The primate cortical LFP exhibits multiple spectral and temporal gradients and widespread task dependence during visual short-term memory.

Although cognitive functions are hypothesized to be mediated by synchronous neuronal interactions in multiple frequency bands among widely distributed cortical areas, we still lack a basic understanding of the distribution and task dependence of oscillatory activity across the cortical map. Here, we ask how the spectral and temporal properties of the local field potential (LFP) vary across the primate cerebral cortex, and how they are modulated during visual short-term memory. We measured the LFP from 55 cortical areas in two macaque monkeys while they performed a visual delayed match to sample task. Analysis of peak frequencies in the LFP power spectra reveals multiple discrete frequency bands between 3 and 80 Hz that differ between the two monkeys. The LFP power in each band, as well as the sample entropy, a measure of signal complexity, display distinct spatial gradients across the cortex, some of which correlate with reported spine counts in cortical pyramidal neurons. Cortical areas can be robustly decoded using a small number of spectral and temporal parameters, and significant task-dependent increases and decreases in spectral power occur in all cortical areas. These findings reveal pronounced, widespread, and spatially organized gradients in the spectral and temporal activity of cortical areas. Task-dependent changes in cortical activity are globally distributed, even for a simple cognitive task.NEW & NOTEWORTHY We recorded extracellular electrophysiological signals from roughly the breadth and depth of a cortical hemisphere in nonhuman primates (NHPs) performing a visual memory task. Analyses of the band-limited local field potential (LFP) power displayed widespread, frequency-dependent cortical gradients in spectral power. Using a machine learning classifier, these features allowed robust cortical area decoding. Further task dependence in LFP power were found to be widespread, indicating large-scale gradients of LFP activity, and task-related activity.

Volumetric mesoscopic electrophysiology: a new imaging modality for the non-human primate.

The primate brain is a densely interconnected organ whose function is best understood by recording from the entire structure in parallel, rather than parts of it in sequence. However, available methods either have limited temporal resolution (functional magnetic resonance imaging), limited spatial resolution (macroscopic electroencephalography), or a limited field of view (microscopic electrophysiology). To address this need, we developed a volumetric, mesoscopic recording approach ( MePhys ) by tessellating the volume of a monkey hemisphere with 992 electrode contacts that were distributed across 62 chronically implanted multi-electrode shafts. We showcase the scientific promise of MePhys by describing the functional interactions of local field potentials between the more than 300,000 simultaneously recorded pairs of electrodes. We find that a subanesthetic dose of ketamine -believed to mimic certain aspects of psychosis- can create a pronounced state of functional disconnection and prevent the formation of stable large-scale intrinsic states. We conclude that MePhys provides a new and fundamentally distinct window into brain function whose unique profile of strengths and weaknesses complements existing approaches in synergistic ways.

The Primate Cortical LFP Exhibits Multiple Spectral and Temporal Gradients and Widespread Task-Dependence During Visual Short-Term Memory.

Although cognitive functions are hypothesized to be mediated by synchronous neuronal interactions in multiple frequency bands among widely distributed cortical areas, we still lack a basic understanding of the distribution and task dependence of oscillatory activity across the cortical map. Here, we ask how the spectral and temporal properties of the local field potential (LFP) vary across the primate cerebral cortex, and how they are modulated during visual short-term memory. We measured the LFP from 55 cortical areas in two macaque monkeys while they performed a visual delayed match to sample task. Analysis of peak frequencies in the LFP power spectra reveals multiple discrete frequency bands between 3-80 Hz that differ between the two monkeys. The LFP power in each band, as well as the Sample Entropy, a measure of signal complexity, display distinct spatial gradients across the cortex, some of which correlate with reported spine counts in layer 3 pyramidal neurons. Cortical areas can be robustly decoded using a small number of spectral and temporal parameters, and significant task dependent increases and decreases in spectral power occur in all cortical areas. These findings reveal pronounced, widespread and spatially organized gradients in the spectral and temporal activity of cortical areas. Task-dependent changes in cortical activity are globally distributed, even for a simple cognitive task.

Feature-Based Visual Short-Term Memory Is Widely Distributed and Hierarchically Organized.

Feature-based visual short-term memory is known to engage both sensory and association cortices. However, the extent of the participating circuit and the neural mechanisms underlying memory maintenance is still a matter of vigorous debate. To address these questions, we recorded neuronal activity from 42 cortical areas in monkeys performing a feature-based visual short-term memory task and an interleaved fixation task. We find that task-dependent differences in firing rates are widely distributed throughout the cortex, while stimulus-specific changes in firing rates are more restricted and hierarchically organized. We also show that microsaccades during the memory delay encode the stimuli held in memory and that units modulated by microsaccades are more likely to exhibit stimulus specificity, suggesting that eye movements contribute to visual short-term memory processes. These results support a framework in which most cortical areas, within a modality, contribute to mnemonic representations at timescales that increase along the cortical hierarchy.

A Large-Scale Semi-Chronic Microdrive Recording System for Non-Human Primates.

Multi-electrode recordings in the non-human primate provide a critical method for measuring the widely distributed activity patterns that underlie brain function. However, common techniques rely on small, often immovable arrays, or microdrives, that are only capable of manipulating a small number of closely spaced probes. These techniques restrict the number of cortical areas that can be simultaneously sampled and are typically not capable of reaching subcortical targets. To overcome these limitations, we developed a large-scale, semi-chronic microdrive recording system with up to 256 independently movable microelectrodes spanning an entire cerebral hemisphere. The microdrive system is hermetically sealed, free of internal connecting wires, and has been used to simultaneously record from up to 37 cortical and subcortical areas in awake behaving monkeys for up to 9 months. As a proof of principle, we demonstrate the capability of this technique to address network-level questions using a graph theoretic analysis of functional connectivity data.

Sorting Overlapping Spike Waveforms from Electrode and Tetrode Recordings.

One of the outstanding problems in the sorting of neuronal spike trains is the resolution of overlapping spikes. Resolving these spikes can significantly improve a range of analyses, such as response variability, correlation, and latency. In this paper, we describe a partially automated method that is capable of resolving overlapping spikes. After constructing template waveforms for well-isolated and distinct single units, we generated pair-wise combinations of those templates at all possible time shifts from each other. Subsequently, overlapping waveforms were identified by cluster analysis, and then assigned to their respective single-unit combinations. We examined the performance of this method using simulated data from an earlier study, and found that we were able to resolve an average of 83% of the overlapping waveforms across various signal-to-noise ratios, an improvement of approximately 32% over the results reported in the earlier study. When applied to additional simulated data sets generated from single-electrode and tetrode recordings, we were able to resolve 91% of the overlapping waveforms with a false positive rate of 0.19% for single-electrode data, and 95% of the overlapping waveforms with a false positive rate of 0.27% for tetrode data. We also applied our method to electrode and tetrode data recorded from the primary visual cortex, and the results obtained for these datasets suggest that our method provides an efficient means of sorting overlapping waveforms. This method can easily be added as an extra step to commonly used spike sorting methods, such as KlustaKwik and MClust software packages, and can be applied to datasets that have already been sorted using these methods.

Experimental observation of phase-flip transitions in the brain.

The phase-flip transition has been demonstrated in a host of coupled nonlinear oscillator models, many pertaining directly to understanding neural dynamics. However, there is little evidence that this phenomenon occurs in the brain. Using simultaneous microelectrode recordings in the nonhuman primate cerebral cortex, we demonstrate the presence of phase-flip transitions between oscillatory narrow-band local field potential signals separated by several centimeters. Specifically, we show that sharp transitions between in-phase and antiphase synchronization are accompanied by a jump in synchronization frequency. These findings are significant for two reasons. First, they validate predictions made by model systems. Second, they have potentially far reaching implications for our understanding of the mechanisms underlying corticocortical communication, which are thought to rely on narrow-band oscillatory synchronization with specific relative phase relationships.

Methods, caveats and the future of large-scale microelectrode recordings in the non-human primate.

Cognitive processes play out on massive brain-wide networks, which produce widely distributed patterns of activity. Capturing these activity patterns requires tools that are able to simultaneously measure activity from many distributed sites with high spatiotemporal resolution. Unfortunately, current techniques with adequate coverage do not provide the requisite spatiotemporal resolution. Large-scale microelectrode recording devices, with dozens to hundreds of microelectrodes capable of simultaneously recording from nearly as many cortical and subcortical areas, provide a potential way to minimize these tradeoffs. However, placing hundreds of microelectrodes into a behaving animal is a highly risky and technically challenging endeavor that has only been pursued by a few groups. Recording activity from multiple electrodes simultaneously also introduces several statistical and conceptual dilemmas, such as the multiple comparisons problem and the uncontrolled stimulus response problem. In this perspective article, we discuss some of the techniques that we, and others, have developed for collecting and analyzing large-scale data sets, and address the future of this emerging field.

Frontoparietal correlation dynamics reveal interplay between integration and segregation during visual working memory.

Working memory requires large-scale cooperation among widespread cortical and subcortical brain regions. Importantly, these processes must achieve an appropriate balance between functional integration and segregation, which are thought to be mediated by task-dependent spatiotemporal patterns of correlated activity. Here, we used cross-correlation analysis to estimate the incidence, magnitude, and relative phase angle of temporally correlated activity from simultaneous local field potential recordings in a network of prefrontal and posterior parietal cortical areas in monkeys performing an oculomotor, delayed match-to-sample task. We found long-range intraparietal and frontoparietal correlations that display a bimodal distribution of relative phase values, centered near 0° and 180°, suggesting a possible basis for functional segregation among distributed networks. Both short- and long-range correlations display striking task-dependent transitions in strength and relative phase, indicating that cognitive events are accompanied by robust changes in the pattern of temporal coordination across the frontoparietal network.

Modeling higher-order correlations within cortical microcolumns.

We statistically characterize the population spiking activity obtained from simultaneous recordings of neurons across all layers of a cortical microcolumn. Three types of models are compared: an Ising model which captures pairwise correlations between units, a Restricted Boltzmann Machine (RBM) which allows for modeling of higher-order correlations, and a semi-Restricted Boltzmann Machine which is a combination of Ising and RBM models. Model parameters were estimated in a fast and efficient manner using minimum probability flow, and log likelihoods were compared using annealed importance sampling. The higher-order models reveal localized activity patterns which reflect the laminar organization of neurons within a cortical column. The higher-order models also outperformed the Ising model in log-likelihood: On populations of 20 cells, the RBM had 10% higher log-likelihood (relative to an independent model) than a pairwise model, increasing to 45% gain in a larger network with 100 spatiotemporal elements, consisting of 10 neurons over 10 time steps. We further removed the need to model stimulus-induced correlations by incorporating a peri-stimulus time histogram term, in which case the higher order models continued to perform best. These results demonstrate the importance of higher-order interactions to describe the structure of correlated activity in cortical networks. Boltzmann Machines with hidden units provide a succinct and effective way to capture these dependencies without increasing the difficulty of model estimation and evaluation.

Optimizing the decoding of movement goals from local field potentials in macaque cortex.

The successful development of motor neuroprosthetic devices hinges on the ability to accurately and reliably decode signals from the brain. Motor neuroprostheses are widely investigated in behaving non-human primates, but technical constraints have limited progress in optimizing performance. In particular, the organization of movement-related neuronal activity across cortical layers remains poorly understood due, in part, to the widespread use of fixed-geometry multielectrode arrays. In this study, we use chronically implanted multielectrode arrays with individually movable electrodes to examine how the encoding of movement goals depends on cortical depth. In a series of recordings spanning several months, we varied the depth of each electrode in the prearcuate gyrus of frontal cortex in two monkeys as they performed memory-guided eye movements. We decode eye movement goals from local field potentials (LFPs) and multiunit spiking activity recorded across a range of depths up to 3 mm from the cortical surface. We show that both LFP and multiunit signals yield the highest decoding performance at superficial sites, within 0.5 mm of the cortical surface, while performance degrades substantially at sites deeper than 1 mm. We also analyze performance by varying bandpass filtering characteristics and simulating changes in microelectrode array channel count and density. The results indicate that the performance of LFP-based neuroprostheses strongly depends on recording configuration and that recording depth is a critical parameter limiting system performance.

Natural movies evoke spike trains with low spike time variability in cat primary visual cortex.

Neuronal responses in primary visual cortex have been found to be highly variable. This has led to the widespread notion that neuronal responses have to be averaged over large numbers of neurons to obtain suitably invariant responses that can be used to reliably encode or represent external stimuli. However, it is possible that the high variability of neuronal responses may result from the use of simple, artificial stimuli and that the visual cortex may respond differently to dynamic, naturalistic images. To investigate this question, we recorded the responses of primary visual cortical neurons in the anesthetized cat under stimulation with time-varying natural movies. We found that cortical neurons on the whole exhibited a high degree of spike count variability, but a surprisingly low degree of spike time variability. The spike count variability was further reduced when all but the first spike in a burst were removed. We also found that responses exhibiting low spike time variability exhibited low spike count variability, suggesting that rate coding and temporal coding might be more compatible than previously thought. In addition, we found the spike time variability to be significantly lower when stimulated by natural movies as compared with stimulation using drifting gratings. Our results indicate that response variability in primary visual cortex is stimulus dependent and significantly lower than previous measurements have indicated.

Dynamics of stimulus-evoked spike timing correlations in the cat lateral geniculate nucleus.

Precisely synchronized neuronal activity has been commonly observed in the mammalian visual pathway. Spike timing correlations in the lateral geniculate nucleus (LGN) often take the form of phase synchronized oscillations in the high gamma frequency range. To study the relations between oscillatory activity, synchrony, and their time-dependent properties, we recorded activity from multiple single units in the cat LGN under stimulation by stationary spots of light. Autocorrelation analysis showed that approximately one third of the cells exhibited oscillatory firing with a mean frequency ∼80 Hz. Cross-correlation analysis showed that 30% of unit pairs showed significant synchronization, and 61% of these pairs consisted of synchronous oscillations. Cross-correlation analysis assumes that synchronous firing is stationary and maintained throughout the period of stimulation. We tested this assumption by applying unitary events analysis (UEA). We found that UEA was more sensitive to weak and transient synchrony than cross-correlation analysis and detected a higher incidence (49% of cell pairs) of significant synchrony (unitary events). In many unit pairs, the unitary events were optimally characterized at a bin width of 1 ms, indicating that neural synchrony has a high degree of temporal precision. We also found that approximately one half of the unit pairs showed nonstationary changes in synchrony that could not be predicted by the modulation of firing rates. Population statistics showed that the onset of synchrony between LGN cells occurred significantly later than that observed between retinal afferents and LGN cells. The synchrony detected among unit pairs recorded on separate tetrodes tended to be more transient and have a later onset than that observed between adjacent units. These findings show that stimulus-evoked synchronous activity within the LGN is often rhythmic, highly nonstationary, and modulated by endogenous processes that are not tightly correlated with firing rate.

Stimulus-selective spiking is driven by the relative timing of synchronous excitation and disinhibition in cat striate neurons in vivo.

What patterns of synaptic input cause cortical neurons to fire action potentials? Are they stochastic in nature, or do action potentials arise from the specific timing of synaptic input? We addressed these questions by measuring the membrane potential fluctuations associated with the generation of visually evoked action potentials in cat striate cortical neurons in vivo. In response to visual stimulation, action potentials occurred at the crest of large-amplitude, transient depolarizations (TDs) riding on sustained depolarization of the membrane potential. The magnitude, duration and rate of depolarization of these transient events were tuned for stimulus orientation. Using numerical simulations, we find that these transient events can arise from the temporal interplay between synchronous excitation and inhibition. To validate these findings, we made conductance measurements, at the preferred stimulus orientation, and showed that the TDs arise either from an increase in excitatory conductance, or from a combination of increased excitatory and decreased inhibitory conductance, both riding on sustained changes in synaptic conductances. The properties of the TDs and their underlying conductance suggest that they arise from a specific temporal interplay between synchronous excitatory and inhibitory synaptic inputs. Our results illustrate a mechanism by which the timing of synaptic inputs determines much of the spiking activity in striate cortical neurons.

Multichannel micromanipulator and chamber system for recording multineuronal activity in alert, non-human primates.

We describe the design and performance of an electromechanical system for conducting multineuron recording experiments in alert non-human primates. The system is based on a simple design, consisting of a microdrive, control electronics, software, and a unique type of recording chamber. The microdrive consists of an aluminum frame, a set of eight linear actuators driven by computer-controlled miniature stepping motors, and two printed circuit boards (PCBs) that provide connectivity to the electrodes and the control electronics. The control circuitry is structured around an Atmel RISC-based microcontroller, which sends commands to as many as eight motor control cards, each capable of controlling eight motors. The microcontroller is programmed in C and uses serial communication to interface with a host computer. The graphical user interface for sending commands is written in C and runs on a conventional personal computer. The recording chamber is low in profile, mounts within a circular craniotomy, and incorporates a removable internal sleeve. A replaceable Sylastic membrane can be stretched across the bottom opening of the sleeve to provide a watertight seal between the cranial cavity and the external environment. This greatly reduces the susceptibility to infection, nearly eliminates the need for routine cleaning, and permits repeated introduction of electrodes into the brain at the same sites while maintaining the watertight seal. The system is reliable, easy to use, and has several advantages over other commercially available systems with similar capabilities.

Heterogeneity in the responses of adjacent neurons to natural stimuli in cat striate cortex.

When presented with simple stimuli like bars and gratings, adjacent neurons in striate cortex exhibit shared selectivity for multiple stimulus dimensions, such as orientation, direction, and spatial frequency. This has led to the idea that local averaging of neuronal responses provides a more reliable representation of stimulus properties. However, when stimulated with complex, time-varying natural scenes (i.e., movies), striate neurons exhibit highly sparse responses. This raises the question of how much response heterogeneity the local population exhibits when stimulated with movies, and how it varies with separation distance between cells. We investigated this question by simultaneously recording the responses of groups of neurons in cat striate cortex to the repeated presentation of movies using silicon probes in a multi-tetrode configuration. We found, first, that the responses of striate neurons to movies are brief (tens of milliseconds), decorrelated, and exhibit high population sparseness. Second, we found that adjacent neurons differed significantly in their peak firing rates even when they responded to the same frames of a movie. Third, pairs of adjacent neurons recorded on the same tetrodes exhibited as much heterogeneity in their responses as pairs recorded by different tetrodes. These findings demonstrate that complex natural scenes evoke highly heterogeneous responses within local populations, suggesting that response redundancy in a cortical column is substantially lower than previously thought.

Electrophysiological classes of cat primary visual cortical neurons in vivo as revealed by quantitative analyses.

To facilitate the characterization of cortical neuronal function, the responses of cells in cat area 17 to intracellular injection of current pulses were quantitatively analyzed. A variety of response variables were used to separate the cells into subtypes using cluster analysis. Four main classes of neurons could be clearly distinguished: regular spiking (RS), fast spiking (FS), intrinsic bursting (IB), and chattering (CH). Each of these contained significant subclasses. RS neurons were characterized by trains of action potentials that exhibited spike frequency adaptation. Morphologically, these cells were spiny stellate cells in layer 4 and pyramidal cells in layers 2, 3, 5, and 6. FS neurons had short-duration action potentials (<0.5 ms at half height), little or no spike frequency adaptation, and a steep relationship between injected current intensity and spike discharge frequency. Morphologically, these cells were sparsely spiny or aspiny nonpyramidal cells. IB neurons typically generated a low frequency (<425 Hz) burst of spikes at the beginning of a depolarizing current pulse followed by a tonic train of action potentials for the remainder of the pulse. These cells were observed in all cortical layers, but were most abundant in layer 5. Finally, CH neurons generated repetitive, high-frequency (350-700 Hz) bursts of short-duration (<0.55 ms) action potentials. Morphologically, these cells were layer 2-4 (mainly layer 3) pyramidal or spiny stellate neurons. These results indicate that firing properties do not form a continuum and that cortical neurons are members of distinct electrophysiological classes and subclasses.

Adaptive coincidence detection and dynamic gain control in visual cortical neurons in vivo.

Several theories have proposed a functional role for response synchronization in sensory perception. Critics of these theories have argued that selective synchronization is physiologically implausible when cortical networks operate at high levels of activity. Using intracellular recordings from visual cortex in vivo, in combination with numerical simulations, we find dynamic changes in spike threshold that reduce cellular sensitivity to slow depolarizations and concurrently increase the relative sensitivity to rapid depolarizations. Consistent with this, we find that spike activity and high-frequency fluctuations in membrane potential are closely correlated and that both are more tightly tuned for stimulus orientation than the mean membrane potential. These findings suggest that under high-input conditions the spike-generating mechanism adaptively enhances the sensitivity to synchronous inputs while simultaneously decreasing the sensitivity to temporally uncorrelated inputs.

Stimulus-Dependent Neuronal Oscillations in Cat Visual Cortex: Inter-Columnar Interaction as Determined by Cross-Correlation Analysis.

We have demonstrated previously that neurons in cat striate cortex, in response to their preferred stimuli, exhibit oscillatory responses in a frequency range of 40 - 60 Hz. Recently, we obtained evidence that such oscillatory responses can synchronize across columns. We have now performed an extensive analysis of this phenomenon for both unit and field potential responses. In addition, we studied the stimulus conditions leading to intercolumnar synchronization. We recorded both multi-unit activity and local field potentials from area 17 of adult cats with arrays of several electrodes. Interelectrode distances ranged from 0.4 to 12 mm. For all pairs of unit (n=200) and field potential (n=174) recordings, we computed auto- and cross-correlation functions. The modulation of the correlograms was quantified by fitting a damped sine wave (Gabor) function to the data. Cross-correlation analysis of the unit data revealed that in 90 out of 200 cases the recorded cells established a constant phase-relationship of their oscillatory responses. This occurred, on average, with no phase difference. If the receptive fields were nonoverlapping, we observed a synchronization primarily between cells with similar orientation preferences. Cells with overlapping receptive fields also showed a high incidence of synchronization if their orientation preferences were different. In this latter group, synchronization occurred even in cases where the stimulus was optimal for only one of the recording sites. Under conditions of monocular instead of binocular stimulation the oscillatory modulation of the responses was attenuated, but the cross-correlogram still indicated a significant interaction. Similar effects were seen with the application of stationary instead of moving stimuli. A synchronization of oscillatory field potential responses was observed in 136 out of 174 paired recordings. At all distances investigated, the probability of synchronization of field potential responses was independent of the orientation preferences of the cells. However, the strength of interaction decreased with increasing spatial separation. Control experiments showed that the synchronization of field potential responses was not due to volume conduction. The results demonstrate that oscillatory responses at separate cortical sites can transiently synchronize. The probability and strength of synchronization are dependent on the spatial separation of the recorded cells and their orientation preferences. In addition, the cross-columnar synchronization is influenced by features of the visual stimulus. It is suggested that this synchronization provides a mechanism for the formation of neuronal assemblies in the visual cortex.

Stimulus-Dependent Neuronal Oscillations in Cat Visual Cortex: Receptive Field Properties and Feature Dependence.

Previously we have demonstrated that neurons in the striate cortex of lightly anaesthetized cats exhibit oscillatory responses at a frequency near 50 Hz in response to their preferred stimuli. Here we have used both single and multiple unit recording techniques to determine: (i) the receptive field properties and laminar distribution of cells exhibiting oscillatory responses; and (ii) the influence of changing stimulus properties on the temporal behaviour of the oscillatory responses. Oscillatory responses were detected and evaluated by computation of the autocorrelation function of the neuronal spike trains. We recorded oscillatory responses in 56% of the standard complex cells and in 12% and 11% of the simple and special complex cells. Cells exhibiting oscillatory responses were located primarily in supra- and infragranular layers. The oscillatory modulation amplitude of the autocorrelation function was enhanced by binocular stimulation (9 out of 16 cells) and reduced by combined stimulation with optimal and orthogonally orientated light bars (16 out of 21 cells). Changing stimulus orientation caused no change in the oscillation frequency of the sampled population of cells, while oscillation frequency increased monotonically with respect to stimulus velocity within the range of 1 - 12 degrees per second (10 out of 11 cells). The oscillatory modulation of the autocorrelation function increased as a function of stimulus length within the boundary of the cell's receptive field (11 out of 11 cells). In 6 out of these 11 cells, the responses did not show an oscillatory modulation if elicited by small moving spots of light. Moving stimuli were much more effective in evoking oscillatory responses than were stationary stimuli (19 out of 20 cells). In no instance, using either stationary or moving stimuli, was the phase of the oscillatory response synchronized with the stimulus. These results demonstrate functional heterogeneity among cells within striate cortex based on their temporal firing patterns and provide evidence that the temporal pattern of oscillatory cellular activity is influenced by changes in stimulus properties.