigate the effects of random background stimulation on the dynamics of network activity patterns and tetanus-induced network plasticity. The simulated model was a “leaky integrate-and-fire” (LIF) neural model with spike timing–dependent plasticity (STDP) and frequency-dependent synaptic depression. Spontaneous and evoked activity patterns were compared with those of living neuronal networks cultures on multi-electrode arrays. To help visualize activity patterns and plasticity in our simulated model, we introduced new population measures called Center of Activity (CA) and Center of Weights (CW) to describe the spatio-temporal dynamics of network-wide firing activity and network-wide synaptic strength, respectively. Without random background stimulation, the network synaptic weights were unstable and often drifted after tetanization. In contrast, with random background stimulation, the network synaptic weights remained close to their values immediately after tetanization. The simulation suggests that the effects of tetanization on network synaptic weights were difficult to control because of ongoing synchronized spontaneous bursts of action potentials, or “barrages.” Random background stimulation helped maintain network synaptic stability after tetanization by reducing the number and thus the influence of spontaneous barrages. We used our simulated network to model the interaction between ongoing neural activity, external stimulation and plasticity, and to guide our choice of sensory-motor mappings for adaptive behavior in hybrid neural-robotic systems or “hybrots.”

A leucine to alanine substitution (L9’A) was introduced in the M2 region of the mouse 4 neuronal nicotinic acetylcholine receptor (nAChR) subunit. Expressed in Xenopus oocytes, α4(L9’A) 2 nAChRs were ≥30-fold more sensitive than wild type (WT) to both ACh and nicotine. We generated knock-in mice with the L9 A mutation and studied their cellular responses, seizure phenotype, and sleep-wake cycle. Seizure studies on α4-mutated animals are relevant to epilepsy research because all known mutations linked to autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE) occur in the M2 region of α4 or β2 subunits. Thalamic cultures and synaptosomes from L9’A mice were hypersensitive to nicotine-induced ion flux. L9’A mice were ~15-fold more sensitive to seizures elicited by nicotine injection than their WT littermates. Seizures in L9’A mice differed qualitatively from those in WT: L9’A seizures started earlier, were prevented by nicotine pretreatment, lacked EEG spike-wave discharges, and consisted of fast repetitive movements. Nicotine-induced seizures in L9 A mice were partial, whereas WT seizures were generalized. When L9’A homozygous mice received a 10 mg/kg nicotine injection, there was temporal and phenomenological separation of mutant and WT-like seizures: an initial seizure ~20 s after injection was clonic and showed no EEG changes. A second seizure began 3–4 min after injection, was tonic-clonic, and had EEG spike-wave activity. No spontaneous seizures were detected in L9 A mice during chronic video/EEG recordings, but their sleep-wake cycle was altered. Our findings show that hypersensitive α4^* nicotinic receptors in mice mediate changes in the sleep-wake cycle and nicotine-induced seizures resembling ADNFLE.

One of the major modes of activity of high-density cultures of dissociated neurons is globally synchronized bursting. Unlike in vivo, neuronal ensembles in culture maintain activity patterns dominated by global bursts for the lifetime of the culture (up to two years). We hypothesize that persistence of bursting is due to a lack of input from other brain areas. To study this hypothesis, we grew small but dense monolayer cultures of cortical neurons and glia from rat embryos on multi-electrode arrays (MEAs), and used electrical stimulation to substitute for afferents. We quantified the burstiness of the cultures’ firing in spontaneous activity and during several stimulation protocols. While slow stimulation through individual electrodes increased burstiness due to burst entrainment, rapid stimulation reduced burstiness. Distributing stimuli across several electrodes, as well as continuously fine-tuning stimulus strength with closed-loop feedback, greatly enhanced burst control. We conclude that externally applied electrical stimulation can substitute for natural inputs to cortical neuronal ensembles in transforming burst-dominated activity to dispersed spiking, more reminiscent of the awake cortex in vivo. This non-pharmacological method of controlling bursts will be a critical tool for exploring the information processing capacities of neuronal ensembles in vitro, and has potential applications for the treatment of epilepsy.

We evolved multiple clones of populations of Digitalia, a type of digital organism, to study the effects of chance, history, and adaptation in evolution. We show that clones adapted to a specific environment can adapt to new environments quickly and efficiently, although their history remains a significant factor in their fitness. Adaptation is most significant (and the effects of history less so) if the old and new environments are dissimilar. For more similar environments, adaptation is slower while history is more prominent. For both similar and dissimilar transfer environments, populations quickly lose the capability to perform computations (the analog of beneficial chemical reactions) that are no longer rewarded in the new environment. Populations that developed few computational “genes” in their original environment were unable to acquire them in the new environment.

This article appeared previously in modified form in the 2000 Proceedings of the 7th International Conference on Artificial Life, M. A. Bedau, J. S. McCaskill, N. H. Packard, & S. Rasmussen (eds), pp 216-220. MIT Press, Cambridge, MA. [Full text available]

Over the last few decades, technology to record through ever increasing numbers of electrodes has become available to electrophysiologists. For the study of distributed neural processing, however, the ability to stimulate through equal numbers of electrodes, and thus to attain bidirectional communication, is of paramount importance. Here, we present a stimulation system for multi-electrode arrays that interfaces with existing commercial recording hardware, and allows stimulation through any electrode in the array, with rapid switching between channels. The system is controlled through real-time Linux, making it extremely flexible: Stimulation sequences can be constructed on-the-fly, and arbitrary stimulus waveforms can be used if desired. A key feature of this design is that it can readily and inexpensively be reproduced in other labs, since it interfaces to standard PC parallel ports and uses only off-the-shelf components. Moreover, adaptation for use with in-vivo multi-electrode probes would be straightforward. In combination with our freely available data acquisition software, MeaBench, this system can provide feedback stimulation in response to recorded action potentials within 15 ms.

Electrical stimulation through multi-electrode arrays is used to evoke activity in dissociated cultures of cortical neurons. We study the efficacies of a variety of pulse shapes under voltage- as well as current-control, and determine useful parameter ranges that optimize efficacy while preventing damage through electrochemistry. For any pulse shape, stimulation is found to be mediated by negative currents. We find that positive-then-negative biphasic voltage-controlled pulses are more effective than any of the other pulse shapes tested, when compared at the same peak voltage. These results suggest that voltage-control, with its inherent control over limiting electrochemistry, may be advantageous in a wide variety of stimulation scenarios, possibly extending to in vivo experiments.

We describe an algorithm for suppression of stimulation artifacts in extracellular micro-electrode array (MEA) recordings. A model of the artifact based on locally fitted cubic polynomials is subtracted from the recording, yielding a flat baseline amenable to spike detection by voltage thresholding. The algorithm, SALPA, reduces the period after stimulation during which action potentials cannot be detected by an order of magnitude, to less than 2 ms. Our implementation is fast enough to process 60-channel data sampled at 25 kHz in real-time on an inexpensive desktop PC. It performs well on a wide range of artifact shapes without re-tuning any parameters, because it accounts for amplifier saturation explicitly and uses a statistic to verify successful artifact suppression immediately after the amplifiers become operational. We demonstrate the algorithm’s effectiveness on recordings from dense monolayer cultures of cortical neurons obtained from rat embryos. SALPA opens up a previously inaccessible window for studying transient neural oscillations and precisely timed dynamics in short-latency responses to electric stimulation.

The brain is perhaps the most advanced and robust computation system known. We are creating a method to study how information is processed and encoded in living cultured neuronal networks by interfacing them to a computer-generated animal, the Neurally-Controlled Animat, within a virtual world. Cortical neurons from rats are dissociated and cultured on a surface containing a grid of electrodes (multi-electrode arrays, or MEAs) capable of both recording and stimulating neural activity. Distributed patterns of neural activity are used to control the behavior of the Animat in a simulated environment. The computer acts as its sensory system providing electrical feedback to the network about the Animat’s movement within its environment. Changes in the Animat’s behavior due to interaction with its surroundings are studied in concert with the biological processes (e.g., neural plasticity) that produced those changes, to understand how information is processed and encoded within a living neural network. Thus, we have created a hybrid real-time processing engine and control system that consists of living, electronic, and simulated components. Eventually this approach may be applied to controlling robotic devices, or lead to better real-time silicon-based information processing and control algorithms that are fault tolerant and can repair themselves.

We culture high-density cortical cultures on multi-electrode arrays (MEAs), which allow us to stimulate and record from thousands of neurons. One of the modes of activity in these high-density cultures is dish-wide synchronized bursting. Unlike in vivo, these synchronized patterns persist for the lifetime of the culture. Such aberrant patterns of activity might be due to the fact that cortical cultures are sensory-deprived and arrested in development. We have devised methods to control this spontaneous activity by multi-electrode electrical stimulation and to study long-term functional neural plasticity, on a background of such burst-quieting stimulation. Here, we investigate whether burst quieting reveals long-term plasticity induced by tetanic stimulation. Spatio-temporal activity patterns (STAPs) that result from probe pulses were clustered and quantified in quieted and non-quieted cultures. Burst-quieted cultures show more tetanus-induced functional change than cultures which are allowed to express spontaneous bursts. The methods developed for this study will help in the understanding of network dynamics and appreciation of their role in long-term plasticity and information processing in the brain.

We present a software suite, MEABench, for data acquisition and online analysis of multi-electrode recordings, especially from micro-electrode arrays. Besides controlling data acquisition hardware, MEABench includes algorithms for real-time stimulation, artifact supression and spike detection, as well as programs for online display of voltage traces from 60 electrodes and continuously updated spike raster plots. MEABench features real-time output streaming, allowing easy integration with stimulator systems. We have been able to generate stimulation sequences in response to live neuronal activity with less than 20 ms lag time. MEABench is open-source software, and is available for free public download at http://www.its.caltech.edu/~pinelab/meabench.html.

We are developing new tools to study the computational properties of living neuronal networks. We are especially interested in the collective, emergent properties at the mesoscopic scale (Freeman 2000) of thousands of brain cells working together to learn, process information, and to control behavior. We grow dissociated monolayer mammalian cortical cultures on multi-electrode arrays. We created the electronics and software necessary for a real-time feedback loop that allows the neurons to trigger their own stimulation. A key part of this loop is a system for re-embodying the in vitro network. We use the neural activity to control either simulated animals (animats) or robots. By using networks of a few thousand neurons and glia, we have tremendous access to the cells, not feasible in vivo. This allows physical and pharmacological manipulation, and continuous imaging at the millisecond and micron scales, to determine the cell- and network-level morphological correlates of learning and memory. We also model the cultured network in software; This helps direct our experiments, which then improves the model. By combining small networks of real brain cells, computer simulations, and robotics into new hybrid neural microsystems (which we call Hybrots), we hope to determine which neural properties are essential for the kinds of collective dynamics that might be used in artificially intelligent systems.

We evolved multiple clones of populations of Digitalia, a type of digital organism, to study the effects of chance, history, and adaptation in evolution. We show that clones adapted to a specific environment can adapt to new environments quickly and efficiently, although their history remains a significant factor in their fitness. Adaptation is most significant (and the effects of history less so) if the old and new environments are dissimilar. For more similar environments, adaptation is slower while history is more prominent. For both similar and dissimilar transfer environments, populations quickly lose the capability to perform computations (the analog of beneficial chemical reactions) that are no longer rewarded in the new environment. Populations that developed few computational “genes” in their original environment were unable to acquire them in the new environment.

Although reward and punishment have opposite valence and promote approach and avoidance respectively, both stimuli can enhance arousal and capture attention to guide appropriate behavioral responses. We have previously demonstrated that dopamine (DA) neurons in the dorsal raphe nucleus (DRN), which project densely to the extended amygdala, are wake-active over sleep states, promote arousal, and show robust activation to salient stimuli, irrespective of their hedonic valence [1]. We further examined whether DRN-DA neurons encode motivational salience [2] (n = 6 TH-Cre mice) with fiber photometry [3]. GCaMP6f-expressing mice were subjected to fear memory acquisition and extinction, where sensory cues of originally neutral context gain and lose motivational salience. Before conditioning, DRN-DA neurons showed small activation to the novel sensory cues (conditioned stimuli, CS; house-light and 65 dB 5 kHz tone) and no significant change across repetitive exposures (Pearson’s r = -0.12, p = 0.75). Throughout learning and repeated pairings (10 trials, random ITIs within [70, 110] seconds) of CS and electric footshock (unconditioned stimuli, US; 0.6 mA for 1 sec), DRN-DA neurons gradually developed phasic response to the CS (Pearson’s r = 0.78, p < 0.01). As previously shown [1], US induced robust activation of DRN-DA neurons, which however decreased across repeated exposures (Pearson’s r = -0.90, p < 0.001). In contrast to the learning phase, evoked response to the CS diminished across extinction trials, where sensory cues lost motivational salience (Pearson’s r = -0.83, p < 0.0001). These findings suggest that phasic activation of DRN-DA groups encodes motivational salience and that they are modulated by expectations at the population level. To discriminate between projection-specific functions at the single-cell level, we built a two-photon microscope for deep brain imaging of projection-identified DRN-DA cell bodies during appetitive and aversive conditioning tasks. Taken together, these data deepen our understanding of the functional properties of DRN-DA neurons. Reference: [1] Cho JR, Treweek JB, Robinson JE, Xiao C, Bremner LR, Greenbaum A, Gradinaru V. (2017) Dorsal raphe dopamine neurons modulate arousal and promote wakefulness by salient stimuli. Neuron in press. [2] Bromberg-Martin ES, Matsumoto M, Hikosaka O. (2010) Dopamine in motivational control: rewarding, aversive, and alerting. Neuron 68, 815-834. [3] Lerner TN, Shilyansky C, Davidson TJ, Evans KE, Beier KT, Zalocusky KA, Crow AK, Malenka RC, Luo L, Tomer R, Deisseroth K. (2015) Intact-brain analyses reveal distinct information carried by SNc dopamine subcircuits. Cell 162, 635-647.

Sensitivity to water waves in aquatic predators greatly facilitates prey location. The behavioral response to visual and mechanical information from water waves is well documented in the leech Hirudo verbana: in response to low-amplitude water waves, they orient themselves and initiate swimming or crawling toward the source of the waves. Here, we begin to quantitatively characterize the neuronal response patterns of the leech to water waves in terms of visual and mechanoreceptive sensitivity. We recorded activity of the S cell, an interneuron that forms a syncytium connecting all midbody ganglia. The S cell is excited by mechanical and visual stimuli and could be critical to coordinating responses across the whole body. Although leeches are behaviorally capable of discerning the direction of water waves, we found that the magnitude of the S-cell response to mechanically cued water waves (in darkness) does not depend on the direction of waves relative to the leech’s orientation: Mechanical waves presented toward the head, toward the tail, and laterally all evoked the same number of action potentials. Remarkably, regardless of wave direction, most action potentials propagated retrogradely (tail-to-head) along the nerve cord, but when waves approached a leech head-first, an initial burst of action potentials propagated anterogradely. The S cell responded to a wide range of wave frequencies, and its sensitivity profile contained multiple peaks. Leeches’ responses to water waves have been linked to hunger state, but here we found that feeding did not immediately impact the mechanosensory response of the S cell. However, sensitivity to high-amplitude, low-frequency waves increased in the weeks following feeding. We explored the leech’s visual response pathways by projecting visual waves in the absence of mechanical cues. In contrast to the high sensitivity to mechanical wave frequency, the magnitude of S-cell responses to bright visual waves was not strongly dependent on wave frequency, although waves with wavelengths longer than the leech’s body evoked more sustained responses. For visual as well as mechanical waves, a majority of action potentials traveled tail to head. Initial bursts of anterograde spike propagation were occasionally seen, but for visual waves were found to be independent of stimulus direction. We are currently exploring the cause of the predominance of retrograde spike propagation in the S-cell system. Initial results indicate that retrograde propagation persists in the absence of the tail brain and even in the absence of most posterior ganglia.

Sensory processing and motor control are functions of intricately organized neuronal ensembles. Therefore, their study could greatly benefit from technologies that simultaneously record the activities of large numbers of individual neurons. We have developed a double-sided microscope for voltage-sensitive dye (VSD) imaging, and used it to record from the medicinal leech Hirudo verbana that is a classic experimental animal for comprehensive analysis of the function of neural circuits. This optical measurement technique realized neuronal recording from the majority of cell bodies located in both ventral and dorsal surfaces of a leech ganglion at high temporal resolution: The dye VoltageFluor VF2.1(OMe).H enabled recording of individual action potentials and excitatory/inhibitory synaptic potentials even in many small cells. By removing vibrational noise from the CCD cameras and algorithmically correcting motion artifacts and global trends, we were able to reconstruct stereotyped electrical activity associated with several leech behaviors. To demonstrate the applicability of these newly developed VSD imaging methods, we addressed the following two questions. (1) How does a leech ganglion discriminate the location of a pressure stimulus based on the population activity of neurons on both surfaces of the ganglion? Using principal component analysis, we demonstrate that neuronal population activity differs markedly depending on which sensory neuron (P cell) was stimulated. (2) To what extent are neural circuit components unique or shared between different behaviors? We simultaneously recorded from nearly all neurons in a ganglion during three fictive behaviors: swimming, crawling and local bending. Coherence analysis of rhythmic activity helped us improve our understanding of the involvement of all ganglionic neurons in each behavior. Using double-sided imaging allowed us, for the first time, to directly analyze functional relationships between neurons on the ventral and dorsal surfaces of the ganglion. Taken together, double-sided VSD imaging is a promising new method for observation of neuronal population dynamics for identification of individual neurons responsible for multiple behaviors, and for the analysis of functional relationships between those neurons.

Vision requires complex integration mechanisms. Investigating those is generally difficult because of the large number of neurons involved and the challenge of identifying specific neurons. Among animals that express complex visual processing mechanisms, the leech Hirudo verbana is a rare example in which all sensory neurons can be readily identified. Hirudo’s previously documented ability to detect the direction of water waves demonstrated that its visual system has the ability to process spatiotemporal patterned visual stimuli. Thus far, however, little is known about the processing mechanisms of the leech’s visual system, which is composed of several pigmented eyes in the head and photosensitive non-pigmented sensilla that are distributed across its entire body. While several interneurons are known to respond to visual stimuli, their response properties remain poorly understood. Among these, the S-cell system is especially intriguing because it is multimodal and spans the entire body of the leech. Thus it could potentially be involved in complex sensory integration. Using a nearly intact leech preparation, and stimulation protocols that tightly controlled area and spectral content of the stimulus as well as background light conditions, we measured the sensitivity of the S-cell system to light stimuli, tested if it performs spatial integration, and whether adaptation is local or global. Two peaks were found in the spectral response of the S-cell system: in green and in UV. The response of the S-cell system was found to be strongly dependent on the size of the area of the leech body that is stimulated. Furthermore, adaptation was local and different areas of the leech’s body contributed differentially to the response depending on their adaptation state. A ‘bleach experiment’ confirmed the recent suggestion that at least two color channels contribute to the response. Their contributions were found to be dependent on the adaptation to background illumination. Taken together, our results show that the response properties of the S-cell system are highly complex, making it an attractive target for future studies of high-level processing of complex visual stimuli in a model animal that is uncommonly amenable to electrophysiological methods.

Studies of neuronal mechanisms of sensory processing and motor control are greatly facilitated by technologies that permit recording of the activity of a large population of neurons at once. The medicinal leech Hirudo verbana is a classic experimental animal for comprehensive analysis of the function of neural circuits. Voltage-sensitive dye (VSD) imaging combined with electrophysiology is an increasingly powerful technology to assist in such studies. One recent dye in particular, the VoltageFluor VF2.1(OMe).H, is sensitive enough to record subthreshold events even in small cells. In the leech ganglion, all cell bodies are located on the surface of a flattened sphere, making all of them potentially accessible to imaging. However, thus far, imaging experiments have been limited to recording from one side of a ganglion. If we were able to simultaneously image the other side, we could record the activity of all the neurons in a ganglion, allowing comprehensive analysis of the neuronal activity in the entire system. We have developed a double-sided microscope to do just this, and compared to last year’s prototype, we greatly improved the ratio of signal to noise by removing vibrational noise from the CCD cameras. From simultaneously imaged neural activity on both (dorsal and ventral) surfaces of isolated leech ganglia, we successfully reconstructed stereotyped electrical activity associated with local bending behavior in response to P-cell stimulation. The technique should extend readily to other situations where researchers require simultaneous recording of activities from two distinct layers of neurons.

One major goal for studying how information is processed by a nervous system is the development of techniques that allow recording the activity of a large number of neurons at once. In this context, the voltage sensitive dye VF2.1.Cl has been proven promising because the dye is very fast, sensitive, and non-toxic. Already, this dye has allowed the simultaneous observation of the activity of most of the neurons on one side of a leech ganglion. To take full advantage of this technique, we now developed a method that allows the simultaneous observation of both surfaces of a ganglion. With our customized double microscope we imaged simultaneously neural activity of both (dorsal and ventral) surfaces of isolated leech ganglia. Optical crosstalk between the layers was substantial, but could be effectively suppressed algorithmically. Stereotyped electrical activity associated with local bending in response to P-cell stimulation could be successfully reconstructed from VSD image sequences from both surfaces simultaneously. The technique should extend readily to other situations in which the activity of two distinct layers of neurons needs to be monitored.

Large-scale multisite electrophysiology recordings with high temporal resolution are essential to discover neural circuitry and elucidate their structure-function relationship. We contribute to this effort by combining multielectrode arrays (MEAs) with through pore arrays in quartz substrates to create multi-suction electrode arrays. The MEA allows for multisite extracellular recordings from neural tissue while the through pore array permits suction to be applied to the tissue to form more intimate contact with the electrodes. We successfully recorded from mouse hippocampi, mouse retina, and leech segmental ganglia. Hippocampus and retina tissue show at least a 50% increase in S/N and twofold increase in detectable spikes following suction. (Interestingly, spiking activity and S/N of spikes in leech ganglia mostly do not increase after applied suction, suggesting sources deeper in the tissue.) Finally, we demonstrate optical imaging through the transparent substrate to visualize the neurons at the electrode interface simultaneously with electrophysiology recordings. This technology will facilitate the combination of optical-based measurements such as voltage-sensitive dye imaging with multisite electrophysiological recordings with high temporal resolution of neuronal networks in a wide range of vertebrate and invertebrate preparations, at the single spike level.

Everyone agrees that good note keeping is essential in science. Yet we often do very little to educate and encourage junior researchers toward good note keeping practices. ELN is a simple and fast electronic lab notebook that, by its very simplicity, does not get in the way of taking notes and makes it very easy to paste in graphics without forcing the user to adhere to arbitrary restrictions. Users have reported taking more extensive and better organized notes after switching to ELN from paper note keeping. ELN notebooks can contain text, tables, and graphics (pasted images or handdrawn within the software). ELN will store pdf copies of articles or other files referred to in the text. ELN features protection against both accidental data loss and intentional post-hoc modification of entries. ELN is open source software that runs on Windows, MacOS and Linux. In combination with standard version control software, ELN notebooks can be accessed from any computer on the internet. ELN is very stable and suitable for daily use. Still, development continues and code contributions are welcomed. Downloads at http:

www.danielwagenaar.net/eln. Live demo at poster.

The medicinal leech Hirudo verbana is an aquatic predator that utilizes water disturbances to localize its prey. To do this it can either use its distributed mechanosensory system comprising an array of nearly 300 receptors sensitive to water movement or its visual system comprising 10 simple eyes in the head and an array of visual sensilla that are colocalized with the mechanoreceptors or even a combination of the two. No matter how the leech senses water disturbances, it must move within the very same water that it uses to sense its prey. How do leeches separate water disturbances induced by prey from water disturbances caused by self-movements? When investigating the answer to this question we found that leeches crawl in a manner directed toward the prey-stimulus, but when they swim they do so essentially blind. Leeches are able to sense and orient toward water movements while crawling, but does crawling behavior increase when a stimulus is present? We noted increases in both the number and duration of crawling bouts when a prey-like stimulus was present. Thus the very presence of sensory information is biasing the leech toward crawling behavior. Crawling behavior can increase in response to stimulation of certain neurons or an increase in dopamine levels. However, which of these mechanisms is involved in increasing crawling when sensory stimuli are present is yet to be elucidated. In the coming months we will attempt to determine how this bias results from changes within the nervous system resulting from sensory stimulation.

The medicinal leech Hirudo verbana is an aquatic predator that utilizes water disturbances to localize its prey. Two sensory systems guide this behavior: a distributed mechanosensory system comprising an array of nearly 300 receptors sensitive to water movement, and a visual system comprising 10 simple eyes in the head and an array of visual sensilla that are colocalized with the mechanoreceptors. In a previous study we determined that leeches identify prey-like stimuli based on the frequency and amplitude of water waves. How do leeches extract directional information from water waves? Furthermore, how do leeches separate water disturbances induced by prey from water disturbances caused by self-movements? The answer to this latter question depends on which mode of locomotion—crawling or swimming—the animal utilizes as it approaches its prey. Our data indicate that crawling is continuously directed by sensory information, whereas during swimming the animal is essentially blind to stimuli. To begin to answer how leeches extract directional information from water waves, we have focused on the respective roles of the eyes in the head and those in the rest of the body. We found that if the head eyes are occluded, the leech fails to localize the source of a visual disturbance. Currently, we are tracing the neuronal projections of individual head eyes and aim to build up a detailed understanding of how signals from the head eyes are combined and used for localizing prey. These and other experiments will contribute to a more complete understanding of locomotor-related decision making.

The medicinal leech is an aquatic predator which uses water disturbances to localize its prey. We have found that it responds preferentially to certain frequencies of stimuli which we believe to represent its prey. Furthermore, as preferred prey change over the course of the leech’s life so does this preferred response band. How is this change accomplished? Leeches are able to sense waves through the use of an array of mechanosensory hairs as well as an array of simple eyes. When examining the capability of these modalities we found that each had its own tuning curve. Responses to visual stimuli reached a peak at much lower frequencies than responses to mechanical stimuli. Interestingly the visual response closely resembled the juvenile tuning curve, and the mechanosensory response closely resembled the tuning curve of adults. We inferred that, rather than shifting the tuning curves over the course of development, leeches modify their reliance on the individual modalities. To test this idea, we challenged juvenile and adult leeches with conflicting stimuli: an optimal visual stimulus in one location, and an optimal mechanical stimulus in another. Indeed, compared to juveniles, adult leeches were attracted to the mechanical stimulus much more strongly than to the visual stimulus. To further understand how this multisensory information is integrated we must understand how information is integrated and processed among sensors within a single modality. Here, we will put most of our focus on the visual modality. Is this information processed in an open loop or closed loop fashion? What properties of visual waves are important? Through the use of projected stimuli we are able to change properties of the visual waves without changing the wave frequency. This allows for us to examine the importance of edges, amplitude, and to ask whether it is the wave’s width or its length that the leech responds to. In addition to these behavioral experiments, we are also beginning to examine information processing among the visual sensors in the brain and ganglia such that we can record neurophysiological responses to such stimuli, a method which will allow us to determine how stimulus direction is encoded. Through using a combination of behavioral and neurophysiological methods we will improve our understanding how this sensory array encodes direction and what aspects of the wave are important to the leech.

Investigations on the biological impact of low levels of millimeter-wave energy date back to the first experiments on the generation and detection of these high frequency signals by Sir Jagadis Chunder Bose at the end of the 19th century. A bit more than a hundred years later, millimeter-wave transmission has become a ubiquitous commercial reality. Despite the widespread use of millimeter-wave transmitters for communications, radar and even non-lethal weapons systems, only a handful of researchers in the USA have funded programs focusing on millimeter-wave interactions with biological systems. As such, there is a growing need for a better understanding of the mechanisms of these interactions and their possible adverse and therapeutic implications. Independent of the health impact of long term exposure to high doses of millimeter-wave energy on whole organisms, there exists the potential for subtle effects on specific tissues or organs which can best be quantified in studies which examine real-time changes in cellular function as energy is applied. In this talk we present a series of experiments that show changes in neuron response with exposure to very low levels of millimeter wave power. The findings have implications for non-contact stimulation and control of neurologic function and might prove useful in a variety of health applications from suppression of peripheral neuropathic pain to the treatment of central neurological disorders.

Medicinal leeches, like many aquatic animals, use water disturbances to localize their prey, so they need to be able to determine if a wave disturbance is created by prey or by another source. Many aquatic predators perform this separation by responding only to those wave frequencies representing their prey. Since leeches’ prey preference changes over the course of development, we examined their responses at three different life stages. We found that juveniles more readily localize wave sources of lower frequencies (2 Hz) than their adult counterparts (8-12 Hz), and that adolescents exhibited elements of both juvenile and adult behavior, readily localizing sources of both low and mid-range frequencies. Leeches are known to be able to localize the source of waves through the use of either mechanical or visual information. We separately characterized their ability to localize various frequencies of stimuli using unimodal cues. For visual or mechanical unimodal stimuli, the frequency response curves of adults and juveniles were virtually indistinguishable. However, the visual and mechanical curves were different from each other: the optimal visual stimulus had a much lower frequency (2 Hz) than the optimal mechanical stimulus (12 Hz), frequencies that matched, respectively, the juvenile and the adult preferred frequency for multimodally sensed waves. This suggests that in the multimodal condition, adult behavior is driven more by mechanosensory information and juvenile behavior more by visual. Indeed, when stimuli of the two modalities were placed in conflict with one another, adult leeches, unlike juveniles, were attracted to the mechanical stimulus much more strongly than to the visual stimulus.

We are studying the neuronal basis of multisensory integration in the medicinal leech, a model animal with a well-characterized and relatively simple nervous system. We aim to discover how the leech’s nervous system combines biologically relevant mechanical and visual sensory cues in order to come to a coherent decision about subsequent motion. One goal is to characterize the neural computation that occurs in the leech ganglion, where processing for most described leech behaviors occurs, while visually and mechanically stimulating the leech preparation. Specifically, we will interrogate all the neurons in the ventral side of the ganglion simultaneously at the single cell level by combining multielectrode array (MEA) recordings with voltage sensitive dye (VSD) imaging. The excellent spatial resolution of the VSDs combined with the temporal resolution of MEAs will provide a high level of detail of this computation in the ganglion. We enhanced the MEA design by fabricating optically transparent multi-suction electrode arrays (MSEAs). Each electrode in an array of 60 is constructed around a microfabricated suction pore. Directed suction holds neurons closer to the electrodes and immobilizes tissue without physical distortion, resulting in more stable recordings. The MSEA fabrication protocol builds upon a previously described batch-fabrication technique to make planar patch-clamp electrodes; thus, an entire array of an arbitrary number of pores and devices can be created simultaneously. We are exploiting this by making devices for our collaborators in parallel with our MSEA devices in order to study network behavior in mouse hippocampus slices, and ion channels in giant unilamellar vesicles (GUVs) and mammalian cells.

We investigate how network dynamics affect interactions between inputs to a neuronal population. We use networks of dissociated cortical neurons which we stimulated electrically through a microelectrode array. This is a model for the interactions between distinct physiological inputs to a neural population in vivo. Interactions of responses of spatially and temporally separated inputs within a population represent a crucial step in the integration of sensory information into a single percept, as well as attentional modulation of information processing. We applied two stimuli from two separated electrodes, and varied the delay between the stimuli upon which we observed varying activity patterns representing different interactions between the two stimuli. We correlated the time dependence of the interaction with different properties of the isolated (i.e., single) responses. To quantify such interaction, we developed a method to evaluate the influence of either stimulus on the response to the pair of stimuli. As possible determinants for interactions we propose the relative size and the time-course of isolated responses. Additionally we quantified the non-linear component of the interaction and we determined the distribution of individual responses. This works aims to further the understanding of neuronal integration of inputs, a topic whose relevance ranges from single-cell short-term plasticity to psychophysics. The relation between stimulus pattern and evoked activity at the network level has been poorly understood, hindering the interpretation of long-term network plasticity studies.

The ability to function within complex natural environments depends on the ability to interpret and encode information about the surroundings. This is particularly true in the case of predators, where survival is dependent on the ability to efficiently and accurately locate prey when the often rare opportunity presents itself. Aquatic predators such as frogs, fish, and leeches find prey by orienting toward disturbances in the water (Elepfandt, 1984; Horridge and Boulton, 1967; Müller and Schwartz, 1982; Young et al., 1981). Our recent findings include that this ability to orient toward water disturbances is frequency dependent. For instance, adult leeches exhibit a higher probability of locating a 6 Hz stimulus than those moving at 1 Hz or at 10 Hz. It would seem that some frequencies are better detected than others. Indeed, neurophysiological recordings have shown that this is the case with both the visual and mechanosensors. Despite that either of those modalities can be used to orient toward waves, (Carlton and McVean, 1993) the visual and mechanosensors respond to different stimulus frequencies. While the visual sensors respond best to low stimulus frequencies, the mechanosensors are much more broadly tuned (Friesen, 1981; Kretz et al., 1976; Peterson, 1984). The possibility exists that these sensory systems alone or in combination may be optimized to function under different conditions or to direct different elements of behavior. To begin to address this issue we are currently examining the leech’s responses when presented with solely visual stimuli. Regardless of the tuning of individual sensors or modalities, the frequency dependence of this behavior suggests that some stimuli are simply more ‘interesting’ than others. A similar case is present in another aquatic predator, the fish Aplocheilus lineatus, which responds specifically to water disturbances representing its prey (Bleckmann and Schwartz, 1982). Interestingly, in leeches, prey preferences change over the course of life (Keim, 1993; Wilkin and Scofield, 1990). This suggests that predatory responses to specific frequencies of water movement may change over the course of life. Indeed, our data suggest that juveniles exhibit a peak prey localization response at lower frequencies (2-4 Hz) than their fully grown counterparts. This modulation is the opposite of what would be expected if it were merely a function of the change in sensor distribution with size. Rather, it suggests a change in tuning which may correspond to changes in prey preference. This primitive nervous system can not only localize prey but likely also is able to tell the difference between prey items simply through sensing water waves.

We are constructing transparent, multi-suction electrode arrays (MSEAs) to study multisensory integration of prey localization in the medicinal leech. The leech is an ideal organism to test novel technologies due the unrivaled accessibility to its nervous system and well characterized behavior. For example, we have recently invented a technique that combines multielectrode arrays (MEAs) with voltage sensitive dye (VSD) imaging to sense neuronal signals from the leech ganglion, the putative site of multisensory integration. The high temporal resolution (<1ms) of the MEA measurements when combined with the high spatial resolution of VSD imaging is a powerful new technique to identify neuronal circuitry and analyze the computations that take place within them. One common problem with VSD measurements in the leech is that a spatial shift of the tissue of even a few micrometers can be detrimental to measurements since the VSD percent signal change is quite small. We solve this problem using custom-designed MSEAs which immobilize leech ganglia during simultaneous MEA/VSD experiments. We have realized quartz suction devices and transparent, low impedance MEAs independently, and are currently combining the two platforms. Further, microfluidics are being incorporated onto the underside of the device to provide negative pressure as well as a means of dye and solution perfusion. Finally, these tools should be amenable to flat tissue systems in higher organisms as well.

Many aquatic predators such as frogs, leeches, and predatory fish are able to localize their prey by orienting toward disturbances in the water which their prey makes (Elepfandt, 1984; Horridge and Boulton, 1967; Müller and Schwartz, 1982; Young et al., 1981). Here, we examine the ability of the leech to perform this task under a variety of conditions. Our data suggests that behavioral responses to waves are frequency dependent. For instance, adult leeches are better able to localize the source of a 4 Hz water disturbance than a 1 Hz disturbance. However, prey preferences in leeches change over the leech’s lifetime (Keim, 1993; Wilkin and Scofield, 1990). While juveniles feed on frogs, adults prefer to feed on mammals and even have higher reproductive viability when they do so (Sawyer, 1983). These animals may not only be able to localize prey by sensing water waves, but also, could use this information to determine the type of prey. Does this change in prey preference correspond to a change in the leech’s response to different frequencies of water waves? We have found that juvenile leeches exhibit a greater behavioral response to lower frequency stimuli than their adult counterparts. Interestingly, leeches have two sensory modalities capable of sensing water waves: mechanoreceptors on their skin, or their array of more than 300 simple eyes. However, these modalities differ in their sensitivity to stimuli of specific frequencies. Neurophysiological recordings have shown that these sensors exhibit a differential response to wave stimuli. While the visual sensors respond best to low stimulus frequencies, the mechanosensors are much more broadly tuned (Friesen, 1981; Kretz et al., 1976; Peterson, 1984). Preliminary behavioral data in adults shows a similar shift in response toward lower frequencies when only visual information is available. In the future, we hope to examine how this information is processed at different levels within the nervous system.

In this paper we describe a setup for combined recordings with extracellular microelectrodes and voltage sensitive dyes. We apply this technique to the Leech ganglion where the role of many individual cells in a variety of circuits has been established. Individuating cells is a traditional drawback of extracellular recording. We show that the extracellular signal is correlated to individual cells from the optical signal. This shows the possibility to eventually for a large number of spike trains simultaneously, assign them to the respective cells as they participate in a neu- ral circuit such as the Leech pattern generator for swimming.

We are constructing transparent, multi–suction electrode arrays (MSEAs) to study multisensory integration of prey localization in the medicinal leech (Figure 1) during simultaneous electrophysiology and voltage sensitive dye (VSD) imaging measurements. However, the VSD percent signal change is small, so a shift of even a few mi- crometers can be detrimental to measurements. Custom-designed multielectrode arrays (MEAs) will be inte- grated with planar quartz suction devices and microfluidics to immobilize leech ganglia during multisensory inte- gration experiments.

Medicinal leeches use both visual and mechanosensory information to guide them toward their prey. We are interested in how they integrate these two streams of information in their nervous system. We approach this question both using behavioural experiments and using recordings from the segmental nervous system. In particular, we are developing a unique combination of custom multielectrode arrays and voltage-sensitive dyes to record in parallel from very many (ultimately perhaps even all) cells in a segmental ganglion.

The traditional glass pipette patch-clamp technique has contributed greatly to fundamental and pharmacological ion channel studies. The success of this serial technique has driven an effort to create wafer-based patch-clamp platforms using materials with inferior dielectric properties than glass and/or using exotic processing techniques to avoid the difficulties inherent to parallel processing of glass. We have developed a material processing scheme that generates ultrasmooth, high aspect ratio pores in fused quartz wafers. These devices are demonstrated here to be superior planar patch-clamp electrodes achieving gigaohm seals in nearly 80% of trials with a mammalian cell line, with the majority of seals over 10 GΩ and as high as 80 GΩ, competing with the best pipette-based patch-clamp measurements. Our method, amenable to batch fabrication technologies, will enable the acquisition of low noise, ion channel measurements in high throughput. We are currently merging the abovementioned devices with voltage sensitive dye (VSD) imaging and multi-electrode array (MEA) recordings in order to study multisensory integration in the medicinal leech. Initially, the planar pores will function to provide precise placement of neurons in the leech ganglion over the MEA’s. The excellent spatial resolution of the VSD’s combined with the temporal resolution of MEA’s will provide much information of all the neurons that respond to visual stimuli in the ganglion. Further studies may employ the planar pores as intracellular electrodes, allowing voltage control and intracellular recordings of individual neurons in the ganglion.

European medicinal leeches, Hirudo verbana, are simultaneous—but not self-fertilizing—hermaphrodites. Prior to copulation, these leeches perform a highly stereotyped periodic twisting to properly align the reproductive segments of potential partners. We previously reported that conopressin, a vasopressin analog derived from the venom of the cone snail Conus imperialis, induces in isolated leeches a behavior strikingly similar to this mating dance. Here, we quantify the similarity and further study the rhythmicity of the behavior in isolated sections of body wall as well as in isolated sections of the nervous system. We find that the typical slow twisting seen in intact animals results from coordination between contractions of left and right longitudinal and circumferential muscles. The slow (approximately 5-minute) period of the behavior is generated not by a gradual variation of muscle tension, but rather by short bursts of motor neuron activity—and hence muscle contractions—that increase and decrease in strength over the period of the behavior.

When a leech is touched anywhere on its body longitudinal and circular muscles locally contract while contralateral muscles relax causing the leech to bend away from the touch location. We quantified the kinematics of this "local bend response" by analyzing the motions of a piece of body wall innervated by a single ganglion in response to force-controlled mechanical stimuli. Blocking GABA-A receptors in the ganglion resulted in a dramatic increase in the amplitude and speed of the local contraction while the contralateral relaxation was reduced abolished or even inverted. This shows an important role for inhibitory signaling in tuning the response.

In response to local pressure applied anywhere on its body, the European medicinal leech contracts muscles ipsilateral to the pressure and relaxes contralateral muscles, thus bending its body away from the touch location. Previous work from our lab has shown that the relative strength of contraction and relaxation around the body circumference directs bending precisely away from the stimulus, despite the fact that only 4 sensory neurons per ganglion respond to pressure stimuli. Here we study the role of inhibition in shaping a well-controlled response.

We used a mechanical stimulator to deliver forces of 0.75–400 mN (corresponding to pressures of about 0.75–400 kPa) to a section of leech body wall innervated by a single ganglion, and recorded the consequent motion of the skin using a computer-attached CCD camera. These measurements were repeated after inhibition was blocked using bath application of bicuculline methiodide (BMI), a blocker of GABA-A receptors. This resulted in a roughly twofold increase in contraction ipsilateral to the stimulus, at all force levels. Contralateral to the stimulus, the (normally weak and brief) initial contractile phase was augmented in amplitude and duration, and the subsequent relaxation was delayed and reduced, or even abolished. This shows that the role of inhibition goes well beyond mediating contralateral relaxation; it strongly tempers the ipsilateral contraction as well. Conversely, contralateral relaxation is tempered by a contractile force which is revealed when inhibition is blocked. While four of the motor neurons involved in the response are inhibitory, applying BMI locally to the muscles had no effect. Thus, the observed effects must be central.

In conclusion, inhibition regulates the dynamic balance between contraction and relaxation in response to contralateral touch, and it controls the amplitude of the contraction in response to ipsilateral touch as well as its spatial extent.

We study stimulus-induced plasticity in cultured cortical networks as an in vitro model for learning and memory. Lasting changes in functional connectivity have been difficult to detect using standard firing-rate metrics. Here we used a simulated model network to compare the ability of 6 different metrics to quantify network plasticity. We then applied the most successful metric, Center of Activity Trajectory (CAT) and two commonly used ones, to measure functional plasticity in living networks grown on multi-electrode arrays (MEAs).

We have made available on the web a comprehensive (>40GB) set of multi-unit spike data on cultures of dissociated embryonic rat cortex. We studied 58 cultures of different densities (3,000 to 50,000 neurons on areas of 30 to 75 mm²) growing on multi-electrode arrays (Multichannel Systems) during the first five weeks of their development [1]. Half-hour recordings of spontaneous activity were made almost daily, as well as network responses evoked by stimulating each electrode sequentially. The dataset also includes 36 long (overnight) recordings.

[1] Wagenaar, D. A., Pine, J., & Potter, S. M. (2006) An extremely rich repertoire of bursting patterns during the development of cortical cultures. BMC Neuroscience 7, 11.

Evidence for recurring precisely-timed spatiotemporal patterns of neural activity has been found both in vivo in behaving animals, and in brain slices. We asked whether such patterns also occur in dissociated cortical networks, which lack most of the intrinsic cortical circuitry present in intact brains. Dissociated cultures were prepared from rat neocortex and plated on multielectrode arrays (MEAs) at high density (1,000 cells/mm²). At 35 days in vitro, one minute of spontaneous multiunit activity was recorded from the MEA’s 59 electrodes (n = 10 cultures from 3 separate platings). A template matching algorithm tracked electrode firing times to locate recurrences of precisely-timed activity patterns. Using a template size of 200 ms and a precision of 10 ms, an average of 1365 ± 281 sequences were found per culture, each repeating 2.5 ± 2.4 times/min (range 2 - 99). On average, the patterns involved 5.5 ± 2.1 electrodes and lasted 127.8 ± 7.6 ms. To evaluate significance, these results were compared to shuffled versions of the same datasets. Two shuffling methods were used: spike shuffling, where action potentials are randomly assigned to new electrodes, and spike swapping, where action potentials are randomly swapped between two electrodes. Since longer sequences are less likely to occur by chance than shorter sequences, we compared to shuffled data the number of recurring sequences where each sequence was weighted by its number of recurrences. Significant results for both shuffling conditions were obtained by this metric (P = 0.002 for spike shuffling and P = 0.025 for spike swapping). The above experiments were repeated at two additional precisions, 1 ms and 0.1 ms, both yielding nonsignificant results when compared to shuffled data. Overall, these results indicate that spontaneously active dissociated cortical networks display recurring spatiotemporal patterns of neural activity with each action potential holding a precision of 10 ms. The discovery of such patterns in dissociated cortical networks suggests that these patterns do not rely on the intact brain’s intrinsic circuitry, but are rather a general property of self-organizing biological neural networks.

Potter lab website: http://www.neuro.gatech.edu/groups/potter

Medicinal leeches (Hirudo medicinalis) move quickly through water by swimming, a behavior that is driven by an elegant and fluid-dynamically efficient whole-body motion. Previous work has characterized the neuronal basis of swimming in detail. We have found that injecting an alpha-conotoxin, Im1, into intact leeches disrupts-but does not destroy-swimming behavior. To understand better how activity in the nervous system generates swimming, we have been studying this disruption. Following injection of Im1, leeches become more likely to swim spontaneously, and they no longer follow their usual straight path, which greatly reduces their forward motion. We have studied the neuronal basis of this effect by bath-applying Im1 to isolated nerve cords and simultaneously recording from up to 7 motor nerves along the ventral nerve cord using suction electrodes. We found that the toxin modified neuronal output during swimming in a least three ways: (1) Bursts of action potentials in cells DE-3-motor neurons that drive the contraction of the dorsal longitudinal muscles-were severely disrupted. Although the median burst duration and period remained largely unchanged, the distribution of durations and periods of individual bursts became much more variable, suggesting a loss of whole-body control. (2) Fictive swim episodes were interspersed with episodes of generally elevated motor neuron activity, a phenomenon never seen in control conditions. (3) Im1 made interganglionic propagation delays less tightly controlled, and shorter on average, which helps to explain the reduction of stroke efficacy. We are currently investigating where in the swim central pattern generator the disruption originates by comparing changes elicited in different phases of the swim cycle as they are reflected in activity recorded from pairs of dorsal and ventral nerves along the nerve cord, and by imaging interneuronal activity using voltage-sensitive dyes.

Leeches (Hirudo medicinalis) are simultaneous hermaphrodites, and although they have been successfully bred in captivity for many years, little is known about the neuronal control of their mating behaviors. We have found that conopressin, a vasopressin analog isolated from cone snail venom, elicits motor behavior that closely resembles mating. The behavior is produced in whole leeches, in semi-intact body walls, and in isolated nerve cord preparations. We have found that ganglia 5 and 6, located in the reproductive segments, are necessary and sufficient to control this behavior. These ganglia control waves of muscle contraction that spread to most of the rest of the body. Using suction electrode recordings and simple spike sorting, several distinct neurons have been found to play a role in the behavior. We have focused on cell DE-3, the excitor of dorsal longitudinal muscles. As in swimming and crawling, the timing of cell DE-3 activity is synchronized between the left and right side of each ganglion. However, the amplitude of the waves varies in cycles of 5-10 minutes, antisymmetrically between left and right sides, corresponding to the period of the longitudinal twisting observed in the intact animals, and likely causing it. Our current efforts focus on imaging ganglia 5 and 6 with voltage sensitive dyes, to study the unknown central pattern generator that we predict underlies the activity patterns of the motor neurons during mating.

Our lab embodies cultured cortical networks in an artificial environment to investigate how associations are formed by the network based on its environmental interactions. This requires creating a sensory feedback stimulation that induces neural plasticity and finding a reliable measure of network state to inform motor commands. Neurons modify their activity through synaptic and intrinsic plasticity mechanisms and temper these changes through homeostatic mechanisms. Here, we alter network activity by adding constant frequency stimulation termed context, a type of “artificial plasticity”, and study the subsequent self-regulatory behavior. A neural response after stimulation is due to a combination of the electrically evoked and ongoing neural activity. By measuring responses to a repeated stimulation termed probe, changes are due solely to changes in network state; the instantaneous stimulation rate was low to avoid transmitter depletion and stimulation during network refractory periods. Early evoked activity from an extra-cellular electrode are precisely timed (jitter less than 1 ms) and detected on multiple electrodes up to 15 ms after the stimulation of a Multi-Electrode Array. These precisely-timed spikes are pre-synaptic as they occur when synaptic activity is blocked (CNQX, APV, Bicuculline). Their latencies changed during and between various paired sets of context and probe stimulations, suggesting plasticity occurred in intrinsic cell excitability and/or in synaptic strengths between the first evoked activity and later precisely-timed activity. The same stimulation done in the presence of blockers did not produce latency changes, suggesting they were synaptic in origin. Thus, this spatio-temporal spike structure is a measure of network state. For the control of a hybrot using an embodied neural network, the context stimulations scale to represent various sensory feedback, and precisely-timed spikes map to motor commands.

We have collected the most comprehensive set to date of multiunit data on dissociated cortical cultures. We followed the first five weeks of the development of 58 cultures of different densities—3000 to 50,000 neurons on 30 to 75 mm²—growing on micro-electrode arrays (MEAs). While the aggregate spike detection rate scaled linearly with density—as expected from the number of cells in proximity to electrodes—dense cultures started to exhibit bursting behavior earlier in development than sparser cultures. Analysis of responses to electrical stimulation suggests that axonal outgrowth likewise occurred faster in dense cultures. After two weeks, the networks’ activity was dominated by population bursts in most cultures. In contrast to previous reports, development continued with changing burst patterns throughout the observation period. Burst patterns were extremely varied, with interburst intervals between 1 and 300 s, different amounts of temporal clustering of bursts, and different firing rate profiles during bursts.

These results are founded on a database consisting of 966 half-hour-long recordings from 58 cultures from 8 plating batches. To encourage more investigation of the rich range of behaviors exhibited by cortical cells in vitro, we are making it available to other researchers in its entirety, together with matlab code to facilitate access.

We developed a device that can deliver arbitrary stimulation sequences to each of 60 electrodes in an in vitro neural interface (MEA). We have used this system to control bursting in ensembles of cortical neurons in high-density dissociated culture. The spontaneous activity of such ensembles is characterized by bursts globally synchronized across the entire culture. These recur at intervals ranging from seconds to a minute, alternating with periods of sparse firing. Single stimuli applied at low frequencies (below 0.2 Hz) typically entrain bursts. We found that the very same stimuli can be combined to reduce or even abolish bursting by increasing the stimulation rate to 10-50 Hz, and distributing the stimuli across up to 25 electrodes. Moreover, we found that a stimulation protocol that explicitly controls tonic firing rates using real-time closed-loop feedback effectively abolished bursting in 50% of cultures using only 10 electrodes. While good burst control was observed in several cases using single electrodes, we found that the use of multiple electrodes greatly increased the reliability of the technique. Perfect burst suppression was achieved by delivering 50 stimuli/s distributed across 25 electrodes.

Our stimulator hardware could easily be adapted for use with in-vivo neural interfaces. We hope that burst control by multi-site stimulation may find application in clinical approaches to the reduction of epileptic seizures by focal stimulation.

The activity of high-density cortical neuronal ensembles in dissociated culture is dominated by globally synchronized bursts, reminiscent of interictal bursts seen in epileptic patients. A few years ago, Latham et al [1] found that networks with a larger fraction of endogenously active cells have lower propensity to bursting. We asked whether using electrical stimulation to increase the tonic firing rate of neuronal ensembles above their spontaneous activity level could be used to quiet bursting. We combined real-time data acquisition and spike detection software [2] with a real-time controlled stimulator [3] to modulate the firing patterns of neuronal ensembles cultured on multi-electrode arrays. We continuously provided 10 stimuli per second to the culture, distributed across 10 electrodes, and adapted the stimulation voltage in closed-loop feedback to match the response rate to a preset target. By increasing the culture-wide firing rate to about 500 spikes/second (summed from 55 electrodes), we could eliminate bursting in over 50% of cultures tested. We hope that this form of non-pharmacological burst control may have applications for the treatment of epilepsy.

[1] P E Latham, B J Richmond, S Nirenberg, and P G Nelson, 2000. Intrinsic dynamics in neuronal networks. II. Experiment. J. Neurophysiol. 83 pp 828-835.

[2]D A Wagenaar, 2002. Meabench. www.its.caltech.edu/~wagenaar/meabench.html

[3]D A Wagenaar, and S M Potter, 2004. A versatile all-channel stimulator for electrode arrays, with real-time control. J. Neural Eng. 1 pp 39-44.

Dissociated cultures are attractive for the study of information processing in cortical networks because by growing them on multi-electrode arrays (MEAs), it is possible to have long-term, high-bandwidth, two-way communication between neurons and a computer. The spontaneous activity of such cultures is characterized by episodes of intense globally synchronized firing with longer periods of low-intensity dispersed activity. This pattern is more reminiscent of spindle waves than of awake cortex in vivo. Clearly, for studies of information processing, a model of awake cortex is more interesting than a model of sleeping cortex, so how can we “wake up” these cultures? In vivo, a key role in raising the cortical arousal level is played by cholinergic (ACh) afferents from the basal forebrain. We tested whether cortical cultures can be woken up by bathing them in 0.2 mM carbachol, a broad-spectrum ACh receptor agonist. This either abolished spontaneous bursting or dramatically reduced burst amplitudes, while increasing the overall culture-wide firing rate. The desynchronizing effect of carbachol in culture resembled the effect of cholinergic stimulation of cortex in vivo: a disruption of sleep waves giving rise to more awake-like patterns. When electrical stimulation was used to evoke activity, carbachol made the response strength much more consistent between applications of identical stimuli. Moreover, the probability that stimulation evokes bursts was reduced tenfold, while the number of spikes in the first 100 ms of the stimulus response was increased. We hypothesize that the more nearly stationary spontaneous activity observed in cortical cultures bathed in low concentrations of carbachol, together with the more consistent response to stimuli, is a much better substrate for information processing than the seizure-like activity of traditional deafferented cortical cultures. We will use these awakened cortical cultures to study learning and for robotic control.

Multi-electrode arrays (MEAs) enable researchers to electrically communicate with large numbers of cells in a neuronal culture. Technology to record from 60 electrodes simultaneously has been commercially available for many years. By contrast, stimulation studies have mostly been limited to small numbers of electrodes. In this talk I will describe a device that allows stimulation of any of 60 electrodes with real-time software control over channel selection. The device can be directly plugged into MultiChannel Systems MEA preamplifiers, is inexpensive to produce, and can easily be modified for use with other recording hardware.

Using this device to stimulate high-density cultures of cortical neurons, we found that any electrode that is in sufficiently close contact with the culture to record activity, can also be used to evoke activity. We quantified the efficacy of a range of stimulus pulse shapes, both under voltage- and current-control, and found that voltage-controlled, biphasic, positive-phase-first pulses are the most effective stimuli for any given peak-voltage quotum. This is good news, because voltage-controlled stimuli have the added advantage that electrochemistry can be explicitly controlled, make their use considerably safer than current-controlled stimuli.

By connecting this stimulator to our freely available data acquisition software, MeaBench, we close the feedback loop between cell culture and computer. The system can generate stimuli in response to recorded action potentials within 15 ms, a timescale that corresponds to only a few typical cortical neuron-to-neuron propagation delays, and similar to the time constant of NMDA channels. Thus, we can now communicate bidirectionally through a 60-channel-wide channel with cortical cultures, at a tempo that matches the neurons’ own.

We study stimulus-induced plasticity and information processing in dense dissociated monolayer cultures of E18 rodent cortical neurons grown on Multi-electrode arrays (MEAs). Dishwide spontaneous bursts, or “barrages” dominate the activity of such networks. We hypothesize that these spontaneous barrages are due to lack of natural input and are wiping out the effects of potential plasticity-inducing stimuli. We compensate for the absence of natural input by applying a continuous stream of weak electrical stimuli at multiple electrodes. With such distributed sequential stimulation, we have successfully reduced the contribution of spontaneous barrages to the total firing rate of the network. The goal of this work is to investigate whether such controlled cultures are more conducive for the induction of plasticity at the network level. A 10-Hz sequence of stimuli applied at 10 electrodes reduced the duration and rate of occurrence of barrages. With this ’quieted’ level of activity as baseline, a tetanic pulse train is applied to two other electrodes, which induces a spatially distributed pattern of LTP and LTD. We study the temporal structure of spike trains and the activity-dependent changes in the reliability and reproducibility of spike patterns evoked by a probe stimulus. We use these patterns in the control of animats or hybrots (hybrid neural-robotic creatures). We find that in cultures controlled by continuous background stimulation throughout the experiment, tetanic stimulation induces a stronger and more sustained change in probed response. Changes in neural plasticity can be mapped to changes in the animat’s behavior, enabling us to study how information is encoded within an embodied living neural network.

Many brain processes, from odor recognition to motion sequence generation, can be described in terms of dynamic attractors. Here we explore the emergence of attractor dynamics in the spiking activity of neuronal cultures growing on multi-electrode arrays (MEAs). We recorded spiking activity through 58 surface electrodes, continuously for 24h periods. Using superparamagnetic clustering (SPC), we were able to isolate in excess of 200 units per culture. The most prominent feature of the spontaneous firing behavior of these cultures is population bursting. In contrast with earlier reports, we find that many cultures generate “superbursts” during development with a complex internal dynamics. While cultures displaying simple population bursts exhibit varying spatio-temporal patterns, superbursts have much more stereotyped dynamics for a given culture: - The order in which different cells are engaged in bursts is highly conserved from burst to burst, and is independent of the firing rate of individual cells. - Principal component analysis (PCA) reveals that consecutive bursts trace similar orbits through activity space. - Burst composition is more conserved across successive superbursts than within a superburst, indicating a superburst level coordination of spike dynamics. These results demonstrate that even in dissociated culture, cortical neurons can form networks that exhibit rich dynamics with recurring structure at timescales far beyond those of individual action potentials. Since networks with attractor dynamics express learning capability, we plan to utilize this feature to control robots (’animats’, or ’hybrots’). Feedback stimulation derived from the environment of the robot will modify the attractor landscape enabling the culture to learn new behavior.

There are two fundamentally different goals for neural interfacing. On the biology side, to interface living neurons to external electronics allows the observation and manipulation of neural circuits to elucidate their fundamental mechanisms. On the engineering side, neural interfaces in animals, people, or in cell culture have the potential to restore missing functionality, or someday, to enhance existing functionality. At the Laboratory for NeuroEngineering at Georgia Tech, we are developing new technologies to help make both goals attainable. We culture dissociated mammalian neurons on multielectrode arrays, and use them as the brain of a ’Hybrot’, or hybrid neural-robotic system. Distributed neural activity patterns are used to control mobile robots. We have created the hardware and software necessary to feed the robots’ sensory inputs back to the cultures in real time, as electrical stimuli. By embodying cultured networks, we study learning and memory at the cellular and network level, using 2-photon laser-scanning microscopy to image plasticity while it happens. We have observed a very rich dynamical landscape of activity patterns in networks of only a few thousand cells. We can alter this landscape via electrical stimuli, and use the hybrot system to study the emergent properties of networks in vitro.

We electrically stimulate cultures of dissociated neurons from rat cortex growing on MEAs, with two goals: 1. We study the influence of patterned stimulation on the development of functional connectivity in living neural networks. 2. We use electrical stimuli to convey an animat’s or hybrot’s sensory input to the culture which is its brain. For both, detailed knowledge of the impact of different stimulation parameters is indispensible. We find that cathodic current pulses are effective stimuli. Both very short but strong (50 uA x 20 us), and very weak but prolonged (5 uA x 1 ms) stimuli elicited network response. No responses to any anodic pulses were observed. While current-controlled pulses are attractive, the required hardware is difficult to implement for many electrodes. Therefore, we also studied voltage-controlled pulses. Biphasic, anodic-first, square waveforms were most effective, due to the cathodic current spike accompanying the voltage transient between the two phases. The response patterns to these stimuli remain stable for several days in mature cultures. On the other hand, by sending in trains of stimuli at frequencies between 0.1 Hz and 100 Hz, we determined that responses are relatively suppressed at rates above 1 Hz. During continued high-frequency stimulation on one electrode, responses to stimulation on another electrode are not reduced, so we think this suppression is due to ionic or vesicle depletion in directly stimulated cells. These findings drive the design of our next generation of stimulators. Moreover, the parameter space of stimuli that retain their efficacy upon repeated presentation is an essential resource for studying the influence of continuous stimulation on development. It could also benefit animal studies involving extracellular stimulation, and have clinical implications for deep brain stimulation and control of epilepsy.

Living neural networks of dissociated rat cortical cells were cultured on a 60 channel multi-electrode array from MultiChannel Systems and interfaced to a robotic body (a Khepera II by K-Team). The spatio-temporal pattern of neural activity was measured in real-time to produce movements of the mobile robot via a custom computer interface. The Khepera’s onboard IR sensors acted as the sensory system, measuring distance from eight IR emitters positioned around a circular pen. This sensory information was then fed back into the neural culture by varying the temporal structure of neural stimulation as a function of distance to each sensor. Because these multi-electrode arrays allow simultaneous electrical, chemical, and optical access to a population of neurons, we can conduct detailed investigations into the mechanisms that produce changes in neural activity as a result of feedback at the microscopic and macroscopic levels. Because we can culture primary cortical neurons for many months, we can examine plasticity in vitro over much longer periods than previously possible. With this simple system we hope one day to develop more advanced computational algorithms in living neural networks, leading to a greater understanding of how these networks can process and encode information, and control behavior.

The goal of spike sorting is to identify the extracellularly recorded multiunit activity with discrete neuronal sources. To perform this task, spike sorting algorithms consist of three independent steps: spike detection, spike projection and clustering. Attempts to accelerate these steps by using unsupervised algorithms have been made but the success of such methods is highly dependent on assumptions related to the statistics of signals relative to the noise component that is specific for the given brain area and recording technique. We present a new method that (1) does not assume spike shapes to follow any specific distribution, (2) it is unsupervised and (3) can be applied during the data collection.

First, we compare different methods of spike detection (threshold, slope, energy, template and wavelet) using the signal detection theory on simulated extracellular multiunit recordings. Second, we introduce a new method using “circular embedding” to project spike shape differences to a multidimensional space. Third, we compare the separability of such projections with that of the principal components and wavelet coefficients. Forth, we employ “superparamagnetic clustering” [1] and compare the results with K-means, and Bayesian clustering methods. Fifth, we test whether the combined method of “circular embedding” and “superparamagnetic clustering” on multi-dimensional projections is suitable to perform “on-line” during the data acquisition. [1] Blatt M, Wiseman S and Domany E (1996) Phys. Rev. Lett. 76: 3251-3254.

Using a novel algorithm for stimulation artifact suppression in micro-electrode array (MEA) recordings based on local regression to artifact shape [1], we are studying short latency responses to electrical stimulation in dense cultures of rat cortical neurons.

In the time window that was previously inaccessible due to artifacts, 1.5-20 ms post-stimulus, we observe spikes (extracellular action potentials) with timing precisions of 0.05-0.15 ms. These spikes are not abolished by blocking glutamatergic synapses, suggesting that they are the result of direct axonal propagation. Such directly evoked responses are observed on electrodes all the way across the array from the stimulation site (1.5 mm distant), and at latencies up to 15 ms. None of these response components are 100% reliable, but many occur in 50-75% of stimulation trials. Evoked responses on different electrode channels are mostly independent, and require different minimum stimulus voltages, indicating that several axons or somata are stimulated by the (single) stimulation electrode.

From 5 ms post-stimulus, spikes are also observed with timing precisions of 1-5 ms. Blocking glutamatergic synaptic transmission confirms that these are mostly postsynaptic from the stimulated neuron(s).

Blocking glutaminergic synapses does modify many of the precisely timed response components, even the very early ones, by sharpening up the timing, shifting the latency or changing the reliability. This indicates that ongoing synaptic activity plays an important modulatory role in the generation of directly evoked firing. We are currently investigating how patterned stimulation may be used to shape ongoing activity and thus to gain control over the modulation, which will be an important step towards realizing the computational capabilities of living neuronal networks [2].

[1] D. A. Wagenaar and S. M. Potter, Stimulus artifact suppression by local polynomial approximation, submitted.

[2] T. B. DeMarse, D. A. Wagenaar, A. W. Blau and S. M. Potter, The neurally controlled animat: biological brains acting with simulated bodies, Autonomous Robots 11 (2001), 305-310.

Our lab pursues the study of learning in vitro by connecting neuronal cultures with simulated bodies in computer generated environments. We connect the output of a set of 60 electrodes embedded in a culture dish on which we grow dense networks of embryonic (E18) rat cortical neurons to the motor control of such artificial animals (or animats). Information from their sensory organs is sent back to the dish in a closed feedback loop by electrically stimulating through the same substrate electrodes. Our recording, processing and stimulation system can provide feedback with a delay of less than 100~ms.

As an essential step towards this goal, we have been studying the response properties of these living neuronal networks to simple electrical stimuli. In absense of stimulation, the cultures’ major mode of activity is global bursts, which show interesting dynamics at the timescale of minutes and large variability at the timescale of days to months. Responses to electrical stimuli consist of spikes in the first 20 ms post-stimulus timed with deep sub-millisecond precision, followed by less precisely timed spikes and occasional induced global bursts. These responses are typically stable over many days of continuous probing.

The study of short-latency responses became possible thanks to an algorithm we recently developed to suppress the very large artifacts that plague recordings shortly after stimulation. The algorithm locally fits polynomials to the shape of the artifact and can remove artifacts ten times the size of action potentials from the recorded trace in real-time on standard PC equipment.

We found that precisely timed responses, unlike other evoked activity, persist in the presence of NMDA and AMPA synapse channel blockers, indicating that they originate directly from the stimulated neuron. However, their reliability and latency do change as a result of blocking synapses, showing that synaptic influences are important. Non-monotonicities in response reliability vs stimulation voltage are further evidence for network influences.

These results will serve to help us to intelligently design sensory-motor mappings for our neurally controlled animats.

Work is currently underway to characterize the response to pairs of stimuli and the associated inter-pulse-interval dependence, which several groups have found to be of major significance in inducing synaptic plasticity.

We are looking for regularities in the firing patterns of cultured cortical networks, which we plan to use to control the behavior of a simulated animal (the Neurally Controlled Animat; Potter et al., SFN2000 abstract 467.20). To this end, neurons and glia from E18 rat cortex were dissociated and densely plated on planar MEAs (multi-electrode arrays) with 60 electrodes. We made daily recordings from each dish for 30 consecutive days starting one day after plating. Recurring dynamic patterns were observed on many timescales, from less than 100 ms through minutes.

Single-cell action potentials were observed from the second day in vitro. Dish-wide bursts occurred from the fifth day, earlier than previously reported. As the cultures matured, bursts became increasingly frequent, and isolated spikes, while increasing in absolute numbers, became an ever smaller part of the dishes’ activity. The dishes produced global bursts between one and 30 times per minute. Often, periodicity was maintained with few interruptions for several minutes.

Most global bursts were found to be immediately preceded (within 50 ms) by elevated activty of neurons near only one or a few electrodes. These initiator sets changed as the cultures developed. Recordings that showed the highest global burst frequencies, often exhibited switching between very low (3/min) and very high (upto 60/min) burst frequencies, in cycles of up to 3 minutes.

These observations will provide the basis for a study of the effects of chronic (continuous) electrical stimulation on cortical networks developing in vitro.

We have developed the first neurally-controlled animat (simulated animal). It consists of a dissociated rat cortical culture in two-way communication with a computer via 60 electrodes embedded in the culture chamber substrate. Spatio-temporal patterns of activity in the culture are used to control the behavior of an animat situated within a computer-generated virtual reality. Sensory inputs to the animat are fed back to the culture as spatio-temporal electrical stimulus patterns in real time (<100 ms). This short-latency feedback loop in which the network can influence its own stimulation provides a new way to study learning, in vitro. “Learning” is typically defined as a change in behavior resulting from experience. We have given a cultured network a body to behave with and an environment to behave in. Thus, it is now possible to observe the changes in network activity and associated behavior as learning in this new model system. Network properties that subserve learning can be studied with unprecedented detail: We apply 2-photon time-lapse fluorescence microscopy to demonstrate submicron morphological changes on the minutes-to-hours time scale, and high-speed imaging with voltage-sensitive dyes to characterize activity patterns on the millisecond time scale. The neurally-controlled animat will help to bridge the gap between bottom-up (biochemical, cellular) and top-down (behavioral) approaches to the study of learning and memory. http:

www.caltech.edu/~pinelab/PotterGroup.htm.

Current studies of learning typically focus on either the biological aspects of learning or on behavioral measures. The goal of the Animat project is to bridge the gap between these two areas by developing a system in which the biology can behave in a virtual world. Using multi-electrode array technology (MEAs), rat cortical neurons cultures on an MEA have been given a simulated body which together form a neurally controlled Animat that can move or “behave” in a virtual environment created by a computer. The computer then acts as the Animat’s senses, providing feedback in the form of electrical stimulation about the effects of interactions between the Animat’s movements and the virtual world (e.g., navigating around a barrier). Because MEAs offer the possibility of the detailed study of neurons in culture, any changes in the Animat’s behavior resulting from experience in the virtual environment can be studied in concert with the biological processes (e.g., neural plasticity) responsible for those changes.

Cortical cells in dissociated culture form densely interconnected networks. Within days after plating, neurons become electrically active, and soon after start to synchronize their activity into culture-wide bursts. By growing cultures on multi-electrode arrays (Petri dishes with a grid of substrate-embedded electrodes), their electrical activity can be recorded non-invasively. I developed software, MEABench, for online visualization and analysis of multi-electrode data, and used it to follow the development of cultures obtained from (E18) embryonic rats. Globally synchronized bursting was observed in all but the most sparsely plated cultures. A remarkable range of bursting behaviors was observed, even in cultures with identical plating parameters. Activity patterns varied widely in terms of the frequency, intensity, duration, and degree of temporal clustering of bursts. During the 2nd week in vitro, bursts in many dense cultures clustered into well-defined trains, separated by long periods without bursts. The number of bursts within these ’superbursts’ and their spatiotemporal structure were found to be stable for hours or days. Cortical cultures on multi-electrode arrays are ideal for studying two-way communication between biological systems and computers. I designed and built hardware to deliver electrical stimuli in arbitrary patterns, developed software to remove stimulation artifacts from recordings, and studied the efficacy of several voltage-defined and current-defined stimulus waveforms. MEABench can control the stimulator in real-time. Thus, stimuli can be made dependent on a culture’s activity with only 15 ms lag-time. We hypothesized that synchronized bursting can dominate activity patterns, because lack of external input puts cultures in a hypersensitive state. Indeed, by feeding cultures a steady stream of stimuli, distributed over many electrodes, bursting could be prevented completely. The number of electrodes required for successful burst control could be reduced by fine-tuning the stimuli with real-time feedback, to make each stimulus evoke the same number of spikes. Burst control could not be achieved with single electrode stimulation. For the final chapter, I tested various protocols for inducing plasticity by tetanic stimulation. In contrast to earlier published reports, I found that none of them induced changes in burst patterns or responses to test pulses that exceeded spontaneously occurring changes.

Natural gradient descent (NGD) learning is compared with ordinary gradient descent (OGD) for Kanter’s bit generator. Analytic results for one or two input bits and one output bit show that the generalization error decreases exponentially with time for NGD learning, while OGD atudents attain an error that only decreases inversely proportional to learning time. Similar results are found numerically for two input bits and more output bits. In some cases students end up in local minima. In this study NGD suffered slightly more from this problem than OGD, but we suspect that in other models it could just as easily work out the other way round. For more than two input bits no analytic results are obtained, and various options for future research with numerical methods are suggested.

A low energy effective action is derived for a single D-particle. We find that at open string tree level, it behaves like a free particle with mass ~ 1/g_c. A system of many D-particles is described by a Yang-Mills action after dimensional reduction to 0+1 dimensions. Physically, the degrees of freedom are the relative positions of the particles, and a set of harmonic oscillators corresponding to strings that may stretch between them. Upon compactification, the quantum mechanics of many D-particles turns out to be equivalent to Yang-Mills theory on a single multiply wound D-string, a fact that is known as T-duality.

Scattering of two D-particles is considered. It is found that supersymmetry is essential, since purely bosonic D-particles are confined by a potential that grows linearly with the distance between them. In the supersymmetric theory, D-particle scattering is found to contain broad resonances, which offers evidence for the existence of bound states of N D-particles, as needed for the conjectured equivalence of M-theory and D-particle quantum mechanics.

A short general introduction to string theory, and to the way D-branes appear in it, is included.

Information geometry is the result of applying non-Euclidean geometry to probability theory. The present work introduces some of the basics of information geometry with an eye on applications in neural network research. The Fisher metric and Amari’s α-connections are introduced and a proof of the uniqueness of the former is sketched. Dual connections and dual coordinate systems are discussed as is the associated divergence. It is shown how information geometry promises to improve upon the learning times for gradient descent learning. Due to the inclusion of an appendix about Riemannian geometry, this text should be mostly self-contained.

This note presents a study of the single wire resolution of the L3 TEC for data taken in 1994. For the main part of the detector the new SMD-based calibration gives a resolution which is 20% to 30% better than the one obtained with a TEC- based calibration. Inside the grid surrounding the anode the improvement is around a factor of two. The degradation of the resolution near the cathode is found to be much smaller than previously assumed. A significant asymmetry is observed between the two halves of each sector.