J Physiol 593.22 (2015) pp 4979–4994 4979 The Journal of Physiology Neuroscience Functional characterization of spikelet activity in the primary visual cortex Benjamin Scholl, Sari Andoni and Nicholas J. Priebe Center for Perceptual Systems, Department of Neuroscience, University of Texas at Austin, 2415 Speedway, Austin, TX 78705, USA Key points In vivo whole-cell patch-clamp recordings in cat visual cortex revealed small deflections in the membrane potential of neurons, termed spikelets. Spikelet statistics and functional properties suggest these deflections originate from a single, nearby cell. Spikelets shared a number sensory selectivities with the principal neuron including orientation selectivity, receptive field location and eye preference. Principal neurons and spikelets did not, however, generally share preferences for depth (binocular disparity). Cross-correlation of spikelet activity and membrane potential revealed direct effects on the membrane potential of some principal neurons, suggesting that these cells were synaptically coupled or received common input from the cortical network. Other spikelet–neuron pairs revealed indirect effects, likely to be the result of correlated network events. Abstract Intracellular recordings in the neocortex reveal not only the membrane potential of neurons, but small unipolar or bipolar deflections that are termed spikelets. Spikelets have been proposed to originate from various sources, including active dendritic mechanisms, gap junctions and extracellular signals. Here we examined the functional characteristics of spikelets measured in neurons from cat primary visual cortex in vivo. Spiking statistics and our functional characterization of spikelet activity indicate that spikelets originate from a separate, nearby cell. Spikelet kinetics and lack of a direct effect on spikelet activity from hyperpolarizing current injection suggest they do not arise from electrical coupling to the principal neuron being recorded. Spikelets exhibited matched orientation tuning preference and ocular dominance to the principal neuron. In contrast, binocular disparity preferences of spikelets and the principal neuron were unrelated. Finally, we examined the impact of spikelets on the principal neuron’s membrane potential; we did observe some records for which spikelets were correlated with the membrane potential of the principal neuron, suggesting that these neurons were synaptically coupled or received common input from the cortical network. (Received 6 May 2015; accepted after revision 20 August 2015; first published online 1 September 2015) Corresponding author N. J. Priebe: Center for Perceptual Systems, Department of Neuroscience, University of Texas at Austin, 2415 Speedway, Austin, TX 78705, USA. Email: [email protected]Abbreviations DSI, disparity selectivity index; F o , mean Fourier amplitude; F 1 , first Fourier harmonic amplitude; FWHM, full-width half-maximum; ISI, inter-spikelet interval; ODI, ocular dominance index; RF, receptive field; STA, spike-triggered average; V1, primary visual cortex; VTA, voltage-triggered average. C 2015 The Authors. The Journal of Physiology C 2015 The Physiological Society DOI: 10.1113/JP270876
Transcript
J Physiol 593.22 (2015) pp 4979–4994 4979
The
Jou
rnal
of
Phys
iolo
gy
Neuroscience Functional characterization of spikelet activity in the
primary visual cortex
Benjamin Scholl, Sari Andoni and Nicholas J. Priebe
Center for Perceptual Systems, Department of Neuroscience, University of Texas at Austin, 2415 Speedway, Austin, TX 78705, USA
Key points
� In vivo whole-cell patch-clamp recordings in cat visual cortex revealed small deflections in themembrane potential of neurons, termed spikelets.
� Spikelet statistics and functional properties suggest these deflections originate from a single,nearby cell.
� Spikelets shared a number sensory selectivities with the principal neuron including orientationselectivity, receptive field location and eye preference.
� Principal neurons and spikelets did not, however, generally share preferences for depth(binocular disparity).
� Cross-correlation of spikelet activity and membrane potential revealed direct effects on themembrane potential of some principal neurons, suggesting that these cells were synapticallycoupled or received common input from the cortical network.
� Other spikelet–neuron pairs revealed indirect effects, likely to be the result of correlated networkevents.
Abstract Intracellular recordings in the neocortex reveal not only the membrane potentialof neurons, but small unipolar or bipolar deflections that are termed spikelets. Spikelets havebeen proposed to originate from various sources, including active dendritic mechanisms, gapjunctions and extracellular signals. Here we examined the functional characteristics of spikeletsmeasured in neurons from cat primary visual cortex in vivo. Spiking statistics and our functionalcharacterization of spikelet activity indicate that spikelets originate from a separate, nearbycell. Spikelet kinetics and lack of a direct effect on spikelet activity from hyperpolarizing currentinjection suggest they do not arise from electrical coupling to the principal neuron being recorded.Spikelets exhibited matched orientation tuning preference and ocular dominance to the principalneuron. In contrast, binocular disparity preferences of spikelets and the principal neuron wereunrelated. Finally, we examined the impact of spikelets on the principal neuron’s membranepotential; we did observe some records for which spikelets were correlated with the membranepotential of the principal neuron, suggesting that these neurons were synaptically coupled orreceived common input from the cortical network.
(Received 6 May 2015; accepted after revision 20 August 2015; first published online 1 September 2015)Corresponding author N. J. Priebe: Center for Perceptual Systems, Department of Neuroscience, University of Texas atAustin, 2415 Speedway, Austin, TX 78705, USA. Email: [email protected]
Action potentials are the primary currency used byneurons in the central nervous system to transmitinformation. Intracellular recordings have been anessential technique for elucidation of the subthresholdmembrane potential fluctuations that lead to actionpotentials (Pel et al. 1991; Ferster & Jagadeesh, 1992)but also reveal the existence of smaller, short membranepotential deflections, called ‘spikelets’, – small unipolaror bipolar waveforms (Spencer & Kandel, 1961; Margrieet al. 2003). Spikelets are distinct from the larger actionpotentials because they are only detectable in sub-threshold records, are less than 10 mV in amplitude,and do not depend directly on the somatic membranepotential. Spikelets could emerge from three distinctsources. One class of spikelets originates from dendriticmechanisms within the recorded neuron. Such spikeletsare reported to be unipolar events and are hypothesized tobe caused by voltage-gated channels (Llinas & Nicholson,1971; Golding & Spruston, 1998; Epsztein et al. 2010;Smith et al. 2013). A second class of spikelets originatesfrom the direct electrical coupling between separate, butnearby, neurons. This coupling is mediated through gapjunctions, membrane pores connecting the cytoplasm oftwo individual cells (Dermietzel & Spray, 1993). Gapjunctions attenuate membrane depolarizations and actionpotentials from the nearby cell, producing spikelets inthe principal neuron, which we define as the patchedneuron (Spencer & Kandel, 1961; MacVicar & Dudek,1981; Taylor & Dudek, 1982; Vigmond et al. 1997; Gibsonet al. 1999). These spikelets are observed in the cerebralcortex (Gibson et al. 1999; Tamas et al. 2000; Margrieet al. 2003), thalamic reticular nucleus (Landisman et al.2002; Landisman & Connors, 2005) and hippocampalinterneuron network (MacVicar & Dudek, 1981; Taylor &Dudek, 1982; Vigmond et al. 1997). Finally, spikelets mayarise from action potentials produced by nearby neurons,recorded extracellularly through the intracellular pipette(Martinez et al. 2014).
We used whole-cell patch-clamp recordings to measurespikelet responses in cortical neurons of cat primary visualcortex (V1). We observed a variety of spikelet shapesand found their waveforms were stable throughout therecording period. Spiking statistics, the brief kinetics ofspikelets, and a lack of effect by polarization of the patchedcell suggest these spikelets originate from a separate,nearby neuron recorded extracellularly. We demonstratethat spikelets in cat V1 neurons share similar responsepreferences to visual stimuli as the patched neuron –the principal neuron – but often have distinctly differentreceptive field properties. Orientation tuning, measuredfrom drifting sinusoidal gratings, is similar betweenspikelets and membrane potential of the principal neuron.The eye preferences of spikelet and principal cell spiking
responses are also matched, but disparity preferences areunmatched. Finally, we measured the impact of spikeletson membrane potential and observed that a subset ofspikelets had a direct effect on the membrane potentialof the principal neuron, suggesting synaptic connectivityor common synaptic input to both neurons.
Methods
Physiology
Experiments were performed as previously describedusing female and male cats (n = 14, 2–5 kg)that were anaesthetized and subject to neuromuscularblockade (Scholl et al. 2013). Anaesthesia was inducedwith ketamine (5–15 mg kg−1) and acepromazine(0.7 mg kg−1), followed by intravenous administrationof a mixture of propofol and sufentanil (Yu &Ferster, 2010). Once a tracheotomy was performed,the animal was placed in a stereotactic frame for theduration of the experiment. Recording stability wasincreased by suspending the thoracic vertebrae from thestereotactic frame and performing a pneumothoracotomy.Eye drift was minimized with intravenous infusion ofvecuronium bromide (0.2 mg kg h−1). Anaesthesiawas maintained during the course of the experimentwith continuous infusion of propofol and sufentanil(6–9 mg kg h−1 and 1–1.5 μg kg h−1, respectively).Body temperature (38.3°C), electrocardiogram, electro-encephalogram, CO2, blood pressure and autonomicsigns were continuously monitored and maintained.Following neuromuscular blockade, electrocardiogram,electroencephalogram, and CO2 were carefully monitoredto maintain a sufficient depth of anaesthesia during thecourse of the experiment. The nictitating membraneswere retracted using phenylephrine hydrochloride, andthe pupils were dilated using topical atropine. Contactlenses were inserted to protect the corneas. Supplementarylenses were selected by direct ophthalmoscopy to focusthe display screen onto the retina. Experiments wereterminated by euthanizing the animal with an overdoseof pentobarbital (100 mg kg−1). All procedures wereapproved by The University of Texas at Austin InstitutionalAnimal Care and Use Committee.
Whole-cell recordings
Blind whole-cell recordings were obtained in vivo (Pelet al. 1991; Ferster & Jagadeesh, 1992; Margrie et al.2003). As a reference electrode, a silver–silver chloridewire was inserted into muscle near the base of theskull, and covered with 4% agarose in normal saline toreduce changes in the surrounding fluid and concomitantchanges in associated junction potentials. The potential
Values are means ± SD (n). The input resistance was measured in all neurons, but we only recorded data for 12 neurons. Only recordswith an inter-spikelet interval (ISI) of 10 ms or less are reported. Two additional records had a minimum ISI greater than 50 ms.
of the cerebral spinal fluid was assumed to be uniformand equal to that of the reference electrode. Pipettes(7–10 M�) were pulled from 1.2 mm outer diameter,0.7 mm inner diameter KG-33 borosilicate glass capillaries(King Precision Glass) on a P-2000 micropipette puller(Sutter Instruments) to record from neurons 250–850μmbelow the cortical surface. To record membrane potentialand spike responses, pipettes were filled with (in mM)135 potassium gluconate, 4 sodium chloride, 0.5 EGTA,2 magnesium-ATP, 10 phosphocreatine disodium, and10 Hepes, pH adjusted to 7.3 with potassium hydro-xide (Sigma-Aldrich). Current clamp recordings wereperformed with a MultiClamp 700B patch clamp amplifier(Molecular Devices, Sunnyvale, CA, USA). Currentflow out of the amplifier into the patch pipette wasconsidered positive. Average resting membrane potential,input resistance, membrane time constant, and recordingduration from whole-cell records with spikelets areshown in Table 1. Acceptable whole-cell recordings wererequired to have series resistances less than 120 M�(mean = 63.8 M�, SD = 25.9 M�), a baseline restingmembrane potential less than −50 mV, and a stable base-line resting membrane potential for at least 15 min. Nojunction potential correction was made for these records.Records were digitized at 15 kHz or higher and saved todisk for offline analysis.
Stimuli
Visual stimuli were generated by a Macintosh computer(Apple) using the Psychophysics Toolbox (Brainard, 1997;Pelli, 1997) for Matlab (The Mathworks, Natick, MA,USA) and presented dichoptically using two Sony videomonitors (GDM-F520) placed 50 cm from the animal’seyes. The video monitors had a non-interlaced refresh rateof 100 Hz and a spatial resolution of 1024 × 768 pixels,which subtended 40 × 30 cm (44 × 34 deg). Thevideo monitors had a mean luminance of 40 cd cm−2.Drifting grating stimuli were presented for 4 s, pre-ceded and followed by 250 ms blank (mean luminance)periods. Spontaneous activity was measured with blankperiods interleaved with drifting grating stimuli of thesame duration. We characterized stimulus orientation,
spatial frequency (0.20–1.0 cycles deg−1), spatial location,size (0.5–2 deg diameter), and eye preference bestevoking a response. Upon isolating a neuron, stimulusparameters were coarsely mapped manually and thenfine-tuned after systematic measurements of orientationand spatial selectivity. Binocular stimuli were presenteddichoptically using the preferred stimulus parameters at2–4 Hz temporal frequency and 90% contrast. A mirrorwas placed directly in front of the contralateral eye toreflect receptive field locations onto a separate monitor.The angle and location of the mirror was adjusted toavoid occlusion of the field of view for the ipsilateraleye. To measure binocular interactions we systematicallychanged the spatial phase of one grating while holdingthe spatial phase of the other grating constant (Ohzawa& Freeman 1986a,b). Relative phase disparities usedranged from −135 to 180 deg. All binocular and mono-cular stimuli were presented during the same block andpseudo-randomly interleaved. One-dimensional noisesequences were presented to measure linear and nonlinearreceptive field components (Mohanty et al. 2012).
Analysis
To compare estimates of subthreshold membranepotential and suprathreshold spikes, raw records werelow-pass filtered with a cutoff at 100 Hz to removespikes. Spikes and spikelets were identified on the basisof their sharp deflections in membrane potential, asmeasured with the first and second derivative. Spike andspikelet times were separated based on peak membranepotential deflection and waveform amplitude. Spikeletwaveform amplitudes were required to be less than10 mV, as defined empirically (see Fig. 1). Principal cellaction potentials are distinct in this sense because theyare initiated at membrane potential threshold. Spikelettimes within 25 ms of a spike time were excluded.Membrane potential, spike and spikelet responses foreach stimulus were cycle-averaged across trials, followingremoval of the first cycle. The Fourier transform was usedto calculate the mean (F0) and modulation amplitude(F1) of each cycle-averaged response. Simple and complexcells were separated by computing the modulation ratio
(F1/F0) for spiking responses to the preferred mono-cular stimulus; neurons with modulation ratios largerthan 1 are considered simple. Peak responses weredefined as the sum of the mean and modulation(F0 + F1). All peak responses are reported aftersubtraction of the mean spontaneous activity. Meanspontaneous activity for spiking activity and membranepotential fluctuations were measured during blank (meanluminance) periods. Error bars represent SEM unlessotherwise indicated. Double Gaussian equation was usedto fit orientation tuning data and define orientation tuningpreference:
Here R(ϴ) is the response of the neuron to differentorientations (ϴ), σ is the width of the tuning curve,spont is the mean spontaneous activity, α and β are peakamplitudes, and ϴpref is the orientation preference. Theocular dominance index was defined by (Scholl et al.2013):
ODI = R contra − R ispi
R contra + R ispi
Here R is the response of the neuron to each monocularvisual stimulus. Binocular disparity tuning curves werefitted to a cosine function to determine phase preference:
R(φ) = α
2
(ei(φ−φpref ) − e−i(φ−φpref )
) + spont
Spikelet FWHM (ms)0 1
Pro
po
rtio
n
0
0.4
Principal AP FWHM (ms)
Sp
ike
let
FW
HM
(m
s)
0
4
Spikelet
AP
A CB
E F
Principal AP Amplitude (mV)
0
80
800Spikelet Amplitude (mV)
Pro
po
rtio
n
0
0.4
0 10
0 4
-1
0
1
2
3
-1
0.5
0
0.5
1
0 2.5 5
-1
0
1
2
3
-2
0
2
4
0 2.5 5
-2
0
2
4
6
-1
0
1
2
3
0 2.5 5
-3
0
3
6
-3
0
3
6
D
10Spikelet Negative
Phase Amplitude (mV)
0
0
10
Sp
ike
let
Po
sitiv
eP
ha
se A
mp
litu
de
(m
V)
StartEnd
Time (ms)
Me
mb
ran
e P
ote
ntia
l (m
V)
-2
0
2
4
G
Time (ms) Time (ms)
10 mV
50 ms
Sp
ike
let
AP
Am
plit
ud
e (
mV
)
Figure 1. Spiking statistics of spikelets andprincipal cell action potentialsA, example waveforms of each spike type. B,distribution of full spikelet amplitude acrosscells. C, plot of spikelet and principal cell actionpotential amplitude across cells. D, distributionof amplitudes for depolarization andhyperpolarization phase of each spikeletwaveform. E, same as in B for full-widthhalf-maximum (FWHM). F, same as in C forFWHM. G, average spikelet waveforms at thebeginning of each recording session (dottedline) and towards the end (continuous line).Each average is composed of 10–15 spikeletwaveforms. Elapsed time between eachaverage indicated for each record.
J Physiol 593.22 Spikelet activity in cortical neurons 4983
Here R(ɸ) is the response of the neuron to binocularphase differences (ɸ), spont is the mean spontaneousactivity, α is the peak amplitudes, and ɸpref is the binocularphase difference preference.
Spatio-temporal receptive fields were extracted fromthe membrane potential or spikelet responses to 1Dnoise stimuli (Park et al., 2013). Membrane potentialand spikelet rate were binned at the frame rate of thestimulus, either 100 or 120 Hz, and the triggered averagewas computed by measuring μ:
μ = 1
N
∑i
s iri
where μ is the spikelet- or voltage-triggered average (STAand VTA, respectively), si is the stimulus at frame i, and ri
is the spiking or subthreshold response to that frame. Afterprojecting out the VTA or STA, the covariance matrix wasconstructed as the following:
C = 1
N − 1
∑i
(si − μ)(si − μ)T
The covariance matrix was decomposed into itseigenvectors, and their significance was determined basedon a bootstrap method of their eigenvalues.
Spikelet-triggered membrane potential averages werecomputed by adding raw, unfiltered membrane potentialat spikelet spike times. During bursting events, only thefinal spikelet time was used. Spikelet times occurringnear principal cell action potentials were not used(± 5 ms). Shuffle corrections were made by shufflingtrials of membrane potential responses for a given set ofspikelet spike times. Mean shuffle corrections were sub-tracted from mean spikelet-triggered averages.
The similarity between spikelet orientation or disparitypreference with the principal cell was quantified bycomputing a circular-correlation coefficient (Batschelet,1981):
corr = 1
N
√(∑cos(ψn − ζn)
)2+
(∑sin(ψn − ζn)
)2
Here ѱ is the individual spikelet preference, ζ is theprincipal cell preference, and N is the total numberof neurons (n). The standard error was computed bybootstrapping and sampling with replacement (Sokal &Rohlf, 1995).
Results
To study the functional characteristics of spikelet activitywe obtained whole-cell patch-clamp recordings in vivofrom neurons in cat V1. In 76 intracellular recordsobtained from 14 cats we observed spikelet activity in21 neurons (number of neurons per animal: 1.5 ± 0.94,
mean ± SD, proportion of neurons with spikelets peranimal: 0.24 ± 0.07, geometric mean ± SD). For 20of those records we obtained enough data to extractfunctional responses of both the principal neuron andthe spikelet. Here we define the principal neuron asthe one patched. Records were considered acceptablebased on their resting membrane potential, stability,evidence of elicited action potentials, and membranebiophysical properties (Table 1, see Methods). The averagemeasured membrane time constant was 15 ± 4.6 ms(range = 6.3–23.1 ms).
Spiking statistics and biophysical properties ofspikelets
Spikelet waveforms were unequivocal deflectionsembedded in membrane potential with distinct size andduration from action potentials elicited by the principalneuron (Fig. 1A). For the example record shown in Fig. 1A,the spikelet amplitude was 4.5 mV, whereas the principalneuron action potential amplitude was 45 mV. Acrossour sample population, principal neuron action potentialswere larger than the full spikelet amplitude (Fig. 1B and C;mean spikelet amplitude = 4.6 ± 2.0 mV, SD, n = 20,mean action potential amplitude = 35.0 ± 14.7 mV,SD, n = 18). Spikelet waveforms exhibited diversitywith different amplitudes of depolarizing and hyper-polarizing phases (Fig. 1G). In our sample populationwe consistently observed a dominant depolarizing phase,but a wide variety of hyperpolarizing phases (Fig. 1D;also compare waveforms in Fig. 1G; mean spikeletpositive phase = 3.2 ± 1.4 mV, SD; mean spikeletnegative phase = −1.4 ± 0.8 mV, SD, n = 20). Thedominant depolarization phase was also evident in theratio of phase amplitudes (depolarized/hyperpolarized)(geometric mean = 2.4, n = 24). The duration of aspikelet was very short, with a width less than 1 ms athalf-maximum amplitude (Fig. 1E and F; mean spikeletfull-width half-maximum (FWHM) = 0.26 ± 0.11 ms, SD,n = 20; mean action potential FWHM = 1.19 ± 0.73 ms,SD, n = 18). Spikelets also had a stereotyped time courseand amplitude throughout the duration of each recording,lasting upwards of 100 min (Fig. 1G).
The fidelity of spikelet waveforms suggests that ineach recording spikelets had a common source, similarto the extracellular waveforms from a single unit. Ifspikelets do indeed originate from a common neuronalsource there should be a spikelet refractory period. Todetermine whether a spikelet refractory period exists, wemeasured the distribution of inter-spikelet intervals (ISIs)(Fig. 2A, left) and identified the minimum ISI or absoluterefractory period for each recording (Fig. 2A, circle). Wefound that for all spikelets examined, there existed anabsolute refractory period lasting between 1.5 and 10 ms
(Fig. 2B, ordinate). For records containing enough actionpotentials to generate an ISI distribution, we also identifiedthe absolute refractory period for comparison (Fig. 2B).The absolute refractory periods for principal cell actionpotentials and their spikelet counterparts were comparable(3.2 ± 1.2 ms and 3.3 ± 1.6 ms, respectively). Therealso exists a relative refractory period in which spikingis less likely than expected from a Poisson distribution,as indicated by the non-exponential distribution of ISIs(Fig. 2A) (Berry & Meister, 1998). Another characteristicof action potentials from a single neuron is a declinein spike amplitude in spike pairs separated by shortISIs or the absolute refractory period, attributed toinactivated sodium channels. For short ISIs, we examinedthe amplitude of the spikelet immediately following a pre-vious spikelet with less than a 2.5 ms delay. In those recordswe observed a small decrease in amplitude in the secondspikelet (mean amplitude difference = −0.15 ± 0.31 mV,mean percentage decrease = 5.0 ± 5.7%, n = 7), butthis was not significant (P = 0.14, Wilcoxon’s signed-ranktest). The fast kinetics of spikelet waveforms, a hyper-polarization phase (Fig. 1), and slight decrease in spikeletamplitude during quick bursts are characteristics ofspikelets being sourced from a single neuron.
Spikelets from a separate, nearby neuron could berecorded through either a passive mechanism, such as gapjunctions (MacVicar & Dudek, 1981; Taylor & Dudek,1982; Vigmond et al. 1997; Gibson et al. 1999) or as anindependent extracellular signal (Martinez et al. 2014).To distinguish between these possibilities, we injectedpulses of hyperpolarizing current (−100 to −240 pA,50–200 ms) into the patched neuron to determine ifsuch hyperpolarization affects the spontaneous spikeletactivity (Fig. 3A). Hyperpolarizing current in thepatched cell caused large negative deflections in recordedmembrane potential, and during periods without currentinjection, spontaneous membrane potential fluctuationsoccasionally eliciting action potentials were observed(Fig. 3A, inset). Spikelets were evident both during hyper-polarizing epochs and during rest periods (Fig. 3A,
inset, asterisk). We then quantified spikelet rate duringhyperpolarization compared to the resting membranepotential. In a subpopulation of recordings with hyper-polarizing current injections (n = 11, Fig. 3B) we observedno change in spikelet activity (hyperpolarized meanrate = 3.5 ± 2.7 spk s−1, Vrest mean rate = 2.7 ± 2.2 spk s−1;P = 0.92, Wilcoxon’s signed-rank test). One potentialcaveat to these data is that the gap junction resistancemay be very high. Hyperpolarizing the principal cell wouldthen only weakly alter the membrane potential of the otherneuron. While such gap-junction gated hyperpolarizationmight be weak, even weak hyperpolarization of thecoupled neuron should dramatically alter the spikeletrate given the known nonlinear relationship betweenmembrane potential and spike rate (Carandini & Ferster,2000; Priebe et al. 2004). Altogether, the fast kinetics,refractory period, stereotyped waveform, and lack of effectfrom current injection suggest that spikelets originate froma separate cell recorded extracellularly.
Functional characterization of spikelets
We characterized the functional response properties ofspikelets and the corresponding principal neuron by pre-senting drifting gratings in the principal neuron’s receptivefield. In V1, both membrane potential and spikingresponses of simple cells are strongly modulated by driftinggratings of preferred orientation, spatial frequency andocular preference (Movshon & Thompson, 1978). Spikeletrecords could also be modulated by drifting gratings(Fig. 4A). To quantify the degree of response modulationfor these records, we measured the modulation ratio,defined as the Fourier amplitude at stimulus temporalfrequency divided by the mean (see Methods) (Skottunet al. 1991; Priebe et al. 2004). Simple cells, due to strongmodulation by a grating’s temporal frequency, have aratio greater than 1, while complex cells, which haveunmodulated responses, have a ratio less than 1. In thisexample both the principal neuron spikes and spikeletshad modulation ratios greater than 1 (F1/F0 = 1.80 and
A
Num
ber
of S
pik
ele
ts
Example Cell
ISI (ms)
102
104
101 102100
0.25
Principal CellMinimum ISI (ms)
10
0
5
10
B
50
Spik
ele
tM
inim
um
IS
I (m
s)
Figure 2. Absolute refractory period ofspikelets and principal cell action potentialsA, inter-spikelet interval (ISI) distribution for anexample cell. Minimum ISI (absolute refractoryperiod) indicated by circle. Note that the minimumordinate value is 1. B, distributions of minimum ISIfor principal cell action potentials (abscissa) andspikelet (ordinate) from each recording.
J Physiol 593.22 Spikelet activity in cortical neurons 4985
1.54, respectively; Fig. 4A and B) indicating that both areclassified as simple. Simple cells in cat V1 are typicallyfound in layer 4 and receive direct thalamocortical input,whereas complex cells are primarily found in superficiallayers of cortex and thought to receive inputs from simplecells (Hubel & Wiesel, 1962; LeVay & Gilbert, 1976; Reid& Alonso, 1995; Ferster et al. 1996; Chung & Ferster, 1998;Usrey et al. 1999; Alonso et al. 2001; Hirsch et al. 2003).Modulation of spikelet and principal neuron responsesis also evident when comparing trial-averaged responsesof spike rate, membrane potential, and spikelet rate(Fig. 4B, middle), which fluctuate at the stimulus temporalfrequency.
To further test simple cells, we measured receptivefields (RFs) using a one-dimensional noise stimulus (seeMethods) (McLean & Palmer, 1989; DeAngelis et al. 1993;McLean et al. 1994; Conway & Livingstone, 2003; Priebe,2005; Mohanty et al. 2012; Park et al. 2013). Simple cellsare characterized by linear filters, whereas complex cellsexhibit dominant nonlinear filters. In this example, onlythe linear voltage-triggered average and spikelet-triggeredaverage displayed significant structure, whereas nonlinearfilters from covariance measurements yielded no structure(Fig. 4C). The dominant linear filters with segregated onand off subregions matched our classification based ongrating responses; membrane potential and spikelet RFscontained segregated on and off subregions, characteristicof classic V1 simple cells (Hubel & Wiesel, 1962; LeVay& Gilbert, 1976; Reid & Alonso, 1995; Ferster et al. 1996;Chung & Ferster, 1998; Usrey et al. 1999; Alonso et al. 2001;Hirsch et al. 2003). Both records also exhibited directionselectivity, which is evident in the slanted spatiotemporalreceptive field (Priebe, 2005). These records demonstratean example of matched selectivity between the principalneuron and the spikelet.
We also observed examples in which membranepotential and spikelets were characterized by differentreceptive field properties, evident both in responsemodulation to gratings and in noise mapping. Forexample, principal neurons that were characterized ascomplex cells could be associated with spikelet activitycharacteristic of a simple cell (Fig. 4D). As in the previousexample, spikelets were modulated by the grating, andyet the recorded membrane potential and accompanyingspike rate showed no phase sensitivity (Fig. 4E).This was evident when comparing the modulationratio of action potentials of the principal neuron withspikelets (F1/F0 = 0.65 and 1.69, respectively). Similarly,one-dimensional noise mapping revealed prominent non-linear filters from voltage-triggered covariance in themembrane potential, while structure only existed in thelinear spikelet-triggered average (Fig. 4F). In this example,the principal neuron is therefore characterized as complex,while the spikelet activity is simple.
All combinations of principal neuron and spikeletRF properties were observed in our example records.Specifically, these included a simple cell with spikelets froma complex cell (Fig. 5A–C) and a complex cell with spikeletsfrom another complex cell (Fig. 5D–F). A comparison ofmodulation ratios of spikelets and spikes from principalneurons across our population revealed a majority ofrecords were simple-to-simple coupling (n = 8/16;Fig. 6). Simple cell spikelets observed in complex cells werethe second most common (n = 4/16), followed by complexcell spikelets observed in simple cells (n = 3/16). Notethat only those records for which there were a minimumof 20 action potentials and spikelets to the preferredstimulus were included in this analysis, as a low number ofspikes systematically alters the F1/F0 ratio (Hietanen et al.2013).
Sp
ike
Ra
te (
spk/
sec)
0
6
12
1sec
0.1sec
10mv
-20
-40
-60
-80
-100
Me
mb
ran
e P
ote
ntia
l (m
V)
Iinject
A B
* *
*
Vrest
Hyperpol.
Iinject
Figure 3. Spikelet activity unaffected by hyperpolarization of principal cellA, example recording depicting principal cell action potentials, membrane potential, and negative current injectionto hyperpolarize membrane potential. Current pulses illustrated below trace (50 ms duration, −210 pA amplitude).Spikelets were evident both at rest and when membrane was hyperpolarized (shaded region, inset at right). Pipetteaccess resistance was not bridge-balanced in this record. The membrane input resistance for this example cell was61 m�. Asterisks indicate presence of a spikelet. B, average spikelet activity and standard error at rest and duringhyperpolarization for each recording.
The drifting gratings and noise stimuli we used toclassify principal neurons and spikelet activity as simpleor complex were optimized for the orientation preferenceof the principal neuron, and thus the robust spikeletactivity observed with this single stimulus suggestedsimilar orientation preferences. We directly comparedthe orientation preference between spikelet activity andthe membrane potential of the principal neuron byvarying the orientation of the presented gratings. Spikeletactivity and membrane potential fluctuations to eachdrifting grating were cycle-averaged to compute thepeak responses (Fo + F1; see Methods). The orientationpreference of membrane potential and spikelets aretightly correlated, as shown in an example record whereboth are tuned to 240–270 deg (Fig. 7A). Spikingresponses of principal neurons were more narrowly tunedthan membrane potential, as reported previously (Finnet al. 2007). Across the population of records thereexisted a strong relationship between orientation pre-ference of membrane potential and spikelets (Fig. 7C)
(circular-correlation = 0.98 ± 0.01, SEM, n = 10,bootstrapped standard error) (Batschelet, 1981; Sokal &Rohlf, 1995). This relationship was also evident in thedistribution of orientation preference difference betweenprincipal neuron membrane potential and spikelets(Fig. 7D; mean |�pref| = 11.0 ± 7.3, SD).
The surprisingly tight relationship in orientationpreference between spikelets and membrane potentialprompted us to explore whether other response propertieswere also shared, in particular, binocular responseproperties. Using a dichoptic presentation of sinusoidaldrifting gratings, we measured ocular preference withmonocular stimulation and disparity sensitivity withbinocular stimulation (Scholl et al. 2013). Disparity pre-ferences were measured by systematically shifting thebinocular phase differences between the two driftinggratings (Ohzawa & Freeman, 1986b; DeAngelis et al.1991; Cumming & Parker, 1997; Cumming & DeAngelis,2001). Unlike orientation preference, disparity preferencewas not shared between the principal neuron and spikelets.
A
D
-80
-60
-40
-20
0
Mem
bra
ne P
ote
ntia
l (m
V)
Rate
(s
pk/
sec)
0
80
0
40
-80
-60
-40
-20
0
Vrest
Vm
(m
V)
-75
-55Vm
Spikelets
Rate
(s
pk/
sec)
Average Covariance
Space (deg)
Tim
e (
ms)
Mem
bra
ne P
ote
ntia
l (m
V)
-72
-60Vm
Vm
(m
V)
-0.4
0.4
-0.2
0.2
-0.2
0.2
-0.2
0.210 mV
5 mV
Average Covariance
Spikelets
B
E
C
F
Spikes
Rate
(s
pk/
sec)
0
10
Spikes
0
10
Rate
(s
pk/
sec)
Figure 4. Functional characterization of simple cell spikeletsA, simple cell spikelets paired with a simple principal cell. Spikelets are embedded in the membrane potential andamongst larger action potentials elicited by the principal neuron (left, inset). B, trial-averaged spike, membranepotential and spikelet responses to preferred sinusoidal grating show modulation to the stimulus temporalfrequency. C, one-dimensional noise stimulus reveals separated on and off subregions in both voltage andspikelet-triggered averages. D, same as in A for simple cell spikelets paired with a complex principal cell. E,same as in B for simple cell spikelets paired with a complex principal cell. Trial-averaged grating response ofmembrane potential and spikes shows no modulation at the stimulus temporal frequency, unlike the spikeletresponses. F, same as in C for simple cell spikelets paired with a complex principal cell. Spikelet-triggered averagerevealed separated receptive field on and off subregions, while complex cell membrane potential covariance hastwo prominent nonlinear filters.
J Physiol 593.22 Spikelet activity in cortical neurons 4987
As shown in an example record (Fig. 8A), the membranepotential is tuned to a binocular phase difference of−180 deg, while the spikelets are well tuned for −45 deg.Since spiking responses from the principal cell arerectified versions of membrane potential, differences intuning between principal cell spikes and spikelets wereeven greater (Priebe, 2008). In several rare cases wedid observe identical tuning across spikes, membranepotential and spikelets (Fig. 8B). Across our population,however, we found a very weak relationship in disparitypreference between membrane potential and spikelets(circular-correlation = 0.27 ± 0.13, SEM, n = 20,bootstrapped standard error) (Batschelet, 1981; Sokal &Rohlf, 1995). This lack of a relationship was also evidentin the differences in disparity preference for individualrecords (mean |�pref| = 103 ± 47.7 deg, SD). Similarity inorientation preference strongly suggests spikelets originatefrom nearby neurons in the same orientation columnas the principal neuron (Hubel & Wiesel, 1963; Essen& Zeki, 1978; Bosking et al. 1997; Ohki et al. 2005,2006; Yu & Ferster, 2013). Like the organization oforientation preference, ocular dominance columns are awell-established feature of binocularity in cat V1 (Hubel& Wiesel, 1962; LeVay et al. 1978; LeVay & Voigt, 1988;Katz & Crowley, 2002). We then compared eye preferencesof spikelets and principal neuron spikes, as measuredby the ocular dominance index (ODI; see Methods)
(Gordon & Stryker, 1996; Scholl et al. 2013). In 80%of our records, we found spikelets and the principalneuron shared eye preference (Fig. 8D, grey shading),reflecting our observations of orientation preference. Thedissimilarity in binocular disparity preference suggeststhat the columnar organization of binocular disparity pre-ference differs from columnar organization for orientationand ocular dominance (Kara & Boyd, 2009).
Correlations between spikelets and membranepotential
While it is clear that there are functional relationshipsbetween the principal neuron and spikelet records, it isunclear whether there is a direct link between spikeletsand the synaptic input onto the principal cells. Toreveal a possible relationship between spikelets andmembrane potential, we computed the cross-correlationor spikelet-triggered membrane potential average (Yu& Ferster, 2013). Since spikelets are embedded inmembrane potential, we carefully extracted spikelettimes during visual stimuli or spontaneous periods (seeMethods, examples from individual trials are shown inFig. 9A, left). Given the large, ongoing fluctuations ofmembrane potential, individual spikelet-related responsesappear noisy. Averaging membrane potential surrounding
A
D
Mem
bra
ne P
ote
ntia
l (m
V)
Rate
(s
pk/
sec)
Vrest
Spikelets
Rate
(s
pk/
sec)
Spikelets
Average Covariance
Space (deg)
Tim
e (
ms)
Mem
bra
ne P
ote
ntia
l (m
V)
Vm
(mV
)
Vm
-75
-45
0
20
Vm
(mV
)
Vm
-70
-50
0
30
-70
-55
-40
-25
-10
-80
-60
-40
-20
0
5 0 5 0
B
E
C
F10 mV
5 mV
Average Covariance
Spikes
Rate
(s
pk/
sec)
0
40
Spikes
Rate
(s
pk/
sec)
0
20
-0.2
0.2
-0.2
0.2
-0.2
0.2
-0.2
0.2
Figure 5. Functional characterization of complex cell spikeletsA–C, complex cell spikelets paired with simple principal cell. Same organization as in Fig. 2. D–F, complex cellspikelets paired with complex principal cell.
spikelets reveals systematic deviations from the meanmembrane potential (blue trace, inset in Fig. 9A,middle). Deviations could be the result of stimulus-evokedresponses, so we isolated correlated activity by shufflingtrials from which spikelets were derived and computeda shuffled spikelet-triggered average (dashed blue trace,inset Fig. 9A, middle). This shuffled average was sub-tracted from the raw spikelet-triggered average to revealchanges in membrane potential surrounding spikelets(Fig. 9A, middle). This shuffle-corrected spikelet-triggeredaverage displays three features: (1) a fast componentdue to the spikelet waveform centred around 0 ms lag,(2) a slower component with a positive cross-correlationlag (�3 ms) (Fig. 9A, middle, asterisk), and (3) a veryslow component that appears common network input tothe spikelet and principal cell. The second componentpossesses a peak amplitude of about 2 mV, exhibiting aneffect of depolarization following the spikelet. We alsoexamined this relationship during spontaneous activity tobe sure this correlation is unrelated to the visual stimulus.This revealed a similar, small positive lag following thespikelet in the shuffle-corrected cross-correlation (Fig. 9A,right) and an even larger slow component that may arisefrom common network input. A similar signature wasobserved in another example (Fig. 9B), although herethe spontaneous activity was less pronounced. Across ourspikelet records we observed small depolarizations witha positive cross-correlation lag in only a subset of cells(n = 4/20). Depolarization amplitudes were small forboth stimulus-driven activity (mean = 0.78 ± 0.51 mV,SD) and spontaneous activity (mean = 0.95 ± 0.73 mV,
F1/F0 Principal Neuron
F1/
F0
Spik
ele
ts
0
1
2
0 1 2
Figure 6. Relationship of principal neuron and spikelet simpleand complex classificationsThe modulation ratio (F1/F0) of spiking responses in principal neuronare plotted with spikelet response modulation ratio in the samerecord. A modulation ratio of greater than 1 is defined as a simplecell, while values less than 1 are complex cells.
SD). Peak depolarization lag times were also greater than0 (stimulus: mean = 5.4 ± 3.24 ms, SD; spontaneous:mean = 5.8 ± 4.0 ms, SD). Data collected fromstimulus-driven activity and spontaneous activity were notsignificantly different for either depolarization amplitudeor peak lag time (P = 0.56 and 0.67, respectively, Student’spaired t test).
In other spikelet records we observed large, slowdeflections in membrane potential occurring aroundthe time of the spikelet, which could be upward of10–15 mV. As shown in an example (Fig. 9C), thespikelet-triggered membrane potential average beganbefore the spikelet (�25 ms) and persisted long after.We observed these types of large, coordinated bumpsof activity in a subset of neurons which did not exhibitthe faster changes in membrane potential shown inthe previous examples (n = 4/20). Unfortunately, thesenetwork events mask smaller correlations between spikeletactivity and the principal neuron membrane potential,so we cannot extract any possible subtle depolarization.In the remaining proportion of spikelet records (12/20)we found no correlation between spikelet times andmembrane potential, as represented by another examplerecord (Fig. 9D). In this recording, there appears tobe no depolarization or hyperpolarization following thespikelet waveform, either during the stimulus period orfor spontaneous activity.
Given that some spikelet-triggered averages showsignatures of synaptic connectivity and others do not, wewondered whether these groups also differed in sharedfunctional selectivity. We found no difference betweengroups in the similarity of orientation and disparity pre-ferences. In records with a depolarization following thespikelet and those without a depolarization, orientationpreferences between principal neuron and spikelet weretightly coupled (mean |�pref| = 15.7 ± 11.4 deg,SD, n = 3; mean |�pref| = 12.7 ± 11.6 deg,SD, n = 7; respectively). These groups were notsignificantly different from one another (P = 0.52,Mann–Whitney test). The same was found for disparityselectivity, where spikelets and the correspondingprincipal neurons did not share similar preferences inrecords with and without cross-correlation signatures(mean |�pref| = 121.4 ± 76.1 deg, SD, n = 4; mean|�pref| = 128.0 ± 80.0 deg, SD, n = 12; respectively).These groups were also not significantly different fromone another (P = 0.95, Mann–Whitney test).
Discussion
Spikelets are observed in intracellular recordings, andyet characterization and a functional analysis of theirreceptive field properties have rarely been undertaken(Spencer & Kandel, 1961; Margrie et al. 2003; Epsztein et al.
J Physiol 593.22 Spikelet activity in cortical neurons 4989
2010, though see Martinez et al. 2014). We demonstratethat a number of properties indicate spikelets originatefrom single, nearby visual cortical neurons, distinct fromthe principal neuron. Spikelet waveforms are extremelyuniform within a recording session and exhibit refractoryperiods expected from a single neuronal source. Otherbiophysical and functional aspects were quite distinct fromthe principal neuron and indicate that spikelets have aseparate origin. First, spikelets are characterized by veryshort waveforms that are less than 2 ms, a time coursethat is too short to be the result of active processeswithin dendrites or passive polarization through gapjunctions (Gibson et al. 2005). Second, spikelet activityis unaffected by current injection into the patched cell,strongly suggesting they are an independent, extracellularsignal. Third, spikelets could be classified as simple orcomplex, independent of the classification of the principalneuron (Fig. 6). Finally, we found that the disparity pre-ferences of spikelets were not related to the disparity pre-ferences of the corresponding principal neurons (Fig. 8).Alongside these differences, however, spikelets sharedresponse properties with the principal neuron, including
orientation selectivity and eye preference (Figs 7 and 8),likely to be a product of the columnar organization in catV1.
A source for spikelets
One interpretation of the spikelet waveforms we observedis they result from electrical connections between neurons,formed via gap junctions (MacVicar & Dudek, 1981;Taylor & Dudek, 1982; Vigmond et al. 1997; Gibson et al.1999; Gibson et al. 2005). However, several aspects of ourdata are inconsistent with direct electrical coupling. First,because these waveforms are stereotyped and exhibit arefractory period, gap junctions would be required toexist only between pairs of neurons and not betweenlarger groups of neurons. Second, we would expect thatthe high resistance of an electrical connection wouldgreatly attenuate action potentials and elongate their timecourse (Gibson et al. 2005), whereas here we observe veryshort waveforms. Finally, injecting negative current into aneuron electrically connected with its neighbours shouldcause a hyperpolarization of membrane potential in those
Figure 7. Orientation selectivity of principal cells and spikeletsA, example orientation tuning measured from principal neuron and spikelets. Mean peak responses from principalcell spikes (top, black), principal cell membrane potential (middle, gray), and spikelets (bottom, blue) are shown.Double Gaussian fits are shown to characterize tuning. Error bars indicate standard error. B, same as A foranother example. C, relationship of orientation preferences of spikelets and principal cell membrane potential. D,distribution of orientation preference difference between spikelet and principal cell membrane potential.
neurons resulting in a reduction in spiking activity, whilehere we observe no effect of hyperpolarization on spikeletactivity.
An alternative interpretation is these waveformsoriginate from a neuron nearby our patch electrode andare recorded as an extracellular signal (Martinez et al.2014). Consistent with this interpretation, spikelet wave-forms appear to have action potential characteristics: theyare extremely rapid and uniform, and have refractory peri-ods consistent with cortical neurons. It is unclear, however,how spikelets were measured through our whole-cellpatch-clamp electrode. During an attempt to achieve awhole-cell patch, the membrane of the principal neuroncould be ruptured and fused with a neighbouring neuron,thereby providing a means for spikelet waveforms to bemeasured on the intracellular pipette. This seems unlikelygiven that hyperpolarization of the principal neuron didnot cause changes in spikelet activity (Fig. 2). Alternatively,the whole-cell patch pipette could record the extracellularsignals of a neighbouring neuron. In this case, spikelets
would be purely an extracellular signal, similar to aloose-cell attached or high impedance extracellular singleunit recording.
The original intent of our recordings was not to measurespikelet activity, and we were surprised to find spikelets insuch a high percentage of our measurements (21/76, 27%)when we performed post hoc examinations of our records.We hypothesize that the use of a higher sampling frequencythan our previous records allowed us to extract these wave-forms despite their short time course and small amplitude(Fig. 1). While we found no relationship between accessresistance and the presence or absence of spikelets, aspectsof the configuration used for blind whole-cell recordingsmay contribute to the ability to measure spikelets fromintracellular records.
Spikelet-membrane potential correlation
We found that spikelets exhibited three distinct effectsof influence on the membrane potential of principal
Spik
ele
tsM
em
bra
ne P
ote
ntia
l
A B C
Spik
es
Spik
e R
ate
(sp
k/se
c)
0
50
Spik
ele
t R
ate
(sp
k/se
c)
0
40
Mem
bra
ne
Pote
ntia
l (m
V)
0
30
Mem
bra
ne
Pote
ntia
l (m
V)
10
Spik
e R
ate
(sp
k/se
c)
0
40
Spik
ele
t R
ate
(sp
k/se
c)
0
16
Binocular Phase Difference (deg)
90 180 LE RE
Binocular Phase Difference (deg)
Principle Cell ODI
Spik
ele
t O
DI
D
r = 0.27
Membrane PotentialDisparity Preference (deg)
-180 0 180
Spik
ele
t D
isparity
Pre
fere
nce
(deg)
-180
0
180Example 1 Example 2
Contr
a
-90 0 90 180 LE RE-90 0
Ipsi
Contra Ipsi
Figure 8. Binocular disparity tuning of principal cells and spikeletsA, example disparity tuning measured from principal neuron and spikelets. Mean peak responses from principal cellspikes (top, black), principal cell membrane potential (middle, grey), and spikelets (bottom, blue) are shown. Bothresponses to binocular (circles) and monocular stimuli (squares) are shown. Cosine fits are used to characterizetuning. Error bars indicate standard error. B, same as A for another example. C, relationship of disparity preferencesof spikelets and principal cell membrane potential. D, relationship of eye preference or ocular dominance index(ODI) for spikelets and principal cell membrane potential.
J Physiol 593.22 Spikelet activity in cortical neurons 4991
cells. In some cases, it appears that spikelets are directlylinked to a depolarization, presumably via synaptic trans-mission (Fig. 9A and B). In other cases, there exists aspikelet-locked depolarization, but the depolarization isslow and extended, over a time scale that is unlikely to berelated to synaptic transmission. Instead, this prolongeddepolarization may be a result of common input drivingboth the spikelet and principal neuron (Yu & Ferster,
2010, 2013). For some records we find no relationshipbetween spikelet activity and membrane potential in theprincipal neuron. In none of these cases was spikeletactivity associated with a hyperpolarization of membranepotential. Finally, although our sample population issmall, none of these different patterns of linkage betweenspikelet and principal cell activity are related to any of thefunctional linkages measured here.
A
B
Exa
mple
1E
xam
ple
3
C
Individual Trials Stimulus Average Spontaneous Average
Exa
mple
4
D
Time from spikelet (ms)Time (ms)
Me
mb
ran
e P
ote
ntia
l (m
V)
Time (ms)
Time (ms)
Me
mb
ran
e P
ote
ntia
l (m
V)
Time (ms)
Time from spikelet (ms)Time (ms)
Me
mb
ran
e P
ote
ntia
l (m
V)
Time (ms)
Time from spikelet (ms) Time (ms)
Me
mb
ran
e P
ote
ntia
l (m
V)
5 mV
5 mV
5 mV
Exa
mple
2
Time from spikelet (ms)
5 mV
-25
Time (ms)
-25 0 250
5
10-55
-50
-25 0 250
5
10-60
-55
-25 0 25
0
5
10 -55
-50
-25 0 25
0
5
10 -55
-50
-25 0 250
10
20-55
-45
-25 0 250
10
20-55
-45
-25 0 25-5
0
5
10-65
-55
-25 0 25-5
0
5
10-60
-50
n = 162 n = 26
n = 2455 n = 19
n = 977 n = 196
n = 1068 n = 7
*
*
**
250
-25 250
-25 250
-25 250
Figure 9. Correlation of spikelet activityand membrane potentialA, individual trials of raw membranepotential with spikelet events shown (left).Stimulus-evoked spikelets aligned andaveraged together reveal a depolarizationlagging the spikelet waveform after shufflecorrection (middle). Inset depicts raw,spikelet-triggered average (continuous lines)and shuffled average (dotted lines). Trialswith action potentials from the principalneuron have been removed to avoidcontamination. Spontaneous spikelets havealso been isolated and averaged, revealing asimilar signature following the spikeletwaveform (right). Grey shading in triggeredaverages is standard error. Asterisks highlightdepolarization following spikelet waveform.B, same as in A for another example. C,example record with a large,stimulus-independent, correlated networkevent occurring simultaneously with thespikelet. Same layout as A. D, example ofspikelet displaying no effect on principal cellmembrane potential during visually evokedor spontaneous activity. Same layout as A.
Functional similarity between spikelet and principalneuron
In comparing the functional selectivity of spikelets andthe neuron recorded, we found orientation preference toclosely match (Fig. 7C and D). We interpret this match asthe consequence of both neurons being co-localized withinthe same cortical orientation column. Since orientationpreference in cat V1 systematically changes across thecortical surface but is similar throughout the layers ofcortex at a single point, a match would be expected ifsignals originated from nearby neurons (Hubel & Wiesel,1963; Essen & Zeki, 1978; Bosking et al. 1997; Ohki et al.2005, 2006; Yu & Ferster, 2013). Likewise, because of oculardominance columns (Hubel & Wiesel, 1962; LeVay et al.1978; LeVay & Voigt, 1988; Katz & Crowley, 2002), eyepreferences should also match if signals are from nearbyneurons, a feature we observed in our records (Fig. 8D).Further, although we did not directly measure spatialfrequency, gratings and one-dimensional noise used tovisually stimulate spikelets and principal neurons wereof a single spatial frequency or bar width. Therefore,strong activation of both signals with a single stimulussuggests both neurons have similar spatial frequencypreferences, another visual feature organized in cat V1(Issa et al. 2000). In contrast, we found disparity pre-ference is not matched between spikelets and principalneurons. This is surprising given a recent report of acolumnar organization for disparity selectivity (Kara &Boyd, 2009). The discrepancy between orientation anddisparity matches could potentially reflect a differentform of columnar organization for these two functionalproperties. As yet it is unclear how separate maps fororientation, disparity and other functional properties,such as spatial frequency and receptive field position, aremaintained along the two-dimensional cortical surfacein cat V1, but our results suggest that the columnarorganizations for these functional properties are distinct.
In addition to a lack of functional similarity for disparitypreference, there was a lack in functional similarity forsimple and complex cell classification (Fig. 6). Ever sincethese V1 cell classes were first described by Hubel &Wiesel (1962), a hierarchy has been proposed in whichmultiple thalamic relay neurons provide drive to simplecells, and subsequently multiple simple cells converge ona complex cell. If spikelets reflect this hierarchal flow ofinformation, it is surprising to find spikelets characterizedas complex associated with principal cells characterized assimple (Fig. 5A–C). One explanation for this associationis that inhibitory complex cells, untuned for orientation(Hirsch et al. 2003), provide a gain control signal to simplecells.
The functional differences between spikelet and prin-cipal cell pairs is surprising given the known large-scalearchitecture in V1. Whether these records reflect gap
junctions or simply nearby neurons, the presenceof distinct receptive field properties could reflect aproblem of representing many stimulus dimensions,including orientation, direction, spatial frequency, oculardominance, and disparity on a two-dimensional surface(Miller, 1996). High resolution imaging at the cellularlevel has revealed subnetworks at very fine spatial scales:segregated subnetworks of neurons exist within thelarger functional network (Yoshimura & Callaway, 2005),while the functional selectivity of neurons can shift overspatial scales of 20 μm (Ohki et al. 2005, 2006). Thedistinct properties in our records may therefore reflect afine-scale functional architecture within a larger columnarorganization for visual cortex.
References
Alonso JM, Usrey WM & Reid RC (2001). Rules of connectivitybetween geniculate cells and simple cells in cat primaryvisual cortex. J Neurosci 21, 4002–4015.
Batschelet E (1981). Circular Statistics in Biology. AcademicPress, London.
Berry MJ & Meister M (1998). Refractoriness and neuralprecision. J Neurosci 18, 2200–2211.
Bosking WH, Zhang Y, Schofield B & Fitzpatrick D (1997).Orientation selectivity and the arrangement of horizontalconnections in tree shrew striate cortex. J Neurosci 17,2112–2127.
Brainard D (1997). The psychophysics toolbox. Spat Vis 10,433–436.
Carandini F & Ferster D (2000). Membrane potential and firingrate in cat primary visual cortex. J Neurosci 20, 470–484.
Chung S & Ferster D (1998). Strength and orientation tuningof the thalamic input to simple cells revealed by electricallyevoked cortical suppression. Neuron 20, 1177–1189.
Conway BR & Livingstone MS (2003). Space-time maps andtwo-bar interactions of different classes of direction-selectivecells in macaque V-1. J Neurophysiol 89, 2726–2742.
Cumming BG & DeAngelis GC (2001). The physiology ofstereopsis. Annu Rev Neurosci 24, 203–238.
Cumming BG & Parker AJ (1997). Responses of primary visualcortical neurons to binocular disparity without depthperception. Nature 389, 280–283.
DeAngelis GC, Ohzawa I & Freeman RD (1991). Depth isencoded in the visual cortex by a specialized receptive fieldstructure. Nature 352, 156–159.
DeAngelis GC, Ohzawa I & Freeman RD (1993).Spatiotemporal organization of simple-cell receptive fields inthe cat’s striate cortex. I. General characteristics andpostnatal development. J Neurophysiol 69, 1091–1117.
Dermietzel R & Spray DC (1993). Gap junctions in the brain:where, what type, how many and why? Trends Neurosci 16,186–192.
Epsztein J, Lee AK, Chorev E & Brecht M (2010). Impact ofspikelets on hippocampal CA1 pyramidal cell activity duringspatial exploration. Science 327, 474–477.
Essen DC & Zeki SM (1978). The topographic organization ofrhesus monkey prestriate cortex. J Physiol 277, 193–226.
J Physiol 593.22 Spikelet activity in cortical neurons 4993
Ferster D, Chung S & Wheat H (1996). Orientation selectivityof thalamic input to simple cells of cat visual cortex. Nature21, 249–252.
Ferster D & Jagadeesh B (1992). EPSP-IPSP interactions in catvisual cortex studied with in vivo whole-cell patch recording.J Neurosci 12, 1262–1274.
Finn IM, Priebe NJ & Ferster D (2007). The emergence ofcontrast-invariant orientation tuning in simple cells of catvisual cortex. Neuron 54, 137–152.
Gibson JR, Beierlein M & Connors BW (1999). Two networksof electrically coupled inhibitory neurons in neocortex.Nature 402, 75–79.
Gibson JR, Beierlein M & Connors BW (2005). Functionalproperties of electrical synapses between inhibitoryinterneurons of neocortical layer 4. J Neurophysiol 93,467–480.
Golding NL & Spruston N (1998). Dendritic sodium spikes arevariable triggers of axonal action potentials in hippocampalCA1 pyramidal neurons. Neuron 21, 1189–1200.
Gordon JA & Stryker MP (1996). Experience-dependentplasticity of binocular responses in the primary visual cortexof mice. J Neurosci 16, 3274–3286.
Hietanen MA, Cloherty SL, Van Kleef JP, Wang C, Dreher B &Ibbotson MR (2013). Phase sensitivity of complex cells inprimary visual cortex. Neuroscience 237, 19–28.
Hirsch JA, Martinez LM, Pillai C & Alonso JM (2003).Functionally distinct inhibitory neurons at the first stage ofvisual cortical processing. Nat Neurosci 6, 1300–1308.
Hubel DH & Wiesel TN (1962). Receptive fields, binocularinteraction and functional architecture in the cat’s visualcortex. J Physiol 160, 106–154.
Hubel DH & Wiesel TN (1963). Shape and arrangement ofcolumns in cat’s striate cortex. J Physiol 165, 559–568.
Issa NP, Trepel C & Stryker MP (2000). Spatial frequency mapsin cat visual cortex. J Neurosci 20, 8504–8514.
Kara P & Boyd JD (2009). A micro-architecture for binoculardisparity and ocular dominance in visual cortex. Nature 458,627–631.
Katz LC & Crowley JC (2002). Development of cortical circuits:lessons from ocular dominance columns. Nat Rev Neurosci 3,34–42.
Landisman CE & Connors BW (2005). Long-term modulationof electrical synapses in the mammalian thalamus. Science310, 1809–1813.
Landisman CE, Long MA, Beierlein M, Deans MR, Paul DL &Connors BW (2002). Electrical synapses in the thalamicreticular nucleus. J Neurosci 22, 1002–1009.
LeVay S & Gilbert CD (1976). Laminar patterns ofgeniculocortical projection in the cat. Brain Res 113, 1–19.
LeVay S, Stryker MP & Shatz CJ (1978). Ocular dominancecolumns and their development in layer IV of the cat’s visualcortex: a quantitative study. J Comp Neuro 179, 223–244.
LeVay S & Voigt T (1988). Ocular dominance and disparitycoding in cat visual cortex. Vis Neurosci 1, 395–414.
Llinas R & Nicholson C (1971). Electrophysiological propertiesof dendrites and somata in alligator Purkinje cells. JNeurophysiol 34, 532–551.
MacVicar BA & Dudek FE (1981). Electrotonic couplingbetween pyramidal cells: a direct demonstration in rathippocampal slices. Science 213, 782–785.
Margrie TW, Meyer AH, Caputi A, Monyer H, Hasan MT,Schaefer AT, Denk W & Brecht M (2003). Targetedwhole-cell recordings in the mammalian brain in vivo.Neuron 39, 911–918.
Martinez LM, Molano-Mazon M, Wang X, Sommer FT &Hirsch JA (2014). Statistical wiring of thalamic receptivefields optimizes spatial sampling of the retinal image. Neuron81, 943–956.
McLean J & Palmer LA (1989). Contribution of linearspatiotemporal receptive field structure to velocity selectivityof simple cells in area 17 of cat. Vis Res 29, 675–679.
McLean J, Raab S & Palmer LA (1994). Contribution of linearmechanisms to the specification of local motion by simplecells in areas 17 and 18 of the cat. Vis Neurosci 11,271–294.
Miller KD (1996). Receptive fields and maps in the visualcortex: Models of ocular dominance and orientationcolumns. In Models of Neural Networks III, ed. Domany E,van Hemmen JL & Schulten K, pp. 55–78. Springer, NewYork.
Mohanty D, Scholl B & Priebe NJ (2012). The accuracy ofmembrane potential reconstruction based on spikingreceptive fields. J Neurophysiol 107, 2143–2153.
Movshon JA & Thompson ID (1978). Spatial summation in thereceptive fields of simple cells in the cat’s striate cortex. JPhysiol 283, 53–77.
Ohki K, Chung S, Ch’ng Y, Kara P & Reid RC (2005).Functional imaging with cellular resolution reveals precisemicro-architecture in visual cortex. Nature 433, 597–603.
Ohki K, Chung S, Kara P, Hubener M, Bonhoeffer T & Reid RC(2006). Highly ordered arrangement of single neurons inorientation pinwheels. Nature 442, 925–928.
Ohzawa I & Freeman RD (1986a). The binocular organizationof complex cells in the cat’s visual cortex. J Neurophysiol 56,243–259.
Ohzawa I & Freeman RD (1986b). The binocular organizationof simple cells in the cat’s visual cortex. J Neurophysiol 56,221–242.
Park I, Archer E, Priebe NJ & Pillow JW (2013). Spectralmethods for neural characterization using generalizedquadratic models. Adv Neural Inf Process Syst 26,2454–2462.
Pel X, Volgushev M, Vidyasagar TR & Creutzfeldt OD (1991).Whole cell recording and conductance measurements in catvisual cortex in-vivo. Neuroreport 2, 485–488.
Pelli D (1997). The VideoToolbox software for visualpsychophysics: Transforming numbers into movies. Spat Vis10, 437–442.
Priebe NJ (2008). The relationship between subthreshold andsuprathreshold ocular dominance in cat primary visualcortex. J Neurosci 28, 8553–8559.
Priebe NJ & Ferster D (2005). Direction selectivity of excitationand inhibition in simple cells of the cat primary visualcortex. Neuron 45, 133–145.
Priebe NJ, Mechler F, Carandini M & Ferster D (2004). Thecontribution of spike threshold to the dichotomy of corticalsimple and complex cells. Nat Neurosci 7, 1113–1122.
Reid RC & Alonso JM (1995). Specificity of monosynapticconnections from thalamus to visual cortex. Nature 378,281–283.
Scholl B, Tan AYY & Priebe NJ (2013). Strabismus disruptsbinocular synaptic integration in primary visual cortex. JNeurosci 33, 17108–17122.
Skottun BC, De Valois RL, Grosof DH, Movshon JA, AlbrechtDG & Bonds AB (1991). Classifying simple and complexcells on the basis of response modulation. Vis Res 31,1079–1086.
Smith SL, Smith IT, Branco T & Hausser M (2013). Dendriticspikes enhance stimulus selectivity in cortical neurons invivo. Nature 503, 115–120.
Spencer WA & Kandel ER (1961). Electrophysiology ofhippocampal neurons: IV. Fast prepotentials. J Neurophysiol24, 272–285.
Sokal RR & Rohlf FJ (1995). The Principles and Practice ofStatistics in Biological Research. W. H. Freeman, New York.
Tamas G, Buhl EH, Lorincz A & Somogyi P (2000). Proximallytargeted GABAergic synapses and gap junctions synchronizecortical interneurons. Nat Neurosci 3, 366–371.
Taylor CP & Dudek FE (1982). Synchronous neuralafterdischarges in rat hippocampal slices without activechemical synapses. Science 218, 810–812.
Yoshimura Y & Callaway EM (2005). Fine-scale specificity ofcortical networks depends on inhibitory cell type andconnectivity. Nat Neurosci 8, 1552–1559.
Yu J & Ferster D (2010). Membrane potential synchrony inprimary visual cortex during sensory stimulation. Neuron68, 1187–1201.
Yu J & Ferster D (2013). Functional coupling from simple tocomplex cells in the visually driven cortical circuit. J Neurosci33, 18855–18866.
Additional information
Competing interests
None of the authors have any conflicts of interests for this paper.
Author contributions
B.S. and N.J.P. conceived and designed experiments. B.S.collected and assembled the data. B.S. and S.A. analysed thedata. B.S. and N.J.P. interpreted the data. B.S. and N.J.P. wroteand revised the manuscript. All experiments were performed atUT Austin in Austin, TX, USA.
Funding
This work was supported by grants from the National Institutesof Health (EY-019288) and the PEW Charitable Trusts.
