MohanpersName, Gallero-SalaspersName, CartapersName, SacramentopersName, LaurenczypersName, SumanovskipersName, de KockpersName, HelmchenpersName (2018date). “Sensory representation of an auditory cued tactile stimulus in the posterior parietal cortex of the mouse.” Scientific Reportstitle·j.
Object of study — how an association cortex represents a tactile stimulus, the sound that announces it, and its absence
Research question — how whisker-based tactile stimuli are represented in the microcircuits of posterior parietal cortex
Contribution — layer 2/3term neurons discriminate not only textureterm graininessterm·var but the presence against the omissionterm of the cued touch, and may encode expectancyterm
phenomenon · claim-type · cross-domain-transfer · mechanism · claim-relation · replication-status · limitation — grounded in the full text| phenomenon | stimulus-omission responses in posterior parietal cortex — “More interestingly, we found that about 30% of trial-active PPC neurons responded to the omission of the texture touch that always arrived after the auditory cues.” | |
| claim-type | existence — “Hence, PPC neurons were able to discriminate not only tactile stimulus features (i.e., texture graininess) but also between the presence and omission of the texture stimulus.” | |
| method — evidence | — “We performed in vivo calcium imaging of layer 2/3 PPC neurons (at 150 to 220 µm depth) through a cranial window using YC-Nano140, a genetically encoded ratiometric calcium indicator that reliably reports action potential discharge 27 (Fig. 2A ).” | |
| object of study — evidence | — “Sensory association cortices receive diverse inputs with their role in representing and integrating multi-sensory content remaining unclear.” | |
| research question — evidence | — “While visual and auditory processing in PPC during behaviour have been extensively studied in mice 7 , 15 , 17 , 18 , 25 , 26 , we know little about how whisker-based tactile stimuli are represented in the neuronal microcircuits of PPC.” | |
| contribution — evidence | — “Notably, altering the flow of sensory events through omission of the cued texture touch elicited large responses in a subset of neurons hardly responsive to or even inhibited by the tactile stimuli. Hence, PPC neurons were able to discriminate not only tactile stimulus features (i.e., texture graininess) but also between the presence and omission of the texture stimulus.” | |
| transfer | from machine learning: A linear SVM classifier (via the scikit-learn python package) is imported as a population read-out model for neural activity, and justified as an operation biological neurons could plausibly implement. — “A linear SVM computes a simple thresholded weighted sum of inputs, an operation that is plausibly implementable by actual biological neurons; high linear classifier performance has been suggested as an indicator that the neural representation is such that it can be easily readout by downstream neurons 62 .” | transfers[0] |
| transfer | from primate parietal physiology: looming responsiveness of multimodal neurons, carried into the mouse as the interpretive frame for the cue — “Apart from one neuron (onset 120 ms after the texture stop tone) the others displayed onset times well before the translation stage came to rest (onset 580 ± 65 ms before texture stop tone, n = 18 neurons). This indicates that the Catch-trial responses were not due to the stop auditory tone and could potentially represent an expectancy response that can be further enhanced by whisking.” | transfers[1] |
| transfer | from machine learning: a linear support vector machine used as a readout model on single-trial population activity — “As a complement to the ideal observer ROC analysis, we trained a linear support vector machine (SVM) classifier to decode texture graininess and texture omission from single-trial population activity, on a frame by frame basis. This simple readout model could effectively decode texture graininess and texture omission during the touch-stimulus window with about 55% and 70% accuracy, respectively (Fig. 4E ). The decoding accuracy was improved when the mean of the touch-stimulus window was used (60.1 ± 3.1% for texture graininess; 73.1 ± 4.3% for texture omission).” | transfers[2] |
| mechanism | The looming-like cue and the ensuing touch inhibit a subset of PPC neurons; when the expected touch fails to arrive these neurons are released from that inhibition, which is why omission produces a large response in cells that texture touch suppresses. — “Interestingly, 55.3% of the Catch-responsive neurons were non-responsive or even suppressed during texture presentation trials, suggesting that they were inhibited by the looming-like auditory cues and/or a subsequent texture touch, but relieved from inhibition when the texture was absent.” | mechanisms[0] |
| mechanism | Thalamic LP/Po input to PPC is gated by top-down feedback, plausibly from the frontal eye field, which activates local inhibitory circuits and supplies the prediction signal of a predictive-coding scheme, so a violated prediction emerges as the omission response. — “This could be in turn modulated through the activation of local inhibitory circuits by top-down feedback, potentially from the FEF, providing the prediction signal in the proposed framework of predictive coding 53 .” | mechanisms[1] |
| mechanism | A looming sound signals an approaching source and thereby creates an expectancy of an imminent touch event, which is the expectancy the omission then violates. — “Looming sounds have been proposed to represent the approach of the sound source in the context of threat detection, either in the form of an attack or collision, and can evoke rapid attention capture and expectancy 38 .” | mechanisms[2] |
| mechanism | Omission-responsive PPC neurons act as mismatch reporters that forward information about unexpected changes in sensory flow to prefrontal cortex, FEF and LP, areas allocating attention. — “Hence these neurons could function as reporters of mismatch signals, which could reliably convey information about unexpected changes in sensory flow to the prefrontal cortex, FEF and LP (the mouse equivalent of the pulvinar), brain areas that are implicated in the allocation of attentional resources 9 .” | mechanisms[3] |
| mechanism | the looming-like sound of the approaching stage may itself signal an incoming touch, so the omission response could be an expectancy that the touch does not satisfy — “Apart from one neuron (onset 120 ms after the texture stop tone) the others displayed onset times well before the translation stage came to rest (onset 580 ± 65 ms before texture stop tone, n = 18 neurons). This indicates that the Catch-trial responses were not due to the stop auditory tone and could potentially represent an expectancy response that can be further enhanced by whisking.” | mechanisms[4] |
| mechanism | two populations contribute: some neurons already respond before any tactile experience, others are recruited only after sound and touch have been coupled — “Our results suggest that while looming-like auditory cues alone can evoke responses in the PPC, coupling with tactile stimuli enhances this response and further recruits more neurons within PPC that can report subsequent texture-touch omissions.” | mechanisms[5] |
| mechanism | the thalamic lateral posterior nucleus and the posterior nucleus are the main thalamic input to this area and a candidate source of the omission response — “LP, along with the posterior nucleus Po, is the main source of thalamic input to the PPC, which does not receive any direct input from primary sensory thalamus 2 , 52 . Hence, LP/Po could be a potential source for the Catch-trial responses that we have observed in layer 2/3 of the PPC. This could be in turn modulated through the activation of local inhibitory circuits by top-down feedback, potentially from the FEF, providing the prediction signal in the proposed framework of predictive coding” | mechanisms[6] |
| mechanism | the responses could report mismatch to frontal and thalamic areas that allocate attention — “Hence these neurons could function as reporters of mismatch signals, which could reliably convey information about unexpected changes in sensory flow to the prefrontal cortex, FEF and LP (the mouse equivalent of the pulvinar), brain areas that are implicated in the allocation of attentional resources” | mechanisms[7] |
| builds-on | prior reports of a temporal associative role of PPC (cited as refs 19, 20) — “These findings are in line with the recently reported temporal associative role of PPC 19 , 20 , which could be due to its long timescale population dynamics allowing the combination of events separated in time 26 .” | claim_relations[0] |
| extends | the reported temporal associative role of posterior parietal cortex in encoding past choices and outcomes — “Hence, neurons within PPC are capable of representing a change in sensory flow, in a manner that is both rapid and stable. This could be essential for the role of PPC in spatial processing, navigation and the allocation of attention, allowing it to rapidly update changes in sensory information in particular to objects that are close to the body.” | claim_relations[1] |
| distinguishes-from | mismatch negativity and oddball paradigms, whose stimulus presentation differs from this one — “Indeed stimulus-omission responses have also been described in the auditory 42 , 43 and visual systems 44 – 46 in the context of mismatch negativity (MMN) and oddball responses. However, the stimulus presentation paradigms employed in MMN are different compared to those used in our study.” | claim_relations[2] |
| builds-on | mismatch responses reported in mouse V1 when visual flow is uncoupled from self-initiated locomotion — “More recently, mismatch responses have been reported in mouse V1 47 , 48 when visual flow is not correlated with self-initiated locomotion. These mismatch signals were also present in LP inputs to layer 1 of V1 49 , and mediated by a balance of motor-related excitation and visually driven inhibition during coupled sensorimotor experience 50 , 51 . While our stimulus paradigm does not involve a conflict with a self-initiated stimulus response, such as visual feedback from locomotion, it nonetheless incorporates a behaviourally salient stimulus (the looming auditory cue) that when coupled with tactile stimuli, can lead to a rapid encoding of omissions in the coupled sensory flow.” | claim_relations[3] |
| supports | the proposal that looming sounds signal an approaching source and evoke attention capture and expectancy — “Apart from one neuron (onset 120 ms after the texture stop tone) the others displayed onset times well before the translation stage came to rest (onset 580 ± 65 ms before texture stop tone, n = 18 neurons). This indicates that the Catch-trial responses were not due to the stop auditory tone and could potentially represent an expectancy response that can be further enhanced by whisking.” | claim_relations[4] |
| sense | a Catch-trial is not a control in the usual sense but the stimulus of interest - everything is present except the touch, so the response measures what the ABSENCE does — “To determine the influence of these stage-generated auditory cues, we interleaved the texture trials with ‘Catch’-trials, in which textures were omitted but with all auditory cues present (start and stop tones), including the sounds generated by the stage translation” | senses[0] |
| sense | the region called PPC here is identified by intrinsic-signal boundaries and then VALIDATED by tracing its inputs and outputs, rather than assumed from coordinates — “In four mice, we carried out retrograde and anterograde labelling experiments with tracer injections targeted to LPtA. The resulting labelling patterns validated that this region receives axonal projections from known PPC input sites (Fig. 1D ), particularly the thalamic lateral posterior nucleus LP (Fig. 1H,I ) and premotor cortex M2 (Fig. 1E-F ), as well as sends its axons to known PPC output sites, including M2, the dorsal striatum (Fig. 1E–G ) and LP (Fig. 1H,I ) 15 , 32 . Hence, we identify this region between barrel cortex and V1 as part of PPC” | senses[1] |
| sense | the auditory cue is not a designed stimulus but a by-product of the apparatus, measured after the fact and confirmed to be looming-like by recording at the animal's head position — “We confirmed that the auditory cue generated by the stage translation was indeed looming-like by performing sound recordings at the location corresponding to the head of the head-fixed mouse” | senses[2] |
| sense | response classes here are outputs of a general linear model fit to calcium traces, so cue-responsive, touch-responsive, mixed and suppressed are model categories rather than observed types — “Using a general linear model (GLM) to fit neuronal calcium traces, we classified the trial-active neurons (n = 126) as cue-responsive (22.2%), touch-responsive (11.9%), mixed-responsive (11.9%) and suppressed (17.5%) based on the dynamics of their calcium responses to the trial period; the remaining 36.5% of ‘unclassified’ active neurons responded neither to auditory cues nor to texture presentation” | senses[3] |
| sense | the omission response was present on the SECOND trial the animal ever experienced, which is what makes it a fact about the cue rather than about learning — “Furthermore, for all 4 FOVs the first Catch-trial was presented as the second trial in the sequence of trials, suggesting that the PPC is able to rapidly encode the omission of the auditory-cued tactile stimulus.” | senses[4] |
| control | every whisker except the one aligned to the imaged barrel column TRIMMED, with the remaining whisker's location mapped by intrinsic optical signal — held — rules out input from surrounding whiskers - a cortical response attributed to the mapped column could otherwise be driven by a neighbour, and barrel columns are close enough that this is not a remote possibility — “to avoid the activation of surrounding whiskers, all whiskers except the γ-barrel-column whisker on the right whisker pad of the mice were trimmed” | controls[0] |
| control | neurons EXCLUDED from the Catch-responsive class when their average calcium transient fell by more than 50 per cent — held — rules out the tail of a large cue-evoked transient being read as a separate response to the catch trial - a slow indicator makes one event look like two — “In order to avoid erroneously classifying cue-responsive neurons with large transients as Catch-responsive, neurons were excluded” | controls[1] |
| differs in | mismatch negativity and oddball paradigms, whose stimulus presentation differs from this one — THE AXIS IS THE STIMULUS PRESENTATION PARADIGM. 'stimulus-omission responses have also been described in the auditory and visual systems in the context of mismatch negativity (MMN) and oddball responses. However, the stimulus presentation paradigms employed in MMN are different compared to those used in our study.' The phenomenon is granted to exist elsewhere and the comparison is declined on a named ground - which is what stops this being a 'same-as' to a large existing literature. — “Indeed stimulus-omission responses have also been described in the auditory 42 , 43 and visual systems 44 – 46 in the context of mismatch negativity (MMN) and oddball responses. However, the stimulus presentation paradigms employed in MMN are different compared to those used in our study.” — “stimulus-omission responses have also been described in the auditory 42 , 43 and visual systems 44 – 46 in the context of mismatch negativity (MMN) and oddball responses. However, the stimulus presentation paradigms employed in MMN are different compared to those used in our study” | claim_relations[2] |
| note | Catch —distinct-from→ texture — everything is held present except the texture - the start and stop tones and the stage sounds all remain - so what is measured is the effect of an ABSENCE — “To determine the influence of these stage-generated auditory cues, we interleaved the texture trials with ‘Catch’-trials, in which textures were omitted but with all auditory cues present (start and stop tones), including the sounds generated by the stage translation” | concept_edges[0] |
| audit | stop tone —distinct-from→ Catch — Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span says the Catch-trial responses 'were not due to the stop auditory tone' - a claim about what caused a response, not a distinction between the tone and 'Catch'. — “Apart from one neuron (onset 120 ms after the texture stop tone) the others displayed onset times well before the translation stage came to rest (onset 580 ± 65 ms before texture stop tone, n = 18 neurons). This indicates that the Catch-trial responses were not due to the stop auditory tone and could potentially represent an expectancy response that can be further enhanced by whisking.” | concept_edges[3] |
| note | stop tone —distinct-from→ Catch — CRITERION: the responses begin about 580 ms BEFORE the stage stops, so the stop tone cannot be what drives them — “Apart from one neuron (onset 120 ms after the texture stop tone) the others displayed onset times well before the translation stage came to rest (onset 580 ± 65 ms before texture stop tone, n = 18 neurons). This indicates that the Catch-trial responses were not due to the stop auditory tone and could potentially represent an expectancy response that can be further enhanced by whisking.” | concept_edges[3] |
| audit | LPtA —same-as→ PPC — Audit 2026-09-28: DOUBTFUL - read against its span and kept, but LPtA is identified as PART of PPC; same-as overstates. — “In four mice, we carried out retrograde and anterograde labelling experiments with tracer injections targeted to LPtA. The resulting labelling patterns validated that this region receives axonal projections from known PPC input sites (Fig. 1D ), particularly the thalamic lateral posterior nucleus LP (Fig. 1H,I ) and premotor cortex M2 (Fig. 1E-F ), as well as sends its axons to known PPC output sites, including M2, the dorsal striatum (Fig. 1E–G ) and LP (Fig. 1H,I ) 15 , 32 . Hence, we identify this region between barrel cortex and V1 as part of PPC” | concept_edges[10] |
| note | LPtA —same-as→ PPC — THE IDENTITY IS EARNED: the anatomical name is matched to the functional one by tracing inputs and outputs, not asserted from coordinates — “In four mice, we carried out retrograde and anterograde labelling experiments with tracer injections targeted to LPtA. The resulting labelling patterns validated that this region receives axonal projections from known PPC input sites (Fig. 1D ), particularly the thalamic lateral posterior nucleus LP (Fig. 1H,I ) and premotor cortex M2 (Fig. 1E-F ), as well as sends its axons to known PPC output sites, including M2, the dorsal striatum (Fig. 1E–G ) and LP (Fig. 1H,I ) 15 , 32 . Hence, we identify this region between barrel cortex and V1 as part of PPC” | concept_edges[10] |
| audit | LP —causes→ PPC — Audit 2026-09-28: DOUBTFUL - read against its span and kept, but LP is the main thalamic INPUT to PPC; causes is the wrong type. — “LP, along with the posterior nucleus Po, is the main source of thalamic input to the PPC, which does not receive any direct input from primary sensory thalamus 2 , 52 . Hence, LP/Po could be a potential source for the Catch-trial responses that we have observed in layer 2/3 of the PPC. This could be in turn modulated through the activation of local inhibitory circuits by top-down feedback, potentially from the FEF, providing the prediction signal in the proposed framework of predictive coding” | concept_edges[11] |
| note | LP —causes→ PPC — the main thalamic input, and the candidate origin of the omission signal — “LP, along with the posterior nucleus Po, is the main source of thalamic input to the PPC, which does not receive any direct input from primary sensory thalamus 2 , 52 . Hence, LP/Po could be a potential source for the Catch-trial responses that we have observed in layer 2/3 of the PPC. This could be in turn modulated through the activation of local inhibitory circuits by top-down feedback, potentially from the FEF, providing the prediction signal in the proposed framework of predictive coding” | concept_edges[11] |
| weight | LP —causes→ PPC — 0.8 — “LP, along with the posterior nucleus Po, is the main source of thalamic input to the PPC, which does not receive any direct input from primary sensory thalamus 2 , 52 . Hence, LP/Po could be a potential source for the Catch-trial responses that we have observed in layer 2/3 of the PPC. This could be in turn modulated through the activation of local inhibitory circuits by top-down feedback, potentially from the FEF, providing the prediction signal in the proposed framework of predictive coding” | concept_edges[11] |
| audit | M2 —causes→ LPtA — Audit 2026-09-28: DOUBTFUL - read against its span and kept, but LPtA receives projections FROM M2; causes is the wrong type. — “In four mice, we carried out retrograde and anterograde labelling experiments with tracer injections targeted to LPtA. The resulting labelling patterns validated that this region receives axonal projections from known PPC input sites (Fig. 1D ), particularly the thalamic lateral posterior nucleus LP (Fig. 1H,I ) and premotor cortex M2 (Fig. 1E-F ), as well as sends its axons to known PPC output sites, including M2, the dorsal striatum (Fig. 1E–G ) and LP (Fig. 1H,I ) 15 , 32 . Hence, we identify this region between barrel cortex and V1 as part of PPC” | concept_edges[12] |
| weight | M2 —causes→ LPtA — 0.7 — “In four mice, we carried out retrograde and anterograde labelling experiments with tracer injections targeted to LPtA. The resulting labelling patterns validated that this region receives axonal projections from known PPC input sites (Fig. 1D ), particularly the thalamic lateral posterior nucleus LP (Fig. 1H,I ) and premotor cortex M2 (Fig. 1E-F ), as well as sends its axons to known PPC output sites, including M2, the dorsal striatum (Fig. 1E–G ) and LP (Fig. 1H,I ) 15 , 32 . Hence, we identify this region between barrel cortex and V1 as part of PPC” | concept_edges[12] |
| outcome (stated by the tagger, no quotation) | control — held | controls[0] |
| outcome (stated by the tagger, no quotation) | control — held | controls[1] |
| for which use | limitation — the three-subset structure: MORE DATA WOULD BE REQUIRED TO DETERMINE IF THESE THREE SUBSETS OF NEURONS REPRESENT FUNCTIONALLY DISTINCT GROUPS — “Due to the low numbers of Catch-responsive neurons detected, however, more data would be required to determine if these three subsets of neurons represent functionally distinct groups.” | limitations[0] |
| for which use | limitation — every texture response reported: WE WERE UNABLE TO REMOVE THE LOOMING-LIKE AUDITORY CUE, AS IT IS AN INTEGRAL PART OF THE TEXTURE PRESENTATION APPARATUS — “Unfortunately, we were unable to remove the looming-like auditory cue, as it is an integral part of the texture presentation apparatus.” | limitations[1] |
| for which use | limitation — the decoding figures: THIS APPROACH PROVIDES A LOWER BOUND ON THE ACTUAL INFORMATION CARRIED BY THE POPULATION, AND DOES NOT PRECLUDE THAT MORE ADVANCED METHODS CANNOT EXTRACT ADDITIONAL INFORMATION — “This approach provides a lower bound on the actual information carried by the population, and does not preclude that more advanced methods cannot extract additional information.” | limitations[2] |
| duplicate of | limitation — limitations[1] — “Unfortunately, we were unable to remove the looming-like auditory cue, as it is an integral part of the texture presentation apparatus.” | limitations[3] |
| for which use | limitation — the same scope as limitations[1] - the confound is produced by the apparatus and cannot be removed — “Unfortunately, we were unable to remove the looming-like auditory cue, as it is an integral part of the texture presentation apparatus.” | limitations[3] |
| for which use | limitation — the same scope as limitations[0] - the subsets are named but not claimed as distinct — “Due to the low numbers of Catch-responsive neurons detected, however, more data would be required to determine if these three subsets of neurons represent functionally distinct groups.” | limitations[4] |
| for which use | limitation — the texture-discrimination result: FEW PPC NEURONS (7.2%) WERE ABLE TO DISCRIMINATE THE TWO DISTINCT TEXTURES, in NAIVE MICE on their second exposure, with NO EXPLICIT ASSOCIATION OF A RELEVANT CONTEXT — “Few PPC neurons (7.2%) were able to discriminate the two distinct textures presented. This low discrimination power is not surprising as the experimental sessions were only the second time that the naïve mice were exposed to these textures (the first being setting the texture stop position, prior to imaging start). No explicit association of a relevant context to the texture graininess as stimulus feature or any reinforcement training had been applied.” | limitations[5] |
| for which use | limitation — the pre-existing-neuron interpretation: IT IS ALSO PLAUSIBLE THAT THE PRE-EXISTING CATCH-RESPONSIVE NEURONS ARE PURELY LOOMING SOUND RESPONSIVE ... AND START TO REPORT TEXTURE OMISSION ONLY AFTER COUPLING WITH TEXTURE TOUCH — “It is also plausible that the pre-existing Catch-responsive neurons are purely looming sound responsive in the Catch-only trials and start to report texture omission only after coupling with texture touch.” | limitations[6] |
| for which use | limitation — the comparison with the omission-response literature: THE STIMULUS PRESENTATION PARADIGMS EMPLOYED IN MMN ARE DIFFERENT COMPARED TO THOSE USED IN OUR STUDY — “Indeed stimulus-omission responses have also been described in the auditory 42 , 43 and visual systems 44 – 46 in the context of mismatch negativity (MMN) and oddball responses. However, the stimulus presentation paradigms employed in MMN are different compared to those used in our study.” | limitations[7] |
| replication | untested | |
| no type for this | Catch-trial responses → whisking — TESTED AND NOT SUPPORTED: Catch-trial responses were observed both in the ABSENCE and PRESENCE of whisking, indicating they were NOT SOLELY DEPENDENT ON WHISKING. Note the hedge the authors chose - 'not solely' leaves a contributory role open, so this is a bounded negative and not a flat denial. No edge type can express either. — “We next verified whether the Catch-trial responses might reflect an efferent copy of a motor command such as whisking because the PPC has strong reciprocal connections with the motor and premotor cortices 1 – 3 . Catch-trial responses were observed both in the absence and presence of whisking behaviour (see Methods) during the touch-stimulus window, indicating that they were not solely dependent on whisking” | unhoused_relations[0] |
| limitation | Too few Catch-responsive neurons were recorded to establish whether the pre-existing, silenced and emergent subsets are genuinely distinct functional groups. — “Due to the low numbers of Catch-responsive neurons detected, however, more data would be required to determine if these three subsets of neurons represent functionally distinct groups.” | limitations[0] |
| limitation | The looming-like auditory cue is inherent to the stimulation apparatus and could not be eliminated as a confound. — “Unfortunately, we were unable to remove the looming-like auditory cue, as it is an integral part of the texture presentation apparatus.” | limitations[1] |
| limitation | The linear SVM decoding gives only a lower bound on the information carried by the population. — “This approach provides a lower bound on the actual information carried by the population, and does not preclude that more advanced methods cannot extract additional information.” | limitations[2] |
| limitation | the looming auditory cue could not be removed because it is generated by the stimulus apparatus itself — “Unfortunately, we were unable to remove the looming-like auditory cue, as it is an integral part of the texture presentation apparatus.” | limitations[3] |
| limitation | the three subsets of omission-responsive neurons are named but explicitly not claimed as functionally distinct groups, for want of numbers — “Due to the low numbers of Catch-responsive neurons detected, however, more data would be required to determine if these three subsets of neurons represent functionally distinct groups.” | limitations[4] |
| limitation | texture discrimination was weak, which the authors attribute to naive animals with no training and no context attached to graininess — “Few PPC neurons (7.2%) were able to discriminate the two distinct textures presented. This low discrimination power is not surprising as the experimental sessions were only the second time that the naïve mice were exposed to these textures (the first being setting the texture stop position, prior to imaging start). No explicit association of a relevant context to the texture graininess as stimulus feature or any reinforcement training had been applied.” | limitations[5] |
| limitation | an alternative reading remains open in which pre-existing neurons are purely sound-driven and only begin reporting omission after coupling — “It is also plausible that the pre-existing Catch-responsive neurons are purely looming sound responsive in the Catch-only trials and start to report texture omission only after coupling with texture touch.” | limitations[6] |
| limitation | omission responses are known from mismatch and oddball work, but with stimulus paradigms different from this one — “Indeed stimulus-omission responses have also been described in the auditory 42 , 43 and visual systems 44 – 46 in the context of mismatch negativity (MMN) and oddball responses. However, the stimulus presentation paradigms employed in MMN are different compared to those used in our study.” | limitations[7] |
finding · quantity · variable · definition · key-term — grounded in the full text| finding | Layer 2/3 PPC neurons split into subgroups that respond to the auditory cue, to the texture touch, to both, or that are suppressed during cue and texture presentation. — “PPC neurons displayed a diversity of trial-related responses with some neurons responding to the auditory cues, to the texture stimuli, or both, while others showed suppressed activity during cue and texture presentation” | content_tags.findings[0] |
| finding | About 30% of trial-active PPC neurons fired to the omission of the texture touch that normally followed the auditory cues. — “More interestingly, we found that about 30% of trial-active PPC neurons responded to the omission of the texture touch that always arrived after the auditory cues.” | content_tags.findings[1] |
| finding | More than twice as many neurons discriminated texture presence from omission (17.5%) as discriminated the two texture grades from each other (7.2%). — “Among the trial-active PPC neurons 7.2% (9/126) discriminated between the graininess of the presented textures whereas 17.5% (22/126) discriminated between texture presence and omission” | content_tags.findings[2] |
| finding | The first Catch-trial was the second trial of the session in every field of view, suggesting PPC can rapidly encode the omission. — “Furthermore, for all 4 FOVs the first Catch-trial was presented as the second trial in the sequence of trials, suggesting that the PPC is able to rapidly encode the omission of the auditory-cued tactile stimulus.” | content_tags.findings[3] |
| finding | PPC neurons reported the texture omission reliably and stably over the session. — “This suggests that PPC neurons are able to report the omission of the texture reliably and stably over time.” | content_tags.findings[4] |
| finding | Omission responses occurred with and without whisking, so they were not solely dependent on whisking. — “Catch-trial responses were observed both in the absence and presence of whisking behaviour (see Methods) during the touch-stimulus window, indicating that they were not solely dependent on whisking” | content_tags.findings[5] |
| finding | Omission responses were not caused by the stop tone and may represent an expectancy response that whisking can enhance. — “This indicates that the Catch-trial responses were not due to the stop auditory tone and could potentially represent an expectancy response that can be further enhanced by whisking.” | content_tags.findings[6] |
| finding | Looming-like auditory cues alone can evoke PPC responses, and coupling them with texture touch enhances the response and recruits more neurons reporting omission. — “Our results suggest that while looming-like auditory cues alone can evoke responses in the PPC, coupling with tactile stimuli enhances this response and further recruits more neurons within PPC that can report subsequent texture-touch omissions.” | content_tags.findings[7] |
| finding | A linear classifier decoded texture omission from single-trial population activity better than texture graininess (about 70% versus about 55% accuracy). — “This simple readout model could effectively decode texture graininess and texture omission during the touch-stimulus window with about 55% and 70% accuracy, respectively” | content_tags.findings[8] |
| finding | Only 7.2% of PPC neurons discriminated the two textures. — “Few PPC neurons (7.2%) were able to discriminate the two distinct textures presented.” | content_tags.findings[9] |
| finding | Playback of the recorded looming cue evoked Catch-like responses whereas a non-looming recording did not. — “In a separate set of experiments in these mice, we were also able to evoke Catch-trial like responses in neurons when presented with the audio playback of the looming-like auditory cue but not when presented with a non-looming recording” | content_tags.findings[10] |
| finding | about 30 percent of trial-active neurons responded when the cued texture touch was omitted — “To our surprise, a subset of these ‘unclassified’ neurons (12 out of 46) as well as a subset of classified neurons (26 out of 80) showed a significant response in the touch-stimulus window during Catch-trials, when the texture was omitted (see for example average traces of neurons 1, 3 and 5, Fig. 2A ). We call these special neurons ‘Catch-responsive’ neurons (see Methods), which overall made up 30.2% of the active neurons” | content_tags.findings[11] |
| finding | 55.3 percent of those omission-responsive neurons were non-responsive to or suppressed by the texture itself — “Interestingly, 55.3% of the Catch-responsive neurons were non-responsive or even suppressed during texture presentation trials, suggesting that they were inhibited by the looming-like auditory cues and/or a subsequent texture touch, but relieved from inhibition when the texture was absent.” | content_tags.findings[12] |
| finding | the trial-active neurons split by GLM into cue-responsive, touch-responsive, mixed, suppressed and unclassified — “Using a general linear model (GLM) to fit neuronal calcium traces, we classified the trial-active neurons (n = 126) as cue-responsive (22.2%), touch-responsive (11.9%), mixed-responsive (11.9%) and suppressed (17.5%) based on the dynamics of their calcium responses to the trial period; the remaining 36.5% of ‘unclassified’ active neurons responded neither to auditory cues nor to texture presentation” | content_tags.findings[13] |
| finding | 126 of 358 neurons were active during trials, imaged across 13 fields of view in 4 mice — “We acquired imaging data from 13 fields of views (FOVs) in 4 mice, recording calcium signals in a total of 358 neurons. About 35% of neurons (126/358) were active during the trials” | content_tags.findings[14] |
| finding | the omission response was comparable at the start and end of a session, p = 0.75 — “We found that the peak amplitudes of the evoked responses in Catch-responsive neurons were comparable at the beginning and end of the imaging session (Fig. 3A,B ; peak ∆ R/R : first 5 trials 9.3 ± 2.4%, last 5 trials 12.1 ± 2.5%; p = 0.75; n = 17 Catch-responsive neurons in 4 FOVs, Wilcoxon signed paired test).” | content_tags.findings[15] |
| finding | LEARNING EXCLUDED BY WHEN IT FIRST APPEARS: the first Catch-trial was the SECOND trial of the session in all four fields of view — “Furthermore, for all 4 FOVs the first Catch-trial was presented as the second trial in the sequence of trials, suggesting that the PPC is able to rapidly encode the omission of the auditory-cued tactile stimulus.” | content_tags.findings[16] |
| finding | the motor-copy account was tested: omission responses occurred with and without whisking, so they are not solely dependent on it — “We next verified whether the Catch-trial responses might reflect an efferent copy of a motor command such as whisking because the PPC has strong reciprocal connections with the motor and premotor cortices 1 – 3 . Catch-trial responses were observed both in the absence and presence of whisking behaviour (see Methods) during the touch-stimulus window, indicating that they were not solely dependent on whisking” | content_tags.findings[17] |
| finding | whisking did enhance the response, 16.4 against 10.1 percent — “A comparison of the peak of the Catch-trial responses in the presence and absence of whisking during this time window revealed that the response was enhanced by whisking (Fig. 3D ; peak ∆ R/R : whisking 16.4 ± 1.4%, non-whisking 10.1 ± 1.0%; p = 1.47 × 10 −7 , n = 34 Catch-responsive neurons in 12 FOVs, Wilcoxon signed paired test; whiskers could not be tracked in one of the 13 FOVs).” | content_tags.findings[18] |
| finding | but response onset time did not differ with whisking, 3.7 against 3.8 s — “However, the onset time of the Catch-trial responses (from trial start) were comparable in the presence and absence of whisking (Fig. 3E ; Catch-response onset time: whisking 3.7 ± 0.1 s, non-whisking 3.8 ± 0.1 s; p = 0.56, n = 34 Catch-responsive neurons in 17.3 ± 0.1% and 82.7 ± 0.1% of Catch-trials respectively, Wilcoxon signed paired test).” | content_tags.findings[19] |
| finding | ACOUSTIC CONFOUND KILLED BY TIMING: omission responses began about 580 ms before the stage stop tone — “Apart from one neuron (onset 120 ms after the texture stop tone) the others displayed onset times well before the translation stage came to rest (onset 580 ± 65 ms before texture stop tone, n = 18 neurons). This indicates that the Catch-trial responses were not due to the stop auditory tone and could potentially represent an expectancy response that can be further enhanced by whisking.” | content_tags.findings[20] |
| finding | omission responses in the touch window were about twice the size of the texture-touch responses — “Catch trial responses in the touch-stimulus window were larger compared to their corresponding texture-touch responses (Fig. 3G ; peak ∆ R/R : P100 4.5 ± 1.0%, P1200 5.5 ± 0.7%, Catch 10.1 ± 1.0%; p = 1.72 × 10 −5 for P100 vs Catch, p = 4.66 × 10 −5 for P1200 vs Catch; n = 34 Catch-responsive neurons in 12 FOVs; Wilcoxon signed paired test).” | content_tags.findings[21] |
| finding | 7.2 percent of active neurons discriminated the two textures while 17.5 percent discriminated presence from omission — “We carried out an ROC analysis to identify the fraction of neurons that could significantly discriminate between the two presented textures as well as the fraction of neurons that could discriminate between texture-presentation and Catch-trials (compared to shuffled controls, Methods; example traces in Fig. 4B ). Among the trial-active PPC neurons 7.2% (9/126) discriminated between the graininess of the presented textures whereas 17.5% (22/126) discriminated between texture presence and omission” | content_tags.findings[22] |
| finding | a population decoder read out graininess at about 60 percent and omission at about 73 percent — “As a complement to the ideal observer ROC analysis, we trained a linear support vector machine (SVM) classifier to decode texture graininess and texture omission from single-trial population activity, on a frame by frame basis. This simple readout model could effectively decode texture graininess and texture omission during the touch-stimulus window with about 55% and 70% accuracy, respectively (Fig. 4E ). The decoding accuracy was improved when the mean of the touch-stimulus window was used (60.1 ± 3.1% for texture graininess; 73.1 ± 4.3% for texture omission).” | content_tags.findings[23] |
| finding | THE CUE DECOUPLED FROM THE TOUCH: three further naive mice received Catch-only trials before any texture was ever presented — “To verify this possibility, we presented an additional 3 naïve mice first with Catch-trials only, effectively decoupling the auditory cues from tactile stimuli.” | content_tags.findings[24] |
| finding | 10 of 74 neurons already responded in the Catch-only condition, and 8 further neurons became responsive only after textures were introduced — “Indeed, 10 out of the 74 active neurons displayed Catch-responses in the Catch-only trials, as identified using the GLM (see Methods). Eight of these neurons were also Catch-responsive in the interleaved trials (‘pre-existing’ Catch responses) whereas 2 neurons were not classified as Catch-responsive anymore (‘silenced’ responses; Fig. 5A,B ). In addition, 8 other neurons were identified as Catch-responsive only after the introduction of tactile stimuli but not in the Catch-only condition” | content_tags.findings[25] |
| finding | CATEGORIES NAMED AND THEN DECLINED: the three subsets are not claimed as functionally distinct groups because the numbers are too low — “Due to the low numbers of Catch-responsive neurons detected, however, more data would be required to determine if these three subsets of neurons represent functionally distinct groups.” | content_tags.findings[26] |
| finding | coupling sound to touch significantly increased the omission response, 3.6 to 5.2 percent — “The Catch-trial response showed a significant increase (touch-stimulus peak ∆ R/R : Catch-only trials 3.6 ± 0.7%, interleaved trials 5.2 ± 0.5%; p = 4.8 × 10 −5 , n = 74 neurons, Wilcoxon signed paired test) after tactile experience with 14.9% of active neurons displaying a significant change in activity within this window (8 cells with an increase, 3 cells with a decrease; significance level set at p < 0.01, Wilcoxon-Mann-Whitney test).” | content_tags.findings[27] |
| finding | so looming sound alone evokes the response, and coupling with touch enlarges it and recruits more neurons — “Our results suggest that while looming-like auditory cues alone can evoke responses in the PPC, coupling with tactile stimuli enhances this response and further recruits more neurons within PPC that can report subsequent texture-touch omissions.” | content_tags.findings[28] |
| finding | THE RIVAL STATED WITH ITS OWN BEST EVIDENCE CONCEDED: more than half of these neurons did not respond to texture at all, which is what an acoustic account predicts — “An alternative interpretation is that Catch-responsive neurons are actually acoustically driven and the presence of texture touches merely suppresses their response. Indeed, more than half of Catch-responsive neurons did not respond to or were suppressed by texture touches. However, these Catch-trial responses had onset times well before the tone signalling stage stop (with the exception of one neuron), suggesting that something else than sound, such as the expectancy of an incoming object, evoked the increase in activity.” | content_tags.findings[29] |
| finding | and then answered: sensory experience changed the number of omission-responsive neurons, which a fixed sensory response should not do — “Our results argue against this notion, however, because sensory experience led to changes in the number of Catch-responsive neurons in subsets of PPC neurons, which should not be the case for a reliable sensory response.” | content_tags.findings[30] |
| finding | playback of the recorded looming cue evoked the response while a non-looming recording did not — “We were also able to evoke Catch-trial like responses in neurons when presented with a playback of this auditory looming-like cue, and not with a non-looming recording in mice with previous tactile experience, effectively eliminating other potential unaccounted stimulus sources for the evoked response.” | content_tags.findings[31] |
| finding | THE CONFOUND CANNOT BE REMOVED: the looming cue is generated by the texture presentation apparatus itself — “Unfortunately, we were unable to remove the looming-like auditory cue, as it is an integral part of the texture presentation apparatus.” | content_tags.findings[32] |
| finding | retrograde and anterograde labelling validated that this region receives projections from known PPC input sites and sends axons to known PPC output sites, which is why it is identified as part of PPC — “In four mice, we carried out retrograde and anterograde labelling experiments with tracer injections targeted to LPtA. The resulting labelling patterns validated that this region receives axonal projections from known PPC input sites (Fig. 1D ), particularly the thalamic lateral posterior nucleus LP (Fig. 1H,I ) and premotor cortex M2 (Fig. 1E-F ), as well as sends its axons to known PPC output sites, including M2, the dorsal striatum (Fig. 1E–G ) and LP (Fig. 1H,I ) 15 , 32 . Hence, we identify this region between barrel cortex and V1 as part of PPC” | content_tags.findings[33] |
| finding | the cue's looming quality was verified by recording sound at the position of the animal's head — “We confirmed that the auditory cue generated by the stage translation was indeed looming-like by performing sound recordings at the location corresponding to the head of the head-fixed mouse” | content_tags.findings[34] |
| finding | weak texture discrimination is attributed to naive untrained animals with no context attached to graininess — “Few PPC neurons (7.2%) were able to discriminate the two distinct textures presented. This low discrimination power is not surprising as the experimental sessions were only the second time that the naïve mice were exposed to these textures (the first being setting the texture stop position, prior to imaging start). No explicit association of a relevant context to the texture graininess as stimulus feature or any reinforcement training had been applied.” | content_tags.findings[35] |
| finding | the paper found evidence for both accounts of the omission response: rapid binding in some neurons and direct cue-driving in others — “This was observed in the very first Catch-trial, suggesting that these neurons either are capable of rapid stimulus binding and association 36 or that they respond directly to the auditory cues generated by the stage translation. We found evidence for both, as Catch-responsive neurons were observed in both the initial Catch-only trials prior to texture touch as well as emerging later in another set of neurons in the subsequent series of interleaved trials.” | content_tags.findings[36] |
| quantity | 358 neurons, 13 FOVs, 4 mice — Size of the main imaging dataset — “We acquired imaging data from 13 fields of views (FOVs) in 4 mice, recording calcium signals in a total of 358 neurons.” | content_tags.quantities[0] |
| quantity | 35% (126/358) — Proportion of imaged neurons that were active during the trials — “About 35% of neurons (126/358) were active during the trials (see Methods).” | content_tags.quantities[1] |
| quantity | 30.2% — Share of active neurons classified as Catch-responsive, i.e. responding when the texture was omitted — “We call these special neurons ‘Catch-responsive’ neurons (see Methods), which overall made up 30.2% of the active neurons” | content_tags.quantities[2] |
| quantity | 55.3% — Share of Catch-responsive neurons that were non-responsive to or suppressed by actual texture presentation — “Interestingly, 55.3% of the Catch-responsive neurons were non-responsive or even suppressed during texture presentation trials, suggesting that they were inhibited by the looming-like auditory cues and/or a subsequent texture touch, but relieved from inhibition when the texture was absent.” | content_tags.quantities[3] |
| quantity | 7.2% (9/126) versus 17.5% (22/126) — Fraction of trial-active neurons discriminating texture graininess versus fraction discriminating texture presence from omission — “Among the trial-active PPC neurons 7.2% (9/126) discriminated between the graininess of the presented textures whereas 17.5% (22/126) discriminated between texture presence and omission” | content_tags.quantities[4] |
| quantity | about 55% and 70% — Single-trial population decoding accuracy for texture graininess and for texture omission respectively — “This simple readout model could effectively decode texture graininess and texture omission during the touch-stimulus window with about 55% and 70% accuracy, respectively” | content_tags.quantities[5] |
| quantity | 0.10 — Average R-squared of the GLM fits to the classified neurons' calcium traces — “We obtained an average R 2 value of 0.10” | content_tags.quantities[6] |
| quantity | 30.2% — share of active neurons responding to texture omission — “To our surprise, a subset of these ‘unclassified’ neurons (12 out of 46) as well as a subset of classified neurons (26 out of 80) showed a significant response in the touch-stimulus window during Catch-trials, when the texture was omitted (see for example average traces of neurons 1, 3 and 5, Fig. 2A ). We call these special neurons ‘Catch-responsive’ neurons (see Methods), which overall made up 30.2% of the active neurons” | content_tags.quantities[7] |
| quantity | 55.3% — of those, the share not responsive to or suppressed by the texture — “Interestingly, 55.3% of the Catch-responsive neurons were non-responsive or even suppressed during texture presentation trials, suggesting that they were inhibited by the looming-like auditory cues and/or a subsequent texture touch, but relieved from inhibition when the texture was absent.” | content_tags.quantities[8] |
| quantity | 580 ± 65 ms before texture stop tone — how early the omission response starts, which excludes the stop tone — “Apart from one neuron (onset 120 ms after the texture stop tone) the others displayed onset times well before the translation stage came to rest (onset 580 ± 65 ms before texture stop tone, n = 18 neurons). This indicates that the Catch-trial responses were not due to the stop auditory tone and could potentially represent an expectancy response that can be further enhanced by whisking.” | content_tags.quantities[9] |
| quantity | 7.2% (9/126) and 17.5% (22/126) — neurons discriminating texture type against those discriminating presence from omission — “We carried out an ROC analysis to identify the fraction of neurons that could significantly discriminate between the two presented textures as well as the fraction of neurons that could discriminate between texture-presentation and Catch-trials (compared to shuffled controls, Methods; example traces in Fig. 4B ). Among the trial-active PPC neurons 7.2% (9/126) discriminated between the graininess of the presented textures whereas 17.5% (22/126) discriminated between texture presence and omission” | content_tags.quantities[10] |
| quantity | 126/358 — neurons active during trials, of all imaged — “We acquired imaging data from 13 fields of views (FOVs) in 4 mice, recording calcium signals in a total of 358 neurons. About 35% of neurons (126/358) were active during the trials” | content_tags.quantities[11] |
| variable | graininess (independent) — “In order to determine how tactile sensory information is represented in the neocortical microcircuit of PPC, we presented one of two textured surfaces (sandpaper of grade P100 and P1200, corresponding to rough and smooth surfaces respectively) to the mouse whisker pad 27 .” | content_tags.variables[0] |
| variable | texture presentation (independent) — “Trials without texture presentation (“Catch”) but with linear stage translation were interleaved” | content_tags.variables[1] |
| variable | whisking during the touch-stimulus window (moderator) — “Catch-trial responses were observed both in the absence and presence of whisking behaviour (see Methods) during the touch-stimulus window, indicating that they were not solely dependent on whisking” | content_tags.variables[3] |
| variable | omission — “We carried out an ROC analysis to identify the fraction of neurons that could significantly discriminate between the two presented textures as well as the fraction of neurons that could discriminate between texture-presentation and Catch-trials (compared to shuffled controls, Methods; example traces in Fig. 4B ). Among the trial-active PPC neurons 7.2% (9/126) discriminated between the graininess of the presented textures whereas 17.5% (22/126) discriminated between texture presence and omission” | content_tags.variables[7] |
| variable | whisking — “A comparison of the peak of the Catch-trial responses in the presence and absence of whisking during this time window revealed that the response was enhanced by whisking (Fig. 3D ; peak ∆ R/R : whisking 16.4 ± 1.4%, non-whisking 10.1 ± 1.0%; p = 1.47 × 10 −7 , n = 34 Catch-responsive neurons in 12 FOVs, Wilcoxon signed paired test; whiskers could not be tracked in one of the 13 FOVs).” | content_tags.variables[8] |
| variable | texture — “To determine the influence of these stage-generated auditory cues, we interleaved the texture trials with ‘Catch’-trials, in which textures were omitted but with all auditory cues present (start and stop tones), including the sounds generated by the stage translation” | content_tags.variables[9] |
| definition | Catch-trial — “To determine the influence of these stage-generated auditory cues, we interleaved the texture trials with ‘Catch’-trials, in which textures were omitted but with all auditory cues present (start and stop tones), including the sounds generated by the stage translation” | content_tags.definitions[0] |
| definition | Catch-responsive neurons — “We call these special neurons ‘Catch-responsive’ neurons (see Methods), which overall made up 30.2% of the active neurons” | content_tags.definitions[1] |
| definition | ‘emergent’ responses — “In addition, 8 other neurons were identified as Catch-responsive only after the introduction of tactile stimuli but not in the Catch-only condition (‘emergent’ responses, Fig. 5A,B ).” | content_tags.definitions[3] |
| definition | whisking amplitude — “Whisking amplitude, defined as the angle between the whisker shaft and sagittal plane from nose tip to the back of the head, was used as a measure to represent both rhythmic and non-rhythmic forms of whisking behaviour.” | content_tags.definitions[4] |
| definition | active neuron — “Active neurons were identified by two-way ANOVA with repeated measures of the neuronal calcium signal against the neuropil signal, defined by all pixels not assigned to neuronal soma of the overall ROI annotation” | content_tags.definitions[5] |
| definition | Catch — “To determine the influence of these stage-generated auditory cues, we interleaved the texture trials with ‘Catch’-trials, in which textures were omitted but with all auditory cues present (start and stop tones), including the sounds generated by the stage translation” | content_tags.definitions[6] |
| definition | GLM — “Using a general linear model (GLM) to fit neuronal calcium traces, we classified the trial-active neurons (n = 126) as cue-responsive (22.2%), touch-responsive (11.9%), mixed-responsive (11.9%) and suppressed (17.5%) based on the dynamics of their calcium responses to the trial period; the remaining 36.5% of ‘unclassified’ active neurons responded neither to auditory cues nor to texture presentation” | content_tags.definitions[7] |
| definition | SVM — “As a complement to the ideal observer ROC analysis, we trained a linear support vector machine (SVM) classifier to decode texture graininess and texture omission from single-trial population activity, on a frame by frame basis. This simple readout model could effectively decode texture graininess and texture omission during the touch-stimulus window with about 55% and 70% accuracy, respectively (Fig. 4E ). The decoding accuracy was improved when the mean of the touch-stimulus window was used (60.1 ± 3.1% for texture graininess; 73.1 ± 4.3% for texture omission).” | content_tags.definitions[8] |
| definition | LPtA — “In order to localize PPC in mice 3 , 6 , we performed intrinsic optical signal (IOS) imaging 28 , 29 to visualize the boundaries between the γ-barrel column in the mouse whisker barrel cortex and the primary visual cortex V1 (see Methods) (Fig. 1C ). This allowed us to select imaging areas in posterior parietal area A 3 , 30 in the lateral parietal association cortex (LPtA) 31 , located in between the somatosensory barrel cortex and V1.” | content_tags.definitions[9] |
| definition | ‘pre-existing’ Catch responses — “Eight of these neurons were also Catch-responsive in the interleaved trials (‘pre-existing’ Catch responses) whereas 2 neurons were not classified as Catch-responsive anymore (‘silenced’ responses; Fig. 5A,B ).” | content_tags.definitions[10] |
| definition | ‘silenced’ responses — “Eight of these neurons were also Catch-responsive in the interleaved trials (‘pre-existing’ Catch responses) whereas 2 neurons were not classified as Catch-responsive anymore (‘silenced’ responses; Fig. 5A,B ).” | content_tags.definitions[11] |
| key-terms | PPC, layer 2/3, Catch, texture, looming, auditory cue, whisking, expectancy, omission, GLM, ROC, SVM, IOS, LP, M2, mismatch, predictive coding |
| Wilcoxon signed-rank test | as written: Wilcoxon signed-rank paired test — “Non-parametric tests were performed for all analyses, and the Wilcoxon signed-rank paired test and the Wilcoxon-Mann-Whitney test for paired and unpaired group comparisons respectively and the Wilcoxon signed-rank test for single distributions.” | content_tags.statistics[0] |
| Mann-Whitney U test | as written: Wilcoxon-Mann-Whitney test — “Non-parametric tests were performed for all analyses, and the Wilcoxon signed-rank paired test and the Wilcoxon-Mann-Whitney test for paired and unpaired group comparisons respectively and the Wilcoxon signed-rank test for single distributions.” | content_tags.statistics[1] |
| repeated-measures ANOVA | as written: two-way ANOVA with repeated measures — “Active neurons were identified by two-way ANOVA with repeated measures of the neuronal calcium signal against the neuropil signal, defined by all pixels not assigned to neuronal soma of the overall ROI annotation” | content_tags.statistics[2] |
| general linear model β | as written: general linear model (GLM) — “A general linear model (GLM) was used to fit the neuronal calcium transients and classify neurons based on their responses” | content_tags.statistics[3] |
| ROC analysis | as written: receiver operating characteristic (ROC) analysis — “To quantify single-neuron discrimination performance we resorted to a receiver operating characteristic (ROC) analysis.” | content_tags.statistics[4] |
| area under the curve AUC | as written: area under the curve (AUC) — “Discrimination performance was then summarized as the obtained area under the curve (AUC).” | content_tags.statistics[5] |
| permutation test N | as written: Monte Carlo permutation test — “Statistical significance was assessed by a Monte Carlo permutation test” | content_tags.statistics[6] |
| support vector machine classification | as written: linear support vector machine (SVM) classifiers — “We trained linear support vector machine (SVM) classifiers to decode texture graininess (P100 vs. P1200) and texture omission from single trial population activity.” | content_tags.statistics[7] |
| held-out test set validation | as written: model parameters and were determined on a training set, consisting of 50% of the recorded trials — “Classification performance was assessed by measuring the accuracy on the remaining 50% of the trials that resulted from the initial split of the dataset and were held aside during training.” | content_tags.statistics[8] |
| standard error of the mean s.e.m. | as written: mean ± s.e.m. — “All data are presented as mean ± s.e.m. unless stated otherwise.” | content_tags.statistics[9] |
| coefficient of determination R 2 | as written: average R 2 value — “We obtained an average R 2 value of 0.10” | content_tags.statistics[10] |
| percentiles | as written: the median, the 25 th and 75 th percentiles in the boxes — “Box plots represent the median, the 25 th and 75 th percentiles in the boxes, whereas the sidebars represent the 5 th and 95 th percentiles of the distribution.” | content_tags.statistics[11] |
| significance level α | as written: p-value was below a significance level of α = 0.01 — “A regression coefficient was considered significantly deviating from baseline β 0 when the corresponding p-value was below a significance level of α = 0.01.” | content_tags.statistics[12] |
| Wilcoxon signed paired test | — “We found that the peak amplitudes of the evoked responses in Catch-responsive neurons were comparable at the beginning and end of the imaging session (Fig. 3A,B ; peak ∆ R/R : first 5 trials 9.3 ± 2.4%, last 5 trials 12.1 ± 2.5%; p = 0.75; n = 17 Catch-responsive neurons in 4 FOVs, Wilcoxon signed paired test).” | content_tags.statistics[13] |
| Wilcoxon-Mann-Whitney test | — “The Catch-trial response showed a significant increase (touch-stimulus peak ∆ R/R : Catch-only trials 3.6 ± 0.7%, interleaved trials 5.2 ± 0.5%; p = 4.8 × 10 −5 , n = 74 neurons, Wilcoxon signed paired test) after tactile experience with 14.9% of active neurons displaying a significant change in activity within this window (8 cells with an increase, 3 cells with a decrease; significance level set at p < 0.01, Wilcoxon-Mann-Whitney test).” | content_tags.statistics[14] |
| ROC analysis | — “We carried out an ROC analysis to identify the fraction of neurons that could significantly discriminate between the two presented textures as well as the fraction of neurons that could discriminate between texture-presentation and Catch-trials (compared to shuffled controls, Methods; example traces in Fig. 4B ). Among the trial-active PPC neurons 7.2% (9/126) discriminated between the graininess of the presented textures whereas 17.5% (22/126) discriminated between texture presence and omission” | content_tags.statistics[15] |
| support vector machine (SVM) classifier | — “As a complement to the ideal observer ROC analysis, we trained a linear support vector machine (SVM) classifier to decode texture graininess and texture omission from single-trial population activity, on a frame by frame basis. This simple readout model could effectively decode texture graininess and texture omission during the touch-stimulus window with about 55% and 70% accuracy, respectively (Fig. 4E ). The decoding accuracy was improved when the mean of the touch-stimulus window was used (60.1 ± 3.1% for texture graininess; 73.1 ± 4.3% for texture omission).” | content_tags.statistics[16] |
| claim type | existence — a second reading reaching the same value; recorded: existence — “These findings indicate that layer 2/3 neuronal populations in PPC potentially encode correlates of expectancy in addition to auditory and tactile stimuli.” | alternatives.claim_type |
| method | two-photon calcium imaging with a ratiometric indicator, GLM classification of response types, ROC and support vector machine decoding — a second reading DIFFERING from the recorded value; recorded: calcium-imaging — “We performed in vivo calcium imaging of layer 2/3 PPC neurons (at 150 to 220 µm depth) through a cranial window using YC-Nano140, a genetically encoded ratiometric calcium indicator that reliably reports action potential discharge” | alternatives.method |
| phenomenon | large responses on trials where the cued texture touch is omitted, in neurons that barely respond to the touch itself — a second reading DIFFERING from the recorded value; recorded: stimulus-omission responses in posterior parietal cortex — “Interestingly, 55.3% of the Catch-responsive neurons were non-responsive or even suppressed during texture presentation trials, suggesting that they were inhibited by the looming-like auditory cues and/or a subsequent texture touch, but relieved from inhibition when the texture was absent.” | 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.
Sensory representation of an auditory cued tactile stimulus in the posterior parietal cortex of the mouse Hemanth MohanpersName Hemanth MohanpersName 1 Brain Research Institute, University of Zurich, Zurich, Switzerland 2 Department of Integrative Neurophysiology, Center for Neurogenomics and Cognitive Research, VU University Amsterdam, Amsterdam, The Netherlands 5 Present Address: Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA 1, 2, 5 , Yasir Gallero-SalaspersName Yasir Gallero-SalaspersName 1 Brain Research Institute, University of Zurich, Zurich, Switzerland 3 Neuroscience Center Zurich, Zurich, Switzerland 1, 3 , Stefano CartapersName Stefano CartapersName 1 Brain Research Institute, University of Zurich, Zurich, Switzerland 1 , João SacramentopersName João SacramentopersName 4 Department of Physiology, University of Bern, Bern, Switzerland 4 , Balazs LaurenczypersName Balazs LaurenczypersName 1 Brain Research Institute, University of Zurich, Zurich, Switzerland 3 Neuroscience Center Zurich, Zurich, Switzerland 1, 3 , Lazar T SumanovskipersName Lazar T SumanovskipersName 1 Brain Research Institute, University of Zurich, Zurich, Switzerland 1 , Christiaan P J de KockpersName Christiaan P J de KockpersName 2 Department of Integrative Neurophysiology, Center for Neurogenomics and Cognitive Research, VU University Amsterdam, Amsterdam, The Netherlands 2 , Fritjof HelmchenpersName Fritjof HelmchenpersName 1 Brain Research Institute, University of Zurich, Zurich, Switzerland 3 Neuroscience Center Zurich, Zurich, Switzerland 1, 3 , Shankar Sachidhanandam Shankar Sachidhanandam 1 Brain Research Institute, University of Zurich, Zurich, Switzerland 4 Department of Physiology, University of Bern, Bern, Switzerland 1, 4, ✉ 1 Brain Research Institute, University of Zurich, Zurich, Switzerland 2 Department of Integrative Neurophysiology, Center for Neurogenomics and Cognitive Research, VU University Amsterdam, Amsterdam, The Netherlands 3 Neuroscience Center Zurich, Zurich, Switzerland 4 Department of Physiology, University of Bern, Bern, Switzerland 5 Present Address: Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA ✉ Corresponding author.
Received 2017 Dec 19; Accepted 2018 Apr 27; Collection date 2018.
© The Author(s) 2018 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: PMC5958066 PMID: 29773806 Abstract Sensory association cortices receive diverse inputs with their role in representing and integrating multi-sensory content remaining unclear. Here we examined the neuronal correlates of an auditory-tactile stimulus sequence in the posterior parietal cortex (PPCterm) using 2-photon calcium imagingterm·mth in awake mice. We find that neuronal subpopulations in layer 2/3term of PPCterm reliably represent texture-touch events, in addition to auditory cuesterm that presage the incoming tactile stimulus. Notably, altering the flow of sensory events through omissionterm of the cued textureterm touch elicited large responses in a subset of neurons hardly responsive to or even inhibited by the tactile stimuli. Hence, PPCterm neurons were able to discriminate not only tactile stimulus features (i.e., textureterm graininessterm·var) but also between the presence and omissionterm of the textureterm stimulus. Whereas some of the neurons responsive to textureterm omissionterm were driven by looming-like auditory sounds others became recruited only with tactile sensory experience. These findings indicate that layer 2/3term neuronal populations in PPCterm potentially encode correlates of expectancyterm in addition to auditory and tactile stimuli.
Introduction Sensory processing carried out by the mammalian neocortex allows for the perception of the external environment. While the primary sensory cortices are efficient at processing their respective sensory modality, higher-order sensory association cortices piece together multiple sensory inputs to generate a more comprehensive representation of the outside world. In rodents, the posterior parietal cortex (PPCterm) is one such association area that lies sandwiched between the primary somatosensory, auditory and visual cortices and has dense reciprocal connections with these regions 1 – 4 . As such, PPCterm is capable of processing somatosensory, visual and auditory information, as well as combined stimuli resulting in multisensory enhancement 5 – 8 . PPCterm is also part of the attention network implicated in top-down stimulus selection, that is distributed across the fronto-parietal-collicular areas 9 . It is reciprocally connected to the orbitofrontal cortex, anterior cingulate cortex and the frontal eye field (FEF) 1 , 10 , cortical areas that are involved in attention.
The capacity of PPCterm to process multimodal stimuli renders it particularly important in coding for peripersonal space, in egocentric and allocentric reference frames 1 , 11 – 13 . In addition to playing a role in spatial navigation, PPCterm is also implicated in sensorimotor decision making in both primates and rodents 14 – 17 and in goal-distance estimation with auditory cuesterm 18 . Recently it has also been shown that PPCterm has a temporal associative role where it encodes the memory of past behavioural choices 19 , as well as choice outcomes and that it can bias action selection based on this history 20 , 21 . Lesion studies show that damage to the PPCterm results in failure to attend and respond to contralesional sensory stimuli. This hemispatial neglect occurs despite intact sensory processing at the primary sensory cortices 22 . It can be alleviated partially by the presentation of auditory cuesterm to direct attention to the contralesional side 23 , effectively bringing the unattended sensory stimulus into conscious perception by activating the fronto-parietal network 24 . While visual and auditory processing in PPCterm during behaviour have been extensively studied in mice 7 , 15 , 17 , 18 , 25 , 26 , we know little about how whisker-based tactile stimuli are represented in the neuronal microcircuits of PPCterm.
Here we exposed naïve awake mice to tactile stimuli in the form of textureterm touches to the intact whisker pad 27 , presaged by auditory cuesterm. We report that layer 2/3term neurons of PPCterm show diverse responses to both the tactile stimuli and the auditory cuesterm that precede their delivery. Moreover, we find that omissionterm of the textureterm stimulus following the auditory cuesterm (in so-called ‘Catch-trialsterm·def’) elicits responses in a subset of neurons that were largely non-responsive to or suppressed by the textureterm stimuli. Some of these neurons became responsive to textureterm omissionterm only after tactile stimulation was coupled with the auditory cuesterm. Others, however, responded to Catch-trialsterm·def in mice that were exposed to the auditory cuesterm without prior texture-touch experience. In these mice the responses apparently arose from the looming-like auditory sounds and were further enhanced to reporting textureterm omissionterm upon coupling with textureterm touch. These findings indicate the presence of potential single-cell neural correlates of expectancyterm in PPCterm.
Results In order to determine how tactile sensory information is represented in the neocortical microcircuit of PPCterm, we presented one of two textured surfaces (sandpaper of grade P100 and P1200, corresponding to rough and smooth surfaces respectively) to the mouse whisker pad 27 . We adopted a motorized linear stage that randomly presented one of the texturesterm to the whisker pad contralateral to the imaging site (Fig. 1A ). We used naïve awake mice pre-implanted with cranial windows and habituated to head restraint in the setup. Before trial initiation and imaging, mice were exposed once to the texturesterm, namely the P100 surface, to set the position of the textureterm presenter. Two auditory tones signalled the start of the trial and when the textureterm was in the final position, respectively. Additional auditory cuesterm were generated by the translation motor as it moved in and out of the stimulus position (Fig. 1A, B ). To determine the influence of these stage-generated auditory cuesterm, we interleaved the textureterm trials with ‘Catchterm’-trials, in which texturesterm were omitted but with all auditory cuesterm present (start and stop tonesterm), including the sounds generated by the stage translation (Fig. 1A ; Methods).
Figure 1.
Stimulation setup and localization of PPCterm in mice. ( A ) Schematic of the texture presentationterm·var set-up that allows for 2-photon imaging of the PPCterm using a cranial window in a head restrained mouse. P100 (rough) and P1200 (smooth) correspond to the graininessterm·var of the sandpaper employed. ( B ) Schematic of the trial structure, where the translation stage presenting the texturesterm slides in and out of position. The loudspeaker symbols indicate the 1-ms auditory tones just before the start and at the end of the stage translation, respectively. The shaded beige rectangles indicate the time windows with auditory cuesterm from the translation stage movement and the green rectangle indicates the time window when the texturesterm are in reach for whisker touches. Cue and touch-stimulus vectors used for the analysis windows are shown as well. ( C ) Location of the γ-whisker barrel in S1 and of V1term, as determined with IOSterm (blue and red overlay, respectively) seen through a 4-mm cranial window. The white rectangle indicates the FOV from Fig. 2A . ( D ) Schematic of known anatomical inputs (red arrows) and outputs (green arrows) to PPCterm. ( E – I ) Fluorescence images of retrogradely labelled cell bodies that project to PPCterm (orange-red) with CTB-Alexa594 (CTB), and anterogradely GFP labelled PPCterm axonal projections (green), along with DAPI-stained cell nuclei (blue) to determine the anatomical connections of the site identified by IOSterm. In ( E ) coronal section of the mouse brain displaying M2term and the dorsal striatum (CP, caudate putamen). Note the concentration of the PPCterm axons in layer 1 of M2term (white arrow). In ( F ) close-up of the region indicated by the white box (upper layer 2/3term) at the top of ( E ) with PPCterm projecting orange-red M2term cell bodies and M2term projecting PPCterm axons in green. In ( G ) close-up of the region indicated by the white box at the bottom of ( E ) with PPCterm axons projecting to the dorsal striatum in green. In ( H ) coronal section displaying the dentate gyrus (DG) and LPterm, with LP-projecting green PPCterm axons. In ( I ) close-up of the region indicated by the white box in ( H ) with PPCterm projecting orange-red LPterm cell bodies intermingled with LP-projecting green PPCterm axons.
In order to localize PPCterm in mice 3 , 6 , we performed intrinsic optical signal (IOSterm) imaging 28 , 29 to visualize the boundaries between the γ-barrel column in the mouse whisker barrel cortexterm and the primary visual cortex V1term (see Methods) (Fig. 1C ). This allowed us to select imaging areas in posterior parietal area A 3 , 30 in the lateral parietal association cortex (LPtAterm) 31 , located in between the somatosensory barrel cortexterm and V1term. In four mice, we carried out retrograde and anterograde labelling experiments with tracer injections targeted to LPtAterm. The resulting labelling patterns validated that this region receives axonal projections from known PPCterm input sites (Fig. 1D ), particularly the thalamic lateral posterior nucleus LPterm (Fig. 1H,I ) and premotor cortex M2term (Fig. 1E-F ), as well as sends its axons to known PPCterm output sites, including M2term, the dorsal striatum (Fig. 1E–G ) and LPterm (Fig. 1H,I ) 15 , 32 . Hence, we identify this region between barrel cortexterm and V1term as part of PPCterm and subsequently used IOSterm imaging to guide our selection of imaging areas accordingly.
We performed in vivo calcium imagingterm·mth of layer 2/3term PPCterm neurons (at 150 to 220 µm depth) through a cranial window using YC-Nano140term, a genetically encoded ratiometric calcium indicator that reliably reports action potential discharge 27 (Fig. 2A ). We acquired imaging data from 13 fields of views (FOVs) in 4 mice, recording calcium signals in a total of 358 neurons. About 35% of neurons (126/358) were active during the trials (see Methods). For each FOV we performed imaging during 100 trials, comprising approximately 40% trials with presentation of each textureterm and 20% Catch-trialsterm·def. PPCterm neurons displayed a diversity of trial-related responses with some neurons responding to the auditory cuesterm, to the textureterm stimuli, or both, while others showed suppressed activity during cue and texture presentationterm·var (Fig. 2A,B ). In particular, the auditory-cue-responsive neurons ramped up their activity as the translation stage approached the mouse. This ramping up of activity resembled loomingterm responses in neurons observed in the superior colliculus 33 and auditory cortex 34 , in response to loomingterm visual and auditory cuesterm, respectively. We confirmed that the auditory cueterm generated by the stage translation was indeed looming-like by performing sound recordings at the location corresponding to the head of the head-fixed mouse (Supplementary Fig. S1 ). Using a general linear model (GLM)term·stat to fit neuronal calcium traces, we classified the trial-active neurons (n = 126) as cue-responsive (22.2%), touch-responsive (11.9%), mixed-responsive (11.9%) and suppressed (17.5%) based on the dynamics of their calcium responses to the trial period; the remaining 36.5% of ‘unclassified’ active neuronsterm·def responded neither to auditory cuesterm nor to texture presentationterm·var (Fig. 2B,C ; see Methods). To our surprise, a subset of these ‘unclassified’ neurons (12 out of 46) as well as a subset of classified neurons (26 out of 80) showed a significant response in the touch-stimulus window during Catch-trialsterm·def, when the textureterm was omitted (see for example average traces of neurons 1, 3 and 5, Fig. 2A ). We call these special neurons ‘Catch-responsive’ neurons (see Methods), which overall made up 30.2% of the active neuronsterm·def (Fig. 2C ). Interestingly, 55.3% of the Catch-responsive neuronsterm·def were non-responsive or even suppressed during texture presentationterm·var trials, suggesting that they were inhibited by the looming-like auditory cuesterm and/or a subsequent textureterm touch, but relieved from inhibition when the textureterm was absent.
Figure 2.
Neural dynamics of sensory representation in the PPCterm. ( A ) Left, 2-photon image of a FOV with neurons expressing the GECI YC-Nano140term. Right, calcium transients from the ROIs selected in the left panel in response to the presented auditory cuesterm and texturesterm, along with the neuropil (NP) signals and the corresponding whisker angles (Wh) in green. The transients represent the very first 5 trials that the mouse has been exposed to. The averaged responses correspond to: textureterm graininessterm·var, P100, n = 57; textureterm graininessterm·var, P1200, n = 26; Catchterm, n = 17 trials. ( B ) Example average calcium transients of neurons classified as cue-responsive, touch-responsive, mixed-responsive, suppressed and unclassified. Traces are averages of 40 to 60 P100 trials and corresponding Catch-trialsterm·def. ( C ) Pie Chart illustrating the percentage-wise classification of trial-active neurons and the corresponding distribution of Catch-responsive neuronsterm·def across the classes (darker slice shade; percentage fractions in brackets). Shaded areas around the mean calcium transients correspond to the standard error of the meanterm·stat (s.e.m.) here and in all subsequent figures.
We went on to further characterize these Catch-trialterm·def responses by examining their neural dynamics during the course of the experiment. To determine if their activity was experience-dependent and gradually developed over time, we compared the mean peak ΔR/R response in the touch-stimulus window between the first 5 and the last 5 Catch-trialsterm·def for Catch-responsive neuronsterm·def. The analysis was restricted to the first session, in which a mouse was confronted with the setup (i.e., we only used data from the first imaging area measured in each mouse). We found that the peak amplitudes of the evoked responses in Catch-responsive neuronsterm·def were comparable at the beginning and end of the imaging session (Fig. 3A,B ; peak ∆ R/R : first 5 trials 9.3 ± 2.4%, last 5 trials 12.1 ± 2.5%; p = 0.75; n = 17 Catch-responsive neuronsterm·def in 4 FOVs, Wilcoxon signed paired testterm·stat). Moreover, the peak calcium transient amplitude in response to the very first Catch-trialterm·def was comparable to the mean peak response for all Catch-trialsterm·def (Fig. 3B ; peak ∆ R/R : first trial 15.2 ± 5.2%, trial mean 10.9 ± 1.9%; p = 0.89, n = 17 Catch-responsive neuronsterm·def in 4 FOVs, Wilcoxon signed paired testterm·stat). This suggests that PPCterm neurons are able to report the omissionterm of the textureterm reliably and stably over time. Furthermore, for all 4 FOVs the first Catch-trialterm·def was presented as the second trial in the sequence of trials, suggesting that the PPCterm is able to rapidly encode the omissionterm of the auditory-cued tactile stimulus.
Figure 3.
Catch-trialterm·def response is robust and stable. ( A ) Example average Catch-trialterm·def responses from the first 5 and last 5 Catch-trialsterm·def. ( B ) Left, mean peak ∆ R/R of the first 5 and last 5 Catch-trialterm·def responses from the first imaging session of each mouse (n = 17 Catch-responsive neuronsterm·def from 4 mice). Right, mean peak ∆ R/R of the very first Catch-trialterm·def and the mean peak Catch-trialterm·def response from all trials of the first imaging session. ( C ) Example single-trial calcium transients of Catch-responsive neuronsterm·def in trials, with the corresponding whisker angles below, and their average traces (n = 24) compared to their corresponding response to P100 trials (n = 54). Trials 2 and 4 display whisking during the touch-stimulus windowterm·var. ( D ) Mean peak ∆ R/R of the Catch-trialterm·def response in trials with whiskingterm and no-whisking. ( E ) Mean Catch-response onset times for the same trials in ( D ). ( F and G ) Mean peak ∆ R/R of Catch-responsive neuronsterm·def to textureterm trials (P100 and P1200) and Catch-trialsterm·def in the auditory cueterm and touch-stimulus windows respectively. Lines represent individual neurons. Filled circles with error bars represent group averages shown as mean ± s.e.m.term·stat Statistical significance is indicated by *** for P < 0.005. No statistically significant differences were found comparing trials in ( B , E and F ).
We next verified whether the Catch-trialterm·def responses might reflect an efferent copy of a motor command such as whiskingterm because the PPCterm has strong reciprocal connections with the motor and premotor cortices 1 – 3 . Catch-trialterm·def responses were observed both in the absence and presence of whiskingterm behaviour (see Methods) during the touch-stimulus window, indicating that they were not solely dependent on whiskingterm (Fig. 3C ). A comparison of the peak of the Catch-trialterm·def responses in the presence and absence of whiskingterm during this time window revealed that the response was enhanced by whiskingterm (Fig. 3D ; peak ∆ R/R : whiskingterm 16.4 ± 1.4%, non-whisking 10.1 ± 1.0%; p = 1.47 × 10 −7 , n = 34 Catch-responsive neuronsterm·def in 12 FOVs, Wilcoxon signed paired testterm·stat; whiskers could not be tracked in one of the 13 FOVs). However, the onset time of the Catch-trialterm·def responses (from trial start) were comparable in the presence and absence of whiskingterm (Fig. 3E ; Catch-response onset time: whiskingterm 3.7 ± 0.1 s, non-whisking 3.8 ± 0.1 s; p = 0.56, n = 34 Catch-responsive neuronsterm·def in 17.3 ± 0.1% and 82.7 ± 0.1% of Catch-trialsterm·def respectively, Wilcoxon signed paired testterm·stat). Whiskingterm onset and the corresponding Catch-trialterm·def onset were also comparable (onset time: whiskingterm 4.1 ± 0.1 s, Catch-response 3.9 ± 0.1 s; p = 0.20, n = 12 FOVs, Wilcoxon signed paired testterm·stat). While Catch-responsive neuronsterm·def showed comparable calcium signals during the auditory-cue window, for both Catch-trialsterm·def and texture presentationterm·var trials (Fig. 3F ; peak ∆ R/R : P100 2.1 ± 0.4%, P1200 2.4 ± 0.3%, Catchterm 3.0 ± 0.6%; p = 0.07 for P100 vs Catchterm, p = 0.72 for P1200 vs Catchterm; n = 34 Catch-responsive neuronsterm·def in 12 FOVs; Wilcoxon signed paired testterm·stat), Catch trialterm·def responses in the touch-stimulus window were larger compared to their corresponding texture-touch responses (Fig. 3G ; peak ∆ R/R : P100 4.5 ± 1.0%, P1200 5.5 ± 0.7%, Catchterm 10.1 ± 1.0%; p = 1.72 × 10 −5 for P100 vs Catchterm, p = 4.66 × 10 −5 for P1200 vs Catchterm; n = 34 Catch-responsive neuronsterm·def in 12 FOVs; Wilcoxon signed paired testterm·stat). To verify whether Catch-trialterm·def responses might have been acoustically driven by the textureterm stop toneterm, we also determined the onset times exclusively in Catch-trialterm·def responsive neurons that were suppressed or non-responsive to texturesterm, avoiding overlap with auditory-cue or texture-touch responses, and analysed only non-whisking trials. Apart from one neuron (onset 120 ms after the textureterm stop toneterm) the others displayed onset times well before the translation stage came to rest (onset 580 ± 65 ms before textureterm stop toneterm, n = 18 neurons). This indicates that the Catch-trialterm·def responses were not due to the stop auditory tone and could potentially represent an expectancyterm response that can be further enhanced by whiskingterm.
We then characterized how the two texturesterm and the absence of texturesterm following the auditory cuesterm are represented in PPCterm. Calcium responses in neurons active during texture presentationterm·var trials were comparable to peak calcium activity in Catch-trialterm·def responsive neurons during the touch-stimulus window (Fig. 4A ; peak ∆ R/R : P100 11.0 ± 2.7%, P1200 10.3 ± 2.2%, Catchterm 9.8 ± 1.0%; p = 0.27 for P100 vs Catchterm and p = 0.13 for P1200 vs Catchterm, n = 31 touch-responsive and mixed neurons, n = 38 Catch-responsive neuronsterm·def; Wilcoxon rank test). We carried out an ROC analysisterm·stat to identify the fraction of neurons that could significantly discriminate between the two presented texturesterm as well as the fraction of neurons that could discriminate between texture-presentation and Catch-trialsterm·def (compared to shuffled controls, Methods; example traces in Fig. 4B ). Among the trial-active PPCterm neurons 7.2% (9/126) discriminated between the graininessterm·var of the presented texturesterm whereas 17.5% (22/126) discriminated between textureterm presence and omissionterm (Fig. 4C ). The subset of Catch-responsive neuronsterm·def discriminative for presence vs. omissionterm of the textureterm discriminated a higher percentage of individual trials compared to texture-type discriminating neurons (Fig. 4D ; percentage of trials discriminated: P100 vs P1200 74.1 ± 1.9%, Catchterm vs P100 81.5 ± 1.8%, Catchterm vs P1200 84.0 ± 1.3%; p = 2.8 × 10 −5 for P100 vs P1200 and Catchterm vs P100 and p = 6.5 × 10 −5 for P100 vs P1200 and Catchterm vs P1200, n = 9 texture-type discriminating neurons, n = 19 neurons discriminating Catchterm vs P100, n = 19 neurons discriminating Catchterm vs P1200; Wilcoxon-Mann-Whitney testterm·stat; see Fig. 4D for their distribution). As a complement to the ideal observer ROC analysisterm·stat, we trained a linear support vector machine (SVM) classifierterm·stat to decode textureterm graininessterm·var and textureterm omissionterm from single-trial population activity, on a frame by frame basis. This simple readout model could effectively decode textureterm graininessterm·var and textureterm omissionterm during the touch-stimulus window with about 55% and 70% accuracy, respectively (Fig. 4E ). The decoding accuracy was improved when the mean of the touch-stimulus window was used (60.1 ± 3.1% for textureterm graininessterm·var; 73.1 ± 4.3% for textureterm omissionterm). Overall, these results suggest that textureterm presence vs omissionterm and to a lesser degree textureterm graininessterm·var are reliably represented in PPCterm in naïve animals.
Figure 4.
Textureterm discrimination and omissionterm detection in the PPCterm. ( A ) Median peak ∆ R/R of textureterm evoked responses (touch-responsive and mixed responsive neurons) and Catch-trialterm·def responses (Catch-responsive neuronsterm·def) respectively. ( B ) Example average calcium transients of neurons that can discriminate texturesterm (top) and those that can report textureterm omissionterm (bottom). Traces are averages of 20 to 50 trials. ( C ) Venn diagram of the distribution of discriminating neurons. ( D ) Performance of neurons discriminating above chance either textureterm type (left) or presence vs. omissionterm of the texture-stimulus (right). ( E ) Population level decoding accuracy of textureterm graininessterm·var and textureterm omissionterm. Horizontal dashed red lines represent chance level. Box plots represent the median, the 25 th and 75 th percentiles in the boxes, whereas the sidebars represent the 5 th and 95 th percentiles of the distribution. Open circles represent individual neurons. Bar graphs with error bars and traces with shaded areas represent group averages shown as mean ± s.e.m.term·stat Statistical significance is indicated by *** for P < 0.005. No statistically significant differences were found comparing trials in ( A ).
While the PPCterm might be capable of rapidly encoding textureterm omissionterm following tactile sensory experience, our findings do not exclude the possibility that Catch-trialterm·def responses are present in PPCterm neurons prior to the presentation of tactile sensory stimuli. In other words, the Catch-trialterm·def response might potentially be generated by the looming-like auditory cueterm produced by the stage. To verify this possibility, we presented an additional 3 naïve mice first with Catch-trialsterm·def only, effectively decoupling the auditory cuesterm from tactile stimuli. We imaged layer 2/3term PPCterm neurons in 2 FOVs per mouse, and imaged 74 trial active neuronsterm·def. The ‘Catch-only’ trials were followed by the presentation of the two texturesterm interleaved with Catch-trialsterm·def as in the earlier experiments, where auditory cuesterm were coupled to textureterm touch (‘Interleaved’ trials). For analysis of this additional data set we excluded trials with obvious whiskingterm behaviour or forepaw and jaw movements, as identified with our behaviour-monitoring camera (see Methods). Indeed, 10 out of the 74 active neuronsterm·def displayed Catch-responses in the Catch-only trials, as identified using the GLMterm (see Methods). Eight of these neurons were also Catch-responsive in the interleaved trials (‘pre-existing’ Catch responsesterm·def) whereas 2 neurons were not classified as Catch-responsive anymore (‘silenced’ responsesterm·def; Fig. 5A,B ). In addition, 8 other neurons were identified as Catch-responsive only after the introduction of tactile stimuli but not in the Catch-only condition (‘emergent’ responsesterm·def, Fig. 5A,B ). Due to the low numbers of Catch-responsive neuronsterm·def detected, however, more data would be required to determine if these three subsets of neurons represent functionally distinct groups. In order to evaluate the influence of tactile sensory coupling on the Catch-trialterm·def response at the single-cell level, we compared the trial-averaged neuronal activity of all neurons in the touch-stimulus window in Catch-only and interleaved trials. The Catch-trialterm·def response showed a significant increase (touch-stimulus peak ∆ R/R : Catch-only trials 3.6 ± 0.7%, interleaved trials 5.2 ± 0.5%; p = 4.8 × 10 −5 , n = 74 neurons, Wilcoxon signed paired testterm·stat) after tactile experience with 14.9% of active neuronsterm·def displaying a significant change in activity within this window (8 cells with an increase, 3 cells with a decrease; significance levelterm·stat set at p < 0.01, Wilcoxon-Mann-Whitney testterm·stat). This was also evident at the population level, where the single-trial population average showed a significant increase in neuronal activity in the touch-stimulus window after tactile experience (Fig. 5C ; touch-stimulus peak ∆ R/R : Catch-only trials 0.95 ± 0.3%, interleaved trials 1.4 ± 0.3%; p = 0.039, n = 74 neurons, Wilcoxon-Mann-Whitney testterm·stat). In a separate set of experiments in these mice, we were also able to evoke Catch-trialterm·def like responses in neurons when presented with the audio playback of the looming-like auditory cueterm but not when presented with a non-looming recording (Supplementary Fig. S2 ). Our results suggest that while looming-like auditory cuesterm alone can evoke responses in the PPCterm, coupling with tactile stimuli enhances this response and further recruits more neurons within PPCterm that can report subsequent texture-touch omissionsterm.
Figure 5.
Encoding of auditory-cued textureterm omissionterm in the PPCterm. ( A ) Example average Catch-trialterm·def responses from pre-existing (brown trace), silenced (red trace) and emergent (blue trace) Catch-responsive neuronsterm·def from the Catch-only and subsequent textureterm and interleaved trials. ( B ) Mean peak ∆ R/R of the Catch-trialterm·def response as observed in the touch-stimulus window of pre-existing (n = 8) and emergent (n = 8) Catch-responsive neuronsterm·def, before (Catch-only) and during (interleaved) tactile sensory experience. The colour codes are identical to ( A ). No mean value is presented for the two suppressed neurons. ( C ) Population Catch-response as observed in the Catch-only trials and after tactile experience in the interleaved trials. Horizontal dashed red and grey lines indicate the mean peak population response in the touch-stimulus window and the baseline, respectively. Lines represent individual neurons. Filled circles with error bars represent group averages shown as mean ± s.e.m.term·stat
Discussion We found that layer 2/3term neuronal populations in the mouse PPCterm display a diversity of responses to an induced textureterm touch, with neurons increasing their activity to the preceding auditory cuesterm and/or tactile stimuli, as well as other neurons being suppressed by these stimuli. Few PPCterm neurons (7.2%) were able to discriminate the two distinct texturesterm presented. This low discrimination power is not surprising as the experimental sessions were only the second time that the naïve mice were exposed to these texturesterm (the first being setting the textureterm stop position, prior to imaging start). No explicit association of a relevant context to the textureterm graininessterm·var as stimulus feature or any reinforcement training had been applied. In the somatosensory barrel cortexterm the proportion of neurons that can discriminate the same textureterm stimuli as applied here in a reward-based go/no-go task is higher in trained compared to naive animals 27 , 35 . More interestingly, we found that about 30% of trial-active PPCterm neurons responded to the omissionterm of the textureterm touch that always arrived after the auditory cuesterm. This was observed in the very first Catch-trialterm·def, suggesting that these neurons either are capable of rapid stimulus binding and association 36 or that they respond directly to the auditory cuesterm generated by the stage translation. We found evidence for both, as Catch-responsive neuronsterm·def were observed in both the initial Catch-only trials prior to textureterm touch as well as emerging later in another set of neurons in the subsequent series of interleaved trials. Coupling tactile stimuli to the auditory cuesterm further enhanced the Catch-trialterm·def response to textureterm omissionsterm. Hence, neurons within PPCterm are capable of representing a change in sensory flow, in a manner that is both rapid and stable. This could be essential for the role of PPCterm in spatial processing, navigation and the allocation of attention, allowing it to rapidly update changes in sensory information in particular to objects that are close to the body.
Our stimulation paradigm incorporated an auditory cueterm that was looming-like, generated by the movement of the translation stage and reported reliably by cue-responsive neurons. Loomingterm stimuli have been shown to be behaviourally salient in humans and primates 37 , 38 , and more recently in mice as well 34 . Loomingterm sounds have been proposed to represent the approach of the sound source in the context of threat detection, either in the form of an attack or collision, and can evoke rapid attention capture and expectancyterm 38 . Indeed the multimodal neurons in the ventral intraparietal area (VIP) in primates have been shown to respond strongly to loomingterm stimuli 39 , 40 . In a more benign context loomingterm sounds can simply indicate the arrival of the source that potentially is associated with a touch event. Hence, in the Catch-only trials (Fig. 5A ) as well as in the audio playbacks of the looming-like auditory cueterm (Supplementary Fig. S2 ) the Catch-trialterm·def responses could potentially represent a neural correlate of expectancyterm, here for a potential touch event. This information is carried over to the period after sensory experience (by pre-existing Catch-responsive neuronsterm·def) and its representation further strengthened upon omissionterm of the cued tactile stimulus (by emerging Catch-responsive neuronsterm·def). It is also plausible that the pre-existing Catch-responsive neuronsterm·def are purely loomingterm sound responsive in the Catch-only trials and start to report textureterm omissionterm only after coupling with textureterm touch. These findings are in line with the recently reported temporal associative role of PPCterm 19 , 20 , which could be due to its long timescale population dynamics allowing the combination of events separated in time 26 .
An alternative interpretation is that Catch-responsive neuronsterm·def are actually acoustically driven and the presence of textureterm touches merely suppresses their response. Indeed, more than half of Catch-responsive neuronsterm·def did not respond to or were suppressed by textureterm touches. However, these Catch-trialterm·def responses had onset times well before the tone signalling stage stop (with the exception of one neuron), suggesting that something else than sound, such as the expectancyterm of an incoming object, evoked the increase in activity. The remaining Catch-responsive neuronsterm·def that also respond to the cue-tone and touch could represent neurons also sensitive to sound-offset 41 . Our results argue against this notion, however, because sensory experience led to changes in the number of Catch-responsive neuronsterm·def in subsets of PPCterm neurons, which should not be the case for a reliable sensory response. Unfortunately, we were unable to remove the looming-like auditory cueterm, as it is an integral part of the texture presentationterm·var apparatus. We were also able to evoke Catch-trialterm·def like responses in neurons when presented with a playback of this auditory looming-like cue, and not with a non-looming recording in mice with previous tactile experience, effectively eliminating other potential unaccounted stimulus sources for the evoked response.
Catch-responsive neuronsterm·def displayed responses that were comparable in amplitude to texture-evoked responses. Furthermore, the omissionterm of textureterm touch could be reliably decoded at the population level. Hence these neurons could function as reporters of mismatchterm signals, which could reliably convey information about unexpected changes in sensory flow to the prefrontal cortex, FEF and LPterm (the mouse equivalent of the pulvinar), brain areas that are implicated in the allocation of attentional resources 9 . Indeed stimulus-omission responses have also been described in the auditory 42 , 43 and visual systems 44 – 46 in the context of mismatchterm negativity (MMN) and oddball responses. However, the stimulus presentation paradigms employed in MMN are different compared to those used in our study. More recently, mismatchterm responses have been reported in mouse V1term 47 , 48 when visual flow is not correlated with self-initiated locomotion. These mismatchterm signals were also present in LPterm inputs to layer 1 of V1term 49 , and mediated by a balance of motor-related excitation and visually driven inhibition during coupled sensorimotor experience 50 , 51 . While our stimulus paradigm does not involve a conflict with a self-initiated stimulus response, such as visual feedback from locomotion, it nonetheless incorporates a behaviourally salient stimulus (the loomingterm auditory cueterm) that when coupled with tactile stimuli, can lead to a rapid encoding of omissionsterm in the coupled sensory flow. LPterm, along with the posterior nucleus Po, is the main source of thalamic input to the PPCterm, which does not receive any direct input from primary sensory thalamus 2 , 52 . Hence, LPterm/Po could be a potential source for the Catch-trialterm·def responses that we have observed in layer 2/3term of the PPCterm. This could be in turn modulated through the activation of local inhibitory circuits by top-down feedback, potentially from the FEF, providing the prediction signal in the proposed framework of predictive codingterm 53 .
Methods Viral and tracer injections Experimental procedures followed the guidelines of the Veterinary Office of Switzerland and were approved by the Cantonal Veterinary Office in Zurich. Intrinsic optical signal (IOSterm) imaging was performed on young adult (P35-42) male wild-type C57/BL6 mice under light isofluorane anaesthesia to identify the location of PPCterm by exclusion. In brief, to avoid the activation of surrounding whiskers, all whiskers except the γ-barrel-column whisker on the right whisker pad of the mice were trimmed and their locations were mapped using IOSterm on the exposed skull under 625 nm illumination during whisker stimulation (rostrocaudal deflections at 10 Hz). The location of primary visual cortex (V1term) was similarly mapped using full-field stimulation with a green LED placed 5 mm in front of the contralateral eye. Subsequent viral or tracer injections were targeted to the non-activated region between these two identified sites, within 500-μm from the border of the IOSterm corresponding to the γ-barrel-column location (light blue overlay in Fig. 1C ). To verify that we are able to localize PPCterm by IOSterm exclusion as described above, we went on to verify the established anatomical input and output connections to PPCterm (n = 4 mice). To identify the sites that send their axons to PPCterm, the retrograde tracer CTB-Alexa594 (Molecular Probes, Invitrogen; 200 nl, 1% wt/vol) was injected into PPCterm ~300 μm and 500 μm below the pial surface targeting layer 2/3term and layer 5, respectively. A similar injection with AAV8-hSyn-Jaws-GFP-ER2 (200 nl, UPenn) was performed in PPCterm to follow anterogradely labelled axons and identify output regions that PPCterm projects to. Note that both injections were performed at the same injection site but spaced by at least 2 weeks (AAV8 followed by CTB), to allow for sufficient expression and uptake respectively. For calcium imagingterm·mth, AAV1- EF1α-YC-Nano14 0 (300 nl, ~1 × 10 9 vg/μl) was injected into the PPCterm (n = 7 mice), targeting layer 2/3term (~300 μm below the pial surface) to induce expression of the ratio-metric genetically encoded calcium indicator yellow cameleon Nano-140 (YC-Nano140term) 54 .
Verification of PPCterm location Following 5–7 days after CTB-Alexa594 injection to allow for uptake 55 , and 3 weeks in the case for virus expression of GFP, mice were anesthetized (ketamine/xylazine; 100/20 mg/kg body weight) and perfused transcardially with 4% paraformaldehyde in phosphate buffer, pH 7.4. Coronal cortical sections (50-µm thickness) were cut using a vibratome (VT100; Leica). Slices were stained with 2.5 μM DAPI for 10 min and mounted on glass slides. Images were acquired with a confocal microscope (Fluoview 1000, 20× and 10× objectives; Olympus) with green (GFP), red (CTB-Alexa594), and blue (DAPI) excitation/emission filters.
Cranial window implantation and habituation In order to carry out long-term in vivo calcium imagingterm·mth, a cranial window was implanted 24 hrs after virus injections over PPCterm as described previously 56 . Briefly, a craniotomy was performed at the injection site. A cover glass (4 mm diameter) was placed directly over the exposed dura mater and sealed to the skull with dental acrylic. A metal post was fixed to the skull with dental acrylic, contralateral to the cranial window, to allow for subsequent head fixation. One week following chronic window implantation, mice were handled daily for 1 week, and gradually habituated to head fixation. This also gave time for the trimmed whiskers to re-grow. Mice were subsequently used for texture presentationterm·var only if their whisker pads were fully intact.
Texture presentationterm·var Behaviour experiments were performed using a data acquisition interface (USB-6008; National Instruments) and custom-written LabVIEW software (National Instruments) to control devices required for texture presentationterm·var. Commercial-grade sandpaper (3 M) was used, with a rough sandpaper (P100) and smooth sandpaper (P1200) as stimuli. Sandpapers were mounted onto panels attached to a stepper motor (T-NM17A04; Zaber) mounted onto a motorized linear stage (T-LSM100A; Zaber) to move texturesterm in and out of reach of whiskers of head-restrained mice. The texturesterm were presented in a randomized order, with no more than 3 consecutive presentations of the same textureterm. Trials without texture presentationterm·var (“Catchterm”) but with linear stage translation were interleaved (Fig. 1A ). In a separate set of experiments, Catch-trialsterm·def were presented first, followed by interleaved textureterm and Catch-trialsterm·def (Fig. 5 ). A trial consisted of a 2-s pre-stimulus period, a 2-s translation period of the linear stage to bring the textureterm in contact with the whiskers where it remained for 1-s before retraction. Two 2093-Hz auditory cueterm tones (1 ms) signalled the start and stop of the stage translation, respectively (Fig. 1B ). Stage translation itself generated a sound, and served as an additional auditory loomingterm cue that indicated the arrival of the incoming textureterm. No sounds were presented during the 1-s when the texturesterm were stationary, in contact with the whiskers, and correspondingly in the Catch-trialsterm·def as well. All experiments were conducted in the dark, apart from illumination of the whisker field with an infrared LED (see below). Mice were not required to perform any task during the trials and were not rewarded during texture presentationterm·var.
Sound recordings and presentation Sound recordings were performed with a microphone (Audio Teck PRO42) placed in the position where the mouse’s head would have been, as well as next to the fixed stage motor that remained stationary while the stage translated, to record the auditory cuesterm generated by stage translation, and amplified with an audio amplifier (SM Pro Audio Q Pre). Note that the 1 ms auditory cueterm tunes (see above) were not played and recorded here. These audio recordings (loomingterm and non-looming respectively) were then played back via loud speakers from a PC, at 55 dB for both recordings, in the absence of the auditory cueterm tones.
Two-photon calcium imagingterm·mth We used a custom-built two-photon microscope controlled by HelioScan 57 , equipped with a Ti:sapphire laser system (~100-fs laser pulses; Mai Tai HP; Newport Spectra Physics), a water-immersion objective (16 × LWDPF, 0.8 NA; Nikon), galvanometric scan mirrors (model 6210; Cambridge Technology), and a Pockel’s Cell (Conoptics) for laser intensity modulation. For calcium imagingterm·mth, we chose the ratio-metric GECI YC-Nano140term as it allows for the correction of movement artefacts in the Z plane typically observed in awake recordings, which can otherwise give rise to the detection of false positive active cells from out of focus planes. Furthermore, YC-Nano140term has a near linear relationship between the amplitude of the calcium transient and the number of action potentials, as verified by single-cell electrophysiology 27 , unlike the GCaMPs. YC-Nano140term was excited at 840 nm and fluorescence collected with blue (480/60 nm) and yellow (542/50 nm) emission filters for CFP and YFP fluorescence detection, respectively. Images were acquired at 15.6 Hz with 128 × 64-pixel resolution. This gave rise to a relatively small field of view, but acquired with a high frame rate to enhance the reliability of detecting neuronal firing. Single trials of 9 to 10-s duration were recorded at a time with 1-s breaks in between trials to allow for hard disk storage during inter-trial interval periods.
Whisker tracking and behaviour monitoring The whisker field was illuminated with 940-nm infrared LED light and movies were acquired at 200 Hz (500 × 500 pixels) using a high-speed CMOS camera (A504k; Basler). The time course of the whisker angle across all imaged whiskers was measured using automated whisker tracking software 58 . Whisking amplitudeterm·def, defined as the angle between the whisker shaft and sagittal plane from nose tip to the back of the head, was used as a measure to represent both rhythmic and non-rhythmic forms of whiskingterm behaviour. We used custom written MATLAB (Mathworks) scripts to extract whiskingterm episodes by first applying a fourier transform on traced theta to identify whiskingterm events with frequences between 4 Hz and 20 Hz. Events with power greater than a predefined threshold were used to segregate significant whiskingterm episodes. Time periods from these whiskingterm episodes was used to define trial events with or without whiskingterm behaviour during the stimulus window. For all trials, the first and last time point for whisker-to-texture contact was quantified manually through visual inspection. In Fig. 5 , the presented data comes from trials free of whiskingterm behaviour, forepaw and jaw movements, as identified visually from the movies that had large field of views (FOVs) covering the entire head and frontal torso of the mouse.
Calcium imagingterm·mth analysis The CFP and YFP fluorescence channel data were imported into MATLAB for processing. First, background was subtracted on each channel (bottom 1 st percentile fluorescence signal across entire video). Motion correction to both channels was performed by using image registration with TurboReg 59 . Regions of interests (ROIs) corresponding to individual neurons were manually selected from the average intensity projections of a single-trial time series using ImageJ (National Institute of Health) and custom scripts in MATLAB. Mean pixel value for each ROI was extracted for both channels. Calcium signals were expressed as relative YFP/CFP ratio change ΔR/R = ( R − R 0 )/ R 0 . R 0 was calculated for each trial as the bottom 12 th percentile of the ratio for the trial. Active neuronsterm·def were identified by two-way ANOVA with repeated measuresterm·stat of the neuronal calcium signal against the neuropil signal, defined by all pixels not assigned to neuronal soma of the overall ROI annotation (significance value, P < 0.01). Additionally, they required mean peak responses larger than 4.5%, corresponding approximately to 2 action potentials 27 . We present peak ΔR /R within certain time windows as an approximate measure of neuronal activity during this period.
GLM-based neuronal response classification A general linear model (GLM)term·stat was used to fit the neuronal calcium transients and classify neurons based on their responses 60 , 61 . The model consisted of four regressors corresponding to the auditory cueterm ( X ac ), texture-stimulation ( X ts ), textureterm omissionterm in Catch-trialsterm·def ( X ch ), and the post-stimulus phase during textureterm retraction ( X ps ), respectively, along with their corresponding regression coefficients β x . We did not distinguish the two types of texturesterm to fit the GLMterm. Each regressor was represented as a vector with the same length as the calcium traces, containing a value of 1 in the frames spanning the respective temporal window and zero-entries otherwise. The textureterm regressor was zero for the Catchterm trials and, vice versa, the Catchterm regressor was zero for texture-stimulation trials. To take into account temporal filtering caused by calcium indicator kinetics, each regressor was convolved with a kernel H ( k ) describing the single-action potential evoked calcium transient of YC-Nano140term as a function of frame number k with signal amplitude A (4.5% ΔR/R), frame rate fr (15.6 Hz), onset time constant τ on (0.19 s), and decay constant τ decay (0.68 s) as parameters 27 : H ( k ) = A ( 1 − e − ( k − 1 ) / ( f r ⋅ τ o n ) ) e − ( k − 1 ) / ( f r ⋅ τ d e c a y ) 1 For each cell, ΔR/R traces were concatenated and then fit with the fitglm function in MATLAB using the following equation: Δ R / R = β 0 + β a c X a c + β t s X t s + β c h X c h + β p s X p s 2 A regression coefficient was considered significantly deviating from baseline β 0 when the corresponding p-value was below a significance levelterm·stat of α = 0.01. Neurons were considered ‘cue-responsive’ if β a c and β p s were both significantly larger than baseline. Similarly, neurons were classified as ‘texture-responsive’ when β t s was significantly increased. Furthermore, neurons with both β a c and β t s significantly greater than β 0 were classified as ‘mixed-responsive’. Neurons with regression coefficients β a c and β t s significant and smaller than β 0 were classified as ‘suppressed’ neurons. The remaining neurons were classed as ‘unclassified’. Of all these neurons, those with β c h significant and greater than β 0 were additionally classified as ‘Catch-responsive’. In order to avoid erroneously classifying cue-responsive neurons with large transients as Catch-responsive, neurons were excluded from being considered ‘Catch-responsive’ if the average calcium transient fell by more than 50% of the peak response within the cue window once the stage stopped moving and failed to show a rise during the touch-stimulus window. Based on this classification scheme, active neuronsterm·def were placed into 1 of 5 classes (cue-responsive, touch-responsive, mixed-responsive, suppressed and unclassified) and could be additionally classified as Catch-responsive neuronsterm·def. We obtained an average R 2 value of 0.10 ± 0.01 for all the classified neurons. The GLMterm was fitted with 100 trials (approximately 80 textureterm trials and 20 Catch-trialsterm·def) per neuron for Fig. 2 and approximately 20 Catch-only and 100 interleaved trials (approximately 80 textureterm trials and 20 Catch-trialsterm·def) in Fig. 5 . With α set to 0.01 the GLMterm potentially misclassifies 1 neuron each, out of the 126 and 74 trial-active neurons in Figs 2 and 5 respectively. In the case of Catch-responsive neuronsterm·def, this would correspond to a false positive probability of 3.3% (1 in 38 neurons, Fig. 2 ), and 10% to 6.3% (1 in 10 and 1 in 16 Catch-responsive neuronsterm·def before and after sensory coupling, Fig. 5 ). Similarly, for the classification of cell response types, this would correspond to a false positive probability of 6.7% to 2.2% (1 out of 15 for touch/mixed responsive neurons and 1 out of 46 for unclassified neurons).
Single-neuron discriminability analysis To quantify single-neuron discrimination performance we resorted to a receiver operating characteristic (ROC) analysisterm·stat. This allowed us to quantify the performance achievable by an ideal observer with access only to the mean calcium transients registered during the time window of interest (e.g., the touch-stimulus period). Briefly, for a given binary discrimination task (e.g., textureterm P100 vs. P1200), an ROCterm curve was generated by varying a decision threshold over the entire range of possible values. Discrimination performance was then summarized as the obtained area under the curve (AUC)term·stat. Statistical significance was assessed by a Monte Carlo permutation testterm·stat ( N = 10000 random permutations), where true trial type labels were shuffled. Neurons whose AUC values fell above the 99.5 th percentile or below the 0.05 th percentile of the shuffled distribution were deemed as discriminating above chance (p < 0.01).
Trial-by-trial population classifiers We trained linear support vector machine (SVM) classifiersterm·stat to decode textureterm graininessterm·var (P100 vs. P1200) and textureterm omissionterm from single trial population activity. A linear SVMterm computes a simple thresholded weighted sum of inputs, an operation that is plausibly implementable by actual biological neurons; high linear classifier performance has been suggested as an indicator that the neural representation is such that it can be easily readout by downstream neurons 62 . Independent classifiers were trained for each time bin on the calcium transients of every neuron recorded on a given session (approximately 25 neurons). To avoid overfitting, model parameters and were determined on a training set, consisting of 50% of the recorded trials. Classification performance was assessed by measuring the accuracy on the remaining 50% of the trials that resulted from the initial split of the dataset and were held aside during training. To deal with class imbalance (e.g., when learning to discriminate textureterm omissionterm trials, which are less numerous) we randomly excluded trials from the class with the largest number of examples. Performance figures were averaged across all FOVs and mice and reported as mean ± s.e.m.term·stat This approach provides a lower bound on the actual information carried by the population, and does not preclude that more advanced methods cannot extract additional information. Performance did not qualitatively improve when using a more complex radial-basis function kernel (data not shown). We used a reference SVMterm implementation provided by the scikit-learn (version 0.18.1) python package 63 .
Statistical analysis All data are presented as mean ± s.e.m.term·stat unless stated otherwise. Non-parametric tests were performed for all analyses, and the Wilcoxon signed-rank paired testterm·stat and the Wilcoxon-Mann-Whitney testterm·stat for paired and unpaired group comparisons respectively and the Wilcoxon signed-rank testterm·stat for single distributions.
Data Availability All relevant datasets are available from the corresponding author upon reasonable request.
Electronic supplementary material Supplementary Information (853.6KB, docx) Acknowledgements We thank A. Gilad for help with analysis. We thank R. Hahnloser and D. Rodrigues for help with the sound recordings. We thank S. Ciocchi, M. Acuna and F. Kasanetz for useful comments on earlier versions of the manuscript. This work was supported by the Swiss National Science Foundation (SNSF; grant 31003A_149858 to F.H.) and the University of Bern. H.M. was supported by the Boehringer Ingelheim Travel Grant and by an Erasmus Mundus Joint Doctorate grant (ENCNetwork c3-p3).
Author Contributions H.M., S.C. and S.S. performed imaging experiments; Y.G. performed sound recordings and analysis; L.S. performed histology; H.M., J.S. and S.S. analysed the data; B.L. developed analytical tools; C.P.J.K. and F.H. provided materials; F.H. and S.S. designed the experiments; S.S. wrote the paper with contributions from all authors.
Competing Interests The authors declare no competing interests.
Footnotes Electronic supplementary material Supplementary information accompanies this paper at 10.1038/s41598-018-25891-x.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.