tmp719B TMP
tmp719B TMP
tmp719B TMP
Systems/Circuits
Acoustic environments are composed of complex overlapping sounds that the auditory system is required to segregate into discrete
perceptual objects. The functions of distinct auditory processing stations in this challenging task are poorly understood. Here we show a
direct role for mouse auditory cortex in detection and segregation of acoustic information. We measured the sensitivity of auditory
cortical neurons to brief tones embedded in masking noise. By altering spectrotemporal characteristics of the masker, we reveal that
sensitivity to pure tone stimuli is strongly enhanced in coherently modulated broadband noise, corresponding to the psychoacoustic
phenomenon comodulation masking release. Improvements in detection were largest following priming periods of noise alone, indicating that cortical segregation is enhanced over time. Transient opsin-mediated silencing of auditory cortex during the priming period
almost completely abolished these improvements, suggesting that cortical processing may play a direct and significant role in detection
of quiet sounds in noisy environments.
Key words: broadband; coherence; masking release; noise; opsin; pyramidal cell
Significance Statement
Auditory systems are adept at detecting and segregating competing sound sources, but there is little direct evidence of how this
process occurs in the mammalian auditory pathway. We demonstrate that coherent broadband noise enhances signal representation in auditory cortex, and that prolonged exposure to noise is necessary to produce this enhancement. Using optogenetic
perturbation to selectively silence auditory cortex during early noise processing, we show that cortical processing plays a crucial
role in the segregation of competing sounds.
Introduction
In the acoustic world, animals are challenged to detect salient
sounds in noisy backgrounds, a process of critical importance in
communication, hunting, and threat-detection. The auditory
system is well suited to the task as it demonstrates a remarkable
spectral (Rosenblith and Stevens, 1953) and temporal resolution
(Klumpp and Eady, 1956; Plomp, 1964), and this acuity is valuable for detecting changes in natural soundscapes (McDermott et
al., 2013; Andreou et al., 2015). Predictable, or coherent, nonrandom features of soundscapes may be exploited by the auditory
Received Feb. 29, 2016; revised Aug. 25, 2016; accepted Aug. 29, 2016.
Author contributions: J.S. and P.C. designed research; J.S. performed research; J.S. and P.C. analyzed data; J.S.
and P.C. wrote the paper.
This work was supported by MRC Career Development Award (G1000512) to P.C. We thank Andrei Kozlov for
feedback on the manuscript; and Matthew Brown and Alexander Morris for discussions and support.
The authors declare no competing financial interests.
This article is freely available online through the J Neurosci Author Open Choice option.
Correspondence should be addressed to Dr. Paul Chadderton, Department of Bioengineering, Imperial College
London, Exhibition Road, London SW7 2AZ, United Kingdom. E-mail: p.chadderton@imperial.ac.uk.
DOI:10.1523/JNEUROSCI.0656-16.2016
Copyright 2016 Sollini and Chadderton
This is an Open Access article distributed under the terms of the Creative Commons Attribution License
Creative Commons Attribution 4.0 International, whichpermitsunrestricteduse,distributionandreproductioninany
medium provided that the original work is properly attributed.
system to improve sound processing (Ulanovsky et al., 2003; Taaseh et al., 2011; Yaron et al., 2012; Teki et al., 2013; Krishnan et
al., 2014; Nelken, 2014). A prevalent feature of natural sound is
coherent amplitude fluctuations across frequency, termed comodulation (CM). CM is present in both environmental sounds
and vocalizations (Nelken et al., 1999). Given its pervasiveness in
nature, CM may be a critical cue for grouping and segregating
overlapping sounds (Nelken et al., 1999; Krishnan et al., 2014).
Comodulation masking release (CMR) is a psychoacoustic phenomenon whereby adding coherently modulated noise to an existing masker makes signals easier to perceive (Hall et al., 1984). This
effect is striking as additional noise energy normally reduces, or does
not change, signal detectability (Fletcher, 1940). CMR encompasses
two separate processes, dependent on the relative frequencies of the
noise and the signal: within-channel CMR (signal and noise similar
in frequency) and across-channel CMR (signal and noise dissimilar
in frequency). Within-channel CMR can be implemented in the
auditory periphery, but across-channel or true CMR (Verhey et
al., 2003) cannot be explained by mechanical processes in the ear and
is sensitive to cues of auditory grouping (Buss et al., 2009; Dau et al.,
2009; Verhey et al., 2012). Across-channel CMR is therefore a result
of brain processing, but the mechanism and location of such processing are not well understood.
Only a small number of studies have directly sought to understand the representation and underlying mechanism(s) of CMR
at the cellular level. In the peripheral auditory system, neuronal
responses to pure tones are enhanced by across-channel CM in a
way consistent with human behavior (Pressnitzer et al., 2001).
However, it is not clear whether this information is inherited or
influenced by processing at later stages of the auditory system. A
CMR correlate has been shown to develop progressively between
the inferior colliculus, medial geniculate body, and auditory cortex (Nelken et al., 1999; Las et al., 2005), although this work
explored both within- and across-channel cues simultaneously.
As such, it remains unclear how much of the observed CMR is
attributable to across-channel processes (Verhey et al., 2003;
Grose et al., 2005a). Neuronal correlates of within-channel CMR
have been observed in the avian auditory forebrain area L2a;
however, when measured in an across-channel configuration
(comparing narrowband [NB] and broadband comodulated
noise), no significant CMR was found (Nieder and Klump, 2001;
Hofer and Klump, 2003).
In this study, we set out to quantify the influence of acrosschannel CM on signal detectability in the neuronal activity of
primary auditory cortex (A1), a critical site in auditory perception (Bizley and Cohen, 2013). We then sought to establish
whether processing by auditory cortical circuits plays a functional
role in the formation of across-channel CMR. We provide the
first quantification of across-channel CMR in A1. Transient inactivation during early sound processing significantly reduces
subsequent masking release, suggesting that auditory cortex may
play a causal role in enhancing signal representation in the presence of comodulated noise.
SR ijk "
Results
Signal- and noise-evoked activity in primary auditory
cortical populations
Our first aim was to establish whether neuronal correlates of
across-channel CMR are present in individual neurons of A1
(Linden et al., 2003). Simultaneous extracellular population recordings were made in anesthetized mice (Fig. 1A), and evoked
UF
A1
AAF
1
2
3
4
A BCCB
DD
Time (s)
Narrowband
(NB)
Incoherent
modulation (IM)
11.8
12.9
14.1
20
28.2
30.8
33.6
20
Coherent
modulation (CM)
Freq (kHz)
Freq (kHz)
Freq (kHz)
11.8
12.9
14.1
20
28.2
30.8
33.6
SNR (dB)
20
15
10
5
0
5
10
Noise
alone
A2
Spikes/s
0.2
Noise alone
5dB SNR
1
Time (s)
Noise alone
5dB SNR
Noise alone
5dB SNR
0.1
0
Spikes/s
0.1
0
0.5
0.75
0.5
0.75
1
Time (s)
0.5
0.75
Figure 1. Signal-evoked activity in mouse primary auditory cortex in the presence of masking noise. A, Experimental setup.
Recordings were made from primary auditory cortex (A1) using 4 # 2 tetrodes. Extracellularly recorded action potentials (spikes;
bottom left) were sorted to isolate evoked single neuron activity (right). BD, Pure tone signals were presented in three different
noise backgrounds: (B) NB (red), (C) broadband with IM (blue), and (D) broadband with comodulated envelopes (CM, black). BD,
Bottom, Raster plots of evoked spiking activity of a single cell. Each mark indicates a spike (30 consecutive trials for each condition).
Noise was presented alone and in the presence of three tone pips at different SNRs (ordinate). Gray dashed lines indicate noise
onset. Pink backgrounds represent the tone presentation window. EG, PSTHs of the same cell for noise alone (light colored lines)
A
Signal response
( STDs)
NB
2
5dB SNR
IM
CM
5dB SNR
5dB SNR
0
0.5
0.75
0.5
0.75
0.5
0.75
Time (s)
B
SNR (dB)
1.3
0.7
0
NB
-10
0.5
0.75
0.5
0.75
Time (s)
0.5
0.75
34
20
10
0
-10
.5 .75 .5 .75 .5 .75 .5 .75 .5 .75 .5 .75 .5 .75
Time (s)
1.6
0.8
0
20
NB
15
IM
10
CM
5
0
10
15
20
25
30
35
Characteristic frequency (kHz)
40
F
0.8
CM
IM
NB
0.6
Signal response
(FR in zscore)
0.1
CM
10
SNR (dB)
IM
Signal response
0.4
0.2
0
10
0
5
SNR (dB)
10
8
6
4
2
0
NB IM CM
Condition
Figure 2. Comodulated flanking noise produces large masking release in signal-tuned neurons. A, Signal-evoked response of
individual cell (shown in Fig. 1BG) in NB, IM, and CM backgrounds at 5 dB SNR (left to right). B, Population mean signal response.
The mean signal response for the NB, IM, and CM (left to right) conditions was calculated across all cells demonstrating a significant
signal response (nNB ( 887, nIM ( 842, and nCM ( 797). C, The effect of frequency selectivity on signal sensitivity. Mean signal
response in CM noise for subpopulations grouped by CF (see Materials and Methods). D, Signal sensitivity of CF-defined subpopulations in different masking conditions. The sensitivity of each subpopulation was defined using 0.25 SD change in firing rate as a
A
0.4
CM
IM
NB
0.2
0.1
0.5
Time (s)
C
10
132
NB (spikes/s)
NB (spikes/s)
10
6
4
2
0
5
CM (spikes/s)
342
10
10
145
IM (spikes/s)
6
4
2
0
5
IM (spikes/s)
336
10
229
8
6
4
2
0
5
CM (spikes/s)
232
10
Figure 3. Broadband noise facilitates ongoing noise-evoked responses. A, Normalized firing rates of cells tuned close to the signal frequency (1523 kHz; ntuned ( 499) for noise-only conditions.
Black represents CM. Blue represents IM. Red represents NB. B, Comparison of individual firing rates (averaged from between 0.1 and 0.4 and 0.6 1 s) for the CM and NB conditions. Red dashed line
indicates when the firing rates are matched. Value in the top left corner ( number of points falling above the line; bottom right ( number of points falling below the line. C, Comparison of individual
firing rates for the IM and NB conditions. D, Comparison of individual firing rates for the IM and CM conditions. Adding spectral energy facilitates ongoing noise responses regardless of temporal
features of the masker.
significant increase in firing rate for both broadband (CM and IM)
conditions relative to the NB condition across this population (Fig.
3; sign test, CM, p '' 0.01, IM, p '' 0.01). Furthermore, in individual neurons, differences between CM and NB noise-evoked responses during the signal window were not correlated with masking
release (Pearsons Product-moment correlation; PPMC, r ( 0.048,
p ( 0.36). This demonstrates that there is no evidence of active
cancellation (or reduction) of neuronal responses to noise at the level
of cortex and the magnitude and relative change of noise-evoked
activity in CM versus NB conditions during the signal window were
not predictive of signal sensitivity. The effect of coherent and incoherent modulation on the ongoing representation of noise responses
was also contrasted (Fig. 3D). There was no significant difference in
spiking responses to the noise between coherent and incoherent
modulation, suggesting that increases in firing rate were spectral and
not related to modulation or across-channel processes (Fig. 3D; sign
test, p ( 0.93).
Psychophysical experiments indicate that properties of the
early portion of the background noise, such as onset synchrony, duration, and spectrotemporal statistics, are critical in
the formation of a large across-channel CMR (McFadden and
Wright, 1990, 1992; Grose and Hall, 1993; Dau et al., 2005;
Grose et al., 2009; Verhey et al., 2012). We therefore looked for
evidence of a role for early noise processing in masking release
at the level of A1. Initially, we compared the relative amplitude
of the early periods (first 100 ms) of broadband (IM and CM)
and NB noise-evoked responses for individual neurons, and
found marked differences in neuronal firing rates between
onset responses for each noise conditions. Evoked responses
to broadband noise were often either suppressed or facilitated
relative to NB noise (Fig. 4A). Differences between CM and
NB noise-evoked responses were quantified with a metric we
termed Across Frequency Interaction (AFI), where positive
values indicated facilitation and negative values indicated suppression in noise-evoked activity to the onset in the broad
relative to narrow-band conditions (Fig. 4 A, B; see Material
and Methods). We investigated the relationship between AFI
and masking release in all cells demonstrating significant responses to the signal (nall ( 1271). The addition of the flanking noise tended to facilitate NB noise onset responses,
demonstrated by slightly positive, yet significant, mean AFI
values (mean CM AFI ( 0.0284, sign test, p '' 0.01, IM AFI (
Signal response
0.2
0.1
Time(s)
FI =
0.2
0.3
I=
AF
0.8
0.3
0
0
0.1
Time(s)
0.2
0.6
0.8
0.8
10
0.3
0.6
0.8
8
6
4
2
0.6
0.8
0.6
0.8
Time(s)
AFI distribution
NB-CM
NB-IM
0.1
0
0.8
MR = 4.77 dB
0.6
20
0.3
0.6
0.2
0.2
Time(s)
0.4
MR = -3.4 dB
0.6
-0.1
MR = -1.07 dB
0.8
10
SNR (dB)
I
AF
0.3
=-
0.4
CM
MR = 1.29 dB
0.6
-0.1
Proportion of cells
I=
AF
-0.5
6
2
0
2
0.8
0.4
AFI binned by CF
0.3
NB-CM
NB-IM
0.2
0.1
0
0.1
4
0.4
0
0.4
Across-frequency interaction
E
Across-frequency Interaction
0.8
IM
20
SNR (dB)
NB
Average change
in FR (z-score)
Average change
in FR (z-score)
0.2
0
0.2
0.4
Across-frequency interaction
10
20
30 40 50
Characteristic frequency (kHz)
Figure 4. Early noise responses are predictive of masking release. A, Noise-evoked responses from two example cells demonstrating different AFIs when comparing between NB and broadband
noise conditions. Red line indicates the response to the NB noise. Blue line indicates the response to broadband IM noise. Black line indicates the response to broadband CM noise. Hatched lines
indicate differences in evoked response between NB and broadband conditions (NB IM and NB CM; blue and black AFI values, respectively; see Materials and Methods). B, Signal response
functions for the same cells in different masker conditions. Mean firing rate changes were used to derive thresholds for each condition (dashed white lines). Differences in thresholds correspond to
masking release (MR) produced by adding broadband comodulated bands, indicated in each panel. C, AFI distribution for all cells in IM (light gray) and CM (dark gray) conditions. CM is associated
with negative AFI values (i.e., reduced evoked activity to noise onset). D, Relationship between AFI and masking release. Individual cells were binned based on their AFI. Negative AFIs values are
associated with increased masking release in broadband maskers. E, Relationship between AFI and frequency tuning. Neurons tuned at or close to signal frequency exhibited negative AFIs in the
presence of broadband CM but not IM noise.
condition (ANOVA, F(5) ( 2.15, p ( 0.0574). In addition, a significant correlation between AFI and masking release was found in
single units for the CM condition (Fig. 4D; PPMC, r ( %0.12, p ''
0.01) but fell short of significance for the IM condition (PPMC, r (
%0.059, p ( 0.067). Therefore, at least for the CM condition, early
noise responses (quantified here with AFI) were predictive of subsequent masking release. Finally, the relationship between AFI and CF
was investigated. AFI varied significantly with CF for both broadband masking conditions (MANOVA, p '' 0.01). Post hoc tests
revealed both CM (ANOVA, F(23) ( 2.22, p ( 0.0008) and IM
(ANOVA, F(23) ( 4.18, p '' 0.01) varied significantly with CF. For
the CM background, the AFI was lowest close to the signal frequency
("20 kHz), where it became negative (i.e., suppressive). Comparison of noise onset firing rates in the NB and CM condition con-
5dB SNR
Signal response
( STDs)
long
0.6
short
-0.4
0.5
0.75
Time (s)
1.3
0
long
0.5
0.1
short
0.75
Time (s)
1.0 0.5
0.75
Time (s)
0.2
0
10
0
5
SNR (dB)
10
4
2
0
0.4
or
0.6
sh
0.8
long
short
ng
1.0
lo
10
0.7
Signal response
10
SNR (dB)
Signal response
(FR in zscore)
Figure 5. Priming noise exposure is necessary for large masking release. A, Signal-evoked
responses in short (yellow) and long (black) maskers. B, Population mean signal response for
short- and long-CM-noise conditions (ntuned-CM ( 358). C, Population signal response function.
D, Population detection thresholds. Cell responses were converted into a population neurometric function from the thresholds estimated using binomial probability ( p ' 0.05 Bonferonni
corrected).
PV-Cre mouse
Signal
Masker
Laser
A1
Anti-PV
UF
A1
A2
Merge
PV+ cells
AAF
EYFP
PV- cells
0.5
SNR (dB)
0.5
0
0
D
1
5dB SNR
0.8
5dB SNR
0.6
0.4
0.2
0
Time (s)
0.7
10
0
Normalised FR
1.3
control
0.5
0.1
A1 inactivation
0.75
0.5
Time (s)
0.75
Signal response
(FR in zscore)
Normalised FR
10
Signal response
control
A1 inactivation
0.8
0.6
0.4
0.2
0
10
0
SNR (dB)
10
Figure 6. Preceding cortical inactivation does not influence subsequent signal-evoked responses. A, Top, AAV-EF1a-DIO-hChR2(H134R)-EYFP was injected into A1 of Pvalb-IRES-Cre mice.
Bottom, Specific infection of PV ! cells was confirmed via coexpression (merged image, right) of ChR2 (green, left) and PV % antibody (red, middle). B, Optogenetic inactivation of auditory cortex
before presentation of noise and signal. Schematic showing approach to functionally measure the role of cortical early sound processing via transient selective inactivation of A1. Activation of
ChR2-expressing PV ! interneurons (teal circles) was used to inhibit ongoing cortical activity. C, Brief (150 ms) laser stimulation of A1 induced strong activation of PV ! cells (left) and inhibition of
PV % cells (right). The fidelity of the approach was confirmed by comparing population PSTH responses with noise presentation in the presence (red) or absence (yellow) of preceding laser
stimulation. While cortical activity was strongly modulated during laser stimulation, subsequent auditory-evoked responses were unchanged in both PV ! and PV % cells. D, Signal-evoked
responses from PV % cells were tested using brief CM noise maskers in the absence (control, yellow) and presence (A1 inactivation, red) of preceding laser stimulation. E, Laser stimulation did not
influence the population signal response (ntuned-CM ( 358). F, Population signal response function under the two conditions. No significant differences were observed.
Discussion
How the brain detects and segregates stimuli is an important
question for understanding sensory perception. Here we have
identified a neuronal correlate of across-channel CMR, an auditory phenomenon that improves signal detection, in A1 of the
mouse. CM increases signal detectability in cortical neurons and
also alters initial noise-evoked responses in a manner predictive
0.6
5dB SNR
long
(A1 inactive)
long
5dB SNR
5dB SNR
0.5
0.5
0.75
Time (s)
Time (s)
-0.4
0.1
long
0.5
0.75
Time (s)
1.0 0.5
0.75
Time (s)
1.0
0.6
0.4
0.2
0
5
0
5
SNR (dB)
10
4
2
0
(A
10
ng
0.8
lo
in n g
ac
tiv
e)
long
long (A1 inactive)
lo
SNR (dB)
0.7
10
1.3
Signal response
Signal response
( STDs)
Figure 7. Processing of the early noise is required for large masking release in auditory cortex. A, The role of A1 in across-channel CMR was tested by inactivating A1 during the early noise period
(left, long; A1 inactive). Blue bar represents laser on. Middle, Single-cell example PSTHs for the normal (top, long, black line) and silenced (bottom, long; A1 inactive, teal) conditions. Right, Signal
response for during long-CM masker, with (teal) and without (black) preceding A1 inactivation. B, Population mean signal response for long, and long-A1 inactive conditions (ntuned-CM ( 358).
C, Population signal response function. D, Population detection thresholds. Cell responses were converted into a population neurometric function from the thresholds estimated using binomial
probability ( p ' 0.05 Bonferonni corrected). Cortical inactivation during early presentation of the noise reduced improvements in signal detection thresholds.
References