Local homogeneity of tonotopic organization in theprimary auditory cortex of marmosetsHuan-huan Zenga,b, Jun-feng Huanga,b, Ming Chena, Yun-qing Wena, Zhi-ming Shena,1, and Mu-ming Pooa,1

aInstitute of Neuroscience, State Key Laboratory of Neuroscience, Key Laboratory of Primate Neurobiology, Center for Excellence in Brain Science andIntelligence Technology, Chinese Academy of Sciences, Shanghai 200031, People’s Republic of China; and bUniversity of Chinese Academy of Sciences,Beijing 100049, People’s Republic of China

Contributed by Mu-ming Poo, December 14, 2018 (sent for review October 3, 2018; reviewed by Mark Hübener and Li Zhang)

Marmoset has emerged as a useful nonhuman primate species forstudying brain structure and function. Previous studies on themouse primary auditory cortex (A1) showed that neurons withpreferential frequency-tuning responses are mixed within localcortical regions, despite a large-scale tonotopic organization. Herewe found that frequency-tuning properties of marmoset A1neurons are highly uniform within local cortical regions. We firstdefined the tonotopic map of A1 using intrinsic optical imagingand then used in vivo two-photon calcium imaging of large neuronalpopulations to examine the tonotopic preference at the single-celllevel. We found that tuning preferences of layer 2/3 neurons werehighly homogeneous over hundreds of micrometers in both horizon-tal and vertical directions. Thus, marmoset A1 neurons are distributedin a tonotopic manner at both macro- and microscopic levels. Suchorganization is likely to be important for the organization of auditorycircuits in the primate brain.

marmoset | primary auditory cortex | tonotopic map | calcium imaging |homogeneity

In the auditory system, the most prominent topographic featureis tonotopic organization, in which adjacent cortical regions

showed preferential responses to pure tones of nearby frequen-cies. In many species, in vivo electrophysiological recordings andimaging techniques have characterized the global tonotopic or-ganization of the auditory cortex, revealing its division intoseparate fields and distinct tonotopic maps within each field (1–9). Although large-scale imaging and recording methods oftenyield global tonotopic maps, recent studies using two-photoncalcium imaging to monitor the activity of individual neuronsin mice showed that local populations of A1 neurons were highlyheterogeneous in their frequency-tuning properties, althoughmacroscopic imaging using intrinsic optical imaging over theentire auditory cortex showed an overall tonotopic organization(10–12). Thus, the macroscopic tonotopic map may reflect anaveraged frequency preference for large populations of neuronswith heterogeneous tuning properties (13–15).Marmoset is a species of New World monkeys known to be

highly vocal and social (16). It has a neocortex much closer tohumans than the commonly used rodent models. The spatialdistribution of cortical neurons with different frequency prefer-ences is important for the organization of neural circuits thatprocesses auditory signals, such as natural sounds comprising acomplex mixture of frequencies. Thus, it is important to de-termine whether the local heterogeneity in neuronal frequencytuning found for mouse A1 neurons is a general property ofmammalian auditory cortices, or alternatively, a property moreunique to rodent brains. To address this issue, we used in vivotwo-photon imaging to monitor pure tone-evoked responses of afew hundred A1 neurons simultaneously in anesthetized com-mon marmosets (Callithrix jacchus). We found that A1 neuronswithin distances of a few hundred micrometers in both horizontaland vertical directions were highly homogeneous in theirfrequency-tuning properties. We also showed that this tonotopicorganization in marmoset A1 is distinctly different from thatfound in rat A1 by the same imaging method. Such microarchitecture

of the auditory cortex may be important for efficient coding of nat-ural sounds in highly vocal animals such as marmosets.

ResultsIn Vivo Two-Photon Calcium Imaging in Marmoset A1. To facilitateidentification of various areas of marmoset auditory cortices, weperformed intrinsic optical imaging (17) in anesthetized mar-mosets (Materials and Methods). We observed that pure tonestimuli could evoke intrinsic optical signals in three primaryauditory regions, previously defined by electrophysiological re-cording and anatomical studies (18, 19) as the primary (A1),rostral field (R), and rostrotemporal field (RT) of the auditorycortex (Fig. 1 A–D and SI Appendix, Fig. S1). For A1 lying on theventral bank of lateral sulcus (LS), we found that gradual in-crements of sound stimuli from low- to high-frequencies evokedresponses in adjacent areas along LS from the the anteroventralregion to the posterodorsal region of A1 (Fig. 1 A–C). Theoverall tonotopic map is shown for one example imaging plane inFig. 1D, with different colors indicating frequency preferences. Asimilar map was observed in two other marmosets (SI Appendix,Fig. S1). This finding is consistent with previous results obtainedby electrophysiological and optical imaging approaches (7, 8).Following obtaining the overall frequency-preference map of

A1 with imaging of intrinsic optical signals, we further performedin vivo two-photon calcium imaging at selected local A1 regionsby bulk loading with the fluorescent calcium indicator Cal-520AM (20). Loading of the indicator reached an apparent plateauat ∼60 min after local injection, when hundreds of fluorescentcells could be detected at imaging depths that cover the majorpart of the cortical layers 2/3. Large-scale imaging over an area

Significance

Marmoset is a NewWorld monkey rich in social interaction andvocal communication. We used in vivo two-photon fluores-cence calcium imaging of sound-evoked responses of largeneuronal populations at single-neuron resolution in the pri-mary auditory cortex (A1) of anesthetized marmosets. Wefound that the pure tone-evoked responses of marmoset A1neurons are highly homogeneous in their frequency preferencewithin local cortical regions, in sharp contrast to that found inrodents. Thus, there is species-specific local tonotopic organi-zation in A1, which imposes distinct neural circuitry constraintsfor the cortical integration of auditory information.

Author contributions: H.-h.Z., Y.-q.W., Z.-m.S., and M.-m.P. designed research; H.-h.Z. andJ.-f.H. performed research; M.C. and Z.-m.S. contributed new reagents/analytic tools;H.-h.Z. analyzed data; and H.-h.Z. and M.-m.P. wrote the paper.

Reviewers: M.H., Max Planck Institute of Neurobiology; and L.Z., University of SouthernCalifornia.

The authors declare no conflict of interest.

Published under the PNAS license.1To whom correspondence may be addressed. Email: zmshen@ion.ac.cn or mpoo@ion.ac.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1816653116/-/DCSupplemental.

Published online February 4, 2019.

www.pnas.org/cgi/doi/10.1073/pnas.1816653116 PNAS | February 19, 2019 | vol. 116 | no. 8 | 3239–3244

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Apr

il 19

, 202

1

of about 1 mm × 1 mm with a 16× objective lens allowed us tomeasure global frequency tuning of A1 subregions, using amethod similar to those reported previously for studying orien-tation and spatial frequency tuning in the primary visual cortex(21, 22). By measuring the fluorescence changes (ΔF) of allpixels within the imaging plane in response to pure tones at sixdiscrete frequencies from 2.8 to 6.7 kHz, we found that activatedregions were organized tonotopically (Fig. 1 E–G). Thus, large-scale imaging using both intrinsic and Ca2+ signals revealedsimilar tonotopic organization of marmoset A1, confirmingprevious electrophysiological findings.

Clustered Distribution of A1 Neurons with Similar Frequency Tuning.Two-photon imaging of Ca2+ fluorescence signals at a higherresolution (with 40× objective, ∼0.6 μm/pixel) allowed us tomonitor pure tone-evoked Cal-520 AM fluorescence changes inindividual A1 cells within a given focal plane (Materials andMethods). We found that most cells exhibited maximal responsesat a specific frequency [defined as the best frequency (BF)]. Thepercentage of responsive cells among all fluorescently loadedcells was 48 ± 6% (SE, n = 5 fields), and cells with similar BFswere found to localize together within the imaged field. In theexample imaging field shown in Fig. 2A, 10 cells sampled within adistance of about 180 μm all showed the BF at 8 kHz. Whenfluorescence signals from all 68 responsive cells were measuredover the entire imaged field, for pure-tone stimuli over discreteincremental frequencies from 0.5 to 32 kHz, we found most cellshad BFs centered around 8 kHz (Fig. 2B). This frequency pref-erence was also shown by the average ΔF/F tuning profile for all68 cells (Fig. 2B, Top Inset). This homogeneity of BFs could alsobe visualized by the tonotopic map with BFs coded by discretecolors (Fig. 2C). As summarized by the histograms for the per-centages of cells exhibiting different BFs, we observed such localuniformity of BFs for five different imaging planes (ranging from180 to 520 μm in width) recorded from three marmosets over thefrequency range from 1 to 9.5 kHz (Fig. 2D). Such homogeneityof tonotopic properties within each imaged area is in sharpcontrast to that found in mice using Ca2+ imaging methods (10,11), where cells within 50- to 100-μm distances showed muchlarger variability in BFs (up to four octaves).We also quantified this clustering of cells with the same BFs by

calculating the nearest-neighbor distance, defined as the distance

of the nearest cell showing the same BF, for all responsive cellsin the imaging plane (Fig. 2E). The cumulative percentage plotof the distribution of nearest-neighbor distances of all five im-aged planes showed steep slopes at small distances, with mediandistance (50%) of 30 ± 7 μm (SE, n = 5). These distributionswere significantly different from the random uniform distribu-tion of nearest-neighbor distances (P < 0.001, Kolmogorov–Smirnov test), consistent with the homogeneity of BFs within theimaged plane.In the experiments above, we have shown a tonotopic orga-

nization of A1 at the millimeter scale using imaging of intrinsicoptical signals and local homogeneous distribution of cells withthe same BFs. To further confirm that the local homogeneous

8 KHz6 KHzC

7 KHz

A1LS

D

VA P

RRT

A

g

4 KHz 4.8 KHz

Vessel mapB

Tonotopic map

A1LS

RT

D

VA P

45678

KHz

D

E3.4 KHzF

5.7KHz

0

max

G0 max

2.8 3.4 4 4.8 5.7 6.7

Fig. 1. Large-scale imaging of neuronal responses in the marmoset auditorycortex. (A) Schematic diagram showing various regions of the marmosetauditory cortex. A1, primary auditory cortex; LS, lateral sulcus; R, rostralfield; RT, rostrotemporal field. (B) Surface blood vessel map within the im-aging window. (C) Intrinsic optical signals evoked by pure-tone stimuli at 6,7, and 8 kHz, respectively. Dark areas show activated regions. (D) Summaryof tonotopic maps (in the left hemisphere) revealed by intrinsic optical imagingusing pure-tone stimuli, color-coded for five discrete frequencies (4–8 kHz).Note the gradual shift of the responses along the LS. (E) Imaging planeobtained using a 16× objective (with 2-μm pixel−1). (F) Pure-tone responses(ΔF) for four different frequencies. Each image represents the average offive repeats. (G) Composite frequency-preference map for the imaging planeas in E. Colors indicate the preferred frequencies, with the brightness in-dicating the magnitude of the frequency selectivity. E–G were obtained withtwo-photon imaging. (Scale bars: 1 mm in B–D; 200 μm in E–G.) EC D

Per

cent

age

of c

ells

31.

2.8.5 2416 32

21 168Best frequency (KHz)

6 8 12

marmoset 1

0100

marmoset 1

4

marmoset 2

marmoset 3

marmoset 3

6 7 8 10 12 Other0 100 200 300

0

100

Per

cent

age

of c

ells

A

13

2

785

94

6

10

B

5 7 104 6 12 16 20 24 32281483.5 1 2

10203040

6050

Cel

l #

Frequency (KHz)

0.5

0

LS

D

VAP

Day 1

Day 10

HF G

1 2 4 8 160Best frequency (KHz)

ROI 1 ROI 2

Per

cent

age

of c

ells

50

1

16

KHz

ROI 2

ROI 1

123456789

1087 10

Frequency (KHz)

Fig. 2. Marmoset A1 cells with similar frequency preference were highlyclustered within local cortical regions. (A, Left) In vivo two-photon image ofCa2+ fluorescence at a focal plane, at ∼150 μm below the pia surface. (Scalebar: 50 μm.) (A, Right) Single trials (gray lines, n = 8) and mean (red line)fluorescence changes (ΔF/F) evoked by pure-tone stimuli (7–10 kHz) at 10sampled cells (marked by red circles on the Left). (Scale bar: 0.2 s and 20%ΔF/F.) (B) ΔF/F with time for 68 pure tone-selective cells during sequentialstepwise application of discrete pure tones from 0.5 to 32 kHz (black bar,tone duration of 0.2 s), with the magnitude of ΔF/F color coded by thecontinuous scale shown on the Right. Red trace on Top shows the averagedΔF/F for all 68 cells. Note that the BF for the vast majority of cells was locatedat 8 kHz. (Scale bar: 0.5 s and 20% ΔF/F.) (C) Spatial distribution of all 68 cellsshown in A, with their BFs (kHz) color coded by the scheme shown on theBottom. (Scale bar: 50 μm.) (D) The percentage of cells with different BFsobserved in five imaged planes from three marmosets. Top corresponds tothe example imaging plane shown in A–C. (E) Cumulative percentage plotfor the distribution of nearest-neighbor distances for all cells in each of fivedifferent imaging planes as that in D (same color coding; black indicates datafor 68 cells shown in A–C and colored indicates data collected from fourother imaging planes observed in three different marmosets). Gray diagonaldashed line depicts theoretically uniform distribution. All curves were sig-nificantly different from the diagonal line (P < 0.001 for all curves, Kolmo-gorov–Smirnov test). (F) Fluorescence images of two adjacent regions inmarmoset A1 after Cal-520 loading of the same marmoset in two experi-ments performed 9 d apart. (Scale bar: 1 mm.) (G) Distribution of cells withdifferent BFs, in the two regions of interest (ROIs) boxed in F. (Scale bar:100 μm.) (H) Histograms of BF distributions for all tone-selective cellsrecorded in the two ROIs. (n = 39 in ROI 1; n = 57 in ROI 2).

3240 | www.pnas.org/cgi/doi/10.1073/pnas.1816653116 Zeng et al.

Dow

nloa

ded

by g

uest

on

Apr

il 19

, 202

1

tonotopy occurred at different A1 regions in the same marmoset,we performed measurements on two adjacent A1 areas whichwere separated by about 500 μm (Fig. 2F), in two separate ex-periments 9 d apart. We observed a clear difference in thedominant BFs of 6.7 and 9.5 kHz in the two regions, respectively,as shown by the BF distribution histograms (Fig. 2H).

Tonotopic Homogeneity in A1 Is Sound-Intensity Invariant. We nextexamined whether the BF of the same A1 cell depends on theintensity of the sound stimuli. Using pure tones of three soundintensities (60, 70, and 80 dB), we found that many A1 cellsexhibited higher responses as the sound intensity was increased,while others showed similar responses at all three intensities, anda few showed reduced responses at higher intensities (Fig. 3 Eand F). Despite the variation in the dependence on sound in-tensity, the vast majority of A1 cells within the same imagingplane were narrowly tuned to the same frequency at all threesound levels, as shown by the cell-based tonotopic maps (Fig. 3A,BFs around 8 kHz). As shown for the example cell in Fig. 3B,similar averaged response profiles and BF distributions wereobserved for the responses at three different sound intensities(all peaked at 8 kHz, Fig. 3C). This is largely consistent with theprevious finding using electrophysiological recording that sound-evoked responses of awake marmoset A1 neurons are sound-level invariant (23). Further analysis indicates that the distribu-tions of nearest-neighbor distances of the same BFs were allsignificantly different from the distribution expected for uniformdistribution (Fig. 3D, P < 0.001, Kolmogorov–Smirnov test).Previous single-unit recordings from A1 of several species

have suggested a patchy organization for intensity tuning alongthe isofrequency axis (24–26). Thus, we have examined the ex-istence of clustering of cells based on the preferred intensity.Although some cells were best activated at different intensities,the numbers of cells activated were similar, with 68, 51, and 54(out of 109) activated at three intensities tested (Fig. 3F), re-spectively. We noted that most cells could be activated by allthree intensities (Fig. 3G; 40 white dots) and a small fraction ofcells showed intensity selectivity (Fig. 3G; 14 red dots, 5 greendots, 2 blue dots). However, there was neither apparent clus-tering of cells that respond selectively to one particular in-tensity nor an apparent gradient of best intensity across theisofrequency axis.

Vertical Organization of A1 Tonotopic Maps at Superficial CorticalLayers. A general organization principle in many sensory corti-ces is that neurons for processing similar functional features areorganized into vertical columns (27, 28). We thus further ex-amine whether tonotopic maps are also homogeneous along thevertical axis of A1. Due to the technical limitation of our two-photon Ca2+ imaging method for the marmoset, we were onlyable to address this issue by exploring the tonotopic properties ofA1 neurons with a depth up to 370 μm from the pial surface,covering the major part of layer 2/3. Fig. 4A shows the spatialdistribution of A1 cells, color coded with their BFs in the ex-ample experiment shown in Fig. 2. Different groups of cells atfive different focal depths (140, 170, 200, 230, and 260 μm) wereall found to exhibit BFs predominately at 8 or 10 kHz, with clearclustering of cells of similar BFs.A composite 3D plot (cell pairs sharing the same BFs from

adjacent imaging planes with a distance <50 μm were connectedby lines) of the cell distribution showed that clusters of cells withthe same BFs are well aligned vertically among imaged planes,although there appeared to be more cells with BFs deviated from10 kHz at deeper cortical regions (Fig. 4B). In a separate ex-periment on a different marmoset, we imaged five planes overthe cortical depths between 250 and 370 μm; similar clusteringand vertical alignment of cells of the same BFs were also ob-served. These findings support the existence of a columnarstructure of A1 tonotopic maps at the microscopic level.

Local Tonotopic Organization Is More Heterogeneous in Rat A1. Theabove studies showed that the tonotopic organization in themarmoset A1 is highly homogeneous. Given the previous find-ings showing marked local heterogeneity in the frequency tuningof mouse A1 neurons, we further examine the tonotopic orga-nization of rat A1 neurons using the same Cal-520 imagingmethod as that used in the marmoset studies above. We foundthat, for anesthetized Sprague-Dawley rats (∼300 g), A1 cellswithin a local area (∼307 × 324 μm) in general showed BFs froma low (2 kHz) to high frequencies (32 kHz), with a range thatcovered four octaves (Fig. 5A). By contrast, over a larger size ofimaging area (∼524 × 554 μm) of marmoset A1, the BFs rangedless than two octaves (Fig. 5B). Rough inspection of the localtonotopic maps of rat A1 revealed many regions exhibited

E

0

0.2

ΔF/

F607080 607080 607080 607080 607080 607080 607080

SPL / dB

.5 8 32Frequency / KHz

.5 8 32 .5 8 32 .5 8 32 .5 8 32 .5 8 32 .5 8 32607080

SP

L / d

B

0

ΔF/F0.2

Cell 1 Cell 7Cell 3Cell 2 Cell 4 Cell 5 Cell 6

A60 dB 70 dB 80 dB

6

7

8

10

12

Other

KHz

B C D

Best frequency (KHz)

Cel

l #

0

30

6

60

7 8 8010 7012 60SPL / dB.5 2 4 8 16 321

Frequency (KHz)

60 dB

70 dB

80 dB ΔF/F50%

1 s

283 5 6 7 10 12 14 20 24 0 60 1200

100

Nearest-Neighbor Distance(μm)

Per

cent

age

of c

ells

60 dB SPL70 dB SPL80 dB SPL

F G60 dB 70 dB 80 dB

ΔF/F 20%60708060/70/8070/8060/8060/70

dB

Fig. 3. Local homogeneity in frequency tuning did not depend on thesound intensity. (A) Sound-evoked responses of the same population of A1cells at three different sound intensities (SPL, sound pressure level; 60, 70,and 80 dB), with BFs of the cells color coded. Data were collected for thesame imaging plane as that shown in Fig. 2. (Scale bar: 50 μm.) (B) The av-eraged frequency-tuning profiles derived from an example cell (indicated byred arrows in A) at three different sound intensities (60, 70, and 80 dB). (C)Total number of cells showing different BFs for three different sound in-tensities. (D) Cumulative percentage plot for the distribution of the nearest-neighbor distance for all cells at three different sound intensities. All curveswere significantly different from the diagonal line (P < 0.001 for all curves,Kolmogorov–Smirnov test). (E) Tonal receptive fields of four example cellsfrom the same imaging plane in A are shown. Plots below depict maximummean fluorescence changes to pure tone stimuli with increasing intensities,showing diverse changes in cellular responses at three different sound in-tensities. (F) Cells in one imaging plane (as in A) that responded to the puretone at a particular sound intensity. The brightness of the dots indicates theresponse magnitude (maximum mean ΔF/F, color coded by the scale bar).(Scale bar: 50 μm.) (G) Color-coded intensity map derived by combining thethree distribution plots shown in F. Colors of cells indicate whether the cellsshowed responses to one, two, or three intensities. Note that most cells wereactivated at all three intensities (white dots).

Zeng et al. PNAS | February 19, 2019 | vol. 116 | no. 8 | 3241

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Apr

il 19

, 202

1

“salt-and-pepper” distributions of BFs (Fig. 5A), although smallregions with clustered distribution of similar BFs could also befound (Fig. 5A).Two types of quantitative analyses were formed to examine the

uniformity in the BF distribution within A1. First, we measuredthe nearest-neighbor distance of the cell with the same BFs forall responsive A1 cells monitored in the same image plane (asthat done earlier for marmoset, Fig. 2). This analysis showed thatthe distribution of nearest-neighbor distances in rat A1 was muchcloser to the random distribution, as shown by the cumulativepercentage curves for the example cases and averaged data fromtwo rats (9 imaging planes) and two marmosets (11 imagingplanes) (Fig. 5C, 9 imaging planes from two rats and 11 imagingplanes from two marmosets). In the second analysis, we mea-sured the ΔBF for all cell pairs within the imaged plane, andplotted the cumulative percentage curves for both rat and mar-moset A1, based on the same dataset as above. This analysisshowed that the distributions of ΔBF for rat A1 was significantlydifferent from those of marmoset (P < 0.001, Kolmogorov–Smirnov test), much closer to the random distribution (diagonalline), although both marmoset and rat distributions were sig-nificantly different from random (P < 0.001, Kolmogorov–Smirnov test) (Fig. 5D). Taken together, our results indicatethat, over distances of hundreds of micrometers, the local A1tonotopic map in rats exhibited a much higher heterogeneitythan that in marmosets.To examine whether the nonuniform local tonotopic organi-

zation exists along the vertical axis of rat A1, we have also im-aged the cellular tuning responses within a local region (∼307 ×324 μm) at different cortical depths (210–350 μm from the piasurface) in three rats examined in this study. We found that for asimilar imaging area in A1, cells with diverse BFs were distrib-uted in similar salt-and-pepper distribution in all three rats andthere was a dominant population of cells with the same BF (2kHz) in all three rats (Fig. 5E). These findings indicate there wasa reproducible tonotopic map among different rats, and the ex-istence of predominant BF in the local A1 region could accountfor the global tonotopic property in rodents, despite the presenceof local heterogeneity in BFs.

Tonotopic Responses of A1 Cells Using Genetically Encoded Ca2+

Indicators. The above results were obtained by using the fluo-rescent Ca2+ indicator Cal-520 that was acutely loaded intoneurons. It is known that due to differential Ca2+ affinities ofdifferent indicators, the recorded neuronal activities may be bi-ased by the extent of subthreshold responses included in the

fluorescence signal (11, 29). Although Cal-520 has a Kd (320 nM)similar to that of Fluo-4 (350 nM), which is known to respondonly to suprathreshold activities (11), we decided to performimaging experiments using a genetically encoded ultrasensitivefluorescent protein GCaMP6f (30), which can reliably detectsingle action potentials in marmoset neurons (31, 32).Marmosets were injected with AAV vectors encoding human

synapsin promoter-driven GCaMP6f in A1 at 3 wk before theimaging experiment (Materials and Methods). In a large focalplane (∼520 × 550 μm, Fig. 6A), A1 cells expressing GCaMP6fshowed robust responses to single pure-tone stimuli, as shown bythe tone-evoked fluorescence changes in four example cells (Fig.6B) as well as by the heat map of florescent changes recordedfrom 20 responsive cells in one imaging plane (Fig. 6C). Theseresults indicate that the BFs were homogeneous over large dis-tances in the imaging plane. The spatial distribution of cells withdifferent BFs also showed homogeneous distribution of BFswithin local regions of A1, as shown in Fig. 6D. In constructingthis distribution map, we have combined the data for cells acti-vated by sound stimuli at three different intensities, due to therelatively low number of cells expressing GCaMP6f, comparedwith that observed with Cal-520 loading. We also examined thetonotopic responses of A1 cells evoked by sounds at various in-tensities and found the BFs were largely invariant over threelevels of sound intensities (40, 60, and 80 dB). When imagingplanes at three different cortical depths (180, 220, and 250 μm)were examined for their tonotopic organization, we found againsimilar clustering of cells with the same BFs at three differentdepths. These results were summarized by the 3D plots, with cellpairs sharing similar BFs from adjacent imaging planes within a

A

Distance (μm)

Dep

th fr

om p

ia (μ

m)

-140

-200

-260

0 100 200

Azimuth = 0°

140 μm 200 μm170 μm 230 μm

-200-260

-140

50

150

15050

Distance (μm)

Distance (μm)

Dep

th fr

om p

ia (μ

m)

Dep

th fr

om p

ia (μ

m)

-140

-200

-260

0 200Distance (μm)

200

Azimuth = 45°

260 μm6781012Other

KHz

B C

Fig. 4. (A) Spatial distribution of A1 cells with distinct BFs at five differentcortical depths from 140 to 260 μm, with the BF of each cell color coded bythe scheme shown on the Right. (B) A 3D composite plot of the distributionof cells with the same BFs, with cells of the same BFs in different depthswithin 50 μm linked by lines of the same color as that coding BFs. (C) Pro-jected 2D maps of the composite plot onto two planes at azimuth angles of0° and 45°, showing vertical alignment of cells with the same BFs at differentdepths. (Scale bar: 50 μm.)

A

2 4KHz

1 8 16

Marmoset

0 100 200Nearest-Neighbor Distance (μm)

Per

cent

age

of c

ells

300 0 1 20

100

48 16KHz

42 32

RatB C D

MarRat

*** ***

Ex. Pooled

P

3Δ BF (octave)

MarRat

*** ***

Ex. Pooled

P

2Best frequency (KHz)

4 8 16 320

50

Per

cent

age

ofce

lls

Best frequency (KHz)2 4 8 16 320

50

Best frequency (KHz)2 4 8 16 320

50

E

100 300

-300

300 100

-200

Distance (μm)Distance (μm)

Dep

th fr

om p

ia(μ

m)

Rat 1

816

KHz

42

32

300100100300

Rat 2-200

-300

100 300100300

Rat 3-200

-300

Fig. 5. Local tonotopic organization in rat A1 is more heterogeneous thanthat in marmoset A1. (A) Spatial distribution of all tone-selective cells in anexample imaging plane in the rat A1, with their BFs (kHz) color coded by thescheme below. (Scale bar: 50 μm.) (B) Spatial distribution of all tone-selectivecells in an example imaging plane from a marmoset. Note the uniform dis-tribution of BFs. (C and D) Cumulative percentage plots for the distributionof nearest-neighbor distances for all cells (in C) and for the distribution ofΔBF for all cell pairs within the same imaging plane (in D), respectively. Red,marmoset; black, rat. Dashed lines indicate data for the example imagingplane as shown in A and B; solid lines indicate pooled data from 9 imagingplanes from two rats and 11 imaging planes from two marmosets; diagonaldashed line indicates theoretically random (uniform) distribution. All curveswere significantly different from the diagonal line, which represents thetheoretically random distribution of neurons with different BFs (P < 0.001for all curves, Kolmogorov–Smirnov test). Curves for rats and marmosetswere significantly different (***P < 0.001 for all curves, Kolmogorov–Smir-nov test). (E) Spatial distribution of A1 cells with various BFs at differentcortical depths from 200 to 350 μm. Data are from three different rats. Thepercentages of cells with different BFs in each imaging region are shown inthe Bottom histograms. Note that in all three rats, similar A1 regions (withdominant BF of 2 kHz) were imaged.

3242 | www.pnas.org/cgi/doi/10.1073/pnas.1816653116 Zeng et al.

Dow

nloa

ded

by g

uest

on

Apr

il 19

, 202

1

distance of <100 μm being connected by lines (Fig. 6E). Takentogether, our experiments using GCaMP6f have largely con-firmed the findings using Cal-520, showing both horizontal andvertical local homogeneity in sound frequency preference andintensity invariance of tonotopic properties in marmoset A1neurons.

DiscussionUsing in vivo two-photon calcium imaging to monitor the ac-tivities of a large population of A1 neurons in responses to puretones of different frequencies, we found that the frequencypreference of marmoset A1 neurons is homogeneous over dis-tances of hundreds of micrometers in both horizontal and ver-tical directions relative to the pia surface. This was shown byusing either acute loading of fluorescence Ca2+ indicator Cal-520or in vivo expression of genetically encoded Ca2+ indicatorGCaMP6f. These results demonstrated the feasibility of simul-taneous recording of large populations of cortical neurons in themarmoset brain at single-cell resolution. Furthermore, macro-scopic tonotopic maps previously observed by electrophysiolog-ical recording (7) and intrinsic optical signals (8) directly reflecthomogeneous neuronal tuning properties within local regions ofmarmoset A1.Neuropil contamination can be a potential problem under

dense labeling of neurons. It is difficult to fully exclude thecontamination because the extent of neuropil contamination insignals recorded at somata is unknown. However, we haveattempted to subtract the neuropil signal by using a methodpreviously reported (33) and found that the local homogeneity oftonotopic organization observed in marmoset A1 was not

affected (SI Appendix, Fig. S2). Furthermore, the clear differ-ence in tonotopic organization between rats and marmosets isunlikely to be caused by neuropil contamination, since such in-terference presumably occurred in both cases.The precision of tonotopic organization at the cellular level

has recently become a controversial issue (34–39), based mostlyon studies from mice. Although a tonotopic map of monkey A1at the cellular level has not been reported, previous studies usingextracellular electrophysiological recording of neuronal activitieshave suggested that the tonotopic preference of cortical cellswithin local regions of A1 is homogeneous, consistent with asmooth tonotopic organization. Since both electrophysiologicalrecording and intrinsic optical imaging have a spatial resolutionof 50–100 μm (7, 8, 40), the smooth tonotopic maps may resultfrom the averaged responses over many neurons. Extracellularrecordings could also be biased toward highly active neurons ineither multi- or single-unit recordings (41). The in vivo two-photon calcium imaging approach offers cellular resolutionthat could unequivocally address the issue of local homogeneityin the frequency tuning properties of A1 neurons.Several recent studies (10, 11, 42, 43) using the in vivo two-

photon imaging approach have cast doubt on the existence ofstrictly tonotopic maps in A1. It was found that neurons in miceA1 with diverse frequency preferences were highly mixed locally,despite the presence of apparent macroscopic tonotopy. Theseprevious studies revealed that the BF for evoking neuronal re-sponses was highly variable among neurons within 50- to 100-μmdistances, with differences in BFs as large as two to four octaves(44). In the present study of marmosets, we found that differ-ences of BFs among cells within 250 μm were less than one oc-tave. To ensure that the discrepancy between our results andprevious findings on mice was not caused by the difference in theCa2+ imaging method, we also examined the local tonotopicproperty of rat A1 cells using our Cal-520 loading method.Quantitative comparison of our marmoset and rat results showedthat rat A1 cells with different BFs were distributed in a muchmore mixed manner than that found in the marmoset A1, im-plicating different A1 organizations in rodents vs. marmosets.Nevertheless, in both rat and marmoset A1, cells with the sameBFs were distributed in a manner that was far from random.Furthermore, we found that in each local rat A1 region, therewas a large majority of cells with a particular BF, with other cellsof diverse BFs interspersed among them. The presence of adominant population of cells with the same BF could account forthe global tonotopic maps found previously in rodents (6, 9, 26)and in our rat results (SI Appendix, Fig. S3), as well as the pre-viously reported “clustered” and salt-and-pepper distributions ofcells in different local regions in mouse A1 (43, 44) and ratA1 (45).In highly visual animals such as monkeys and cats, neurons

with similar receptive field properties in the primary visual cortex(V1) are well organized into local columns. By contrast, in ro-dents with poor vision, neurons with different receptive fieldproperties are mixed locally in a salt-and-pepper manner (21, 22,46–48). By analogy, highly uniform tuning properties of A1neurons locally in marmosets may reflect an organization prin-ciple favorable for auditory processing in animal species that arerich in vocal communication and social interaction. Naturalsounds with syllables in marmoset calls need to be first decom-posed into frequency-specific signals and then reintegrated forauditory perception. Previous studies have shown that someneurons in marmoset A1 selectively respond to the “twitter”sound syllables, but not to the time-reversed one (49, 50). In-tegration of frequency-specific signals into syllables may dependon intracortical connections among A1 neurons, in addition topotential contributions from thalamocortical inputs. Given thelocal homogeneity in the frequency tuning of A1 neurons ofmarmoset, long-range intracortical circuitry among neurons indifferent tonotopic A1 domains may play a significant role in theintegration of auditory signals of different frequencies. The

Frequency (KHz)10.8 1.2

1

10.8 1.2

2

3

4

4 1680.5 1 2Frequency (KHz)

20

Cel

l #

1 0.1

0

12

34

A

2 4KHz

10.5 8 16

DB

C

E

-220-250

-180

200600

200

40 dB

400

0.5-1

1.2-2

> 2KHz

200600

200

60 dB

400 200600

200

80 dB

400

Fig. 6. Two-photon imaging of marmoset A1 using genetically encoded Caindicator GCaMP6f. (A, Top) Fluorescence image of a region in marmoset A1,3 wk after AAV-GCaMP6f injection. (A, Bottom) Two-photon fluorescenceimage of a ∼520 × 550 μm imaging area in A1 superficial layer at a depth of220 μm, corresponding to that marked by the red box at the Top. (Scale bars:100 μm.) (B) Mean fluorescence changes (ΔF/F) evoked by three pure-tonestimuli (0.8, 1.0, and 1.2 kHz; duration 0.2 s, eight trials each) for four ex-ample cells (marked by red circles in A). (Scale bar: 0.5 s and 20% ΔF/F.) (C)ΔF/F with time for 20 pure tone-responsive cells during sequential stepwiseapplication of pure tones from 0.5 to 16 kHz (black bar, tone duration 0.2 s),with the magnitude of ΔF/F color coded by the continuous scale shown onthe Right. Red trace on Top shows the averaged ΔF/F for all 20 cells. Notethat the BF for the vast majority of cells was located at 1.2 kHz. (Scale bar:0.5 s and 20% ΔF/F.) (D) Spatial distribution of all tone-selective cells in theexample imaging plane as that shown in A, Bottom, including data recordedat three different sound levels (40, 60, and 80 dB) of the same frequency. (E)Three-dimensional distribution of A1 cells with similar BFs at three differentcortical depths, with sound stimuli of three different intensities applied.

Zeng et al. PNAS | February 19, 2019 | vol. 116 | no. 8 | 3243

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Apr

il 19

, 202

1

current results thus pave the way for further analysis of A1 cir-cuitry underlying natural sound processing in marmosets.

Materials and MethodsAnimal care and experimental procedures were approved by the Animal CareCommittee of Shanghai Institutes for Biological Sciences, Chinese Academy ofSciences (Shanghai, China). Four adult common marmosets (C. jacchus; onemale and three females; body weight: 260–400 g) obtained from the non-human primate facility of the Institute of Neuroscience were used in thisstudy. All acoustic stimuli were generated using MATLAB (MathWorks) andthe sound delivery system was calibrated using a B&K (2669-L) calibrator. Fortwo-photon imaging, the duration of pure-tone stimuli was 0.2 s (including5-ms ON and 5-ms OFF linear ramps), with an interstimulus interval of 1–1.5s. Cal-520 AM (AAT Bioquest) was injected into layer 2/3 auditory cortex aspreviously described (51). Cells were identified manually on the basis of size,shape, and brightness. Fluorescence changes with time of individual cellswere extracted by averaging pixel intensity values within each cell in eachframe. Correction for neuropil contamination was not applied (see howeverSI Appendix, Fig. S2). Average fluorescence intensity level (Ft) evoked by

each stimulus (measurement window 0.4 s; 0.2-s stimulus duration and 0.2-spoststimulus duration) was compared with average prestimulus baselinefluorescence (F0) over a 0.5-s window. Cells showing Ft that was significantlylarger than F0 (five to eight repeats, P < 0.05, ANOVA) were defined as“responsive cells.” Of these, BF of the responsive cell was determined by thetone frequency that evoked the largest responses over the frequency rangetested. In our dataset, nearly all responsive cells were selective to pure tones.Error bars indicate SEM. Frequency vector maps were calculated using thevector-summation method (52). Full description of materials and methodscan be found in SI Appendix, Materials and Methods.

ACKNOWLEDGMENTS. We thank Yang Dan, Ninglong Xu, and Siyu Zhangfor suggestions and comments on the manuscript; and Neng Gong, Hao Li,and Xuebo Li for technical support. This work was supported by the StrategicPriority Research Program of the Chinese Academy of Sciences (CAS) (GrantXDBS0100000), the Shanghai Municipal Government Bureau of Science andTechnology (Grant 16Jc1420200), the National Natural Science Foundationof China (Grant 31571101), and the Youth Innovation Promotion Associationof CAS (Grant 2015223 to Z.-m.S.).

1. Stiebler I, Neulist R, Fichtel I, Ehret G (1997) The auditory cortex of the house mouse:Left-right differences, tonotopic organization and quantitative analysis of frequencyrepresentation. J Comp Physiol A Neuroethol Sens Neural Behav Physiol 181:559–571.

2. Bizley JK, Nodal FR, Nelken I, King AJ (2005) Functional organization of ferret audi-tory cortex. Cereb Cortex 15:1637–1653.

3. Nelken I, et al. (2004) Large-scale organization of ferret auditory cortex revealedusing continuous acquisition of intrinsic optical signals. J Neurophysiol 92:2574–2588.

4. Merzenich MM, Knight PL, Roth GL (1973) Cochleotopic organization of primaryauditory cortex in the cat. Brain Res 63:343–346.

5. Recanzone GH, Schreiner CE, Sutter ML, Beitel RE, Merzenich MM (1999) Functionalorganization of spectral receptive fields in the primary auditory cortex of the owlmonkey. J Comp Neurol 415:460–481.

6. Kalatsky VA, Polley DB, Merzenich MM, Schreiner CE, Stryker MP (2005) Fine func-tional organization of auditory cortex revealed by Fourier optical imaging. Proc NatlAcad Sci USA 102:13325–13330.

7. Bendor D, Wang X (2008) Neural response properties of primary, rostral, and ros-trotemporal core fields in the auditory cortex of marmoset monkeys. J Neurophysiol100:888–906.

8. Tani T, et al. (2018) Sound frequency representation in the auditory cortex of thecommon marmoset visualized using optical intrinsic signal imaging. eNeuro 5:ENEURO.0078-18.2018.

9. Wu GK, Li P, Tao HW, Zhang LI (2006) Nonmonotonic synaptic excitation and im-balanced inhibition underlying cortical intensity tuning. Neuron 52:705–715.

10. Rothschild G, Nelken I, Mizrahi A (2010) Functional organization and populationdynamics in the mouse primary auditory cortex. Nat Neurosci 13:353–360.

11. Bandyopadhyay S, Shamma SA, Kanold PO (2010) Dichotomy of functional organi-zation in the mouse auditory cortex. Nat Neurosci 13:361–368.

12. Teichert M, Bolz J (2017) Simultaneous intrinsic signal imaging of auditory and visualcortex reveals profound effects of acute hearing loss on visual processing.Neuroimage 159:459–472.

13. Kanold PO, Nelken I, Polley DB (2014) Local versus global scales of organization inauditory cortex. Trends Neurosci 37:502–510.

14. Schreiner CE (1995) Order and disorder in auditory cortical maps. Curr Opin Neurobiol5:489–496.

15. Rothschild G, Mizrahi A (2015) Global order and local disorder in brain maps. AnnuRev Neurosci 38:247–268.

16. Miller CT, et al. (2016) Marmosets: A neuroscientific model of human social behavior.Neuron 90:219–233.

17. Grinvald A, Frostig RD, Lieke E, Hildesheim R (1988) Optical imaging of neuronalactivity. Physiol Rev 68:1285–1366.

18. Morel A, Kaas JH (1992) Subdivisions and connections of auditory cortex in owlmonkeys. J Comp Neurol 318:27–63.

19. Morel A, Garraghty PE, Kaas JH (1993) Tonotopic organization, architectonic fields,and connections of auditory cortex in macaque monkeys. J Comp Neurol 335:437–459.

20. Tada M, Takeuchi A, Hashizume M, Kitamura K, Kano M (2014) A highly sensitivefluorescent indicator dye for calcium imaging of neural activity in vitro and in vivo.Eur J Neurosci 39:1720–1728.

21. Ohki K, Chung S, Ch’ng YH, Kara P, Reid RC (2005) Functional imaging with cellularresolution reveals precise micro-architecture in visual cortex. Nature 433:597–603.

22. Nauhaus I, Nielsen KJ, Disney AA, Callaway EM (2012) Orthogonal micro-organizationof orientation and spatial frequency in primate primary visual cortex. Nat Neurosci 15:1683–1690.

23. Sadagopan S, Wang X (2008) Level invariant representation of sounds by populationsof neurons in primary auditory cortex. J Neurosci 28:3415–3426.

24. Schreiner CE, Winer JA (2007) Auditory cortex mapmaking: Principles, projections,and plasticity. Neuron 56:356–365.

25. Schreiner CE, Read HL, Sutter ML (2000) Modular organization of frequency in-tegration in primary auditory cortex. Annu Rev Neurosci 23:501–529.

26. Polley DB, Read HL, Storace DA, Merzenich MM (2007) Multiparametric auditory re-ceptive field organization across five cortical fields in the albino rat. J Neurophysiol97:3621–3638.

27. Mountcastle VB (1997) The columnar organization of the neocortex. Brain 120:701–722.

28. Hubel DH, Wiesel TN (1977) Ferrier lecture. Functional architecture of macaquemonkey visual cortex. Proc R Soc Lond B Biol Sci 198:1–59.

29. Sato TR, Gray NW, Mainen ZF, Svoboda K (2007) The functional microarchitecture ofthe mouse barrel cortex. PLoS Biol 5:e189.

30. Chen TW, et al. (2013) Ultrasensitive fluorescent proteins for imaging neuronal ac-tivity. Nature 499:295–300.

31. Sadakane O, et al. (2015) Long-term two-photon calcium imaging of neuronal pop-ulations with subcellular resolution in adult non-human primates. Cell Rep 13:1989–1999.

32. Li M, Liu F, Jiang H, Lee TS, Tang S (2017) Long-term two-photon imaging in awakemacaque monkey. Neuron 93:1049–1057.e3.

33. Kerlin AM, Andermann ML, Berezovskii VK, Reid RC (2010) Broadly tuned responseproperties of diverse inhibitory neuron subtypes in mouse visual cortex. Neuron 67:858–871.

34. Hromádka T, Deweese MR, Zador AM (2008) Sparse representation of sounds in theunanesthetized auditory cortex. PLoS Biol 6:e16.

35. Evans EF, Ross HF, Whitfield IC (1965) The spatial distribution of unit characteristicfrequency in the primary auditory cortex of the cat. J Physiol 179:238–247.

36. Goldstein MH, Jr, Abeles M, Daly RL, McIntosh J (1970) Functional architecture in catprimary auditory cortex: Tonotopic organization. J Neurophysiol 33:188–197.

37. Merzenich MM, Knight PL, Roth GL (1975) Representation of cochlea within primaryauditory cortex in the cat. J Neurophysiol 38:231–249.

38. Schreiner CE, Sutter ML (1992) Topography of excitatory bandwidth in cat primaryauditory cortex: Single-neuron versus multiple-neuron recordings. J Neurophysiol 68:1487–1502.

39. Liang F, et al. (October 11, 2018) Sparse representation in awake auditory cortex: Cell-type dependence, synaptic mechanisms, developmental emergence, and modulation.Cereb Cortex, 10.1093/cercor/bhy260.

40. Nishimura M, Takemoto M, Song WJ (2018) Organization of auditory areas in thesuperior temporal gyrus of marmoset monkeys revealed by real-time optical imaging.Brain Struct Funct 223:1599–1614.

41. Margrie TW, Brecht M, Sakmann B (2002) In vivo, low-resistance, whole-cell record-ings from neurons in the anaesthetized and awake mammalian brain. Pflugers Arch444:491–498.

42. Li J, et al. (2017) Functional imaging of neuronal activity of auditory cortex by usingCal-520 in anesthetized and awake mice. Biomed Opt Express 8:2599–2610.

43. Issa JB, et al. (2014) Multiscale optical Ca2+ imaging of tonal organization in mouseauditory cortex. Neuron 83:944–959.

44. Winkowski DE, Kanold PO (2013) Laminar transformation of frequency organizationin auditory cortex. J Neurosci 33:1498–1508.

45. Tao C, et al. (2017) Diversity in excitation-inhibition mismatch underlies local func-tional heterogeneity in the rat auditory cortex. Cell Rep 19:521–531.

46. Ohki K, et al. (2006) Highly ordered arrangement of single neurons in orientationpinwheels. Nature 442:925–928.

47. Bonin V, Histed MH, Yurgenson S, Reid RC (2011) Local diversity and fine-scale or-ganization of receptive fields in mouse visual cortex. J Neurosci 31:18506–18521.

48. Kara P, Boyd JD (2009) A micro-architecture for binocular disparity and ocular dom-inance in visual cortex. Nature 458:627–631.

49. Wang X, Merzenich MM, Beitel R, Schreiner CE (1995) Representation of a species-specific vocalization in the primary auditory cortex of the common marmoset: Tem-poral and spectral characteristics. J Neurophysiol 74:2685–2706.

50. Wang X, Kadia SC (2001) Differential representation of species-specific primate vocal-izations in the auditory cortices of marmoset and cat. J Neurophysiol 86:2616–2620.

51. O’Herron P, et al. (2012) Targeted labeling of neurons in a specific functional micro-domain of the neocortex by combining intrinsic signal and two-photon imaging. J VisExp e50025.

52. Bosking WH, Zhang Y, Schofield B, Fitzpatrick D (1997) Orientation selectivity and thearrangement of horizontal connections in tree shrew striate cortex. J Neurosci 17:2112–2127.

3244 | www.pnas.org/cgi/doi/10.1073/pnas.1816653116 Zeng et al.

Dow

nloa

ded

by g

uest

on

Apr

il 19

, 202

1