Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception

Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception

Report Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception Highlights d Temporal spike patterns may shape freq...

859KB Sizes 0 Downloads 15 Views

Report

Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception Highlights d

Temporal spike patterns may shape frequency perception regardless of spike count

d

Periodicity is not the most salient temporal cue for vibrotactile frequency

d

Tactile frequency is determined by duration of the silent gap between spike bursts

d

This new code is well suited to signal naturalistic complex vibratory patterns

Birznieks & Vickery, 2017, Current Biology 27, 1–6 May 22, 2017 ª 2017 Elsevier Ltd. http://dx.doi.org/10.1016/j.cub.2017.04.011

Authors Ingvars Birznieks, Richard M. Vickery

Correspondence [email protected]

In Brief By controlling spike timing in tactile afferents, Birznieks and Vickery demonstrate that temporal spiking patterns may shape frequency perception regardless of spike count. The most salient temporal feature for vibrotactile frequency was the duration of the silent gap between bursts of neural activity and not the periodicity as previously expected.

Please cite this article in press as: Birznieks and Vickery, Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception, Current Biology (2017), http://dx.doi.org/10.1016/j.cub.2017.04.011

Current Biology

Report Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception Ingvars Birznieks1,2,3,4,* and Richard M. Vickery1,2,3 1School

of Medical Sciences, Faculty of Medicine, UNSW Sydney, Sydney, NSW 2052, Australia Research Australia, Barker Street, Randwick, NSW 2031, Australia 3These authors contributed equally 4Lead Contact *Correspondence: [email protected] http://dx.doi.org/10.1016/j.cub.2017.04.011 2Neuroscience

SUMMARY

Skin vibrations sensed by tactile receptors contribute significantly to the perception of object properties during tactile exploration [1–4] and to sensorimotor control during object manipulation [5]. Sustained low-frequency skin vibration (<60 Hz) evokes a distinct tactile sensation referred to as flutter whose frequency can be clearly perceived [6]. How afferent spiking activity translates into the perception of frequency is still unknown. Measures based on mean spike rates of neurons in the primary somatosensory cortex are sufficient to explain performance in some frequency discrimination tasks [7–11]; however, there is emerging evidence that stimuli can be distinguished based also on temporal features of neural activity [12, 13]. Our study’s advance is to demonstrate that temporal features are fundamental for vibrotactile frequency perception. Pulsatile mechanical stimuli were used to elicit specified temporal spike train patterns in tactile afferents, and subsequently psychophysical methods were employed to characterize human frequency perception. Remarkably, the most salient temporal feature determining vibrotactile frequency was not the underlying periodicity but, rather, the duration of the silent gap between successive bursts of neural activity. This burst gap code for frequency represents a previously unknown form of neural coding in the tactile sensory system, which parallels auditory pitch perception mechanisms based on purely temporal information where longer inter-pulse intervals receive higher perceptual weights than short intervals [14]. Our study also demonstrates that human perception of stimuli can be determined exclusively by temporal features of spike trains independent of the mean spike rate and without contribution from population response factors. RESULTS The current study consists of a series of three linked experiments that first address a fundamental sensory encoding question

about whether the temporal structure of spike trains determines perceived frequency of vibrotactile stimuli and then tests a possible encoding mechanism in two subsequent experiments. Is Temporal Structure of Spike Trains a Key Factor in Determining Frequency Perception? The aim of the first experiment was to investigate how the temporal structure of spike trains that consist of periodic clusters of multiple spikes (resembling responses to high-amplitude vibration) influences the perception of vibrotactile frequency. It addresses the sensory processing challenge of how afferent inputs are interpreted so as to achieve constancy of frequency perception, regardless of variation in mean spike rates of single afferents and the afferent population. The spike rate in a single afferent depends on the vibration amplitude [15, 16]; at amplitudes above the tuning threshold, the afferent becomes entrained and generates one spike per sinusoidal vibratory cycle (1:1), and so the mean spike rate would indeed represent the frequency; however, it may instead be double or triple that rate at higher vibration amplitudes of the same frequency [15, 16]. Thus, it is unlikely that the rate code alone would be sufficient to explain vibrotactile frequency encoding. Nevertheless, until now, due largely to methodological limitations, it has not been possible to demonstrate how the temporal structure of spike trains is interpreted by the nervous system. The main obstacle has been that changes in the frequency of a sinusoidal stimulus concurrently entail complex changes in the population of afferents responding, making it impossible to tease out the effect of temporal features of the afferent response and assess quantitatively how they influence the perceived frequency of vibrotactile stimuli. We were able to investigate this question by using brief pulsatile mechanical stimuli (see, for example, [17]) to generate an arbitrary pattern of spike trains in the responding afferents without affecting any other parameters of the afferent population response. Four different stimuli were tested (Figure 1A, 1–4) designed to evoke spike trains consisting of periodic bursts of two to four spikes spaced 4.35 ms (stimuli 1–3) or 8.70 ms (stimulus 4) apart. In the current study we focused on the perception of vibrotactile stimuli within the frequency range referred to as a flutter (5–60 Hz) [16]; higher-frequency ranges were not considered to avoid possible confounding effects by primary afferent refractoriness and adaptation over time of testing. Due to the stereotyped nature of the pulsatile mechanical stimuli, and their short duration, which is comparable to the refractory period of the Current Biology 27, 1–6, May 22, 2017 ª 2017 Elsevier Ltd. 1

Please cite this article in press as: Birznieks and Vickery, Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception, Current Biology (2017), http://dx.doi.org/10.1016/j.cub.2017.04.011

A

Stimuli (bursts)

Period 43.5 ms; 23 Hz

#1

tL

#2

tL

39.15 ms 25.5 Hz

tL

#3

34.80 ms 28.7 Hz 30.45 ms 32.8 Hz

tL

#4

26.10 ms 38.3 Hz

23 bursts/s 46 spikes/s 23 bursts/s 69 spikes/s 23 bursts/s 92 spikes/s 23 bursts/s 69 spikes/s

Period

B

Spike trains evoked in human FA-I afferent Count 25

#1 0 25

#2 0 25

#3 0 25

#4

10 μV

20 ms 0

0 Time 30 (ms)

Figure 1. Time-Controlled Afferent Spike Trains Created by Short Pulsatile Mechanical Stimuli (A) Schematic representation of a set of four pulsatile vibrotactile stimuli intended to evoke spike trains with the same repetition rate (base periodicity) but strikingly different mean spike rates. Each vertical line indicates the time of the mechanical pulse. (B) Illustrative recording of spike trains evoked in a single human FA-I tactile afferent, demonstrating a perfect match with the temporal pattern of pulsatile stimuli as shown in (A) (one spike per stimulus pulse). Identified single spikes are highlighted in red. Cycle histograms on the right show binned spike count over the 1-s stimulation period, recorded in four consecutive 1-s stimulation periods each consisting of 23 bursts. Each mechanical pulse generated a single action potential without failing on a single occasion. The dashed horizontal line corresponds to the count of 23 spikes. One bin corresponds to 1/40th of the period (1.1 ms; the cycle histogram is truncated at 30 ms). The onset latency is the time from when the command was sent to the stimulator at the beginning of each burst to when the spike was recorded from the median nerve at the wrist level; spike jitter was less than 1 ms. See also Figure S1.

action potential, each mechanical stimulation event generated only a single time-controlled spike in each responding peripheral afferent regardless of repetition rate (see STAR Methods). To illustrate this, we obtained single tactile afferent recordings in human subjects using microneurography (Figure 1B, left panels; Figure S1). Cycle histograms showing binned spike count (right panels in Figure 1B) demonstrate the reliable firing of the 2 Current Biology 27, 1–6, May 22, 2017

afferent, reproducing the precise stimulus pattern within each vibrotactile stimulation cycle, over the 1 s stimulus period. Were the mean spike rate to determine the perceived frequency, there would be significant differences between the perceived frequencies for the four stimuli (ranging from 46 to 92 Hz). Alternatively, if a temporal feature of the spike train related to its periodicity such as the burst rate determines the perceived frequency, then all four stimuli would be perceived to have equal frequency (23.0 Hz, the burst rate). Perceived frequency for the four stimuli was measured in psychophysical experiments as the point of subjective equality (PSE) in a forced-choice comparison with regular pulses designed to evoke spike trains with uniform inter-spike intervals (equivalent to one spike per cycle; see STAR Methods and Figure S2). In contrast to our expectations, the perceived frequency of the four stimuli (Figure 2A, boxplots) could be explained by neither the mean spike rate (green arrowheads), nor the burst rate (purple arrowheads). For example, stimulus 1, which consisted of two spikes per burst, resembling the afferent response when two spikes are generated per sinusoidal cycle, did not cause a doubling of the perceived frequency, which instead only increased by about 10% (25.3 Hz; Table S1). The most dramatic difference between the perceived frequency and that predicted by the mean spike rate was with stimulus 3, which had a quadrupled spike count. If these spikes were evenly spaced, the perceived frequency would be 92.0 Hz, but the PSE in our experiment was only 32.5 Hz—almost 3-fold lower (Figure 2A, stimulus 3). This dramatic effect presents the first direct unequivocal evidence that the temporal structure of the spike train in tactile afferents does determine the perception of vibrotactile frequency within the flutter range in humans. The only stimulus property that matched the perceptual experience was the reciprocal of the longest inter-spike interval tL (Figure 2A, blue arrowheads). This inter-spike interval is the silent period between successive bursts of neural activity (indicated as tL in Figure 1A) and is henceforth referred to as the burst gap. There was a very close match between the perceived frequency and predicted frequency based on the duration of the burst gap (Figure 2B; R2 = 0.93, df = 4 for the goodness of fit to the identity line, slope = 1, intercept = 0). Does Spike Number within a Burst Determine the Perceived Frequency? Although the subjects reported that frequency perception was clear and that they could make a judgment regardless of other cues, we had to exclude the possibility that the variation in the number of spikes may have acted as an intensity cue and confounded frequency perception. The aim of experiment 2 was to determine whether the number of spikes in a burst biased the subjects making their frequency judgments. Stimuli 1d–4d (see Figure 3A) had the duration of the silent period between bursts identical to stimuli used in experiment 1 (Figure 1), but the number of spikes in a burst was fixed at two (a doublet). We used the same psychophysical methods to determine the perceived frequencies of this set of stimuli and found they were no different from corresponding stimuli with multiple spikes in experiment 1 (Figure 3B; Table S1). Two-way repeated-measures ANOVA (excluding stimulus 1 and 1d as they were identical between the experiments) indicated that there was a statistically significant main effect by the duration of the burst gap

Please cite this article in press as: Birznieks and Vickery, Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception, Current Biology (2017), http://dx.doi.org/10.1016/j.cub.2017.04.011

Frequency (Hz)

A 100

Perceived frequency

B 50

Perceived frequency

Figure 2. Spike Rate Does Not Predict the Perceived Frequency in Psychophysical Tests of Complex Tactile Stimuli

80

PSE (Hz)

(A) Boxplots with whiskers represent the point of subjective equality (PSE) data obtained with stimuli illustrated in Figure 1. The box extends from the 25th to 60 75th percentiles; the whiskers extend to minimum and maximum values, and circle symbol indicates median. 40 Burst gap (1/tL) 30 The horizontal dashed line indicates mean (n = 12). Predicted PSE values if the frequency judgment was Burst rate 20 based on mean spike rate code are indicated using green arrowheads; burst rate, purple arrowheads, 20 0 reciprocal of burst gap 1/tL, blue arrowheads. The #1 #2 #3 #4 20 30 40 50 experimentally obtained PSE values indicate that Reciprocal of tL (Hz) neither the mean spike rate code nor burst rate code Stimulus can explain the perceived frequency; the closest match is the reciprocal of tL. (B) The reciprocal of burst gap tL provides good match with PSEs. The solid black line connects data points obtained by averaging all of the individual subject data, which are shown by thin gray lines (n = 12). The dotted red line represents the line of identity. See also Figure S2. Spike rate

40

(p < 0.0001, F2,44 = 104, accounting for 66% of total variation), but no effect by the number of spikes in the burst (p = 0.23, F1,22 = 1.52, accounting for 1% of total variation), with no interaction effect (p = 0.43, F2,44 = 0.87). Thus, corresponding stimuli with different numbers of spikes in a burst yielded the same PSEs, which indicates that the number of spikes within bursts does not determine the perceived frequency. What Defines a Burst? We next investigated how closely two spikes must follow each other to be treated as a burst within the flutter frequency range. In experiment 3, we used a set of stimuli that created spike trains consisting of repeating patterns of one doublet with a short interspike interval followed by one long inter-spike interval. Different durations of the shorter inter-spike interval (tS) ranging from 4.35 to 65.25 ms were tested. The duration of the longest inter-spike interval was the same (tL: 87.00 ms; representing the burst gap) for all stimuli in the set (Figure 4A). The prediction from the proposed burst gap code is that, while two spikes in a doublet follow each other closely enough to be treated as a burst, the duration of the shorter inter-spike interval will have no effect on the perceived frequency, which will be solely determined by the duration of the burst gap and equal to 11.5 Hz (Figure 4B, blue

Stimuli (doublets)

Period 43.5 ms; 23 Hz

#1d

tL

#2d

tL

#3d

tL

#4d

46 spikes/s

B

39.15 ms 25.5 Hz

tL Period

34.80 ms 28.7 Hz 30.45 ms 32.8 Hz 26.10 ms 38.3 Hz

PSE (Hz)

A

line). The experimental data showed that there was a complex relationship between the duration of the shorter inter-spike interval (tS) and the perceived frequency (Figure 4B, orange line). The joinpoint regression model [18] algorithm (see STAR Methods) determined that the psychophysical data were best explained by a curve divided into three linear segments, the first extending from data points at 4.35 to 13.05 ms, the second from 13.05 to 30.45 ms, the last from 30.45 to 64.95 ms. The initial segment had a slope not significantly different from zero (p = 0.71); the other two segments had slopes significantly different from zero (0.26 and 0.07; p < 0.0002). We interpret this to mean that while the inter-spike interval between successive spikes is short (up to about 15 ms) they are grouped into a burst and the burst gap code (Figure 4B, blue line) provides a satisfactory explanation of the data. As the duration of the shorter inter-spike interval increases beyond 15 ms, the propensity for two consecutive spikes to be regarded as a burst gradually decreases, and eventually each individual spike becomes attributed to a separate vibrotactile sensory event that contributes to the perceived frequency. Consequently, the inter-spike interval between these individual spikes beyond 30 ms becomes essentially analogous to the burst gap yielding a result equivalent to the mean spike rate (Figure 4B, green line). Figure 3. Number of Spikes within the Burst Does Not Affect the Perceived Frequency

Perceived frequency 50 45 40 35 30 25 20 15

bursts 2-4 spikes doublets

25

30

35

Reciprocal of tL (Hz)

40

(A) Schematic representation of the stimuli used in the second experiment. The index ‘‘d,’’ standing for ‘‘doublet,’’ is added to the stimulus number to distinguish corresponding stimuli from those in experiment 1. (B) PSE values for the perceived frequency of stimuli shown in (A) (blue line with diamonds; with 95% confidence intervals are shown; n = 12). PSEs for the stimuli depicted in Figure 1 are plotted for comparison (orange line with squares, slightly horizontally shifted). The dashed black lines indicate PSE values predicted by the burst gap code model. Despite the different number of spikes in the two sets of corresponding stimuli, the perceived frequencies were similar. See also Table S1.

Current Biology 27, 1–6, May 22, 2017 3

Please cite this article in press as: Birznieks and Vickery, Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception, Current Biology (2017), http://dx.doi.org/10.1016/j.cub.2017.04.011

Stimuli

tS

tL

tS

B

tL

Frequency (Hz)

A

Perceived frequency

20

15

10

Psychophysical data Mean spike rate Reciprocal of tL

5 tL = 87 ms (fL = 11.5 Hz) tS = n x 4.35 ms

0

5 10 15 20 25 30 35 40 45 50 55 60 65 70

(A) Schematic representation of pulsatile vibrotactile stimuli consisting of a repeating pattern of one shorter interval constituting a doublet followed by an 87-ms-long interval. (B) PSE values obtained in psychophysical tests shown with 95% confidence level shaded in gray (n = 12). Individual data points indicated by orange squares and three linear segments as determined by joinpoint regression model indicated by three connected orange lines. Predicted PSE values if the frequency judgment was based on: mean spike rate code (green line); the burst gap code (blue line).

Shortest inter-spike interval ts (ms)

DISCUSSION The current study contributes to the mounting evidence that the temporal structure of spike trains is essential for vibrotactile frequency perception. Our experiments clearly demonstrate that features of stimuli reflected in the temporal pattern of afferent responses do determine perception and thus must be utilized at some levels of neural processing. This remains true, regardless of any possible subsequent conversion to other coding schemes. Most importantly, for the first time it reveals some essential principles about how this temporal information is interpreted in the tactile system shaping perception. Previous experimental evidence suggested that, along the path from primary afferents [19] to cuneate nucleus [20], thalamus [21], S1 [22], and higher centers, temporal and rate coding mechanisms may both contribute to stimulus discrimination. Experiments using periodic and aperiodic vibrotactile stimuli have demonstrated that neurometric curves based on the mean spike rates of cortical neurons were sufficient to explain trained monkeys’ performance in a frequency discrimination task [7–11]. More recent work suggests that the mean spike rate may operate over relatively short time periods such as 200 ms, enabling discrimination between certain patterns of neural activity that have identical rates over a 1 s period [23]. On the other hand, in the human ability to discriminate stimuli such as diharmonic and noise waveforms, the timing of individual spikes plays a significant role, both at the level of peripheral afferents [13] and neurons in the somatosensory cortex [12, 24]. Our new findings point to a way to reconcile the literature in regard to the importance of temporal features of the spiking pattern versus the mean discharge rate in different experimental contexts. For example, studies by Romo’s group that showed a predominant role for the mean spike rate code used 20-ms-long single-cycle mechanical sinusoid pulses to stimulate fingertips with a minimum inter-stimulus interval of 20 ms [8, 10], which accords with the time interval in which we found the burst gap code begins to provide frequency perception similar to the mean rate code. Conversely, we note that the work demonstrating a significant role for temporal coding used frequencies of 50–1,000 Hz [13], corresponding to inter-spike intervals shorter than 20 ms. We have also confirmed that perception 4 Current Biology 27, 1–6, May 22, 2017

Figure 4. Relationship between the Shortest Inter-spike Interval Duration and the Perceived Vibrotactile Frequency

is not primarily determined by spiking pattern periodicity [8] but instead is determined by duration of individual inter-spike intervals [13]. A vibrotactile frequency perception mechanism based on the longest inter-spike interval does not detect the true periodicity of skin vibrations, which, for example, a burst rate code might do. However, unlike periodicity-based codes, this might represent a more universal method of encoding naturalistic complex vibratory patterns during surface-structure exploration or when detecting and acting upon mechanical events during object manipulation. Although experimental stimuli used to study the nervous system and in clinical diagnostic tests rely on pure frequency vibration, the majority of naturally occurring vibrotactile stimuli have a complex frequency composition. The advantage of the coding scheme that relies on individual inter-spike intervals is that it is potentially suitable for application to various types of stimuli with complex frequency compositions, or even those without fixed periodicity. The significance of temporal features of spiking activity appears to be a shared characteristic between sensory systems [25]. It has been demonstrated that temporal frequency channels are linked across audition and touch [26], and several tactile analysis mechanisms are envisaged to be analogous with the auditory system [27]. Indeed, the phenomenon that inter-spike intervals of longer duration receive higher weights than short ones has been previously documented in the auditory system [14, 28]. The experimental techniques largely resembled our approach: pulse trains were presented acoustically to normal listeners and electrically to users of a cochlear implant, while place-of-excitation was held constant. The similarities of this new temporal coding mechanism with those reported in the auditory system also seems consistent with the substantial evidence that spike timing plays an important role across a variety of neural systems such as in encoding of complex stimulus features [4, 19, 22, 29–31], underpinning essential processes in learning [32], and contributing to neuronal information processing and pattern recognition [33–37]. Finding and studying the neural circuits that give rise to the performance observed in the current study may help further our understanding of the general significance of this form of sensory processing in somatosensory and auditory systems [27].

Please cite this article in press as: Birznieks and Vickery, Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception, Current Biology (2017), http://dx.doi.org/10.1016/j.cub.2017.04.011

STAR+METHODS Detailed methods are provided in the online version of this paper and include the following: d d d d

d

KEY RESOURCES TABLE CONTACT FOR REAGENT AND RESOURCE SHARING EXPERIMENTAL MODEL AND SUBJECT DETAILS METHOD DETAILS B Experimental design B Spike train generation B Microneurographic experiments B Psychophysical experiments QUANTIFICATION AND STATISTICAL ANALYSIS

SUPPLEMENTAL INFORMATION Supplemental Information includes two figures and one table and can be found with this article online at http://dx.doi.org/10.1016/j.cub.2017.04.011.

AUTHOR CONTRIBUTIONS I.B. and R.M.V. conceived and conducted the study, analyzed data, and prepared the manuscript.

ACKNOWLEDGMENTS

8. Salinas, E., Hernandez, A., Zainos, A., and Romo, R. (2000). Periodicity and firing rate as candidate neural codes for the frequency of vibrotactile stimuli. J. Neurosci. 20, 5503–5515. 9. Romo, R., Herna´ndez, A., Zainos, A., and Salinas, E. (1998). Somatosensory discrimination based on cortical microstimulation. Nature 392, 387–390. 10. Romo, R., and Salinas, E. (2003). Flutter discrimination: neural codes, perception, memory and decision making. Nat. Rev. Neurosci. 4, 203–218. 11. de Lafuente, V., and Romo, R. (2005). Neuronal correlates of subjective sensory experience. Nat. Neurosci. 8, 1698–1703. 12. Harvey, M.A., Saal, H.P., Dammann, J.F., 3rd, and Bensmaia, S.J. (2013). Multiplexing stimulus information through rate and temporal codes in primate somatosensory cortex. PLoS Biol. 11, e1001558. 13. Mackevicius, E.L., Best, M.D., Saal, H.P., and Bensmaia, S.J. (2012). Millisecond precision spike timing shapes tactile perception. J. Neurosci. 32, 15309–15317. 14. Carlyon, R.P., van Wieringen, A., Long, C.J., Deeks, J.M., and Wouters, J. (2002). Temporal pitch mechanisms in acoustic and electric hearing. J. Acoust. Soc. Am. 112, 621–633. 15. Johnson, K.O. (1974). Reconstruction of population response to a vibratory stimulus in quickly adapting mechanoreceptive afferent fiber population innervating glabrous skin of the monkey. J. Neurophysiol. 37, 48–72. 16. Talbot, W.H., Darian-Smith, I., Kornhuber, H.H., and Mountcastle, V.B. (1968). The sense of flutter-vibration: comparison of the human capacity with response patterns of mechanoreceptive afferents from the monkey hand. J. Neurophysiol. 31, 301–334. 17. Rothenberg, M., Verrillo, R.T., Zahorian, S.A., Brachman, M.L., and Bolanowski, S.J., Jr. (1977). Vibrotactile frequency for encoding a speech parameter. J. Acoust. Soc. Am. 62, 1003–1012.

We would like to thank Mr. Edward Crawford (UNSW Sydney) for developing hardware and software to interface the Optacon stimulator, Mr. Hilary Carter (NeuRA) for mechanical works, and Ms. Barbara Toson (NeuRA) for statistical advice. We thank Mr. Kevin Ng, Mr. Ian Tse, Ms. Sara Nilsson, and Ms. Josefin Andersen for assistance with data collection. This work was supported by NHMRC project grant APP1028284 to I.B. and R.M.V.

18. Kim, H.J., Fay, M.P., Feuer, E.J., and Midthune, D.N. (2000). Permutation tests for joinpoint regression with applications to cancer rates. Stat. Med. 19, 335–351.

Received: January 18, 2017 Revised: March 9, 2017 Accepted: April 10, 2017 Published: May 4, 2017

20. Jo¨rntell, H., Bengtsson, F., Geborek, P., Spanne, A., Terekhov, A.V., and Hayward, V. (2014). Segregation of tactile input features in neurons of the cuneate nucleus. Neuron 83, 1444–1452.

REFERENCES 1. Hollins, M., Bensmaı¨a, S.J., and Roy, E.A. (2002). Vibrotaction and texture perception. Behav. Brain Res. 135, 51–56. geas, G. (2009). The role 2. Scheibert, J., Leurent, S., Prevost, A., and Debre of fingerprints in the coding of tactile information probed with a biomimetic sensor. Science 323, 1503–1506. 3. Manfredi, L.R., Saal, H.P., Brown, K.J., Zielinski, M.C., Dammann, J.F., 3rd, Polashock, V.S., and Bensmaia, S.J. (2014). Natural scenes in tactile texture. J. Neurophysiol. 111, 1792–1802.

19. Johansson, R.S., and Birznieks, I. (2004). First spikes in ensembles of human tactile afferents code complex spatial fingertip events. Nat. Neurosci. 7, 170–177.

21. Ahissar, E., Sosnik, R., and Haidarliu, S. (2000). Transformation from temporal to rate coding in a somatosensory thalamocortical pathway. Nature 406, 302–306. 22. Zuo, Y., Safaai, H., Notaro, G., Mazzoni, A., Panzeri, S., and Diamond, M.E. (2015). Complementary contributions of spike timing and spike rate to perceptual decisions in rat S1 and S2 cortex. Curr. Biol. 25, 357–363. 23. Rossi-Pool, R., Salinas, E., Zainos, A., Alvarez, M., Vergara, J., Parga, N., and Romo, R. (2016). Emergence of an abstract categorical code enabling the discrimination of temporally structured tactile stimuli. Proc. Natl. Acad. Sci. USA 113, E7966–E7975. €rinen, J. (1969). 24. Mountcastle, V.B., Talbot, W.H., Sakata, H., and Hyva Cortical neuronal mechanisms in flutter-vibration studied in unanesthetized monkeys. Neuronal periodicity and frequency discrimination. J. Neurophysiol. 32, 452–484.

4. Weber, A.I., Saal, H.P., Lieber, J.D., Cheng, J.W., Manfredi, L.R., Dammann, J.F., 3rd, and Bensmaia, S.J. (2013). Spatial and temporal codes mediate the tactile perception of natural textures. Proc. Natl. Acad. Sci. USA 110, 17107–17112.

25. VanRullen, R., Guyonneau, R., and Thorpe, S.J. (2005). Spike times make sense. Trends Neurosci. 28, 1–4.

5. Johansson, R.S., and Flanagan, J.R. (2009). Coding and use of tactile signals from the fingertips in object manipulation tasks. Nat. Rev. Neurosci. 10, 345–359.

26. Yau, J.M., Olenczak, J.B., Dammann, J.F., and Bensmaia, S.J. (2009). Temporal frequency channels are linked across audition and touch. Curr. Biol. 19, 561–566.

6. Mountcastle, V.B., Talbot, W.H., Darian-Smith, I., and Kornhuber, H.H. (1967). Neural basis of the sense of flutter-vibration. Science 155, 597–600.

27. Saal, H.P., Wang, X., and Bensmaia, S.J. (2016). Importance of spike timing in touch: an analogy with hearing? Curr. Opin. Neurobiol. 40, 142–149.

7. Luna, R., Herna´ndez, A., Brody, C.D., and Romo, R. (2005). Neural codes for perceptual discrimination in primary somatosensory cortex. Nat. Neurosci. 8, 1210–1219.

, A., and Winter, I.M. (2004). Physiological cor28. Pressnitzer, D., Cheveigne relates of the perceptual pitch shift for sounds with similar waveform autocorrelation. Acoust. Res. Lett. Online 5, 1–6.

Current Biology 27, 1–6, May 22, 2017 5

Please cite this article in press as: Birznieks and Vickery, Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception, Current Biology (2017), http://dx.doi.org/10.1016/j.cub.2017.04.011

29. Petersen, R.S., Panzeri, S., and Diamond, M.E. (2002). The role of individual spikes and spike patterns in population coding of stimulus location in rat somatosensory cortex. Biosystems 67, 187–193.

35. Panzeri, S., and Diamond, M.E. (2010). Information carried by population spike times in the whisker sensory cortex can be decoded without knowledge of stimulus time. Front. Synaptic Neurosci. 2, 17.

30. Gollisch, T., and Meister, M. (2008). Rapid neural coding in the retina with relative spike latencies. Science 319, 1108–1111.

36. Tiesinga, P., Fellous, J.M., and Sejnowski, T.J. (2008). Regulation of spike timing in visual cortical circuits. Nat. Rev. Neurosci. 9, 97–107.

31. Pruszynski, J.A., and Johansson, R.S. (2014). Edge-orientation processing in first-order tactile neurons. Nat. Neurosci. 17, 1404–1409. 32. Caporale, N., and Dan, Y. (2008). Spike timing-dependent plasticity: a Hebbian learning rule. Annu. Rev. Neurosci. 31, 25–46.

37. VanRullen, R., and Thorpe, S.J. (2002). Surfing a spike wave down the ventral stream. Vision Res. 42, 2593–2615.

33. Masquelier, T., Hugues, E., Deco, G., and Thorpe, S.J. (2009). Oscillations, phase-of-firing coding, and spike timing-dependent plasticity: An efficient learning scheme. J. Neurosci. 29, 13484–13493.

38. Gardner, E.P., and Palmer, C.I. (1989). Simulation of motion on the skin. I. Receptive fields and temporal frequency coding by cutaneous mechanoreceptors of OPTACON pulses delivered to the hand. J. Neurophysiol. 62, 1410–1436.

34. Gu¨tig, R., and Sompolinsky, H. (2006). The tempotron: a neuron that learns spike timing-based decisions. Nat. Neurosci. 9, 420–428.

39. Vickery, R.M., Morley, J.W., and Rowe, M.J. (1993). The role of single touch domes in tactile perception. Exp. Brain Res. 93, 332–334.

6 Current Biology 27, 1–6, May 22, 2017

Please cite this article in press as: Birznieks and Vickery, Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception, Current Biology (2017), http://dx.doi.org/10.1016/j.cub.2017.04.011

STAR+METHODS KEY RESOURCES TABLE

REAGENT or RESOURCE

SOURCE

IDENTIFIER

Excel 2010

Microsoft

N/A

GraphPad Prism 7

GraphPad Software

N/A

LabChart 8

ADInstruments,

N/A

Joinpoint Trend Analysis Software v 4.2.0.2

National Cancer Institute

https://surveillance.cancer.gov/joinpoint/download

Software and Algorithms

CONTACT FOR REAGENT AND RESOURCE SHARING Further information and requests for resources should be directed to and will be fulfilled by the Lead Contact, Ingvars Birznieks ([email protected]). EXPERIMENTAL MODEL AND SUBJECT DETAILS Research subjects were healthy volunteers aged 20 to 43 without any known history of neurological disorders which would affect the somatosensory system. Ethics approval was obtained from the UNSW Human Research Ethics Committee, and all subjects signed a consent form. Twelve different subjects were tested in each of three Experiments (1-3). The gender breakdown in three experiments was 7/5, 7/5 and 9/3 male/female, respectively. In Experiments 1 and 2 two subjects were the authors, while the other subjects were naive to the research objective. The sample size was determined by pilot studies to determine effect size, and according to accepted practice in psychophysical experiments. The microneurography experiments were performed on 6 subjects (4 males, 2 females). METHOD DETAILS Experimental design The pre-specified hypothesis was that the perceived frequency would not be directly related to the rate of impulse activity in the peripheral afferents. The results from Experiment 1 suggested an additional hypothesis: that the perceived frequency is determined by the burst gap. The study was a controlled laboratory experiment involving behavioral measurements of the ability to discriminate different vibrotactile frequencies. Measurements were made by recording button presses indicating which of the test-comparison stimulus pair was perceived to have the higher frequency. Spike train generation Spike trains with desired temporal features were generated using short pulsatile mechanical stimuli delivered using an Optacon piezoelectric bimorph pin-array stimulator (Telesensory Systems, Silicon Valley, USA). Each mechanical pulse is a reproducible and uniform event which ensures that the same population of afferents will be excited regardless of the rate at which these pulses are repeated. Optacon pins were controlled by a custom-built interface that enabled switching individual pins on/off during each 4.35 ms long cycle set by the internal circuits of the device. The block containing the pin array was removed from the body of Optacon and mounted on a cantilevered arm to maintain 35 g contact force on the fingertip. The Optacon amplitude dial was set to maximum, which produces 60 mm pin excursion under no-load condition. Stimuli were delivered by two active pins 2 mm apart in the same row to the tip of the index finger of the dominant hand. Microneurographic experiments Previous studies have demonstrated that the Optacon evokes one spike per protraction in fast adapting type 1 (FA-I) and type 2 (FA-II) tactile afferents without activating slowly adapting (SA) afferents [38]. We confirmed those observations and verified our methodological approach by microneurographic single afferent recordings. Nerve impulses (spikes) were recorded from single tactile afferents with tungsten needle electrodes inserted percutaneously into the median nerve of human subjects. The median nerve was located at the wrist by cathodal stimulation (1.0–4.5 mA, 0.2 ms, 1 Hz) delivered by an optically isolated constant-current stimulator (Stimulus Isolator, ADInstruments, Sydney, Australia). An insulated tungsten microelectrode (FHC, Bowdoin, ME) was then inserted percutaneously into the nerve at the wrist; an uninsulated microelectrode inserted subcutaneously 1 cm away served as the reference electrode. Neural activity was amplified (320,000, 0.3–5.0 kHz) using a low-noise headstage (NeuroAmpEX, ADInstruments) and single mechanoreceptor afferent responses isolated by small adjustments to the electrode position. Isolated

Current Biology 27, 1–6.e1–e2, May 22, 2017 e1

Please cite this article in press as: Birznieks and Vickery, Spike Timing Matters in Novel Neuronal Code Involved in Vibrotactile Frequency Perception, Current Biology (2017), http://dx.doi.org/10.1016/j.cub.2017.04.011

afferents were identified by their receptive field location and by their afferent class, based on the response to sustained indentation and mechanical thresholds measured with monofilaments (Semmes-Weinstein Aesthesiometers, Stoelting, Chicago, IL). The recordings were obtained from 16 tactile afferents - six fast adapting afferents (5 FA-I and 1 FA-II) and 10 slowly adapting afferents (8 SA-I and 2 SA-II). Various stimulation patterns were tested. When optimally positioned, the mechanical pulse evoked a single spike in responding FA afferents while no response was observed in SA afferents. The responding afferents faithfully reproduced the stimulation pattern over prolonged periods of testing compatible with the duration of our psychophysical experiments. We tested how reliably the afferent is capable of generating spikes at high repetition rates, as shown in Figure S1 which has inter-spike interval histograms for one FA afferent (different from afferent shown in Figure 1B). The data shown are obtained stimulating the afferent at the two highest repetition rates used in the current study. When stimulation pulses were applied only 4.35 ms apart (230 Hz) for 10 s, 2300 successive mechanical pulses generated 2300 time-locked spikes with minimal jitter due to the short duration of the stimulation pulse. In comparison, the highest number of successive pulses 4.35 ms apart used in experiments described in the current paper was only 4 (compared with the 2300 exemplified in Figure S1A). Figure S1B shows the inter-spike interval histogram for the same afferent excited by mechanical pulses 8.7 ms apart applied for 10 s (1150 pulses in total). Psychophysical experiments PSE for each stimulus was obtained by a two-interval forced choice paradigm. Test stimuli were compared with regular trains of pulses, applied for 1 s each, in random order, separated by 0.5 s. The subject had 3 s to indicate which stimulus had higher frequency by pressing one of two buttons. Noise delivered via headphones masked auditory cues. To obtain the psychometric curve, test stimuli were compared 20 times each against six regular train stimuli with uniform inter-stimulus intervals. Data were recorded by button press using PowerLab/LabChart (ADInstruments, Sydney, Australia), then exported to Excel (Microsoft, USA). For each comparison frequency, we calculated the proportion of times the participant responded that it was higher in frequency than the test stimulus (PH). The logit transformation (ln(PH/(1-PH))) was applied to produce a linear psychometric function using standard methods [39] (see example in Figure S1). Perceived frequency is represented as the PSE (point of subjective equality) which is the value of the comparison frequency that is equally likely to be judged higher as judged lower than the test stimulus; it was determined as the frequency at the zero crossing of the logit axis determined from a regression line fitted to the logit transformed data. QUANTIFICATION AND STATISTICAL ANALYSIS Linear regression was performed on the data shown in Figure 2B to determine if the burst gap model provided a satisfactory explanation of the experimental data (R2 and df reported in the text; n values representing number of subjects given in Figure 2 legend; mean, median, min, max, 25 and 75 percentiles shown in Figure 2A). Two-way repeated-measures ANOVA was used on the data in Figure 3B to determine if the data were better explained by the number of spikes or by the burst gap or an interaction of these factors (p and F values including degrees of freedom are reported in the text; n values representing number of subjects given in Figure 3 legend; mean PSEs with 95% confidence intervals shown in Figure 3B and given in Table S1). Prism 7 (GraphPad Software, USA) was used for these analyses. We used the joinpoint regression model [18] implemented by Joinpoint Trend Analysis Software (v 4.2.0.2, National Cancer Institute, USA) to determine the appropriate number of linear segments to fit to the data of Figure 4B. After determining that two joinpoints gave the best explanation, the joinpoint fit for the slope values was based on 13 observations, with 6 parameters - three slopes and three intercepts (p values and significant slopes reported in the text; n values representing number of subjects given in Figure 4 legend; mean PSEs and confidence intervals shown in Figure 4B).

e2 Current Biology 27, 1–6.e1–e2, May 22, 2017