WangpersName, LipersName, LipersName, ZhangpersName, ChenpersName, ZengpersName, LiaopersName (2018date). “Frequency selectivity of echo responses in the mouse primary auditory cortex.” Scientific Reportstitle·j.
Object of study — echo responsesterm - activity at the anticipated time after a rhythmic tone train ends - in layer 2/3 neurons of mouse A1term
Research question — whether a neuron's echo responseterm depends on which frequency was used, that is on its own frequency tuning
Contribution — shows echo responsesterm occur preferentially at each neuron's best frequency, in both anaesthetised and awake mice, while their spatial arrangement stays locally heterogeneous
phenomenon · claim-type · cross-domain-transfer · mechanism · claim-relation · replication-status · limitation — grounded in the full text| phenomenon | echo responses (omitted stimulus-evoked responses) in primary auditory cortex — “Specifically, whether the echo responses of the neurons in the auditory cortex are related to their frequency selectivity remains unknown.” | |
| claim-type | existence — “We confirmed the presence of echo responses in a subgroup of mouse Layer 2/3 A1 neurons following a train of rhythmic pure tone stimulation.” | |
| method — evidence | — “Therefore, we used in vivo two-photon Ca 2+ imaging to record the neuronal activities in the mouse A1 to elucidate the relationship between their echo responses and frequency selectivity.” | |
| object of study — evidence | — “In the primary auditory cortex (A1), neuronal ensembles are activated relative to anticipated sound events following rhythmic stimulation, but whether the echo responses of the neurons are related to their frequency selectivity remains unknown.” | |
| research question — evidence | — “Although the prediction correlated neural signals have already been observed in multiple sensory regions in different mammals, the stimulus type for the preferentially evoked echo responses is unclear. Specifically, whether the echo responses of the neurons in the auditory cortex are related to their frequency selectivity remains unknown.” | |
| contribution — evidence | — “After testing with a range of frequencies, we found that these echo responses occurred preferentially close to the best frequencies of the neurons.” | |
| transfer | from machine learning and Bayesian statistics: predictive coding and iterative error minimisation, imported as the interpretive frame for an omitted-stimulus response — “Theoretically, predictive coding suggests that the brain is constantly making efforts to ‘fit’ internal models in perceiving through inferring the incoming sensory input. The discrepancies between the actual sensory data and the predicted sensory data (‘error’) are used to refine the model in an iterative process until the model optimized” | transfers[0] |
| transfer | from network science: scale-free topology and hub nodes, imported as the circuit explanation for rhythm following — “We propose that, there is scale-free topology in the cortical network, which contains hub neurons in L2/3 and many low-degree loops with different lengths in L5 (Fig. 5 ; modified from the previous study 5 ). These low-degree loops enable the neural system to follow the external input rhythm, while hub neurons can trigger synchronization in the whole network. During learning, the connectivity from the L5 loops to the L2/3 hubs is highly strengthened, which allows the signals to flow across the entire network. After learning through a train of rhythmic sensory stimulation, a loop with certain length is selected to match the input rhythm. Finally, the network is able to generate echo responses with the same rhythm.” | transfers[1] |
| mechanism | A scale-free cortical network in which layer 5 carries many low-degree loops of different lengths and layer 2/3 carries hub neurons — “We propose that, there is scale-free topology in the cortical network, which contains hub neurons in L2/3 and many low-degree loops with different lengths in L5” | mechanisms[0] |
| mechanism | During rhythmic stimulation a layer 5 loop whose length matches the stimulus interval is selected, and after the sound stops that loop keeps regenerating the rhythm through the network — “After learning through a train of rhythmic sensory stimulation, a loop with certain length is selected to match the input rhythm. Finally, the network is able to generate echo responses with the same rhythm.” | mechanisms[1] |
| mechanism | Low-degree loops track the external rhythm while hub neurons synchronise the whole network so the echo is expressed population-wide — “These low-degree loops enable the neural system to follow the external input rhythm, while hub neurons can trigger synchronization in the whole network.” | mechanisms[2] |
| mechanism | a proposed scale-free cortical topology: low-degree L5 loops of differing length follow the input rhythm, one loop is selected to match it, and L2/3 hub neurons synchronise the network — “We propose that, there is scale-free topology in the cortical network, which contains hub neurons in L2/3 and many low-degree loops with different lengths in L5 (Fig. 5 ; modified from the previous study 5 ). These low-degree loops enable the neural system to follow the external input rhythm, while hub neurons can trigger synchronization in the whole network. During learning, the connectivity from the L5 loops to the L2/3 hubs is highly strengthened, which allows the signals to flow across the entire network. After learning through a train of rhythmic sensory stimulation, a loop with certain length is selected to match the input rhythm. Finally, the network is able to generate echo responses with the same rhythm.” | mechanisms[3] |
| mechanism | the echo is treated as an intrinsic property of cortical circuits responding to the absence of an expected sound, not as a residue of the sound response — “Therefore, it implies that echo response is an intrinsic property of cortical circuits in response to the absence of an expected sound, and echo response may reflect a distinct neural process in comparison to sound-evoked responses.” | mechanisms[4] |
| mechanism | longer latency and lower amplitude of echoes attributed to across-neuron variation in estimating the input rhythm — “Those differences could be attributed to a variation in neurons for the estimation of the input rhythm, and it thus leads to a longer latency and a lower amplitude of the echo responses on average.” | mechanisms[5] |
| mechanism | the imaged region is identified as A1 by its thalamic connectivity, using retrograde tracer to MGBv, rather than by the response property under study — “To verify that the imaged cortical regions were located in the A1, we used the criterion 44 , 45 that the ventral part of the lateral medial geniculate body (MGBv) is connecting with A1.” | mechanisms[6] |
| mechanism | indicator sensitivity is decisive because most cortical neurons fire only one or two spikes to a sound, so a dye that misses single spikes misses the population — “The sensitivity plays the most important role in studying the organization of the A1, because most of cortical neurons respond with only one or two spikes to sound stimulation” | mechanisms[7] |
| mechanism | predictive coding framing: the brain fits an internal model and minimises the discrepancy between predicted and actual input — “Theoretically, predictive coding suggests that the brain is constantly making efforts to ‘fit’ internal models in perceiving through inferring the incoming sensory input. The discrepancies between the actual sensory data and the predicted sensory data (‘error’) are used to refine the model in an iterative process until the model optimized” | mechanisms[8] |
| extends | the authors' earlier report of predictive sound information in A1 using broadband noise stimulation (ref. 5) — “This is consistent with the representations of predictive sound information in the A1 reported in a previous study that used broadband noise stimulation 5 .” | claim_relations[0] |
| builds-on | the network model of Mi et al. (ref. 51) for rhythmic synchronous firing — “For the possible mechanisms of how echo responses arise, a model proposed by Mi et al . 51 matches our previous 5 and current results.” | claim_relations[1] |
| extends | the earlier report of echo responses to broadband noise, by showing the effect is frequency-specific — “After the end of the rhythmic tone stimulation, these neurons exhibited 1–5 additional responses at the expected times for the same ISI (Fig. 2b , red), which is consistent with the neuronal echo response that has been described recently” | claim_relations[2] |
| supports | Mi et al's network model of rhythmic synchronous firing — “We propose that, there is scale-free topology in the cortical network, which contains hub neurons in L2/3 and many low-degree loops with different lengths in L5 (Fig. 5 ; modified from the previous study 5 ). These low-degree loops enable the neural system to follow the external input rhythm, while hub neurons can trigger synchronization in the whole network. During learning, the connectivity from the L5 loops to the L2/3 hubs is highly strengthened, which allows the signals to flow across the entire network. After learning through a train of rhythmic sensory stimulation, a loop with certain length is selected to match the input rhythm. Finally, the network is able to generate echo responses with the same rhythm.” | claim_relations[3] |
| supports | reports of local functional heterogeneity in A1 below 200 microns — “Therefore, these studies demonstrate that a rough tonotopic organization emerges at large scales, but local heterogeneity emerges at fine scales (<250 μm) and is embedded in the global tonotopic order” | claim_relations[4] |
| qualifies | studies of tonotopic organisation conducted under anaesthesia — “Early studies have shown that tonal selectivity in cats 59 and response variability in rats 60 were increased under anesthesia, suggesting anesthetics would alter tonotopic map to be stricter and more regular.” | claim_relations[5] |
| partially-supports | predictive coding as an account of cortical function — “Our studies based on unexpectedly omitted stimuli can give support for the prospective and predictive property of the expectancy outcomes, which would suggest the echo response as possibly the signature finding of predictive coding, as it is even able to be observed when sensory input is absent.” | claim_relations[6] |
| contradicts | the assumption that GCaMP3 and OGB-1 are adequate for mapping A1 organisation — “However, both of these indicators fail to detect Ca 2+ transients caused by single spikes in living animals because of insufficient sensitivity” | claim_relations[7] |
| sense | echo response here means a response at the anticipated time AFTER the sound stops - it has nothing to do with acoustic echoes; the corpus also contains echo suppression in Muniak2014, where the echo is a real reflected sound — “These studies have shown that repetitive stimulation at a regular inter-stimulus interval (ISI) can result in neuronal activity at the anticipated time interval after stimulation, a phenomenon known as “omitted stimulus-evoked responses” or “echo responses” | senses[0] |
| sense | tonotopic organization is scale-dependent in this paper: present at about 1 mm, absent below 200 microns, so the same claim is true or false depending on the scale at which it is asserted — “The precise tonotopic organization in auditory cortex is a long debated issue 20 , 21 , 26 and remains unresolved. On the macroscopic scale level (~1 mm), the tonotopic map in the A1 arises from MGBv. However, the frequency selectivity of neighboring neurons can be very different, and no reliable tonotopic organization could be observed over spatial scales of <200 µm” | senses[1] |
| sense | A1 is identified by thalamic connectivity from MGBv, not by the frequency responses being studied - an external criterion, as in Humphries2010's core/belt boundary — “To verify that the imaged cortical regions were located in the A1, we used the criterion 44 , 45 that the ventral part of the lateral medial geniculate body (MGBv) is connecting with A1.” | senses[2] |
| sense | best frequency here is read off a Gaussian fit to 11 log-spaced frequencies between 2 and 40 kHz, each repeated 20 times — “To test the frequency tuning properties of single neurons, 11 frequencies logarithmically spaced in the range of 2–40 kHz were used in our experiment. Every frequency of stimulation was repeated 20 times, with the inter-stimulus interval (ISI) of 2 s.” | senses[3] |
| sense | local heterogeneity - neighbouring neurons with very different tuning - is presented as embedded within the global tonotopic order rather than as its refutation — “Therefore, these studies demonstrate that a rough tonotopic organization emerges at large scales, but local heterogeneity emerges at fine scales (<250 μm) and is embedded in the global tonotopic order” | senses[4] |
| control | retrograde tracing was used to establish that the imaged region really is A1, using the MGBv-connectivity criterion — held — rules out that the recordings come from a neighbouring auditory field rather than from A1 — “To verify that the imaged cortical regions were located in the A1, we used the criterion 44 , 45 that the ventral part of the lateral medial geniculate body (MGBv) is connecting with A1.” | controls[0] |
| note | echo responses —same-as→ omitted stimulus-evoked responses — given as alternate names in one clause. NOTE for the join key: echo here is NOT an acoustic echo - Muniak2014 in this same corpus uses echo suppression for real reflected sound — “These studies have shown that repetitive stimulation at a regular inter-stimulus interval (ISI) can result in neuronal activity at the anticipated time interval after stimulation, a phenomenon known as “omitted stimulus-evoked responses” or “echo responses” | concept_edges[0] |
| note | EVIDENCE CORRECTED (RE-GROUNDED) · echo responses —distinct-from→ sound response — distinguished by amplitude and latency, and argued to be a distinct neural process — “These results indicate that there are clear discrepancies between the echo response and the sound response with regarding to the activation strength and temporal precision.” | concept_edges[1] |
| retracted | EVIDENCE CORRECTED (RE-GROUNDED) · echo responses —distinct-from→ sound response — RE-GROUNDED. The old span asserted discrepancies without naming the comparison. This sentence names both and gives the measured difference - 'the average amplitude of the sound responses (0.083 +/- 0.002 df/f) ... was significantly higher'. — “These results indicate that there are clear discrepancies between the echo response and the sound response with regarding to the activation strength and temporal precision.” — “These results indicate that there are clear discrepancies” | concept_edges[1] |
| weight | EVIDENCE CORRECTED (RE-GROUNDED) · echo responses —distinct-from→ sound response — 0.85 — “These results indicate that there are clear discrepancies between the echo response and the sound response with regarding to the activation strength and temporal precision.” | concept_edges[1] |
| note | local heterogeneity —part-of→ global tonotopic order — embedded in - the paper explicitly declines to treat the heterogeneity as refuting the map — “Therefore, these studies demonstrate that a rough tonotopic organization emerges at large scales, but local heterogeneity emerges at fine scales (<250 μm) and is embedded in the global tonotopic order” | concept_edges[2] |
| weight | local heterogeneity —part-of→ global tonotopic order — 0.9 — “Therefore, these studies demonstrate that a rough tonotopic organization emerges at large scales, but local heterogeneity emerges at fine scales (<250 μm) and is embedded in the global tonotopic order” | concept_edges[2] |
| note | tonotopic organization —contrasted-with→ local heterogeneity — the contrast is purely one of scale: large scales versus below 250 microns. the same edge is true or false depending on a number the schema cannot record — “Therefore, these studies demonstrate that a rough tonotopic organization emerges at large scales, but local heterogeneity emerges at fine scales (<250 μm) and is embedded in the global tonotopic order” | concept_edges[3] |
| audit | anesthetics —causes→ tonotopic map — Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says anesthetics WOULD alter maps, as a suggestion; causes is hedged. — “Early studies have shown that tonal selectivity in cats 59 and response variability in rats 60 were increased under anesthesia, suggesting anesthetics would alter tonotopic map to be stricter and more regular.” | concept_edges[4] |
| note | anesthetics —causes→ tonotopic map — VOICE,: this is Wang reporting what early cat and rat studies suggest, not Wang asserting it. the weight carries the hedge; nothing carries the attribution — “Early studies have shown that tonal selectivity in cats 59 and response variability in rats 60 were increased under anesthesia, suggesting anesthetics would alter tonotopic map to be stricter and more regular.” | concept_edges[4] |
| weight | anesthetics —causes→ tonotopic map — 0.7 — “Early studies have shown that tonal selectivity in cats 59 and response variability in rats 60 were increased under anesthesia, suggesting anesthetics would alter tonotopic map to be stricter and more regular.” | concept_edges[4] |
| edge | two-photon Ca 2+ imaging —contrasted-with→ electrophysiological recordings — “Compared to electrophysiological recordings, two-photon Ca 2+ imaging has less tissue damage and is able to detect subthreshold neuronal activity” | concept_edges[5] |
| audit | predictive coding —operationalised-as→ echo responses — Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says the organisation 'could be observed', suggesting predictive coding; operationalised-as is hedged. — “In our results, the frequency selectivity profiles of echo responses (Figs 3 and 4 ) are similar between anesthetized and awake mice, suggesting that the local heterogeneous organization for predictive coding could be observed under awake and light anesthesia conditions.” | concept_edges[6] |
| note | predictive coding —operationalised-as→ echo responses — hedged twice - would suggest, and possibly the signature finding. the weight carries both — “In our results, the frequency selectivity profiles of echo responses (Figs 3 and 4 ) are similar between anesthetized and awake mice, suggesting that the local heterogeneous organization for predictive coding could be observed under awake and light anesthesia conditions.” | concept_edges[6] |
| weight | predictive coding —operationalised-as→ echo responses — 0.6 — “In our results, the frequency selectivity profiles of echo responses (Figs 3 and 4 ) are similar between anesthetized and awake mice, suggesting that the local heterogeneous organization for predictive coding could be observed under awake and light anesthesia conditions.” | concept_edges[6] |
| outcome (stated by the tagger, no quotation) | control — held | controls[0] |
| duplicate of | limitation — limitations[3] — “However, the precise effects of anesthesia on A1 neurons of mouse are still largely unknown and hence need further investigation.” | limitations[1] |
| evidential status | mechanism — the paper's own words: We propose that — “We propose that, there is scale-free topology in the cortical network, which contains hub neurons in L2/3 and many low-degree loops with different lengths in L5” | mechanisms[0] |
| evidential status | mechanism — the paper's own words: may reflect — “Therefore, it implies that echo response is an intrinsic property of cortical circuits in response to the absence of an expected sound, and echo response may reflect a distinct neural process in comparison to sound-evoked responses.” | mechanisms[4] |
| replication | untested | |
| limitation | Evidence in the auditory field is still insufficient to decide whether prediction and error neurons occupy separate cortical layers. — “However, evidences by far are still deficient in the auditory field in accordance with the notion that prediction and error neurons reside separately in distinct cortical layers 56 , which shall be studied further.” | limitations[0] |
| limitation | How anaesthesia affects mouse A1 neurons remains largely unknown, so the anaesthetised results carry that uncertainty. — “However, the precise effects of anesthesia on A1 neurons of mouse are still largely unknown and hence need further investigation.” | limitations[1] |
| limitation | The behavioural relevance of the heterogeneous echo-response organisation was not established. — “However, the underlying mechanisms of contribution to behavior are still largely unknown and warrant investigation in future studies.” | limitations[2] |
| limitation | the precise effects of anaesthesia on mouse A1 neurons are still largely unknown, and the awake replication is the only guard — “However, the precise effects of anesthesia on A1 neurons of mouse are still largely unknown and hence need further investigation.” | limitations[3] |
| limitation | no behavioural measure, so the contribution of these responses to behaviour is untested — “However, the underlying mechanisms of contribution to behavior are still largely unknown and warrant investigation in future studies.” | limitations[4] |
| limitation | the layer-specific prediction of predictive coding is stated to be inadequately evidenced in the auditory field, including by this study — “However, evidences by far are still deficient in the auditory field in accordance with the notion that prediction and error neurons reside separately in distinct cortical layers 56 , which shall be studied further.” | limitations[5] |
| limitation | male mice only, with no analysis by sex — “C57BL/6J mice (male, 2–3 months old) were supplied by the Laboratory Animal Center of the Third Military Medical University.” | limitations[6] |
| limitation | tuning curves are Gaussian-fitted for visibility, so the plotted shape is a model and not the measurement — “To improve visibility here, the frequency tuning curves were fitted by Gaussian function.” | limitations[7] |
| limitation | data are available on request rather than deposited — “The data that support the findings of this study are available from the corresponding author upon reasonable request.” | limitations[8] |
| limitation | overall tuning across echo neurons was broad at the sound level used, which weakens the frequency-specificity claim it is set against — “Although the overall tuning for sound-evoked responses across all of the echo neurons (Fig. 3c ) was relatively broad given the sound level used here, the echo responses in most of the neurons occurred at the preferred frequencies.” | limitations[9] |
finding · quantity · variable · definition · key-term — grounded in the full text| finding | A subgroup of layer 2/3 neurons in mouse A1 fired at the anticipated time interval after a train of rhythmic pure tones had stopped. — “a subgroup of neurons in L2/3 exhibited echo responses at the anticipated time interval following a train of rhythmic pure tone stimulation.” | content_tags.findings[0] |
| finding | Echo responses were elicited preferentially by tones at or close to each neuron's own best frequency. — “After testing with a range of frequencies, we found that these echo responses occurred preferentially close to the best frequencies of the neurons.” | content_tags.findings[1] |
| finding | The same match between echo-evoking frequency and preferred frequency held in awake mice, so it is not an artefact of anaesthesia. — “By comparing the preferred frequencies between sound responses and echo responses (92 neurons from 4 awake mice), we found that the frequency that evoked the echo response also closely corresponded to the preferred frequency to tone stimulation in awake mice” | content_tags.findings[2] |
| finding | Echo neurons tuned to very different frequencies sat side by side within a small imaging field, so the local map of echo responses is heterogeneous rather than tonotopically ordered. — “adjacent neurons were capable of displaying echo responses to very different frequencies we employed here” | content_tags.findings[3] |
| finding | Within one small cortical patch, neurons echoed to every frequency tested, which the authors read as heterogeneous local organisation for coding prediction. — “neurons showed echo responses to any of the tested frequencies, suggesting heterogeneous organization of mouse A1 neurons in the coding of predictive information.” | content_tags.findings[4] |
| finding | Echo responses were on average later and weaker than sound-evoked responses in the same neurons. — “Those differences could be attributed to a variation in neurons for the estimation of the input rhythm, and it thus leads to a longer latency and a lower amplitude of the echo responses on average.” | content_tags.findings[5] |
| finding | Frequency-selectivity profiles of echo responses were comparable under light anaesthesia and in the awake state. — “are similar between anesthetized and awake mice, suggesting that the local heterogeneous organization for predictive coding could be observed under awake and light anesthesia conditions.” | content_tags.findings[6] |
| finding | The echo response is read as an intrinsic property of cortical circuits responding to the absence of an expected sound. — “it implies that echo response is an intrinsic property of cortical circuits in response to the absence of an expected sound” | content_tags.findings[7] |
| finding | 15% of imaged L2/3 neurons showed echo responses after a rhythmic tone train — “We found that 15% of the imaged neurons (15.7% ± 1.3%) showed echo responses.” | content_tags.findings[8] |
| finding | echo responses occur preferentially at each neuron's own best frequency — “A comparison of the properties of the sound responses and the echo responses revealed that the frequency that evoked the echo response in a given neuron closely corresponded to its preferred frequency to tone stimulation in anesthetized mice” | content_tags.findings[9] |
| finding | the same frequency correspondence holds in awake mice, so it is not an anaesthesia artifact — “By comparing the preferred frequencies between sound responses and echo responses (92 neurons from 4 awake mice), we found that the frequency that evoked the echo response also closely corresponded to the preferred frequency to tone stimulation in awake mice” | content_tags.findings[10] |
| finding | echoes are smaller and later than sound-evoked responses — “When comparing the sound response with the echo response, we found the average amplitude of the sound responses (0.083 ± 0.002 Δf/f) for all of the imaged neurons in L2/3 was significantly higher than that of the echo responses (0.055 ± 0.003 Δf/f, n = 777 neurons from 8 mice, P < 0.001; Fig. 2e ), and the latency of the sound responses (0.083 ± 0.001 s) was significantly shorter than that of the echo responses (0.237 ± 0.024 s, n = 777 neurons from 8 mice, P < 0.001; Fig. 2f ).” | content_tags.findings[11] |
| finding | echo neurons tuned to very different frequencies sit side by side within 200 microns — “Echo neurons in response to different frequencies were spatially mixed over a scale about 200 µm × 200 µm: adjacent neurons were capable of displaying echo responses to very different frequencies we employed here” | content_tags.findings[12] |
| finding | no reliable tonotopic organisation is observable below 200 microns, though the map is present at about 1 mm — “The precise tonotopic organization in auditory cortex is a long debated issue 20 , 21 , 26 and remains unresolved. On the macroscopic scale level (~1 mm), the tonotopic map in the A1 arises from MGBv. However, the frequency selectivity of neighboring neurons can be very different, and no reliable tonotopic organization could be observed over spatial scales of <200 µm” | content_tags.findings[13] |
| finding | retrograde labelling to MGBv confirmed the imaged site was A1 — “In addition, we performed a post hoc histological experiment with a retrograde neuronal tracer, CTB, which was injected into the imaged area (Fig. 1d ) and we found retrogradely-labeled cortical projecting neurons mainly located in the ventral part of the lateral medial geniculate body (MGBv) (Fig. 1e ), confirming that the recorded cortical site was located on the A1.” | content_tags.findings[14] |
| quantity | 8 mice (777 neurons) ... 4 mice (194 neurons) — samples imaged in the anaesthetised and awake conditions respectively — “In our experiments, 8 mice (777 neurons) and 4 mice (194 neurons) were used for Ca 2+ imaging in anesthetized and awake conditions, respectively.” | content_tags.quantities[0] |
| quantity | 15% — proportion of imaged layer 2/3 A1 neurons that produced echo responses — “We found that 15% of the imaged neurons” | content_tags.quantities[1] |
| quantity | 1–5 — number of extra responses individual neurons produced at the expected times after the tone train ended — “these neurons exhibited 1–5 additional responses at the expected times for the same ISI” | content_tags.quantities[2] |
| quantity | 185 — echo neurons from 6 anaesthetised mice entering the averaged tone-evoked tuning curve — “The average tone-evoked frequency tuning property (the data fitted with Gaussian function) of 185 echo neurons from 6 anesthetized mice.” | content_tags.quantities[3] |
| quantity | 8 mice and 777 neurons anaesthetised, 4 mice and 194 neurons awake — the sample — “In our experiments, 8 mice (777 neurons) and 4 mice (194 neurons) were used for Ca 2+ imaging in anesthetized and awake conditions, respectively.” | content_tags.quantities[4] |
| quantity | 15.7% plus or minus 1.3% — proportion of imaged neurons showing echo responses — “We found that 15% of the imaged neurons (15.7% ± 1.3%) showed echo responses.” | content_tags.quantities[5] |
| quantity | echo 0.055 plus or minus 0.003 df/f versus sound 0.083 plus or minus 0.002 df/f — response amplitudes, P < 0.001 — “When comparing the sound response with the echo response, we found the average amplitude of the sound responses (0.083 ± 0.002 Δf/f) for all of the imaged neurons in L2/3 was significantly higher than that of the echo responses (0.055 ± 0.003 Δf/f, n = 777 neurons from 8 mice, P < 0.001; Fig. 2e ), and the latency of the sound responses (0.083 ± 0.001 s) was significantly shorter than that of the echo responses (0.237 ± 0.024 s, n = 777 neurons from 8 mice, P < 0.001; Fig. 2f ).” | content_tags.quantities[6] |
| quantity | echo latency 0.237 plus or minus 0.024 s versus sound 0.083 plus or minus 0.001 s — response latencies, P < 0.001 — “When comparing the sound response with the echo response, we found the average amplitude of the sound responses (0.083 ± 0.002 Δf/f) for all of the imaged neurons in L2/3 was significantly higher than that of the echo responses (0.055 ± 0.003 Δf/f, n = 777 neurons from 8 mice, P < 0.001; Fig. 2e ), and the latency of the sound responses (0.083 ± 0.001 s) was significantly shorter than that of the echo responses (0.237 ± 0.024 s, n = 777 neurons from 8 mice, P < 0.001; Fig. 2f ).” | content_tags.quantities[7] |
| quantity | 11 frequencies log-spaced from 2 to 40 kHz, 20 repeats, ISI 2 s — the tuning protocol — “To test the frequency tuning properties of single neurons, 11 frequencies logarithmically spaced in the range of 2–40 kHz were used in our experiment. Every frequency of stimulation was repeated 20 times, with the inter-stimulus interval (ISI) of 2 s.” | content_tags.quantities[8] |
| quantity | about 1 mm versus below 200 microns — the two scales at which tonotopy is present and absent — “The precise tonotopic organization in auditory cortex is a long debated issue 20 , 21 , 26 and remains unresolved. On the macroscopic scale level (~1 mm), the tonotopic map in the A1 arises from MGBv. However, the frequency selectivity of neighboring neurons can be very different, and no reliable tonotopic organization could be observed over spatial scales of <200 µm” | content_tags.quantities[9] |
| quantity | below 250 microns — the fine scale at which local heterogeneity emerges, per the studies reviewed — “Therefore, these studies demonstrate that a rough tonotopic organization emerges at large scales, but local heterogeneity emerges at fine scales (<250 μm) and is embedded in the global tonotopic order” | content_tags.quantities[10] |
| quantity | isoflurane 1 to 2% for surgery, reduced to 0.4 to 0.8% for imaging — the anaesthetic depth, which the paper's own discussion says alters the map — “In brief, we applied isoflurane at the level of 1–2% in pure oxygen for anesthesia and placed the animals onto a heating plate at a temperate of 37.5–38 °C. After local anesthesia with lidocaine, we removed the skin and muscle over the A1. Then the skull was glued with a customized plastic chamber using cyanoacrylic glue (UHU) and a ~2 × 2 mm small craniotomy was made centered at 2.5 mm posterior to bregma and 4.5 mm lateral to midline. After that, we reduced the isoflurane level to 0.4–0.8% and transferred the animal to the imaging system.” | content_tags.quantities[11] |
| variable | relative changes of fluorescence (dependent) — “neuronal Ca 2+ signals were expressed as the relative changes of fluorescence” | content_tags.variables[2] |
| variable | inter-stimulus interval (control) — “Every frequency of stimulation was repeated 20 times, with the inter-stimulus interval (ISI) of 2” | content_tags.variables[3] |
| variable | tone frequency — “To test the frequency tuning properties of single neurons, 11 frequencies logarithmically spaced in the range of 2–40 kHz were used in our experiment. Every frequency of stimulation was repeated 20 times, with the inter-stimulus interval (ISI) of 2 s.” | content_tags.variables[5] |
| variable | amplitude of the echo responses — “The average amplitude of the echo responses (0.055 ± 0.003 Δf/f) for all of the imaged neurons in L2/3 showed a significant increase comparing to that of the baseline activity (0.014 ± 0.001 Δf/f, n = 777 neurons from 8 mice, P < 0.001; Fig. 2d ).” | content_tags.variables[6] |
| variable | difference in preferred frequency — “A comparison of the properties of the sound responses and the echo responses revealed that the frequency that evoked the echo response in a given neuron closely corresponded to its preferred frequency to tone stimulation in anesthetized mice” | content_tags.variables[7] |
| definition | omitted stimulus-evoked responses — “These studies have shown that repetitive stimulation at a regular inter-stimulus interval (ISI) can result in neuronal activity at the anticipated time interval after stimulation, a phenomenon known as “omitted stimulus-evoked responses” or “echo responses”” | content_tags.definitions[0] |
| definition | primary auditory field (A1) — “Located in the caudal part of the auditory cortex, the primary auditory field (A1) displays a regularly tonotopic layout with low best frequencies positioned caudally and high best frequencies dorsoventrally” | content_tags.definitions[1] |
| definition | predictive coding — “Theoretically, predictive coding suggests that the brain is constantly making efforts to ‘fit’ internal models in perceiving through inferring the incoming sensory input.” | content_tags.definitions[2] |
| definition | echo responses — “These studies have shown that repetitive stimulation at a regular inter-stimulus interval (ISI) can result in neuronal activity at the anticipated time interval after stimulation, a phenomenon known as “omitted stimulus-evoked responses” or “echo responses” | content_tags.definitions[3] |
| definition | predictive coding — “Theoretically, predictive coding suggests that the brain is constantly making efforts to ‘fit’ internal models in perceiving through inferring the incoming sensory input. The discrepancies between the actual sensory data and the predicted sensory data (‘error’) are used to refine the model in an iterative process until the model optimized” | content_tags.definitions[4] |
| key-terms | echo responses, omitted stimulus-evoked responses, predictive coding, tonotopic organization, local heterogeneity, best frequencies, primary auditory cortex, MGBv, two-photon, Cal-520, GCaMP3, OGB-1, isoflurane, anesthesia, inter-stimulus interval, scale-free topology, hub neurons, spectro-temporal filters, receptive fields, Wilcoxon |
| Wilcoxon signed-rank test | — “We used Wilcoxon signed rank test and Wilcoxon rank sum test to determine statistical significances for paired and unpaired cases, respectively.” | content_tags.statistics[0] |
| Wilcoxon rank-sum test | — “We used Wilcoxon signed rank test and Wilcoxon rank sum test to determine statistical significances for paired and unpaired cases, respectively.” | content_tags.statistics[1] |
| statistical-significance | as written: considered statistically significant — “P < 0.05 was considered statistically significant.” | content_tags.statistics[2] |
| mean | as written: presented as mean — “Summarized data are presented as mean” | content_tags.statistics[3] |
| standard error of the mean SEM | as written: standard error of the mean (SEM) — “standard error of the mean (SEM) in figures.” | content_tags.statistics[4] |
| curve fitting | as written: fitted by Gaussian function — “To improve visibility here, the frequency tuning curves were fitted by Gaussian function.” | content_tags.statistics[5] |
| percentile | as written: The 25th percentile — “The 25th percentile of the entire fluorescence changes in single-trial was calculated as the baseline f0” | content_tags.statistics[6] |
| Wilcoxon signed rank test | — “We used Wilcoxon signed rank test and Wilcoxon rank sum test to determine statistical significances for paired and unpaired cases, respectively. P < 0.05 was considered statistically significant.” | content_tags.statistics[7] |
| Wilcoxon rank sum test | — “We used Wilcoxon signed rank test and Wilcoxon rank sum test to determine statistical significances for paired and unpaired cases, respectively. P < 0.05 was considered statistically significant.” | content_tags.statistics[8] |
| Gaussian function fit | — “To improve visibility here, the frequency tuning curves were fitted by Gaussian function.” | content_tags.statistics[9] |
| relative fluorescence change | — “The 25th percentile of the entire fluorescence changes in single-trial was calculated as the baseline f0, and then neuronal Ca 2+ signals were expressed as the relative changes of fluorescence” | content_tags.statistics[10] |
| claim type | existence — a second reading reaching the same value; recorded: existence — “Therefore, these results suggested that the echo responses of neurons within A1 are highly related to their frequency selectivity.” | alternatives.claim_type |
| method | in vivo two-photon Ca2+ imaging with the high-sensitivity indicator Cal-520 AM, 11 log-spaced tone frequencies from 2 to 40 kHz, repeated in awake animals as a control on anaesthesia — a second reading DIFFERING from the recorded value; recorded: electrophysiology — “Therefore, we used in vivo two-photon Ca 2+ imaging to record the neuronal activities in the mouse A1 to elucidate the relationship between their echo responses and frequency selectivity.” | alternatives.method |
| phenomenon | predictive or omitted-stimulus responses in sensory cortex — a second reading DIFFERING from the recorded value; recorded: echo responses (omitted stimulus-evoked responses) in primary auditory cortex — “These studies have shown that repetitive stimulation at a regular inter-stimulus interval (ISI) can result in neuronal activity at the anticipated time interval after stimulation, a phenomenon known as “omitted stimulus-evoked responses” or “echo responses” | alternatives.phenomenon |
Full text — tagged
The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.
Frequency selectivity of echo responsesterm in the mouse primary auditory cortexterm·key Meng WangpersName Meng WangpersName 1 Britton Chance Center for Biomedical Photonics, Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan, 430074 China 2 Brain Research Center and State Key Laboratory of Trauma, Burns, and Combined Injury, Third Military Medical University, Chongqing, 400038 China 1, 2, # , Ruijie Li Ruijie Li 2 Brain Research Center and State Key Laboratory of Trauma, Burns, and Combined Injury, Third Military Medical University, Chongqing, 400038 China 2, # , Jingcheng Li Jingcheng Li 2 Brain Research Center and State Key Laboratory of Trauma, Burns, and Combined Injury, Third Military Medical University, Chongqing, 400038 China 2 , Jianxiong ZhangpersName Jianxiong ZhangpersName 2 Brain Research Center and State Key Laboratory of Trauma, Burns, and Combined Injury, Third Military Medical University, Chongqing, 400038 China 2 , Xiaowei ChenpersName Xiaowei ChenpersName 2 Brain Research Center and State Key Laboratory of Trauma, Burns, and Combined Injury, Third Military Medical University, Chongqing, 400038 China 3 CAS Center for Excellence in Brain Science and Intelligence Technology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai, 200031 China 2, 3, ✉ , Shaoqun ZengpersName Shaoqun ZengpersName 1 Britton Chance Center for Biomedical Photonics, Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan, 430074 China 1, ✉ , Xiang LiaopersName Xiang LiaopersName 2 Brain Research Center and State Key Laboratory of Trauma, Burns, and Combined Injury, Third Military Medical University, Chongqing, 400038 China 2, ✉ 1 Britton Chance Center for Biomedical Photonics, Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan, 430074 China 2 Brain Research Center and State Key Laboratory of Trauma, Burns, and Combined Injury, Third Military Medical University, Chongqing, 400038 China 3 CAS Center for Excellence in Brain Science and Intelligence Technology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai, 200031 China ✉ Corresponding author.
# Contributed equally.
Received 2017 May 23; Accepted 2017 Dec 12; Collection date 2018.
© The Author(s) 2017 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ .
PMCID: PMC5758803 PMID: 29311673 Abstract In the primary auditory cortexterm·key (A1term), neuronal ensembles are activated relative to anticipated sound events following rhythmic stimulation, but whether the echo responsesterm of the neurons are related to their frequency selectivity remains unknown. Therefore, we used in vivo two-photon Ca 2+ imagingterm to record the neuronal activities in the mouse A1term to elucidate the relationship between their echo responsesterm and frequency selectivity. We confirmed the presence of echo responsesterm in a subgroup of mouse Layer 2/3 A1term neurons following a train of rhythmic pure tone stimulation. After testing with a range of frequencies, we found that these echo responsesterm occurred preferentially close to the best frequenciesterm of the neurons. The local organization of the echo responsesterm of the neurons was heterogeneous in the A1term. Therefore, these results indicate that the observed echo responsesterm of neurons within A1term are highly related to their frequency selectivity.
Introduction When humans and other animals interact with the natural environment, the use of sensory cues and previous experience are critical to form predictions that guide behavior and ultimately survival in the world 1 , 2 . Recent studies have shown that anticipatory events are found at the levels of both brain activity and behavior. Sumbre et al . found that a zebrafish larva was capable of estimating the timing of a specific impending stimulus through repetitive visual stimulation 3 . Located in the caudal part of the auditory cortex, the primary auditory field (A1)term·def displays a regularly tonotopic layout with low best frequenciesterm positioned caudally and high best frequenciesterm dorsoventrally 4 . The neuronal activities in the mouse A1term have also been found to contribute coding sound information and driving motor responses to expected events 5 . These studies have shown that repetitive stimulation at a regular inter-stimulus intervalterm·key (ISI) can result in neuronal activity at the anticipated time interval after stimulation, a phenomenon known as “omitted stimulus-evoked responsesterm” or “echo responsesterm” 5 – 11 . Although the prediction correlated neural signals have already been observed in multiple sensory regions in different mammals, the stimulus type for the preferentially evoked echo responsesterm is unclear. Specifically, whether the echo responsesterm of the neurons in the auditory cortex are related to their frequency selectivity remains unknown.
The processing and representation of behaviorally relevant information have been suggested to be selectively enhanced by predictive sensing in an efficient manner. For example, a recent nonhuman primate study demonstrated that high excitability was confined both in time and across neuronal ensembles in the auditory cortex when auditory stimuli were predictable in both timing and pitch 12 . As spectro-temporal filtersterm·key, the features of the responses in the brain were restricted in both time and space, which is an efficient strategy to minimize the energy expenditure required to maintain a specific cortical state 13 , 14 .
In the auditory cortex, the response profiles of neurons have often been shown to characterize sound features such as sound intensity and frequency 15 – 17 . Electrophysiological recordingsterm and large-scale imaging techniques have revealed the global tonotopic mapterm in auditory cortex and the coarse division of the separate fields in the cortex 18 , 19 . Although large-scale recording techniques normally demonstrate smooth tonotopic organizationterm of the neuronal response patterns, the response properties of neurons may vary remarkably, even in neighboring neurons 17 , 20 . By elucidating the response properties of neurons to sound frequencies in a small brain region, the local populations in the A1term are revealed to be highly heterogeneous, which means that neighboring neurons may present similar or extraordinarily different frequency tunings properties 21 , 22 . Therefore, these studies demonstrate that a rough tonotopic organizationterm emerges at large scales, but local heterogeneityterm emerges at fine scales (<250 μm) and is embedded in the global tonotopic orderterm 22 .
The two-photonterm·key microscopy combined with Ca 2+ imaging has allowed simultaneously sampling the activity of hundreds of neurons in the intact brain at cellular resolution 23 , 24 . Compared to electrophysiological recordingsterm, two-photon Ca 2+ imagingterm has less tissue damage and is able to detect subthreshold neuronal activity 23 , 25 . Furthermore, two-photon Ca 2+ imagingterm can readily monitor activity of specific types of cells and precisely locate their positions within neural circuits 24 . For these reasons, in vivo two-photon Ca 2+ imagingterm provided us with a powerful tool for searching for the population of neurons in A1term participating in the processing of expected sound events. Issa et al . previously used a genetically encoded Ca 2+ indicator, GCaMP3term, to address the issue of topographic organization at multiple scales in the A1term 26 . Bandyopadhyay et al . used the synthetic indicator OGB-1term, the most commonly-used Ca 2+ indicator dye, to study the activity of individual neurons in the A1term during sound stimulation 25 . However, both of these indicators fail to detect Ca 2+ transients caused by single spikes in living animals because of insufficient sensitivity 27 – 30 . The sensitivity plays the most important role in studying the organization of the A1term, because most of cortical neurons respond with only one or two spikes to sound stimulation 20 . In contrast, recent studies showed that Cal-520term 31 with a high sensitivity was a more suitable Ca 2+ indicator for functional mapping experiments in the A1term 32 of living brains.
Since in vivo two-photon Ca 2+ imagingterm has become widely used for the study of the functional organization of neuronal populations in the cortex 22 , 25 , 33 , we applied this technique to investigate the relationship between the echo responsesterm of the neurons and their frequency tuning in the A1term. As stable two-photonterm·key imaging can be easily achieved under anesthesiaterm·key, we initiated our experiments using anesthetized preparations. However, it has been suggested that general anesthesiaterm·key can affect many aspects of spontaneous activity and sensory processing in the cortical circuits 34 , 35 . To confirm whether anesthesiaterm·key could also affect predictive information processing in our conditions, we repeated these experiments in awake mice. Similar to broadband noise stimulation, as reported previously 5 , following a train of rhythmic pure tone stimulation, we found that a subgroup of mouse Layer 2/3 (L2/3) auditory cortex neurons showed one or multiple times of the echo responsesterm at the anticipated time interval in the absence of sound. Testing with a series of pure tone stimuli, we revealed that these echo responsesterm were typically elicited following stimulation at the neurons’ best frequenciesterm. The local organization of the echo responsesterm of the neurons was heterogeneous.
Materials and Methods Animals C57BL/6J mice (male, 2–3 months old) were supplied by the Laboratory Animal Center of the Third Military Medical University. These mice are a suitable and common animal model for auditory experiments 25 , 36 . All experimental protocols were carried out based on institutional animal welfare guidelines with approval of the Third Military Medical University Animal Care and Use Committee. The mice were housed in a cycle of 12 h light/dark (lights off at 19:00), free of accessing food and drinking water. In our experiments, 8 mice (777 neurons) and 4 mice (194 neurons) were used for Ca 2+ imaging in anesthetized and awake conditions, respectively.
In vivo two-photon Ca 2+ imagingterm and electrophysiologyterm·mth The right A1term was exposed to two-photonterm·key imaging in anesthetized mice as elsewhere 21 , 37 – 39 . In brief, we applied isofluraneterm·key at the level of 1–2% in pure oxygen for anesthesiaterm·key and placed the animals onto a heating plate at a temperate of 37.5–38 °C. After local anesthesiaterm·key with lidocaine, we removed the skin and muscle over the A1term. Then the skull was glued with a customized plastic chamber using cyanoacrylic glue (UHU) and a ~2 × 2 mm small craniotomy was made centered at 2.5 mm posterior to bregma and 4.5 mm lateral to midline. After that, we reduced the isofluraneterm·key level to 0.4–0.8% and transferred the animal to the imaging system. The multi-cell bolus loading procedure with Cal-520term AM was similar to the previous studies 23 , 31 . The head fixation and training procedures were similar to our recent studies 5 , 32 . When performing two-photonterm·key imaging in awake mouse, an infrared camera (frame rate: 30 Hz) was used to collect the videos of the mouse behavior. Two-photonterm·key imaging was carried out with a resonant scanner-based “LotosScan 1.0” system (Suzhou Institute of Biomedical Engineering and Technology), as described in details previously 40 , 41 . To conduct targeted in vivo whole-cell recordings in neurons of A1term, we used the previously described “shadow-patching” procedure 37 – 39 , 42 , 43 .
Retrograde tracing To verify that the imaged cortical regions were located in the A1term, we used the criterion 44 , 45 that the ventral part of the lateral medial geniculate body (MGBvterm) is connecting with A1term. Hence, we first determined the center of the craniotomy window in a stereotactic way (70% of bregma-lambda and ventral: ~2 mm, or lateral: ~4.4 mm) 21 , 25 , and then performed post hoc histology with all mice after imaging experiments. For labeling cortico-thalamic projections, we used a glass electrode, which has a tip diameter of 20–30 μm, and filled it with neural tracer solution. In the experiment, we inserted the electrode into the cortical region at a depth of ~500 μm below the surface. We used Alexa Fluor 488-conjugated cholera toxin subunit B (CTB) as the neural tracer, and injected the fluorescent CTB solution with 0.5% in phosphate buffer by pressure (700 mbar) for 15 min. Seven days after the fluorescent CTB injection, the mice were anesthetized with pentobarbital (1.0 g/kg ip). The brain was first dissected out and then it was immersed with 4% paraformaldehyde overnight. To visualize fluorescent tracers, a consecutive series of coronal or horizontal sections (40 μm thick) were prepared using a sliding cryotome, and then all sections were mounted onto glass slides and imaged with a stereoscope (Olympus).
Auditory stimulation As shown previously 37 , the sounds were presented through a free-field ES1 electrostatic speaker and an ED1 speaker driver (Tucker Davis Technologies). The distance was 6 cm between the speaker and the contralateral ear of the mouse. The tone duration we applied in this experiment was 100 ms, containing 10 ms on and off linear ramps. To test the frequency tuning properties of single neurons, 11 frequencies logarithmically spaced in the range of 2–40 kHz were used in our experiment. Every frequency of stimulation was repeated 20 times, with the inter-stimulus intervalterm·key (ISI) of 2 s.
All sound levels were measured with a microphone placed ~6 cm away from the speaker. The sound levels were calibrated by a pre-polarized condenser microphone (377A01 microphone, PCB Piezotronics Inc) as published previously 46 . The data were recorded at 1 MHz by using a data acquisition device (USB-6361, National Instruments) and then analyzed by our customized LabVIEW program. During pure tone stimulation experiments, the use of the sound level was equal to the levels of ~79 dB SPL (1 kHz to 10 kHz) and ~81 dB SPL (10 kHz to 40 kHz). The major components of the background noise (~60 dB SPL) in the experiments were low-frequencies, similar to the equipment used previously 37 , 38 . The peak of background noise, with a spectral density about 33 dB/sqrt(Hz), is located below 1 kHz. This component was out of the hearing range of mice (from 1 kHz to about 100 kHz) 47 , 48 and therefore should not be concerned, as also we described in our previous studies 5 , 32 .
Data analysis We analyzed our data with software including MATLAB 2014a (MathWorks), Igor Pro 5.0 (Wavemetrics) and LabVIEW 2014 (National Instruments). We identified individual neurons in two-photonterm·key imaging data and performed the regions of interest (ROIs) drawing for each of them. In each image frame, fluorescence changes (f) were calculated by averaging the image intensities within each ROI. The 25th percentile of the entire fluorescence changes in single-trial was calculated as the baseline f0, and then neuronal Ca 2+ signals were expressed as the relative changes of fluorescenceterm·var Δf/f = (f − f0)/f0.
As described in our previous studies 5 , 32 , Ca 2+ transient detection was performed based on thresholding criteria about peak amplitude and rising rate, which is similar to previously described peeling algorithm 49 . For the analysis of frequency tuning, we constructed the frequency tuning curves of individual neurons, and performed curve normalization and averaging as previously described 37 . To improve visibility here, the frequency tuning curves were fitted by Gaussian functionterm·stat.
Summarized data are presented as meanterm·stat ± standard error of the mean (SEM)term·stat in figures. We used Wilcoxon signed rank testterm·stat and Wilcoxon rank sum testterm·stat to determine statistical significances for paired and unpaired cases, respectively. P < 0.05 was considered statistically significantterm·stat.
Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request.
Results Spontaneous and pure tone-evoked responses in A1term L2/3 neurons We used two-photon Ca 2+ imagingterm 23 in combination with the fluorescent Ca 2+ indicator Cal-520term AM 31 to measure the activity of L2/3 neurons in the A1term in isoflurane-anesthetized mice (Fig. 1a ). Our data set included 777 neurons ( n = 8 mice), ranging from 47 to 80 neurons (average number: 60 neurons) for each imaging plane (e.g., Fig. 1b ). The activities of many neurons were simultaneously imaged during the presentation of rhythmically repeated sound stimuli, and both randomly occurring spontaneous (Fig. 1c , left) and pure tone-evoked (Fig. 1c , right) Ca 2+ transients were observed in the imaged neurons. In addition, we performed a post hoc histological experiment with a retrograde neuronal tracer, CTB, which was injected into the imaged area (Fig. 1d ) and we found retrogradely-labeled cortical projecting neurons mainly located in the ventral part of the lateral medial geniculate body (MGBvterm) (Fig. 1e ), confirming that the recorded cortical site was located on the A1term. This result is consistent with the connectivity between MGB and A1term, as some previous studies have reported 5 , 44 , 45 , 50 .
Figure 1.
Spontaneous and sound-evoked activities of A1term neurons in anesthetized mice. ( a ) Side view of the experimental arrangement. ( b ) Two-photonterm·key image of Cal-520term AM-labeled neurons at 260 μm depth below the pial surface in an anesthetized mouse. Four exemplary neurons are labeled 1 through 4. ( c ) Four representative neurons marked in panel b show spontaneous (left) and sound-evoked activities (right). The vertical gray bars indicate tone stimuli. The tone frequencyterm·var we used in the example was 4913 Hz. ( d ) Coronal slices show the injection site of Cal-520term in mouse A1term (left) and an enlarged image (right) of the region marked with a red box (left). The regions marked with a white box were imaging regions (right). ( e ) Coronal slices show the injection site of CTB-green, i.e., the retrograde neuronal tracer, in the A1term (left) and the retrograde-labeled cortical projecting neurons mainly in the lateral MGBvterm (right).
Rhythmic tone stimulation induces echo responsesterm in the A1term L2/3 neurons To study the echo responseterm to tonal stimulation, we presented rhythmic pure tones at each imaging plane. Figure 2a shows two representative neurons recorded in L2/3. The responses of these neurons to 20 repetitive tone stimuli (30 kHz) with an inter-stimulus-interval (ISI) of 2 s are presented in Fig. 2b . After the end of the rhythmic tone stimulation, these neurons exhibited 1–5 additional responses at the expected times for the same ISI (Fig. 2b , red), which is consistent with the neuronal echo responseterm that has been described recently 5 . The observed echo responsesterm were derived from a subgroup of the imaged neurons, as the neurons (Fig. 2a ) represented in red (Fig. 2c ). We found that 15% of the imaged neurons (15.7% ± 1.3%) showed echo responsesterm. The average amplitude of the echo responsesterm·var (0.055 ± 0.003 Δf/f) for all of the imaged neurons in L2/3 showed a significant increase comparing to that of the baseline activity (0.014 ± 0.001 Δf/f, n = 777 neurons from 8 mice, P < 0.001; Fig. 2d ). When comparing the sound responseterm with the echo responseterm, we found the average amplitude of the sound responsesterm (0.083 ± 0.002 Δf/f) for all of the imaged neurons in L2/3 was significantly higher than that of the echo responsesterm (0.055 ± 0.003 Δf/f, n = 777 neurons from 8 mice, P < 0.001; Fig. 2e ), and the latency of the sound responsesterm (0.083 ± 0.001 s) was significantly shorter than that of the echo responsesterm (0.237 ± 0.024 s, n = 777 neurons from 8 mice, P < 0.001; Fig. 2f ). These results indicate that there are clear discrepancies between the echo responseterm and the sound responseterm with regarding to the activation strength and temporal precision.
Figure 2.
Neurons in L2/3 of A1term of anesthetized mice generate echo responsesterm with rhythmic tone stimulation. ( a ) Two-photonterm·key image of Cal-520term AM-labeled neurons in L2/3. ( b ) Two representative neurons marked in the image of panel a show Ca 2+ responses before, during and after tone stimuli (frequency: 30 kHz) under anesthesiaterm·key. The echo responsesterm of the neurons are marked in red. ( c ) Summary of neurons (red) that generated echo responsesterm with rhythmic tone stimulation on this imaging plane. ( d ) The average amplitudes of echo responseterm and baseline activity. Wilcoxon signed rank testterm·stat, *** P < 0.001, n = 8 mice, n = 777 imaged neurons. ( e ) The average amplitudes of sound-evoked response and echo responseterm. Wilcoxon rank sum testterm·stat, *** P < 0.001, n = 8 mice, n = 777 imaged neurons. ( f ) The average latencies of sound-evoked response and echo responseterm. Wilcoxon rank sum testterm·stat, *** P < 0.001, n = 8 mice, n = 777 imaged neurons. Error bars indicate SEM.
Frequency selectivity of echo responsesterm in the A1term in anesthetized mice To investigate whether the echo responsesterm of the neurons were related to their frequency selectivity, we evoked echo responsesterm with pure tones over a series of 2 to 40 kHz (every frequency was repeated 20 times with an ISI of 2 s) in anesthetized mice. Figure 3a shows an example neuron recorded in an anesthetized mouse that could be activated by sound stimulation at several frequencies, and the frequency evoked the greatest responses was 8.9 kHz. Remarkably, the echo responsesterm were evoked in this neuron when the stimulation sequence consisted of tones close to 8.9 kHz (Fig. 3b ). Although the overall tuning for sound-evoked responses across all of the echo neurons (Fig. 3c ) was relatively broad given the sound level used here, the echo responsesterm in most of the neurons occurred at the preferred frequencies. A comparison of the properties of the sound responsesterm and the echo responsesterm revealed that the frequency that evoked the echo responseterm in a given neuron closely corresponded to its preferred frequency to tone stimulation in anesthetized mice (Fig. 3d ). We also analyzed the spatial organization of the frequency-tuned echo neurons. Echo neurons in response to different frequencies were spatially mixed over a scale about 200 µm × 200 µm: adjacent neurons were capable of displaying echo responsesterm to very different frequencies we employed here (Fig. 3e ).
Figure 3.
A1term neurons preferentially show echo responsesterm to their best frequenciesterm in anesthetized mice. ( a ) Left: two-photonterm·key imaging of neurons in L2/3; right: tone-evoked responses (upper, average of 20 trials) and echo responsesterm (lower, single trials) in one example neuron outlined in the left panel. Eleven frequencies of pure tones were applied. Each frequency of stimulation was presented repeatedly for 20 times with an interval of 2 s. The echo responseterm of the neuron is marked in red. ( b ) Frequency tuning of the neuron presented in panel a for the tone-evoked responses (purple) and the echo responsesterm (orange). The points of the curves are the integrals of the Ca 2+ responses within 500 ms (the data fitted with Gaussian function). ( c ) The average tone-evoked frequency tuning property (the data fitted with Gaussian function) of 185 echo neurons from 6 anesthetized mice. ( d ) Distribution of the difference in preferred frequencyterm·var (ΔFrequency) between the tone response and the echo responseterm (the data fitted with Gaussian function) in 6 anesthetized mice. ( e ) Spatial distribution of the echo neurons in two example imaging planes from two anesthetized mice. The frequencies are those that evoked echo responsesterm.
Frequency selectivity of echo responsesterm in the A1term in awake mice It is widely believed that general anesthesiaterm·key will have an impact on lots of aspects of cortical activities. So we repeated these experiments in awake mice. We exploited a specially-made chamber and a fixing setup (Fig. 4a , left and middle), thus keeping the surface of the A1term vertical to the microscope objective. From the camera recording, we could know the mice were awake by observing the behavior such as eye opening, whisking and struggling (Fig. 4a , right). Whole-cell current-clamp recording in a L2/3 neuron of A1term showed that high-frequency but small membrane potential variations with a unimodal distribution further proved the mouse was in the awake condition (Fig. 4b ). These membrane potential fluctuations differed from the slow but large fluctuations recording in anesthetized mice. We then repeated the experiments to study the frequency selectivity of echo responsesterm in awake mice. By comparing the preferred frequencies between sound responsesterm and echo responsesterm (92 neurons from 4 awake mice), we found that the frequency that evoked the echo responseterm also closely corresponded to the preferred frequency to tone stimulation in awake mice (Fig. 4c ). The spatial distribution of the preferred frequency for evoking an echo responseterm in neurons was also mixed at a small scale (Fig. 4d ). Therefore, these results suggested that the echo responsesterm of neurons within A1term are highly related to their frequency selectivity.
Figure 4.
A1term neurons preferentially show echo responsesterm to their best frequenciesterm in awake mice. ( a ) Left and middle: a custom-made chamber for an awake mouse and its fixing apparatus; right: infrared image of a head-fixed mouse in the two-photonterm·key imaging setup. ( b ) Left: whole-cell recording in an example neuron from an awake mouse; right: the membrane potential (binned at 1 mV) distribution from the neuron presented in the left panel. ( c ) Distribution of the difference in preferred frequencyterm·var (ΔFrequency) between the tone-evoked response and the echo responseterm (the data fitted with Gaussian function) in 4 awake mice. ( d ) Spatial distribution of the echo neurons in two example imaging planes from two awake mice. The frequencies are those that evoked echo responsesterm.
Discussion Using two-photon Ca 2+ imagingterm in combination with the improved fluorescent Ca 2+ indicator Cal-520term AM, we confirmed that neuronal responses representing predictive information can be observed in the mouse A1term. The results showed that a subgroup of neurons in L2/3 exhibited echo responsesterm at the anticipated time interval following a train of rhythmic pure tone stimulation. This is consistent with the representations of predictive sound information in the A1term reported in a previous study that used broadband noise stimulation 5 . Furthermore, we found that these echo responsesterm are preferentially observed near the preferred frequencies of the neurons during sound stimulation. In an imaged region of approximately 200 µm × 200 µm, neurons showed echo responsesterm to any of the tested frequencies, suggesting heterogeneous organization of mouse A1term neurons in the coding of predictive information.
In previous studies, it has been shown that temporal intervals of visual inputs can be memorized by neural systems in the order of seconds 3 . Thus, understanding the mechanisms of how temporal information is processed in neural systems is crucial for elucidating brain functions 1 , such as for speech recognition and music appreciation. For the possible mechanisms of how echo responsesterm arise, a model proposed by Mi et al . 51 matches our previous 5 and current results. The simulation of the model reproduced the experimentally observed rhythmic synchronous firing, and it provides supports for the idea that temporal information can be processed by the dynamics of distributed circuits in the neural system. In this study, we believe that this model can also be used for explaining the mechanism for generating echo responseterm following rhythmic tone stimulation. We propose that, there is scale-free topologyterm in the cortical network, which contains hub neuronsterm in L2/3 and many low-degree loops with different lengths in L5 (Fig. 5 ; modified from the previous study 5 ). These low-degree loops enable the neural system to follow the external input rhythm, while hub neuronsterm can trigger synchronization in the whole network. During learning, the connectivity from the L5 loops to the L2/3 hubs is highly strengthened, which allows the signals to flow across the entire network. After learning through a train of rhythmic sensory stimulation, a loop with certain length is selected to match the input rhythm. Finally, the network is able to generate echo responsesterm with the same rhythm. Therefore, it implies that echo responseterm is an intrinsic property of cortical circuits in response to the absence of an expected sound, and echo responseterm may reflect a distinct neural process in comparison to sound-evoked responses. In our result, we found significant differences in the amplitudes (Fig. 2e ) and the latencies (Fig. 2f ) between sound-evoked responses and echo responsesterm. Those differences could be attributed to a variation in neurons for the estimation of the input rhythm, and it thus leads to a longer latency and a lower amplitude of the echo responsesterm·var on average.
Figure 5.
Working model for generating echo responseterm in cortical networks. After rhythmic sensory stimulation, a loop of low-degree neurons in L5 is formed with a certain length in the neural network to match the interval of the external input rhythm. Meanwhile, the connections from L5 low-degree neurons to L2/3 hub neuronsterm have been strengthened for transmitting signals from the selected L5 loop (pacemaker-like loop) to L2/3. Finally, the neural network is able to generate echo responsesterm with the same period as the rhythmic input.
A presumption in neuroscience shows that a brain is capable of effectively performing genuine Bayesian inferences 52 . It proposes that predictive codingterm may give an explanation on how Bayesian inference could be performed in the cerebral cortex from the neural perspective 53 , 54 . Theoretically, predictive codingterm suggests that the brain is constantly making efforts to ‘fit’ internal models in perceiving through inferring the incoming sensory input. The discrepancies between the actual sensory data and the predicted sensory data (‘error’) are used to refine the model in an iterative process until the model optimized (‘error’ minimized). Our studies based on unexpectedly omitted stimuli can give support for the prospective and predictive property of the expectancy outcomes, which would suggest the echo responseterm as possibly the signature finding of predictive codingterm, as it is even able to be observed when sensory input is absent. Furthermore, predictive codingterm theory assumes that neural responses are shaped by expectations, which are hierarchically organized. For instance, a simple anatomical assumption is that prediction and error computations are implemented by separate neurons that reside in different cortical layers 55 . However, evidences by far are still deficient in the auditory field in accordance with the notion that prediction and error neurons reside separately in distinct cortical layers 56 , which shall be studied further.
Receptive fieldsterm·key of neurons represent how sensory information is encoded and mapped for the guidance of perception and behavior 57 . The neurons in the A1term can form spectro-temporal filtersterm·key to enhance the auditory representation of stimuli along two-feature dimensions, incorporating both the frequency and the timing information of the relevant auditory stimulus sequence 12 . For example, neurons in the A1term had consistent spectral and temporal features to maintain stable receptive fieldsterm·key for several hours 58 . In our results, we found that the preferred frequency of echo responsesterm was very close to the preferred frequencies of the sound-triggered responses of the neurons in mice under both awake and anestheticterm conditions. These results suggest that neurons in the A1term maintained stable spectral receptive fieldsterm·key to form a reliable, consistent representation of tonal information for the prediction.
The precise tonotopic organizationterm in auditory cortex is a long debated issue 20 , 21 , 26 and remains unresolved. On the macroscopic scale level (~1 mm), the tonotopic mapterm in the A1term arises from MGBvterm. However, the frequency selectivity of neighboring neurons can be very different, and no reliable tonotopic organizationterm could be observed over spatial scales of <200 µm 22 . Early studies have shown that tonal selectivity in cats 59 and response variability in rats 60 were increased under anesthesiaterm·key, suggesting anestheticsterm would alter tonotopic mapterm to be stricter and more regular. As neuronal dynamics are modulated by the type and depth of anesthesiaterm·key, the observed topographic organization can be altered and thus there were a great number of experiments under light anesthesiaterm·key conditions. Previous studies of using two-photon Ca 2+ imagingterm in the anesthetized mouse brain demonstrated that the response profiles of nearby neurons were quite different within local circuit in A1term 21 , 25 . Moreover, a recent study also showed a sound-evoked local heterogeneityterm map of the functional organization in the A1term in both lightly anesthetized and awake mice 32 . Similarly, in this study we found local heterogeneityterm of echo responsesterm (Figs 3e and 4d ) in L2/3 in both the awake mice and the isofluorane (0.4–0.8%) anesthetized mice. In our results, the frequency selectivity profiles of echo responsesterm (Figs 3 and 4 ) are similar between anesthetized and awake mice, suggesting that the local heterogeneous organization for predictive codingterm could be observed under awake and light anesthesiaterm·key conditions. However, the precise effects of anesthesiaterm·key on A1term neurons of mouse are still largely unknown and hence need further investigation.
Therefore, our findings provide evidence for the heterogeneity of tonal organization for the coding of prediction in the A1term. The fact that the local topographic organization in L2/3 of A1term was heterogeneous may support the hypothesis that neurons with different frequency selectivity interact to represent the predictive pattern of complex sounds 21 . However, the underlying mechanisms of contribution to behavior are still largely unknown and warrant investigation in future studies.
Acknowledgements We thank Jia Lou for assisting in the composition of the figures. This work was supported by grants from the 1000 Talents Program for Young Scholars, the National Natural Science Foundation of China (No. 81671106, 31400933, 31700933), the National Basic Research Program of China (“973 Program”: 2015CB759500) and the China Postdoctoral Science Foundation (2015M572680). X.C. is a junior fellow of the CAS Center for Excellence in Brain Science and Intelligence Technology.
Author Contributions M.W., R.L., J.L., X.C., S.Z. and X.L. contributed to the design of the study and the interpretation of the data. M.W., R.L., J.L. and J.Z. performed the experiments and acquired the data. M.W., R.L., J.L. and X.L. processed and analyzed the data. M.W., R.L., X.C., S.Z. and X.L. wrote the manuscript with help from all of the other authors.
Competing Interests The authors declare that they have no competing interests.
Footnotes Meng WangpersName and Ruijie Li contributed equally to this work.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information Xiaowei ChenpersName, Email: xiaowei_chenpersName@tmmu.edu.cn.
Shaoqun ZengpersName, Email: sqzeng@mail.hust.edu.cn.
Xiang LiaopersName, Email: xiang.liaopersName@aliyun.com.