Neuron

Article

The Precise Temporal Pattern of PrehearingSpontaneous Activity Is Necessaryfor Tonotopic Map RefinementAmanda Clause,1,2,3,6,7 Gunsoo Kim,2,3,6,8,* Mandy Sonntag,4 Catherine J.C. Weisz,1 Douglas E. Vetter,5

Rudolf R}ubsamen,4 and Karl Kandler1,2,3,*1Department of Otolaryngology, University of Pittsburgh School of Medicine, Pittsburgh, PA 15261, USA2Department of Neurobiology, University of Pittsburgh School of Medicine, Pittsburgh, PA 15261, USA3Center for the Neural Basis of Cognition, University of Pittsburgh School of Medicine, Pittsburgh, PA 15261, USA4Faculty of Biosciences, Pharmacy and Psychology, University of Leipzig, 04103 Leipzig, Germany5Department of Neurobiology and Anatomical Sciences, University of Mississippi Medical Center, Jackson, MS 39216, USA6Co-first author7Present address: Eaton-Peabody Laboratory, Massachusetts Eye and Ear Infirmary, Department of Otology and Laryngology, Harvard

Medical School, Boston, MA 02114, USA8Present address: Department of Physiology, Center for Integrative Neuroscience, University of California San Francisco, San Francisco,

CA 94143-0444, USA*Correspondence: gkim@phy.ucsf.edu (G.K.), kkarl@pitt.edu (K.K.)

http://dx.doi.org/10.1016/j.neuron.2014.04.001

SUMMARY

Patterned spontaneous activity is a hallmark ofdeveloping sensory systems. In the auditory system,rhythmic bursts of spontaneous activity are gener-ated in cochlear hair cells and propagated alongcentral auditory pathways. The role of these activitypatterns in the development of central auditorycircuits has remained speculative. Here we demon-strate that blocking efferent cholinergic neurotrans-mission to developing hair cells in mice that lackthe a9 subunit of nicotinic acetylcholine receptors(a9 KO mice) altered the temporal fine structureof spontaneous activity without changing activitylevels. KO mice showed a severe impairment in thefunctional and structural sharpening of an inhibitorytonotopic map, as evidenced by deficits in synapticstrengthening and silencing of connections and anabsence in axonal pruning. These results provideevidence that the precise temporal pattern of spon-taneous activity before hearing onset is crucial forthe establishment of precise tonotopy, the majororganizing principle of central auditory pathways.

INTRODUCTION

Before the developing brain responds to external stimuli, the

dominant activity in neuronal pathways consists of spontane-

ously generated action potentials. This spontaneous activity

is typically characterized by rhythmic bursts of high levels of

activity separated by periods of quiescence (Hanson and Land-

messer, 2003; Kirkby et al., 2013; Meister et al., 1991). Similar

burst-like activity is also present in the developing auditory sys-

822 Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc.

tem before the onset of hearing (i.e., sensitivity to airborne

sound; Jones et al., 2007; Kotak and Sanes, 1995; Lippe,

1994; Sonntag et al., 2009; Tritsch et al., 2010). Prehearing activ-

ity bursts originate in cochlear inner hair cells (IHCs), which

fire trains of calcium action potentials (Glowatzki and Fuchs,

2000; Johnson et al., 2011; Kros et al., 1998; Tritsch et al.,

2007) that are transmitted to spiral ganglion cells and are

faithfully propagated along ascending central auditory pathways

(Tritsch et al., 2010).

Before hearing onset, IHCs are transiently innervated by the

efferent axons of medial olivocochlear neurons, a cholinergic

cell group located in the ventral brainstem (Simmons et al.,

1996; Warr and Guinan, 1979). At hair cells, acetylcholine

activates nicotinic acetylcholine receptors (AChRs) that contain

calcium-permeable a9 and a10 subunits (Elgoyhen et al., 1994;

Vetter et al., 1999). Calcium influx through these a9-containing

AChRs rapidly activates small-conductance potassium chan-

nels, resulting in the hyperpolarization of IHCs and an inhibition

of calcium spike generation (Glowatzki and Fuchs, 2000; Katz

et al., 2004). The transient cholinergic modulation of immature

IHCs may be a mechanism that modulates the level or temporal

pattern of cochlea-generated prehearing activity (Glowatzki and

Fuchs, 2000; Johnson et al., 2011).

In analogy to other neuronal systems (Hanson and Land-

messer, 2004; Kirkby et al., 2013), it has been widely assumed

that cochlea-generated patterns of spontaneous activity play

an important role in the development of the auditory system.

Whereas the rate of IHC spikes is important for the maturation

of vesicle fusion at IHC synapses (Johnson et al., 2013), a causal

link between patterned activity and the developmental organi-

zation of central auditory circuits has remained speculative due

to difficulties in experimentally altering the temporal patterns of

spontaneous activity without also severely changing the overall

levels of cochlea-generated activity. For instance, blocking

cochlea-generated activity before hearing onset leads to the

degeneration of spiral ganglion neurons and their postsynaptic

Neuron

Spontaneous Activity and Tonotopic Refinement

targets in the ventral cochlear nucleus (Hashisaki and Rubel,

1989; Hirtz et al., 2011; Seal et al., 2008), and it interferes with

the maturation of neuronal and synaptic properties in higher or-

der auditory neurons (Cao et al., 2008; Couchman et al., 2011;

Kotak and Sanes, 1996; Leao et al., 2006; Youssoufian et al.,

2005). Conflicting results have been obtained as to whether pre-

hearing activity plays a role in the formation of precise tonotopic

maps, themajor organizational principle of auditory pathways. In

congenitally deaf mice, the tonotopic organization of central

auditory pathways appears normal before hearing onset (Cao

et al., 2008; Noh et al., 2010; Rubel and Fritzsch, 2002; Youssou-

fian et al., 2008), whereas in neonatally deafened cats (Leake

et al., 2006) or gerbils (Sanes and Takacs, 1993), tonotopic orga-

nization of brainstem pathways is less precise.

In this study, we investigated whether changes in the temporal

pattern of spontaneous activity affect the development of a cen-

tral tonotopic map. We hypothesized that spontaneous activity

patterns would be altered in mice in which the a9 AChR subunit

has been genetically deleted (a9 knockout [KO] mice) (Vetter

et al., 1999). Because a9-containing AChRs are not expressed

in the brain (Allen Institute for Brain Science, 2012; Vetter et al.,

1999; Zuo et al., 1999), cholinergic transmission in a9 KO mice

is abolished in cochlear hair cells while remaining normal in cen-

tral auditory pathways. Single unit recordings from inhibitory

neurons in the medial nucleus of the trapezoid body (MNTB) re-

vealed that a9 KO mice exhibit altered temporal spike patterns

while having normal levels of spontaneous activity. To determine

whether this affects the development of tonotopic maps, we

characterized the synaptic and anatomical refinement of the

inhibitory pathway from the MNTB to the lateral superior olive

(LSO), an inhibitory pathway in the mammalian sound localiza-

tion system whose development has been well characterized

both physiologically and anatomically (Friauf, 2004; Kandler

and Gillespie, 2005). Our results demonstrate that, compared

to wild-type (WT) mice, in a9 KO mice, the strengthening and

silencing of inhibitory MNTB-LSO connections before hearing

onset was impaired, leading to a reduced sharpening of func-

tional topography. In addition, a9 KO mice have severe deficits

in the axonal pruning that occurs in normal mice during the first

week after hearing onset. These results provide evidence that

the precise temporal pattern of spontaneous activity before

hearing onset is crucial for the development of a precise tono-

topic map.

RESULTS

Altered Temporal Structure of Spontaneous Activity ina9 KO MiceTo test whether a loss of cholinergic innervation of IHCs alters

the level or the temporal patterns of spontaneous activity in cen-

tral auditory pathways, we recorded single units from the MNTB

in WT and a9 KO mice before hearing onset (Figure 1A). We

chose the MNTB because the spontaneous firing of MNTB neu-

rons faithfully follows the spike patterns of spiral ganglion neu-

rons (Tritsch et al., 2010) and because these recordings directly

reveal the presynaptic activity in the MNTB-LSO pathway. In WT

mice, spontaneous action potentials (APs) occurred in bursts

that were separated by periods of low, sporadic activity (Fig-

ure 1B; Figure S1 available online). Single events exhibited the

characteristic complex waveform composed of a pre- and post-

synaptic component typical for MNTB neurons (Sonntag et al.,

2009). Bursts occurred five to six times per minute (5.6 ± 0.3

bursts/min, n = 21 cells), with a mean burst duration of 5.5 ±

0.8 s and a mean firing rate during bursts of 8.4 ± 0.7 Hz (Figures

1G–1J). During bursts, spikes did not occur randomly, as indi-

cated by a coefficient of variation of interspike intervals (ISIs)

significantly greater than one (1.9 ± 0.1, n = 21 cells). Instead,

most spikes in a burst were clustered into ‘‘mini-bursts,’’ which

occurred at approximately 2 Hz (Figure 1K) and consisted of

two to nine APs (2.3 ± 0.1 APs/mini-burst) occurring at high rates

(383 ± 13 Hz; range, 275–481 Hz). The clustering of spikes into

bursts and mini-bursts resulted in ISI histograms with distinct

peaks (Figures 1D and 1F) very similar to those previously

described for spiral ganglion neurons and MNTB neurons in

rats (Tritsch et al., 2010) and spiral ganglion neurons in neonatal

cats (Jones et al., 2007).

In a9 KO mice, the mean firing rate of MNTB neurons was

indistinguishable from the rate in WT mice (Figure 1G; p = 0.58,

Student’s t test). In addition, bursts occurred with a similar fre-

quency as inWTmice (Figure 1H; p = 0.31). However, burst dura-

tion was 49% shorter in a9 KO mice than in WT mice (Figure 1I;

p = 0.0036), and during bursts themean firing rate was increased

by 83% (Figure 1J; p = 0.00002). Accordingly, the mean fre-

quency of mini-bursts and single spikes was 71% and 82%

higher, respectively, in KO mice (Figures 1K and 1L; p = 0.0018

and p = 0.00034), whereas the fraction of spikes that belonged

to mini-bursts remained the same (58% both in WT and KO,

p = 0.98). These changes resulted in marked differences in

the population ISI histograms (Figures 1E and 1F). Whereas the

first peak of the ISI histogram was similar in KO and WT mice

(peaks at �4 ms), the second peak, which reflects the intervals

between mini-bursts, was significantly shifted toward shorter

values (from �120 ms in WT to �40 ms in a9 KO mice). Interest-

ingly, the higher firing rates during bursts in a9 KO mice had

no effect on the lengthening-shorting pattern of ‘‘spike cluster’’

intervals (mini-bursts and single APs) that occurs during bursts

(Tritsch et al., 2010), indicating that the temporal dynamics of

spike clusters were unaffected in a9 KOmice (Figure 1M). Taken

together, loss of functional cholinergic transmission to hair cells

was associated with relatively subtle changes in spontaneous

activity as overall spike rate or the burst-like nature of the activity

of MNTB neurons remained unchanged, while the duration of

bursts was reduced and intraburst firing was increased.

Refinement of Functional Connectivity of theMNTB-LSOPathway before HearingOnset Is Impaired in a9KOMiceWe next investigated whether the topographic sharpening of

MNTB-LSO maps and the synaptic refinement of MNTB-LSO

connections before hearing onset is affected by the altered activ-

ity pattern in a9 KOmice. We first used focal glutamate uncaging

to characterize the development of functional MNTB-LSOmaps.

In WT mice, there was a developmental decrease in the area of

theMNTB that provides synaptic input to individual LSO neurons

(MNTB input area), indicating a sharpening of the functional

MNTB-LSO map. Between postnatal days (P) 1 and 12, the

average MNTB input area decreased by 63% (Figures 2A and

Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc. 823

Figure 1. Temporal Patterns of Spontaneous

Prehearing Activity Differ between MNTB

Neurons of WT and a9 KO Mice

(A) Scheme of ascending neuronal pathways to

the LSO. Excitatory pathways, releasing glutamate,

are in green; inhibitory pathways, coreleasing gly-

cine, GABA, and glutamate are in red. VCN, ventral

cochlear nucleus. The tonotopic organization is

reflected by shading (low-frequency areas are dark,

high-frequency areas are white).

(B and C) Example in vivo recordings of the

spontaneous spiking activity of MNTB neurons in

an 8-day-old WT (B) and a9 KO mouse (C). Neural

traces are shown on increasingly finer time scales

from left to right, as indicated by the scale bars.

Blue bars indicate activity bursts. Red brackets

mark the portion of the traces shown at higher

resolution on the right. Individual events exhibited

complex waveforms composed of pre- (Pr) and

postsynaptic components (Po) typical for MNTB

neurons.

(D and E) Interspike interval (ISI) histograms plotted

on a log scale for the neurons shown in (B) and (C),

respectively.

(F) Normalized population ISI histograms of WT

(black, n = 21 cells) and a9 KO mice (red, n = 21

cells). In a9 KO mice, the second peak is shifted to

shorter intervals.

(G–L) Quantification of spiking activity in WT (n =

21 cells) and a9 KO mice (n = 21 cells). The mean

overall firing rate (G) and the frequency of bursts

(H) did not differ between the two genotypes. The

duration of bursts (I) was reduced in a9 KO mice

(WT, 5.5 ± 0.8 s; KO, 2.9 ± 0.3 s; p = 0.0036,

Student’s t test). The mean firing rate during bursts

(J) was increased in a9 KO mice (WT, 8.4 ± 0.7 Hz;

KO, 15.4 ± 1.3 Hz; p = 0.000026, Student’s t test),

as was the mean frequency of mini-bursts (K;

WT, 2.1 ± 0.2 Hz; KO, 3.6 ± 0.4 Hz; p = 0.0018,

Student’s t test) and single spikes that did not

form mini-bursts (L; WT, 3.3 ± 0.3 Hz; KO, 6.0 ±

0.4; p = 0.00034).

(M) Mean intervals between spike clusters (mini-

bursts and single spikes together) plotted as a

function of position within a burst (black: WT, 214

bursts; red: KO, 216 bursts). Positions 1 through 5

denote the first five intervals at the beginning of a

burst, whereas positions �5 through �1 denote the

last five intervals at the end of a burst. All error bars

represent SEM.

See also Figure S1.

Neuron

Spontaneous Activity and Tonotopic Refinement

2B). Along the tonotopic (mediolateral) axis of the MNTB, the

width of input areas decreased by 45% (Figure 2C), indicating

an approximate 2-fold increase in the topographic precision of

connectivity along the tonotopic axis. These results are in line

with previous results from other mouse strains and from rats

(Kim and Kandler, 2003; Noh et al., 2010).

In newborn a9 KO mice, the size of MNTB input areas was

similar to those of WT mice (32% ± 3%, P1–P3, n = 6, p =

0.46, Student’s t test; Figures 2A and 2B), indicating that the

initial formation of MNTB-LSO topography occurs normally in

824 Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc.

a9 KO mice. Over the next 7–10 days, the size of MNTB input

areas in a9 KO mice decreased by 44% (to 18% ± 2% at P10–

P12, n = 16), but this decrease was significantly smaller than

that in WT mice (p < 0.01, Student’s t test). Therefore, at hearing

onset, MNTB input areas were on average 80% larger in a9 KO

mice than in WT mice (18% ± 2% in a9 KO versus 10% ± 2%

in WT at P10–P12; p = 0.01, Student’s t test, Figure 2B). These

larger input areas spread over 27% of the mediolateral axis of

the MNTB, a 50% increase compared to WT mice (Figure 2C).

The larger input maps in a9 KO mice were not due to a smaller

Figure 2. Topographic Refinement of Functional MNTB-LSO Con-

nectivity Is Impaired in a9 KO Mice

(A) Examples of MNTB input maps in brain slices obtained from WT mice

(upper row) and a9 KO mice (lower row). Example traces show postsynaptic

currents (PSCs) in LSO neurons elicited by uncaging glutamate at corre-

sponding MNTB sites. The amplitudes and transferred charge of the PSCs are

quantified in Figure S3 and Table S2. The MNTB is outlined in black. Stimu-

lation locations are depicted as squares and locations in theMNTB fromwhich

responses could be elicited are shown filled in red (see Experimental Pro-

cedures for details).

(B) Mean input areas normalized to the corresponding MNTB area in WT and

a9 KO mice aged P1–P3 and P10–P12. In WT mice, normalized MNTB input

areas decreased by 63% (from 27% ± 5% at P1–P3 [n = 7] to 10% ± 2% at

P10–P12 [n = 14]; p < 0.01, Student’s t test), whereas in KO mice, input areas

decreased by 44% (from 32% ± 3% at P1–P3 [n = 6] to 18 ± 2% at P10–P12

[n = 16]; p < 0.01). At P10–P12, input maps in KOmice were 80% larger than in

WT mice (p = 0.01).

(C) Developmental changes in the normalized mediolateral width of MNTB

input maps in WT and a9 KO mice. In WT mice, the width decreased by 45%

(from 33% ± 5% at P1–P3 [n = 7], to 18% ± 2% at P10–P12 [n = 14]; p < 0.01),

whereas in a9 KOmice the width decreased by 34% (from 41%± 3%at P1–P3

to 27% ± 3% at P10–P12; p < 0.01). At P10–P12, the tonotopic width of input

maps was 50% larger in a9 KO mice than in WT mice (p = 0.01). All error bars

represent SEM.

See also Figure S2 and Table S1.

Neuron

Spontaneous Activity and Tonotopic Refinement

cross-sectional MNTB area (WT: 69 3 103 ± 3.2 3 103 mm; a9

KO: 63.5 3 103 ± 1.8 3 103 mm; p = 0.15, Student’s t test) or

to differences in the spatial resolution of glutamate uncaging

(Figure S2) that may be caused by altered neuronal excitability

or glutamate sensitivity of MNTB neurons in a9 KO mice.

We next examined whether the decrease in MNTB-LSO

convergence that occurs during the first 2 postnatal weeks

was altered in a9 KO mice. We used minimal and maximal stim-

ulation techniques, which provide a measure of the strength

of single MNTB-LSO fiber connections and of all converging

MNTB fibers (Gil et al., 1999; Hirtz et al., 2012; Kim and Kandler,

2003, 2010). In P1–P3 mice, the maximal response amplitudes

did not differ between WT and a9 KO mice (Figures 3A, 3B,

and 3E). The developmental increase of maximal response

amplitude between P1 and P12 was also similar in both geno-

types with an approximate 6-fold increase in WT mice and an

approximate 5-fold increase in a9 KO mice (Figure 3E), resulting

in maximum responses at P10–P12 that did not differ signifi-

cantly between genotypes (p = 0.21, Student’s t test). Thus,

despite receiving inputs from a significantly larger MNTB area,

in a9 KOmice the total amount of inhibition received by individual

LSO neurons at hearing onset was unchanged.

In contrast to a normal increase in the maximal responses, in

a9 KO mice the increase in single-fiber responses was sig-

nificantly impaired (Figures 3A–3D). In WT mice, we observed a

16-fold increase in single-fiber response amplitude between

P1–P3 and P10–P12, whereas in a9 KO mice, this increase

was only 7-fold. Consequently, at P10-12 single-fiber response

amplitudes were significantly smaller in a9 KO mice than in WT

mice (p < 0.01, Figure 3E). The amplitude distributions and cu-

mulative histograms of single-fiber postsynaptic currents

(PSCs) indicate that the smaller average PSC amplitude in a9

KO mice is due to a higher prevalence of weak inputs

(<100 pA; Figure 3G; p = 0.001, Kolmogorov-Smirnov test).

This conclusion is supported by the smaller peak amplitudes eli-

cited from individual uncaging sites in P10–P12 a9 KO mice in

the mapping experiments (Figure S3 and Table S1). In contrast

to the reduced amplitudes of single-fiber responses, the kinetics

of single-fiber responses, such as the decay and rise times, did

not differ between WT and a9 KO mice (Figure S4), suggesting

that other basic properties of synaptic transmission at the

MNTB-LSO synapse, such as release synchronicity, the type

of neurotransmitters released, and postsynaptic receptor

composition, develop normally in a9 KO mice.

The observed decrease in single-fiber response amplitudes in

a9 KO mice may reflect more distal dendritic locations of inputs

rather than a reduction of their actual strength. This possibility,

however, seems unlikely because the rise times (which correlate

with the electrotonic distance between recording site and syn-

apse site) not differ between the genotypes (Figure S4) and

the rise times were also not correlated with the amplitude of

the synaptic currents (WT: r = �0.15, p = 0.33; KO: r = �0.09,

p = 0.51). Therefore, smaller single-fiber responses combined

with unchanged maximal responses suggest that LSO neurons

in a9 KO mice are contacted by a larger number of MNTB neu-

rons, each of which forms weaker connections.

To estimate the convergence of MNTB inputs to LSO neu-

rons, we used bootstrap resampling of the mean single-fiber

and maximal PSC amplitudes (see Supplemental Experimental

Procedures). Between P1 and P12, the MNTB-LSO conver-

gence decreased by 57% in WT mice (from 23 ± 6 inputs to

10 ± 2 inputs per LSO neuron), whereas in a9 KO mice this

decrease was only 22% (27 ± 7 inputs to 21 ± 6 inputs; Fig-

ure 3F). Thus, in a9 KO mice, individual LSO neurons receive

inputs from twice as many MNTB neurons compared to LSO

neurons in WT mice (p = 0.03, permutation test). Taken

together, both functional mapping and minimal/maximal stimu-

lation experiments demonstrate that the strengthening and

elimination (silencing) of individual MNTB inputs are impaired

Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc. 825

Figure 3. Developmental Strengthening and Elimination of MNTB-LSO Connections Are Impaired in a9 KO Mice

(A and B) Examples of synaptic responses in LSO neurons elicited by minimal (min) and maximal (max) stimulation of MNTB fibers in neonatal (P1–P3) WT (A) and

a9 KO mice (B). Insets represent 5–30 superimposed individual current traces.

(C and D) Examples of minimal and maximal synaptic responses in LSO neurons of P10–P11 WT (C) and a9 KO mice (D).

(E) Average amplitudes of synaptic responses elicited by minimal and maximal stimulation. In WT animals, maximal responses increased from 0.75 ± 0.12 nA

at P1–P3 (n = 26) to 4.23 ± 0.54 nA at P10–P12 (n = 42), and minimal responses increased from 30 ± 4 pA at P1–P3 (n = 22) to 470 ± 93 pA at P10–P12 (n = 43).

In a9 KO mice, maximal responses increased from 0.82 ± 0.16 nA at P1–P3 (n = 18) to 3.93 ± 0.36 nA at P10–P12 (n = 57) and minimal responses from 29 ± 9 pA

at P1–P3 (n = 18) to 200 ± 52 pA at P10–P12 (n = 57). The amplitude of minimal responses at P10–P12 was significantly smaller in a9 KO mice compared to WT

mice (*p < 0.01, Student’s t test).

(F) Bootstrap analysis of MNTB-LSO convergence. In WT mice, the mean convergence decreased from 23 ± 6 MNTB fibers per LSO neuron at P1–P3 to 10 ± 2

MNTB fibers per LSO neuron at P10–P12. In KO mice, the mean convergence was 27 ± 7 MNTB fibers per LSO neuron at P1–P3 (not significantly different from

WT mice, p = 0.66, permutation test) and 21 ± 6 MNTB fibers/ LSO neuron at P10–P12, significantly higher than WT mice (p = 0.03, permutation test).

(G) Amplitude and cumulative histograms of minimal stimulation responses in WT and a9 KO mice (*p = 0.001, Kolmogorov-Smirnov test). All error bars

represent SEM.

See also Figure S4.

Neuron

Spontaneous Activity and Tonotopic Refinement

in a9 KO mice, resulting in a deficit in the sharpening of func-

tional MNTB-LSO maps.

Functional Refinement Is Accompanied by AxonalGrowth in Both WT and a9 KO MiceTopographic refinement of the MNTB-LSO pathway also

involves the pruning of MNTB axons and LSO dendrites

(Rietzel and Friauf, 1998; Sanes and Siverls, 1991). We therefore

826 Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc.

analyzed the axonal termination patterns of individual MNTB

axons in the developing LSO to investigate whether axonal prun-

ing is affected in a9 KO mice. To this end, individual MNTB

neurons in brain slices were filled with biocytin and their axonal

projections to the LSO reconstructed (Figure S5). Because

the axonal termination field and the magnitude of pruning of

MNTB-LSO axons depend on the tonotopic (mediolateral)

location (Sanes and Siverls, 1991), we restricted our analysis to

Figure 4. Growth of MNTB Axon Terminals

and Formation of New Boutons before

Hearing Onset

(A and B) Example of reconstructed MNTB axon

terminals in the medial LSO (outlined in blue) in

WT mice (A) and a9 KO mice (B). Lower portions

illustrate bouton locations of corresponding axons.

Bouton clouds are fit with an ellipse that maximizes

enclosed bouton density (see Experimental Pro-

cedures). D, dorsal; L, lateral. Scale bar represents

100 mm.

(C) Increase in the average number of boutons per

MNTB axon in the LSO before hearing onset is

similar in WT and a9 KO mice (P2–P4: 127 ± 13 in

WT versus 112 ± 12 in KO, p = 0.39, Student’s t test;

P12–P14: 301 ± 29 in WT versus 324 ± 26 in KO,

p = 0.55, Student’s t test).

(D) The average bouton area, normalized to LSO

cross-sectional area, does not change before

hearing onset and does not differ between WT and

a9 KOmice at either age (P2–P4: 19.7 ± 1.7% in WT

versus 15.5 ± 2.2% in KO, p = 0.14, Student’s t test;

P12–P14: 19.6 ± 1.1% in WT versus 21.3 ± 2.6% in

KO, p = 0.54, Student’s t test).

(E) The average width of boutons along the LSO

tonotopic axis, normalized to LSO length, does not

change before hearing onset and does not differ

between WT and a9 KO mice at either age (P2–P4:

23.3 ± 2.3% in WT versus 20.6 ± 1.5% in KO, p = 0.33, Student’s t test; P12–P14: 20.7 ± 1.1% in WT versus 23.1 ± 2.1% in KO, p = 0.28, Student’s t test).

Errors bars represent SEM. P2–P4: n = 9 axons from six WT animals and nine axons from eight a9 KO animals. P12–P14: n = 10 axons from eight WT animals and

eight axons from six a9 KO animals. See Figure S6 for absolute values.

Neuron

Spontaneous Activity and Tonotopic Refinement

MNTB neurons that were located in the medial, high-frequency

part of the MNTB (Table S2).

During the first 2 postnatal weeks, there was a substantial

growth of MNTB-LSO axons in WT mice, and the average num-

ber of boutons per MNTB-LSO axon increased approximately

2.5-fold (Figure 4). To quantitatively and objectively estimate

the topographic specificity of axon terminals, an ellipse that

maximized the density of enclosed boutons was fit to the two-

dimensional projection of the boutons. The ellipse area repre-

sents the area innervated by a single MNTB axon (bouton

area), whereas the length of the short axis of the ellipse provides

a measure of the spread of boutons along the tonotopic axis

(bouton spread).

In WT mice, the bouton area increased almost 2-fold before

hearing onset (p < 0.01, Student’s t test; Figure S6A), and bouton

spread along the tonotopic axis increased approximately 1.4-

fold (p < 0.01, Student’s t test; Figure S6B). This expansion of

the bouton area, however, was matched by an almost 2-fold

expansion of the cross-sectional area of the LSO (Figure S6C)

and as a result, bouton area and bouton spread, when normal-

ized to the LSO cross-sectional area and length, respectively,

remained unchanged before hearing onset (Figures 4D and 4E;

p = 0.99 for normalized area; p = 0.29 for normalized spread,

Student’s t test). Thus, despite the considerable refinement of

functional MNTB-LSO maps, on an anatomical level, the degree

of topographic accuracy remains constant before hearing onset.

In a9 KOmice, the average number of boutons perMNTB-LSO

axon increased 2.9-fold before hearing onset, similar to the

increase in WT mice, resulting in bouton numbers at P12–P14

that did not differ significantly between genotypes (p = 0.55; Fig-

ures 4A–4C). In addition, the cross-sectional area of the LSO,

normalized bouton area, and normalized bouton spread at either

P2–P4 or P12–P14 did not differ between genotypes (Figures 4

and S6; p > 0.1). This indicates that the growth of MNTB axons

and the expansion of the LSO before hearing onset were normal

in a9 KO mice despite their deficits in functional refinement.

To investigate whether the addition of new boutons occurred

in a random or in a spatially guided manner, we determined the

density of boutons across the termination field during prehearing

development. In P2–P4 mice, the average bouton density was

relatively flat across the termination area, with only a slight in-

crease toward the center (bouton centroid; Figures 5A and 5B).

Over the next 2 weeks, new boutons were added preferentially

to the center of the termination field (Figure 5B), resulting in a

prominent bouton-dense peak at hearing onset. A similar pattern

of bouton formation was also present in a9 KO mice, yielding

a spatial distribution of boutons that closely resembles that of

WT mice at hearing onset (Figures 5C and 5D). These results

indicate that boutons are primarily added to the tonotopic center

of the termination area and that the spatial pattern of bouton

addition is not altered in a9 KO mice.

Axonal Pruning Occurs after Hearing Onset in WT Miceand Is Absent in a9 KO MiceDuring the first week after hearing onset, the number of boutons

per MNTB axon decreased by �50% in WT mice (p < 0.001;

Figure 6). This pruning was associated with a 50% decrease in

bouton area (p < 0.001; Figures 6 and 7A) and a 35% decrease

Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc. 827

Figure 5. Changes in the Spatial Distribution

of Boutons Formed by Single Axons before

Hearing Onset

(A) Heat maps of the average density of boutons

of individual MNTB axons in the LSO. Color bar

represents average number of boutons per axon

per bin (0.08% LSO area). Scale bars represent

20% LSO length and 40% dorsoventral LSO

height.

(B) Histograms of the average distribution of

boutons per MNTB axon along the LSO frequency

axis in WT mice at P2–P4 (gray) and P12–P14

(black). The x-axis represents bouton positions

with respect to the centroid of the bouton fields,

and the bin size was 2% of the LSO length.

Asterisks (*) mark the bins where the mean number

of boutons significantly differed (p < 0.05, Stu-

dent’s t test) between the P2–P4 (n = 9 axons) and

P12–P14 (n = 10 axons) WT groups. Error bars

represent SEM.

(C) Heat maps for a9 KO mice.

(D) Histograms for a9 KO mice. Asterisks (*) mark

the bins that are significantly different between

the P2–P4 (magenta, n = 9 axons) and P12–P14 (red, n = 8 axons) KO groups. The histogram for WT P12–P14 from (B) is also plotted for comparison (black). The

mean bouton numbers did not significantly differ between genotypes at P12–P14 (black versus red) at any bins (p > 0.1, Student’s t test).

See Figure S6 for absolute values.

Neuron

Spontaneous Activity and Tonotopic Refinement

in bouton spread along the tonotopic axis (p = 0.002). The bouton

elimination occurred not only at the periphery, but also at the

very center of the distribution (Figure 7B, gray versus black). In

fact, shifting the distribution curve at P12–P14 by subtracting

a constant number of boutons yielded a close match for the dis-

tribution at P19–P21 (Figure 7B, thick gray line versus black;

correlation coefficient = 0.84). These results indicate that bouton

elimination is not biased toward ‘‘inappropriate’’ boutons at

the periphery of the termination area, but rather occurs homo-

geneously across the entire termination field, a process best

described by a ‘‘sinking iceberg’’ scenario.

Surprisingly, in a9 KO mice, which have no obvious hearing

deficits (May et al., 2002; Vetter et al., 1999), the pruning of

MNTB-LSO boutons after hearing onset was absent (Figures

6B–6E). Bouton areas in a9 KO mice were over twice as large

as in WT mice and boutons were distributed over twice the

distance along the LSO frequency axis (Figures 6 and 7; area:

p < 0.0001, width: p = 0.002, Student’s t test). The spatial distri-

bution of boutons along the tonotopic axis further demonstrates

the difference between genotypes at P19–P21 (Figure 7D, black

versus red). The cross-sectional area and length of the LSO in a9

KO mice was not different from that in WT mice (Figure S6C;

area, p = 0.16; length, p = 0.69; Student’s t test). Thus, in a9

KO mice, sharpening of the structural MNTB-LSO map was

virtually absent, resulting in the maintenance of an immature,

less precise tonotopic map.

Although a9 KO mice have seemingly normal hearing as

defined by threshold and gain of compound action potentials

(Vetter et al., 1999), auditory brainstem responses, tone detec-

tion, and intensity discrimination (May et al., 2002), neurons in

the cochlear nucleus, MNTB or LSO of a9 KO mice may still

have subtle changes in their spontaneous or sound-evoked

activity, which may interfere with the pruning process. We thus

reasoned that if pruning of the MNTB-LSO pathway is indeed

828 Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc.

sensitive to slight changes in activity patterns, then pruning

should be altered in mice in which sound-evoked activity is sub-

stantially altered. To test this possibility, we raised mouse pups

from P8 until P19–P21 in pulsed white noise, an acoustic envi-

ronment that severely disrupts the maturation of tonotopic

maps in the primary auditory cortex (Zhang et al., 2002) and

the frequency tuning of neurons in the inferior colliculus (Sanes

and Constantine-Paton, 1983). As shown in Figure 8, pulsed

noise-reared mice displayed normal pruning of MNTB-LSO

that was indistinguishable from that in WT mice. The distance

of filled MNTB neurons to the medial edge of the MNTB did

not differ between WT and noise-reared mice (p = 0.31,

Student’s t test), indicating that we sampled MNTB neurons

from a similar frequency region. These results further support

the notion that pruning of MNTB-LSO axons does not depend

on normal spatiotemporal patterns of sound-evoked activity

and argues against the possibility that subtle changes in post-

hearing activity in a9 KO mice underlie the absence of pruning

in these mice.

No Obvious Deficits in Cochlear Nucleus Projectionsto the MNTB in a9 KO MiceFinally, we addressed the possibility that in a9 KO mice the

projection from the cochlear nucleus (CN) to the MNTB is

disturbed, which then may ‘‘propagate’’ to the MNTB-LSO

pathway.MNTB neurons receive their excitatory input from glob-

ular bushy cells that are located exclusively in the contralateral

cochlear nucleus (Kuwabara et al., 1991; Tolbert et al., 1982).

Globular bushy cell axons give rise to a single, large glutamater-

gic synapse, the calyx of Held, which contacts a single MNTB

neuron. This typical connection pattern persists in congenitally

deaf mice (Youssoufian et al., 2008), but cochlear ablation in

neonatal animals disrupts the strict contralateral innervation of

the MNTB (Kitzes et al., 1995). Using in vitro biocytin tracing

Figure 6. Pruning of MNTB Axon Terminals

and Elimination of Boutons after Hearing

Onset

(A) Example of reconstructed MNTB axon terminals

in the LSO (outlined in blue) in WT mice and (B) a9

KO mice. Lower portions illustrate bouton locations

of corresponding axons. D, dorsal; L, lateral. Scale

bar represents 100 mm.

(C) Average number of boutons per MNTB axon

in the LSO decreases in WT mice, but not in a9

KO mice (WT: 302 ± 29 boutons/MNTB axon at

P12–P14 to 137 ± 21 boutons at P19–P21, p <

0.001, Student’s t test; KO: 325 ± 26 boutons/MNTB

axon at P12–P14 to 278 ± 28 at P19–P21, p = 0.25,

Student’s t test; WT versus KO at P19–P21: p =

0.001, Student’s t test).

(D) Average bouton area normalized to LSO cross-

sectional area decreases in WT mice, but not in a9

KO mice (WT: 20% ± 1% LSO cross-sectional area

at P12–P14 to 10% ± 2% at P19–P21, p < 0.001,

Student’s t test; KO: 21% ± 3% LSO cross-

sectional area at P12–P14 to 22%±1%at P19–P21,

p = 0.75, Student’s t test; WT versus KO at P19–

P21: p < 0.0001).

(E) Average width of bouton area along LSO tono-

topic axis decreases in WT mice, but not in a9 KO

mice (WT: 21% ± 1% LSO length at P12–P14 to

14% ± 2% at P19–P21, p = 0.002, Student’s t test;

KO: 23% ± 2% LSO length at P12–P14 to 27% ± 3% at P19–P21, p = 0.34, Student’s t test; WT versus KO at P19–P21: p = 0.002).

Errors bars represent SEM. At P12–P14, n = 10 axons from eight animals in WT mice and n = 10 axons from six animals in a9 KO mice. At P19–P21, n = 9 axons

from eight animals in WT mice and n = 9 axons from eight animals in a9 KO mice. See Figure S6 for absolute values.

Neuron

Spontaneous Activity and Tonotopic Refinement

(P10–P14; 17 KO and 9 WT mice), we found no evidence for

abnormal CN projections or altered calyx morphology in a9 KO

mice (Figure 9). In both WT and a9 KO mice, CN axons inner-

vated the ipsilateral but not contralateral LSO, and calyces

were present exclusively in the contralateral MNTB. In both

genotypes, we observed the occasional retrogradely labeled

MNTB neuron as reported previously in other species (Schofield,

1995; Winter et al., 1989). Calyx morphology in a9 KO mice ap-

peared normal, showing the typical engulfment of postsynaptic

neurons with several typical finger-like processes (Figure 9B;

Morest, 1968; Kandler and Friauf, 1993; Hoffpauir et al., 2006).

DISCUSSION

Before hearing onset, neuronal activity in the auditory system is

characterized by stereotypical patterns of spiking activity, yet

the role of these patterns in the formation of precise auditory

circuits has remained speculative. In this study, we addressed

this question by characterizing the functional and anatomical

emergence of precise tonotopy in the inhibitory MNTB-LSO

pathway of a9 KO mice, which lack the expression of a9 AChR

subunits in cochlear hair cells, and thus lack classic olivoco-

chlear feedback activity (Vetter et al., 1999). In vivo recordings

from MNTB neurons in a9 KO mice demonstrated that the

functional denervation of efferent cholinergic transmission had

no effect on average spontaneous spike rates or overall burst-

activity but that it distinctively changed the temporal pattern of

spontaneous spikes. Compared to WT mice, LSO neurons in

a9 KO mice exhibited reduced strengthening of single-axon

MNTB-LSO connections, retained more MNTB inputs, and had

an impaired refinement of functional MNTB-LSO maps before

hearing onset. Although axonal growth and the formation of

boutons were normal in a9 KO mice before hearing onset, the

pruning of MNTB axon branches, which in WT mice occurred

after hearing onset, was virtually absent in a9 KOmice. These re-

sults provide strong evidence that the precise temporal pattern

of spontaneous prehearing activity is crucial for the formation

of precise tonotopy in a central auditory pathway.

Cholinergic Modulation of Spontaneous ActivityImmature IHCs are able to generate calcium action potentials

(Kros et al., 1998; Marcotti, 2012), which are elicited or modu-

lated by ATP release from Kolliker’s organ, a transient structure

in the immature cochlea (Johnson et al., 2011, 2012; Tritsch

et al., 2007). The firing of IHCs is also modulated by acetylcho-

line, which is released from the terminals of medial olivocochlear

neurons, a cholinergic cell group located in the brainstem that

transiently innervates IHCs (Simmons et al., 1996; Warr and

Guinan, 1979). Acetylcholine hyperpolarizes IHC by activating

a9 and a10 subunit-containing nicotinic AChRs (Elgoyhen

et al., 1994; Katz et al., 2004; Vetter et al., 1999), leading to

calcium influx and rapid activation of calcium-dependent small

conductance potassium (SK2) channels (Fuchs and Murrow,

1992; Glowatzki and Fuchs, 2000; Johnson et al., 2011; Kros

et al., 1998). The role of the transient cholinergic innervation

of immature IHCs has remained poorly understood, although

it has been suggested that it modulates spontaneous activity

levels or patterns (Glowatzki and Fuchs, 2000; Johnson et al.,

Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc. 829

Figure 7. Changes in the Spatial Distribution

of Boutons after Hearing Onset

(A) Heat maps of average density of the boutons

of individual MNTB axons in the LSO of WT mice.

Color bar represents the average number of bou-

tons per axon per bin (0.08% LSO area). Scale bars

represent 20% LSO length and 40% dorsoventral

LSO height.

(B) Histograms of the average distribution of

boutons per MNTB axon along the LSO frequency

axis at P12–P14 (gray circles) and P19–P21 (black

squares). The x-axis represents bouton positions

with respect to the centroid of the bouton fields, and

the bin size was 2% of the LSO length. Asterisks (*)

mark the bins where the mean number of boutons

significantly differed (p < 0.05, Student’s t test)

between the P12–P14 (n = 10 axons) and P19–P21

(n = 9 axons) WT groups. Also shown is the down-

shifted P12–P14 histogram (thick gray line) to match

the peak of the P19–P21 distribution. Error bars

represent SEM.

(C) Heat maps for a9 KO mice.

(D) Histograms for a9 KOmice showing the P12–P14 (magenta, n = 8 axons) and P19–P21 (red, n = 9 axons) groups, which did not differ significantly at any bins.

The histogram for the WT P19–P21 group from (B) is also plotted for comparison (black). The asterisks on the black curve mark the bins where the mean bouton

numbers significantly differed between genotypes at P19–P21 (black versus red; Student’s t test, p < 0.05).

Neuron

Spontaneous Activity and Tonotopic Refinement

2011). Results from cochlear explants have provided evidence

both for and against this hypothesis (Johnson et al., 2011; Tritsch

and Bergles, 2010; Tritsch et al., 2007). Our in vivo recordings

from MNTB neurons in a9 KO mice revealed normal average

spike rates, typical bursting patterns, and normal acceleration-

deceleration spike train dynamics (Figure 1), indicating these

properties emerge independent of cholinergic transmission

(Tritsch and Bergles, 2010; Tritsch et al., 2007). Our results,

however, reveal that acetylcholine modulates burst duration

(50% decrease in a9 KO mice) and intraburst firing rates

(70%–80% increase in in a9 KOmice). The mechanisms respon-

sible for these changes remain to be determined but may include

the loss of a tonic acetylcholine-mediated hyperpolarization of

IHCs, resulting in an increased spiking during bursts (Johnson

et al., 2011) or a deficit in the maturation of vesicle fusion in

IHCs (Johnson et al., 2013).

One could argue that the altered spike patterns of MNTB

neurons in a9 KO mice result from changes in the CN or the

MNTB itself, rather than from changes in the cochlea. However,

this possibility seems rather unlikely due to the following argu-

ments: First, the pattern of spontaneous activity in the MNTB

of WT mice (Figure 1) is strikingly similar to the activity of

MNTB neurons in rats, where it closely resembles the spike

pattern of spiral ganglion cells (Tritsch et al., 2010), indicating a

faithful transmission of spikes from the auditory nerve to the

MNTB even at the young ages examined. Second, nicotinic a9

AChR subunits are not expressed in the brain and, within the

auditory system, are expressed only in the cochlea (Allen Insti-

tute for Brain Science, 2012; Elgoyhen et al., 1994; Zuo et al.,

1999), arguing against altered cholinergic transmission at CN

neurons (Brown et al., 1988; Brown and Vetter, 2009; Fujino

and Oertel, 2001) in a9 KO mice. Third, glutamate sensitivity

and glutamate-elicited spike patterns of MNTB neurons in a9

KO mice were normal (Figure S2), as were the afferent pathways

(Figure 9). Taken together, these results make it highly likely that

830 Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc.

the specific changes in the temporal firing pattern of MNTB neu-

rons in a9 KOmice reflect changes in the activity pattern of spiral

ganglion neurons.

Impaired Sharpening of Functional MNTB-LSO Maps ina9 KO MiceThe present study provides strong evidence that even subtle

changes in the temporal pattern of spontaneous activity before

hearing onset lead to profound deficits in the tonotopic precision

of an inhibitory auditory map. Although changes beyond altered

activity patterns within the cochlea cannot be excluded at this

point, the most likely way that a genetic manipulation specific

to cochlea hair cells could influence connections in the LSO

(four synapses downstream from IHCs) is through activity prop-

agated along the auditory pathway.

It remains to be shown why shorter bursts with higher spike

frequencies interfere with the developmental strengthening of

MNTB-LSO connections. Because the rise and decay times of

postsynaptic currents were normal in a9 KO mice, it is unlikely

that the smaller single-fiber responses are caused bymore distal

locations of dendritic inputs, reduced synchronicity of neuro-

transmitter release, or changes in GABA and glycine response

components (Hirtz et al., 2012; Kapfer et al., 2002; Leao et al.,

2004). In contrast to single-fiber responses, maximal MNTB-

LSO responses increased normally in a9 KO mice, suggesting

that the developmental increase in the amount of inhibition in

the LSO is resilient to abnormal temporal activity patterns. In

line with this, the amount of inhibition in deaf otoferlin mice

(Longo-Guess et al., 2007; Noh et al., 2010; Roux et al., 2006)

or dn/dn mice (Couchman et al., 2011; Leao et al., 2006) is

unchanged. However, in these genetically deaf mice, both

cochleae are affected equally, leaving the excitation-inhibition

balance in the LSO unchanged. In contrast, disturbing the exci-

tation-inhibition balance, as is the case after unilateral cochlear

ablation (Kotak and Sanes, 1996) or by blocking glutamate

Figure 8. RearingMice in PulsedWhite Noise

Does Not Prevent Axonal Pruning after Hear-

ing Onset

(A) Example of reconstructed MNTB axon terminals

in the LSO in WT mice (left; same axon shown in

Figure 6) and noise-reared WT (WT-NR) mice (right)

at P19–P21. Lower portions illustrate bouton loca-

tions of corresponding axons. D, dorsal; L, lateral.

Scale bar represents 100 mm.

(B) Average number of boutons per MNTB axon in

the LSO does not differ between WT and WT-NR

mice at P19–P21 (137 ± 21 boutons/MNTB axon in

WT compared to 166 ± 14 boutons/MNTB axon in

WT-NR, p = 0.26, Student’s t test), but is signifi-

cantly reduced compared to WT mice at P12–P14

(301 ± 29 boutons/MNTB axon, p < 0.0005 for both

WT and WT-NR).

(C) Average bouton area normalized to LSO cross-

sectional area also does not differ between WT and

WT-NR mice at P19–P21 (10% ± 2% LSO area in

WT compared to 11%± 1%LSO area inWT-NR, p =

0.34, Student’s t test), but is significantly reduced

compared to WT mice at P12–P14 (20% ± 1% LSO

area, p < 0.0005 for both WT and WT-NR).

(D) Average width of bouton area along LSO tono-

topic axis also does not differ between WT and

WT-NR mice at P19–P21 (14% ± 2% LSO length in

WT compared to 14% ± 1% LSO length in WT-NR,

p = 0.71, Student’s t test), but is significantly

reduced compared to WT mice at P12–P14 (21% ±

1% LSO length, p < 0.002 for both WT andWT-NR).

Errors bars represent SEM. n = 9 axons from eight animals in WT mice and n = 14 axons from 11 animals in WT-NR mice.

(E) Heat maps of average density of the boutons of individual MNTB axons in the LSO of WT (left) and WT-NR (right) mice at P19–P21. Color bar represents the

average number of boutons per axon per bin (0.08% LSO area). Scale bars represent 20% LSO length and 40% dorsoventral LSO height.

(F) Histograms of the average distribution of boutons per MNTB axon along the LSO frequency axis at P19–P21 in WT (black circles) andWT-NR (green squares).

The x-axis represents bouton positions with respect to the centroid of the bouton fields, and the bin size was 2%of the LSO length. Error bars represent SEM. The

mean bouton numbers did not significantly differ between WT and WT-NR at any of the bins (Student’s t test, p > 0.05).

Neuron

Spontaneous Activity and Tonotopic Refinement

cotransmission at MNTB-LSO synapses (Noh et al., 2010), leads

to changes in the overall amount of inhibition in LSO neurons.

These results support the notion that the amount of inhibition

in the LSO is regulated by homeostatic plasticity rules whereas

the strength of single-fiber inhibition is adjusted by the precise

timing of inhibitory synaptic activity.

Distinct Developmental Periods for Functional andStructural RefinementActivity-dependent changes in synaptic function usually are

translated rapidly into changes in structural connectivity (Anto-

nini and Stryker, 1993; Colman et al., 1997; Keck et al., 2011;

Knott et al., 2002; Mikuni et al., 2013; Trachtenberg and

Stryker, 2001; Yamahachi et al., 2009). In the MNTB-LSO

pathway, however, functional and structural refinement are

clearly segregated and occur during distinct developmental

stages. Before the onset of hearing, synaptic silencing and

strengthening sharpen functional connectivity, yet at the same

time MNTB-LSO axons grow extensively to compensate for a

growth of the LSO, maintaining a constant level of topographic

precision on a structural level. This scenario predicts that

silenced boutons remain present until the onset of hearing

and that the majority of newly formed boutons remain silent.

In contrast, axonal pruning and sharpening occur exclusively

after hearing onset (Figure 6), when synaptic strengthening

and silencing has already ceased (Walcher et al., 2011). The

segregation of functional and structural refinement may indi-

cate that functional and anatomical refinement each requires

synaptic properties or activity patterns that are only present

before or after hearing onset such as corelease of glutamate

or GABA from MNTB terminals (Case and Gillespie, 2011;

Gillespie et al., 2005; Kotak et al., 1998; Nabekura et al.,

2004, Noh et al., 2010) or an excitatory action of GABA/glycine

(Kakazu et al., 1999; Kandler and Friauf, 1995; Kullmann et al.,

2002). Linking these transient properties to specific processes

in the functional and structural refinement of the MNTB-LSO

pathway will be important for understanding tonotopic refine-

ment of auditory pathways and the developmental emergence

of precise inhibitory circuits in other brain regions.

Precise MNTB-LSO Maps Emerge through DirectedAxon Growth followed by Nondirected Axonal PruningPrecise neuronal maps generally emerge via a two-stage pro-

cess consisting of an initial exuberant growth of axons followed

by the specific pruning of ‘‘incorrect’’ axonal branches, which is

guided by patterned neuronal activity (Nakamura and O’Leary,

1989; Shatz, 1996). Topographic sharpening of the MNTB-LSO

pathway deviates from this scheme because the addition of

Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc. 831

Figure 9. Projections from the Cochlear Nu-

cleus to the Superior Olivary Complex Appear

Normal in a9 KO Mice

(A) Pattern of cochlear nucleus projections to the

ipsilateral and contralateral superior olivary com-

plex in a P13 WT mouse and P12 a9 KO mouse

following insertion of a biocytin crystal into the left

cochlear nucleus. In both genotypes, labeled axon

terminals were predominantly found in the ipsilateral

LSO and contralateral MNTB. A fewMNTB neurons,

but never calyces, were observed in the ipsilateral

MNTB in both genotypes.

(B) Normal morphology of calyx of Held in a9 KO

mice. An example of rhodamine dextran-labeled

calyces of Held in P14WT and a9 KOmice is shown.

Each image is a maximum intensity projection of a

stack of two-photon optical sections. Scale bars

represent 5 mm.

Neuron

Spontaneous Activity and Tonotopic Refinement

new boutons before hearing onset is directed to the ‘‘correct’’

area (center of the termination field; Figure 5) whereas subse-

quent bouton elimination is nonspecific, following a ‘‘sinking

iceberg’’ mechanism (Figure 7). Because nondiscriminatory

pruning does not necessitate guidance by instructive signals

derived from sound-evoked activity, pruning should be resilient

to abnormal patterns of hearing-evoked activity. In support of

this, noise-reared animals show normal pruning (Figure 8). It

remains to be shown whether directed growth and unguided

pruning are features unique to the LSO or related to the inhibitory

nature of the MNTB-LSO pathway.

Structural Refinement in a9 KO MiceAxonal growth and bouton addition were normal a9 KO mice,

indicating that the amount and spatial specificity of axon

growth is insensitive to the altered temporal pattern of sponta-

neous activity in a9 KO mice. Because functional map refine-

ment was impaired in a9 KO mice, it appears that functional

and structural refinement are differentially regulated by sponta-

neous activity before hearing onset. Axonal pruning after hear-

ing onset, however, was severely impaired in a9 KO mice

(Figures 6 and 7). While it remains to be shown whether

these deficits persist throughout adulthood, the lack of pruning

was unexpected because a9 KO mice lack any obvious hearing

832 Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc.

deficits (Vetter et al., 1999, May et al.,

2002). It is possible that neurons in the

MNTB-LSO pathway or in upstream areas

have subtle changes in sound-evoked or

spontaneous activity, which may prevent

pruning. However, because pruning was

normal in noise-reared animals, possible

changes in sound-evoked or spontaneous

activity a9 KO mice after hearing onset

would need to be dramatic to disrupt

pruning and thus would have disrupted

normal hearing (Vetter et al., 1999, May

et al., 2002). Our results rather support

the interpretation that structural refine-

ment after hearing onset depends on the

refinement of synaptic strength and functional maps before

hearing onset.

EXPERIMENTAL PROCEDURES

Animals and Preparation

Experiments were performed on a9 KO mice (on background 129S6/SvEv)

or WT mice (129S6/SvEv) (Vetter et al., 1999). Experimental procedures

were in accordance with National Institutes of Health guidelines and were

approved by the Institutional Animal Care and Use Committee at the

University of Pittsburgh and by the Saxonian District Government Leipzig

(TVV 06/09).

In Vivo Recording and Burst Analysis

Single units were recorded from the MNTB in P8 mice under xylazine hydro-

chloride anesthesia (Sonntag et al., 2009). MNTB principal neurons were

identified by the complex waveform and recording sites were also verified

histologically. Mini-bursts and bursts were defined based on the population

ISI histogram (Figure S1). Within a burst, spikes that did not form mini-bursts

were analyzed separately as single spikes (Figures 1K and 1L). In Figure 1M,

we refer to mini-bursts and single spikes together as ‘‘spike clusters’’ to

analyze them together in a single temporal sequence.

In Vitro Electrophysiological Recordings and Functional Mapping

Whole cell recordings were obtained from neurons in the medial LSO in

300 mm thick slices with recording electrodes (2–3 MOhm) containing (in

mM) 76 Cs-methanesulfonate, 56 CsCl, 10 EGTA, 1 MgCl2, 1 CaCl2, 2

Neuron

Spontaneous Activity and Tonotopic Refinement

ATP-Mg, 0.3 GTP-Na, 5 Na2-phosphocreatine, and 10 HEPES (Kim and Kan-

dler, 2003). For current clamp recordings (Figure S2), Cs-methanesulfonate

was replaced with K-gluconate. MNTB-LSO connections were mapped by

photolysis of p-hydroxyphenacyl-glutamate (Givens et al., 1997) or CNB-

glu (150–200 mM; Invitrogen) using a fiber optic-based system (Kandler

et al., 2013) and 20 ms (P1–P3) or 100 ms (P10–P12) light pulses. For minimal

and maximal stimulation, MNTB axons were stimulated with constant current

pulses (0.2 ms) at a rate of 0.2 Hz. The experimenter was blind to the geno-

type of the animals until data analysis was completed. Convergence ratios

were determined using bootstrap resampling of maximal and minimal PSC

amplitudes and statistical significance was determined with a permutation

test (Figure 3F).

Noise-Rearing Protocol

Mice were reared from P8 until P19–P21 in pulsed white noise (70 dB SPL) in a

sound-attenuating chamber (Coulbourn Instruments).White noise pulse length

was pseudorandomly set to 700–1400 ms, and pulses were pseudorandomly

presented at a rate of �36 per minute.

Axon and Bouton Analysis

MNTB neurons were filled with patch pipettes containing 0.1 mM Alexa Fluor

568 hydrazide (Invitrogen) and 0.5% biocytin, followed by standard biocytin

histochemistry. Axons and boutons within the LSO were reconstructed using

a Neurolucida system (MBF Bioscience). Cells with cut axon branches within

the LSO were discarded from analysis. Bouton area was determined by fitting

an ellipse that maximized the density of enclosed boutons. Analysis using

bouton areas determined by conventional approaches (connecting peripheral

boutons) yielded qualitatively and statistically similar results. Density heat

maps were created by overlaying boutons with a two-dimensional grid

comprised of 56.8 mm2 (at P2–P4) or 100 mm2 (at P12–P14 and P19–P21)

bins. The chosen bin size encompassed a constant proportion of LSO area

(�0.08%) across ages. Bouton numbers per bin were calculated for each

axon and then averaged across axons. Throughout the paper, results are

presented as mean ± SEM, and a Student’s t test was used unless stated

otherwise.

In Vitro Tracing

Following decapitation and removal of the brain, a small crystal of biocytin

(Sigma-Aldrich) was inserted into the cochlear nucleus (Kandler and Friauf,

1993). Brains were incubated in artificial cerebrospinal fluid for 3–4 hr, fixed

for 2–5 days (4% parafomaldehyde), and processed for biocytin (Vectastain

ABC kit). Calyceal axons were labeled in 300 mm thick living slices by in-

serting a crystal of rhodamin dextrane (Sigma-Aldrich) into the acoustic

decussation between both MNTBs. After 30–60 min, slices of the calyces

were imaged with a two-photon microscope (VIVO 2-photon, Intelligent

Imaging Innovations) with excitation at 810 nm and z axis spacing of

0.5 mm.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures,

six figures, and two tables and can be found with this article online at http://

dx.doi.org/10.1016/j.neuron.2014.04.001.

AUTHOR CONTRIBUTIONS

A.C. conducted and analyzed anatomical experiments shown in Figures 4, 5,

6, 7, 8, S5, and S6. G.K. performed and analyzed all in vitro recordings shown

in Figures 2, 3, and S1–S4. M.S. and R.R. performed all in vivo single unit

recordings. G.K, M.S., and R.R analyzed the in vivo data shown in Figures 1

and S1. D.E.V. provided a9 KO founder mice and consulted on breeding and

genotyping. C.J.C.W. performed and analyzed anatomical experiments shown

in Figure 9. All authors contributed to corresponding sections of the manu-

script. G.K., A.C, and K.K. conceptualized and planned the study and wrote

the paper.

ACKNOWLEDGMENTS

We thank Dr. Richard Givens (Department of Chemistry, University of Kansas)

for his generous gift of p-hydroxyphenacyl-glutamate, Jessica Garver and

Xinyan Gu for technical support, and Dr. Tuan Nguyen for help with the

noise-rearing experiments. The work was supported by US National Institute

on Deafness and Other Communication Disorders grant 04199 (to K.K.) and

06258 (to D.E.V.), NIH Basic Neuroscience Predoctoral Training grant T32

NS007433 (to A.C.), National Science Foundation IGERT Training grant DGE

0549352 (to A.C.), National Institute on Deafness and Other Communication

Disorders Institutional Training grant T32 DC011499 (to C.J.C.W.), a Pennsyl-

vania Lions Hearing Research Foundation grant (to K.K.), Deutsche For-

schungsgemeinschaft Ru 390/19-1 (to R.R.), and GK 1097 (to M.S.).

Accepted: March 24, 2014

Published: May 21, 2014

REFERENCES

Allen Institute for Brain Science. (2012). Allen developing mouse brain atlas.

http://www.brain-map.org.

Antonini, A., and Stryker, M.P. (1993). Rapid remodeling of axonal arbors in the

visual cortex. Science 260, 1819–1821.

Brown, M.C., and Vetter, D.E. (2009). Olivocochlear neuron central anatomy is

normal in alpha 9 knockout mice. J. Assoc. Res. Otolaryngol. 10, 64–75.

Brown, M.C., Liberman, M.C., Benson, T.E., and Ryugo, D.K. (1988).

Brainstem branches from olivocochlear axons in cats and rodents. J. Comp.

Neurol. 278, 591–603.

Cao, X.J., McGinley, M.J., and Oertel, D. (2008). Connections and synaptic

function in the posteroventral cochlear nucleus of deaf jerker mice. J. Comp.

Neurol. 510, 297–308.

Case, D.T., and Gillespie, D.C. (2011). Pre- and postsynaptic properties of glu-

tamatergic transmission in the immature inhibitory MNTB-LSO pathway.

J. Neurophysiol. 106, 2570–2579.

Colman, H., Nabekura, J., and Lichtman, J.W. (1997). Alterations in synaptic

strength preceding axon withdrawal. Science 275, 356–361.

Couchman, K., Garrett, A., Deardorff, A.S., Rattay, F., Resatz, S., Fyffe, R.,

Walmsley, B., and Leao, R.N. (2011). Lateral superior olive function in congen-

ital deafness. Hear. Res. 277, 163–175.

Elgoyhen, A.B., Johnson, D.S., Boulter, J., Vetter, D.E., and Heinemann, S.

(1994). Alpha 9: an acetylcholine receptor with novel pharmacological proper-

ties expressed in rat cochlear hair cells. Cell 79, 705–715.

Friauf, E. (2004). Developmental changes and cellular plasticity in the superior

olivary complex. In Plasticity of the Auditory System, T.N. Parks, E.W. Rubel,

R.R. Fay, and A.N. Popper, eds. (New York: Springer), pp. 49–95.

Fuchs, P.A., and Murrow, B.W. (1992). Cholinergic inhibition of short (outer)

hair cells of the chick’s cochlea. J. Neurosci. 12, 800–809.

Fujino, K., and Oertel, D. (2001). Cholinergic modulation of stellate cells in the

mammalian ventral cochlear nucleus. J. Neurosci. 21, 7372–7383.

Gil, Z., Connors, B.W., and Amitai, Y. (1999). Efficacy of thalamocortical

and intracortical synaptic connections: quanta, innervation, and reliability.

Neuron 23, 385–397.

Gillespie, D.C., Kim, G., and Kandler, K. (2005). Inhibitory synapses in the

developing auditory system are glutamatergic. Nat. Neurosci. 8, 332–338.

Givens, R.S., Jung, A., Park, C.H., Weber, J., and Bartlett, W. (1997). New

photoactivated protecting groups. 7. p-hydroxyphenacyl - a phototrigger for

excitatory amino acids and peptides. J. Am. Chem. Soc. 119, 8369–8370.

Glowatzki, E., and Fuchs, P.A. (2000). Cholinergic synaptic inhibition of inner

hair cells in the neonatal mammalian cochlea. Science 288, 2366–2368.

Hanson, M.G., and Landmesser, L.T. (2003). Characterization of the circuits

that generate spontaneous episodes of activity in the early embryonic mouse

spinal cord. J. Neurosci. 23, 587–600.

Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc. 833

Neuron

Spontaneous Activity and Tonotopic Refinement

Hanson, M.G., and Landmesser, L.T. (2004). Normal patterns of spontaneous

activity are required for correct motor axon guidance and the expression of

specific guidance molecules. Neuron 43, 687–701.

Hashisaki, G.T., and Rubel, E.W. (1989). Effects of unilateral cochlea removal

on anteroventral cochlear nucleus neurons in developing gerbils. J. Comp.

Neurol. 283, 5–73.

Hirtz, J.J., Boesen, M., Braun, N., Deitmer, J.W., Kramer, F., Lohr, C., Muller,

B., Nothwang, H.G., Striessnig, J., Lohrke, S., and Friauf, E. (2011). Cav1.3 cal-

cium channels are required for normal development of the auditory brainstem.

J. Neurosci. 31, 8280–8294.

Hirtz, J.J., Braun, N., Griesemer, D., Hannes, C., Janz, K., Lohrke, S., Muller,

B., and Friauf, E. (2012). Synaptic refinement of an inhibitory topographic

map in the auditory brainstem requires functional Cav1.3 calcium channels.

J. Neurosci. 32, 14602–14616.

Hoffpauir, B.K., Grimes, J.L., Mathers, P.H., and Spirou, G.A. (2006).

Synaptogenesis of the calyx of Held: rapid onset of function and one-to-one

morphological innervation. J. Neurosci. 26, 5511–5523.

Johnson, S.L., Eckrich, T., Kuhn, S., Zampini, V., Franz, C., Ranatunga, K.M.,

Roberts, T.P., Masetto, S., Knipper, M., Kros, C.J., and Marcotti, W. (2011).

Position-dependent patterning of spontaneous action potentials in immature

cochlear inner hair cells. Nat. Neurosci. 14, 711–717.

Johnson, S.L., Kennedy, H.J., Holley, M.C., Fettiplace, R., and Marcotti, W.

(2012). The resting transducer current drives spontaneous activity in prehear-

ing mammalian cochlear inner hair cells. J. Neurosci. 32, 10479–10483.

Johnson, S.L., Kuhn, S., Franz, C., Ingham, N., Furness, D.N., Knipper, M.,

Steel, K.P., Adelman, J.P., Holley, M.C., and Marcotti, W. (2013).

Presynaptic maturation in auditory hair cells requires a critical period of sen-

sory-independent spiking activity. Proc. Natl. Acad. Sci. USA 110, 8720–8725.

Jones, T.A., Leake, P.A., Snyder, R.L., Stakhovskaya, O., and Bonham, B.

(2007). Spontaneous discharge patterns in cochlear spiral ganglion cells

before the onset of hearing in cats. J. Neurophysiol. 98, 1898–1908.

Kakazu, Y., Akaike, N., Komiyama, S., and Nabekura, J. (1999). Regulation

of intracellular chloride by cotransporters in developing lateral superior olive

neurons. J. Neurosci. 19, 2843–2851.

Kandler, K., and Friauf, E. (1993). Pre- and postnatal development of efferent

connections of the cochlear nucleus in the rat. J. Comp. Neurol. 328, 161–184.

Kandler, K., and Friauf, E. (1995). Development of glycinergic and glutama-

tergic synaptic transmission in the auditory brainstem of perinatal rats.

J. Neurosci. 15, 6890–6904.

Kandler, K., and Gillespie, D.C. (2005). Developmental refinement of inhibitory

sound-localization circuits. Trends Neurosci. 28, 290–296.

Kandler, K., Nguyen, T., Noh, J., and Givens, R.S. (2013). An optical fiber-

based uncaging system. Cold Spring Harbor Protoc. 2013, 118–121.

Kapfer, C., Seidl, A.H., Schweizer, H., and Grothe, B. (2002). Experience-

dependent refinement of inhibitory inputs to auditory coincidence-detector

neurons. Nat. Neurosci. 5, 247–253.

Katz, E., Elgoyhen, A.B., Gomez-Casati, M.E., Knipper, M., Vetter, D.E., Fuchs,

P.A., and Glowatzki, E. (2004). Developmental regulation of nicotinic synapses

on cochlear inner hair cells. J. Neurosci. 24, 7814–7820.

Keck, T., Scheuss, V., Jacobsen, R.I., Wierenga, C.J., Eysel, U.T., Bonhoeffer,

T., and Hubener, M. (2011). Loss of sensory input causes rapid structural

changesof inhibitory neurons in adultmousevisual cortex.Neuron71, 869–882.

Kim, G., and Kandler, K. (2003). Elimination and strengthening of glycinergic/

GABAergic connections during tonotopic map formation. Nat. Neurosci. 6,

282–290.

Kim, G., and Kandler, K. (2010). Synaptic changes underlying the strength-

ening of GABA/glycinergic connections in the developing lateral superior olive.

Neuroscience 171, 924–933.

Kirkby, L.A., Sack, G.S., Firl, A., and Feller, M.B. (2013). A role for correlated

spontaneous activity in the assembly of neural circuits. Neuron 80, 1129–1144.

Kitzes, L.M., Kageyama, G.H., Semple, M.N., and Kil, J. (1995). Development

of ectopic projections from the ventral cochlear nucleus to the superior olivary

834 Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc.

complex induced by neonatal ablation of the contralateral cochlea. J. Comp.

Neurol. 353, 341–363.

Knott, G.W., Quairiaux, C., Genoud, C., and Welker, E. (2002). Formation of

dendritic spines with GABAergic synapses induced by whisker stimulation in

adult mice. Neuron 34, 265–273.

Kotak, V.C., and Sanes, D.H. (1995). Synaptically evoked prolonged depolar-

izations in the developing auditory system. J. Neurophysiol. 74, 1611–1620.

Kotak, V.C., and Sanes, D.H. (1996). Developmental influence of glycinergic

transmission: regulation of NMDA receptor-mediated EPSPs. J. Neurosci.

16, 1836–1843.

Kotak, V.C., Korada, S., Schwartz, I.R., and Sanes, D.H. (1998). A develop-

mental shift from GABAergic to glycinergic transmission in the central auditory

system. J. Neurosci. 18, 4646–4655.

Kros, C.J., Ruppersberg, J.P., and Rusch, A. (1998). Expression of a potas-

sium current in inner hair cells during development of hearing in mice.

Nature 394, 281–284.

Kullmann, P.H., Ene, F.A., and Kandler, K. (2002). Glycinergic and GABAergic

calcium responses in the developing lateral superior olive. Eur. J. Neurosci. 15,

1093–1104.

Kuwabara, N., DiCaprio, R.A., and Zook, J.M. (1991). Afferents to the medial

nucleus of the trapezoid body and their collateral projections. J. Comp.

Neurol. 314, 684–706.

Leake, P.A., Hradek, G.T., Chair, L., and Snyder, R.L. (2006). Neonatal deaf-

ness results in degraded topographic specificity of auditory nerve projections

to the cochlear nucleus in cats. J. Comp. Neurol. 497, 13–31.

Leao, R.N., Oleskevich, S., Sun, H., Bautista, M., Fyffe, R.E., andWalmsley, B.

(2004). Differences in glycinergic mIPSCs in the auditory brain stem of normal

and congenitally deaf neonatal mice. J. Neurophysiol. 91, 1006–1012.

Leao, R.N., Sun, H., Svahn, K., Berntson, A., Youssoufian, M., Paolini, A.G.,

Fyffe, R.E., and Walmsley, B. (2006). Topographic organization in the auditory

brainstem of juvenile mice is disrupted in congenital deafness. J. Physiol. 571,

563–578.

Lippe, W.R. (1994). Rhythmic spontaneous activity in the developing avian

auditory system. J. Neurosci. 14, 1486–1495.

Longo-Guess, C., Gagnon, L.H., Bergstrom, D.E., and Johnson, K.R. (2007). A

missense mutation in the conserved C2B domain of otoferlin causes deafness

in a new mouse model of DFNB9. Hear. Res. 234, 21–28.

Marcotti, W. (2012). Functional assembly of mammalian cochlear hair cells.

Exp. Physiol. 97, 438–451.

May, B.J., Prosen, C.A., Weiss, D., and Vetter, D. (2002). Behavioral investiga-

tion of some possible effects of the central olivocochlear pathways in trans-

genic mice. Hear. Res. 171, 142–157.

Meister, M., Wong, R.O., Baylor, D.A., and Shatz, C.J. (1991). Synchronous

bursts of action potentials in ganglion cells of the developing mammalian

retina. Science 252, 939–943.

Mikuni, T., Uesaka, N., Okuno, H., Hirai, H., Deisseroth, K., Bito, H., and Kano,

M. (2013). Arc/Arg3.1 is a postsynaptic mediator of activity-dependent syn-

apse elimination in the developing cerebellum. Neuron 78, 1024–1035.

Morest, D.K. (1968). The growth of synaptic endings in the mammalian brain: a

study of the calyces of the trapezoid body. Z. Anat. Entwicklungsgesch. 127,

201–220.

Nabekura, J., Katsurabayashi, S., Kakazu, Y., Shibata, S., Matsubara, A.,

Jinno, S., Mizoguchi, Y., Sasaki, A., and Ishibashi, H. (2004). Developmental

switch from GABA to glycine release in single central synaptic terminals.

Nat. Neurosci. 7, 17–23.

Nakamura, H., and O’Leary, D.D. (1989). Inaccuracies in initial growth and

arborization of chick retinotectal axons followed by course corrections and

axon remodeling to develop topographic order. J. Neurosci. 9, 3776–3795.

Noh, J., Seal, R.P., Garver, J.A., Edwards, R.H., and Kandler, K. (2010).

Glutamate co-release at GABA/glycinergic synapses is crucial for the refine-

ment of an inhibitory map. Nat. Neurosci. 13, 232–238.

Neuron

Spontaneous Activity and Tonotopic Refinement

Rietzel, H.J., and Friauf, E. (1998). Neuron types in the rat lateral superior olive

and developmental changes in the complexity of their dendritic arbors.

J. Comp. Neurol. 390, 20–40.

Roux, I., Safieddine, S., Nouvian, R., Grati, M., Simmler, M.C., Bahloul, A.,

Perfettini, I., Le Gall, M., Rostaing, P., Hamard, G., et al. (2006). Otoferlin,

defective in a human deafness form, is essential for exocytosis at the auditory

ribbon synapse. Cell 127, 277–289.

Rubel, E.W., and Fritzsch, B. (2002). Auditory system development: primary

auditory neurons and their targets. Annu. Rev. Neurosci. 25, 51–101.

Sanes, D.H., and Constantine-Paton, M. (1983). Altered activity patterns

during development reduce neural tuning. Science 221, 1183–1185.

Sanes, D.H., and Siverls, V. (1991). Development and specificity of inhibitory

terminal arborizations in the central nervous system. J. Neurobiol. 22,

837–854.

Sanes, D.H., and Takacs, C. (1993). Activity-dependent refinement of inhibi-

tory connections. Eur. J. Neurosci. 5, 570–574.

Schofield, B.R. (1995). Projections from the cochlear nucleus to the superior

paraolivary nucleus in guinea pigs. J. Comp. Neurol. 360, 135–149.

Seal, R.P., Akil, O., Yi, E., Weber, C.M., Grant, L., Yoo, J., Clause, A., Kandler,

K., Noebels, J.L., Glowatzki, E., et al. (2008). Sensorineural deafness and sei-

zures in mice lacking vesicular glutamate transporter 3. Neuron 57, 263–275.

Shatz, C.J. (1996). Emergence of order in visual system development. Proc.

Natl. Acad. Sci. USA 93, 602–608.

Simmons, D.D., Mansdorf, N.B., and Kim, J.H. (1996). Olivocochlear innerva-

tion of inner and outer hair cells during postnatal maturation: evidence for a

waiting period. J. Comp. Neurol. 370, 551–562.

Sonntag, M., Englitz, B., Kopp-Scheinpflug, C., and Rubsamen, R. (2009).

Early postnatal development of spontaneous and acoustically evoked

discharge activity of principal cells of the medial nucleus of the trapezoid

body: an in vivo study in mice. J. Neurosci. 29, 9510–9520.

Tolbert, L.P., Morest, D.K., and Yurgelun-Todd, D.A. (1982). The neuronal ar-

chitecture of the anteroventral cochlear nucleus of the cat in the region of the

cochlear nerve root: horseradish peroxidase labelling of identified cell types.

Neuroscience 7, 3031–3052.

Trachtenberg, J.T., and Stryker, M.P. (2001). Rapid anatomical plasticity of

horizontal connections in the developing visual cortex. J. Neurosci. 21,

3476–3482.

Tritsch, N.X., and Bergles, D.E. (2010). Developmental regulation of sponta-

neous activity in the Mammalian cochlea. J. Neurosci. 30, 1539–1550.

Tritsch, N.X., Yi, E., Gale, J.E., Glowatzki, E., and Bergles, D.E. (2007). The

origin of spontaneous activity in the developing auditory system. Nature 450,

50–55.

Tritsch, N.X., Rodrıguez-Contreras, A., Crins, T.T., Wang, H.C., Borst, J.G.,

and Bergles, D.E. (2010). Calcium action potentials in hair cells pattern audi-

tory neuron activity before hearing onset. Nat. Neurosci. 13, 1050–1052.

Vetter, D.E., Liberman, M.C., Mann, J., Barhanin, J., Boulter, J., Brown, M.C.,

Saffiote-Kolman, J., Heinemann, S.F., and Elgoyhen, A.B. (1999). Role of

alpha9 nicotinic ACh receptor subunits in the development and function of

cochlear efferent innervation. Neuron 23, 93–103.

Walcher, J., Hassfurth, B., Grothe, B., and Koch, U. (2011). Comparative post-

hearing development of inhibitory inputs to the lateral superior olive in gerbils

and mice. J. Neurophysiol. 106, 1443–1453.

Warr, W.B., and Guinan, J.J., Jr. (1979). Efferent innervation of the organ of

corti: two separate systems. Brain Res. 173, 152–155.

Winter, I.M., Robertson, D., and Cole, K.S. (1989). Descending projections

from auditory brainstem nuclei to the cochlea and cochlear nucleus of the

guinea pig. J. Comp. Neurol. 280, 143–157.

Yamahachi, H., Marik, S.A., McManus, J.N., Denk, W., and Gilbert, C.D.

(2009). Rapid axonal sprouting and pruning accompany functional reorganiza-

tion in primary visual cortex. Neuron 64, 719–729.

Youssoufian, M., Oleskevich, S., and Walmsley, B. (2005). Development of a

robust central auditory synapse in congenital deafness. J. Neurophysiol. 94,

3168–3180.

Youssoufian, M., Couchman, K., Shivdasani, M.N., Paolini, A.G., and

Walmsley, B. (2008). Maturation of auditory brainstem projections and calyces

in the congenitally deaf (dn/dn) mouse. J. Comp. Neurol. 506, 442–451.

Zhang, L.I., Bao, S., and Merzenich, M.M. (2002). Disruption of primary audi-

tory cortex by synchronous auditory inputs during a critical period. Proc.

Natl. Acad. Sci. USA 99, 2309–2314.

Zuo, J., Treadaway, J., Buckner, T.W., and Fritzsch, B. (1999). Visualization of

alpha9 acetylcholine receptor expression in hair cells of transgenic mice con-

taining a modified bacterial artificial chromosome. Proc. Natl. Acad. Sci. USA

96, 14100–14105.

Neuron 82, 822–835, May 21, 2014 ª2014 Elsevier Inc. 835

Neuron, Volume 82

Supplemental Information

The Precise Temporal Pattern of Prehearing

Spontaneous Activity Is Necessary

for Tonotopic Map Refinement

Amanda Clause, Gunsoo Kim, Mandy Sonntag, Catherine J.C. Weisz, Douglas E. Vetter,

Rudolf Rűbsamen, and Karl Kandler

  1  

Supplemental Items Figure S1, related to Figure 1 Boundary ISI values used to define mini-bursts and bursts (the 2 vertical red lines). A population ISI histogram in WT mice was fit with a sum of 3 gaussians using a maximum likelihood estimation algorithm (MATLAB). Blue curves show the 3 gaussians, while the green curve is the sum of the 3 blue curves. The ISI value at the minimum between the first and the second gaussians defined the boundary for minibursts (red dashed line on the left; ISI=12 msec). The ISI value at the intersection between the second and the third gaussians defined the boundary for bursts (red dashed line on the right; ISI=1888 msec).

Rel

ativ

e fre

quen

cy

ISI (msec)100

0

0.5

1.0

101 102 103 104 105

  2  

Figure S2, related to Figure 2 and Experimental Procedures. Spatial resolution of uncaging is independent of age and genotype. (A) Examples of direct stimulation of MNTB neurons by uncaging glutamate. Open squares indicate uncaging sites with no responses, red squares indicate locations where spikes were evoked, and blue squares indicate subthrehold stimulus locations. Spikes could only be elicited by uncaging glutamate directly over the cell bodies of recorded MNTB neurons despite using longer light-pulse durations at older ages (Kim and Kandler, 2003). (B) Summary plot of spatial resolution of uncaging expressed as spike-eliciting distance from soma. Distance was measured in steps of 50 µm, and in all cases one step away from the location of the cell body failed to elicit spikes; therefore no error bars are shown. The numbers within the bars indicate the number of MNTB neurons recorded for each group. The same numbers apply to (C). (C) Average number of spikes elicited did not differ between the genotypes at both age groups (P1-3: 2.4 ± 0.2 (WT) vs 2.4 ± 0.5 (KO), p = 0.96; P10-12: 10.7 ± 1.9 (WT) vs 8.8 ± 2.6 (KO), p = 0.53).

100 m

WT

9 KO

MNTBA

B

Spik

e-e

licitin

g d

ista

nce (

µm)

P1-3 P10-12

WT

9 KO

C

Num

ber

of spik

es

P1-3 P10-12

WT

9 KO

5 7 6 12

  3  

Figure S3, related to  Figures 2 and 3. Quantification of synaptic responses evoked by glutamate uncaging shown in Figure 2. (A-B) Comparison of peak amplitudes and charge transfer of the synaptic responses between WT and KO mice at P1-3 (A) and P10-12 (B). “Per map” measurements show amplitudes and charge transfer summed over all stimulation sites of an input map. “Per stimulation location” measurements show the values at individual stimulation sites.

C

A

B

First

pe

ak a

mp

litu

de

(p

A)

WT KO

Pe

ak a

mp

litu

de

(p

A)

Ch

arg

e t

ran

sfe

r (p

C)

Pe

ak a

mp

litu

de

(p

A)

Ch

arg

e t

ran

sfe

r (p

C)

per map per stim location

Pe

ak a

mp

litu

de

(p

A)

Ch

arg

e t

ran

sfe

r (p

C)

WT KO

P1-3

WT KO WT KO WT KO

P10-12

WT KO WT KO WT KO WT KO

P10-12

Pe

ak a

mp

litu

de

(p

A)

Ch

arg

e t

ran

sfe

r (p

C)

per map per stim location

First

pe

ak a

mp

litu

de

(p

A)

per map per stim location

*

WT KO

(C) For the P10-12 group, amplitudes of the first peak of the synaptic responses are also quantified. Values for all the bar graphs are presented in Table S2 along with p-values (Student’s t test). Results: At P1-3, peak amplitudes and charge transfer did not differ between the genotypes, either when summed over stimulation sites of each map (“per map”), or not (“per stimulation location”). This is consistent with the results from electrical stimulation that amplitudes of synaptic responses to both maximal and minimal stimulations were similar between the genotypes (Figure 3E). At P10-12, peak amplitudes and charge transfer per map did not differ between the genotypes (Figure S3B), consistent with the maximal stimulation results (Figure 3E). Peak amplitudes per stimulation site did not significantly differ (Figure S3B), unlike the minimal stimulation results. In contrast to minimal stimulation, however, our glutamate uncaging likely drives multiple connected neurons to fire even at a single site, whose inputs can add up. Indeed, amplitudes of the first peak of the compound synaptic responses, which is more likely to represent an input from a single MNTB neuron, were significantly reduced in KOs, consistent with the minimal stimulation results (Figure S3C; p = 0.47). Transferred charge per stimulation site did not differ between the genotypes (Figure S3B). Even though inputs from individual MNTB neurons are smaller in KOs (Figure 3E), if uncaging tends to activate more neurons owing to a higher convergence in KOs, the summation of synaptic currents over time can increase the amount of transferred charge. Because we did not observe differences in the mean number spikes in single MNTB neurons evoked by uncaging (Figure S2C) or in the decay time of single-fiber inputs (Figure S4), it seems likely that enhanced temporal summation of MNTB inputs make up for reduced single-fiber inputs in KOs (Figure 3E and Figure S3C). In summary, our analysis of synaptic responses to uncaging generally agrees with the results from electrical stimulation. Methods: To quantify the amplitude and transferred charge of uncaging responses, synaptic currents during a 400 (P1-3) or 200 msec (P10-12) window from the onset of the UV flash were analyzed. Longer window size was necessary for the P1-3 groups because response latency and duration were often longer presumably due to a combination of factors including longer latencies of the evoked spikes, slower decay of synaptic currents, and less effective glutamate clearance mechanisms. Peak amplitudes and charge transfer (measured as the area under the synaptic traces) were measured for each response (averaged over 1-3 trials per stimulation site). The amplitude of the first peaks was measured only for the P10-12 groups because it was often not possible to unambiguously identify the first peak of the compound synaptic currents at P1-3.

  4  

  5  

Figure S4, related to Figure 3. No differences in the kinetics of single-fiber PSCs in WT and α9 KO mice. (A-B) Example traces of single-fiber PSCs in LSO neurons (averaged over 5-15 PSCs) at P2 (A) and P10 (B). The traces on the left are from WT mice, and those on the right are from KO mice. Single exponential fits for the decay phase are also shown (red). (C) 20-80% rise times of single-fiber PSCs (P1-3: WT 0.83 ± 0.07 msec, KO 0.92 ± 0.04 msec, p = 0.28, Student’s t test; P10-12: WT 0.37 ± 0.02 msec, KO 0.40 ± 0.02 msec, p = 0.29, Student’s t test). (D) Decay time constants obtained by single exponential fits did not differ between WT and KO mice (P1-3: WT 14.7 ± 1.3 msec, KO 14.6± 1.4 msec; p = 0.99, Student’s t test; P10-12: WT 3.3± 0.4 msec, KO 3.8± 0.4 msec; p = 0.36, Student’s t test).

P1-3 P10-12

De

ca

y t

ime

(m

se

c) WT

9 KO WT

9 KO

C D

0

0.2

0.4

0.6

0.8

1

22 18 43 56

P1-3 P10-12

20

-80

% r

ise

tim

e (

mse

c)

0

5

10

15

20

22 18 42 56

100 pA

10 msec

200 pA

τ = 3.6 msec τ = 3.5 msec

20 pA

10 msec

τ = 15.8 msec

20 pA

10 msec

τ = 14.7 msec

WT KOA

B

P2 P2

P10 P10

  6  

Figure S5, related to Figure 4 and Experimental Procedures. Reconstruction of single MNTB-LSO axons (A) Photomicrographs (top) and reconstruction (bottom) of MNTB-LSO axons in a P19 WT mouse. Reconstructions also include axons contained in neighboring sections (not shown). Blue arrows point to putative axonal boutons, marked in red in the high-magnification reconstruction. (B) same as (A) but from a P21 α9 KO mouse. Scale bars are 250 µm (left) and 10 µm (right)

A

B

  7  

Figure S6, related to Figures 4,6, and 8. Non-normalized data of MNTB axon morphology. (A) Average area covered by boutons in the LSO. At P1-21 bouton area is significantly greater in α9 KO mice (11.1 ± 1.8 x 103µm in WTvs.26.6 ± 1.0 x 103 µm in KO, p < 0.001, Student’s t-test). At P2-4, the smaller bouton area in α9 KO mice is due to a reduction in the dorsoventral spread of boutons, rather than a reduction in the spread along the tonotopic axis (13.3 ± 1.6 x 103 µm in WT vs. 8.0 ± 0.9 x 103 µm in KO, p = 0.01, Student’s t--‐test). (B) Average spread of boutons along the LSO tonotopic axis. At P19-21 the spread of boutons is significantly greater in α9 KO mice (72 ± 8 µm in WT vs. 147 ± 21 µm in KO, p < 0.005, Student’s t--‐test). (C) Average crosssectional area of the LSO does not differ between WT and α9 KO mice at any of the ages studied (p > 0.1 at all ages, Student’s t--‐test). Data originate from the same axons as the data presented in Figs. 4-7. N = 9 axons from 6 WT animals and 9 axons from 8 α9 KO animals. Error bars, SEM. *p < 0.05, Student’s t-test against age-matched.

  8  

Table S1. Quantification of synaptic responses to glutamate uncaging (related to Figure S3)

Peak amplitude (pA)

Charge transfer (pC)

First peak amplitude (pA)

P1-3

map WT (n = 6) 1655 ± 1240 129.1 ± 105.0 - KO (n = 6) 1799 ± 500 122.6 ± 41.5 -

p-value 0.92 0.95 -

stimulation site

WT (n = 31) 320 ± 82 25.0 ± 7.0 - KO (n = 42) 257 ± 38 17.5 ± 3.0 -

p-value 0.45 0.29 -

P10-12

map WT (n = 14) 1147 ± 336 24.1 ± 8.4 700 ± 186 KO (n = 16) 1343 ± 281 41.3 ± 8.2 707 ± 128

p-value 0.66 0.16 0.97

stimulation site

WT (n = 38) 423 ± 62 8.9 ± 1.7 258 ± 46 KO (n = 74) 290 ± 52 8.9 ± 1.5 153 ± 29

p-value 0.13 0.99 0.47

  9  

Table S2. Related to Figure 4 and 6 and Experimental procedures. Location of somas within the MNTB and bouton centroids within the LSO.

Supplemental  Experimental  Procedures    In  vivo  recording  and  spike  sorting  In   vivo   single   unit   recordings   from   MNTB   neurons   (P8)   were   performed   at   the  Neurobiology   Laboratories   of   the   Institute   of   Biology   at   the  University   of   Leipzig.  Experimental   procedures   are  described   in  detail   elsewhere   (Sonntag   et   al.,   2009).  Mice   were   anesthetized   by   an   initial   intraperitoneal   injection   of   a   mixture   of  ketamine  hydrochloride  (0.1  mg/g  body  weight;  Pfizer)  and  xylazine  hydrochloride  (0.005  mg/g   body  weight;   Bayer)   and  by   successive   injections   of   one-­‐third   of   the  initial   dose   every   120–230   min.     Previously   it   was   shown   that   the   spontaneous  firing   rates   of   MNTB   neurons   does   not   correlate   with   the   length   of   anesthesia  (Sonntag  et  al.,  2009).  Recordings  were  performed  in  a  sound-­‐attenuation  chamber  (A400,   Industrial  Acoustics).  Body   temperature  was  maintained  at  38°C.  The  head  was   secured   via   a   metal   post   on   the   skull.   The   MNTB   was   approached   dorsally  through  a  hole  (diameter  ~  0.5  mm)  at  the  midline  placed  1.8–2  mm  caudal  to  the  lambda  suture.  High  impedance  glass  micropipettes  (5-­‐12  MΩ;  Harvard  Apparatus,  GC150F-­‐10)   filled  with  3  M  KCl  were  advanced   into   the  MNTB,  approximately  4-­‐5  mm  deep  from  the  skull  surface.  Voltage  signals  were  bandpass  filtered  (0.3–7  kHz),  amplified   (PC1,   TDT),   and   digitized   (RP2.1,   TDT)   at   97.7   kHz.   MNTB   principal  neurons  were  identified  by  the  characteristic  complex  waveform  that  consists  of  the  calyceal   potential   (“prepotential”)   preceding   the   postsynaptic   action   potential  (Guinan  and  Li,  1990)(Figure  1B).  Recording  sites  were  also  verified  histologically  by  iontophoretic  injection  (4  μA,  10  min)  of  fluorogold.       Only   stable   and   well-­‐isolated   MNTB   units   were   included   in   the   analysis.  Generally,  the  spikes  of  a  single  neuron  could  be  identified  by  simple  thresholding,  but   we   performed   further   spike   sorting   offline   using   custom   routines   written   in  MATLAB7.5   (MathWorks).   Thresholded   spike   waveforms   were   separated   from  noise  by  clustering  based  on  principal  components  analysis.  Although  the  amplitude  of   MNTB   spikes   can   decrease   during   high-­‐frequency   bursting,   they   still   formed   a  cluster  that  was  clearly  separable  from  noise.  We  also  checked  the  amplitude  of  the  prepotentials,   which   remains   constant   even   when   postsynaptic   spike   waveforms  vary   (Sonntag   et   al.,   2009).   A   lack   of   refractory   period   violations   (ISI   <   1   msec)  further  ensured  the  single  unit  nature  of  the  activity.        Burst  analysis  To   quantitatively   compare   the   patterns   of   spontaneous   activity   between   the   two  genotypes,  we  defined  mini-­‐bursts  and  bursts  based  on  the  population  ISI  histogram  (Figure   S1).   The   population   ISI   histogram   of   WT   mice   was   fit   with   a   sum   of   3  Gaussians  using  maximum  likelihood  estimation  ("mle"  function  in  MATLAB).  Based  on   the   intersections   of   the   Gaussian   fits,   mini-­‐bursts   were   defined   as   successive  spikes  with   ISIs   less   than  12  msec,  whereas   bursts  were  defined   as   ISIs   less   than  1888   msec.   Within   a   burst,   spikes   that   did   not   form   mini-­‐bursts   (i.e.,   spikes   that  were   preceded   and   followed   by   intervals   greater   than   12   msec)   were   analyzed  separately   as   single   spikes     (Figure   1K   and   L).   The   same   boundary   values   were  used  for  both  WT  and  KO  mice  to  make  the  comparison  of  activity  patterns  possible.  The   difference   between   the   genotypes   shown   in   Figure   1   is   not   a   result   of   the  

  10  

particular   choices   of   boundaries   because  we  obtained  qualitatively   similar   results  with  a  range  of  boundary  values  (10,  12,  or  14  msec  for  mini-­‐bursts;  1600,  1888,  or  2200  msec   for  bursts).   For   the   interval   analysis   shown   in  Figure   1M,  we   refer   to  mini-­‐bursts  and  single  spikes  together  as  “spike  clusters”  to  analyze  them  together  in  a  single  temporal  sequence.  Only  bursts  that  contain  at  least  11  spike  clusters  (or  10   intervals)   were   analyzed.   The   interval   between   a   neighboring   pair   of   spike  clusters  was  defined  as   the   time   from  the   last   spike  of   the   first   cluster   to   the   first  spike  of  the  second.        In  vitro  electrophysiological  recordings  and  functional  mapping  Brainstem   slices   (300   µm)   were   prepared   as   previously   described   (Kim   and  Kandler,   2003).   Whole   cell   recordings   were   obtained   from   visually   identified,  putative   principal   cells   of   the   medial   LSO.   Recording   electrodes   (2-­‐3   MΟhm)  contained  (in  mM)  76  Cs-­‐methanesulfonate,  56  CsCl,  10  EGTA,  1  MgCl2,  1  CaCl2,  2  ATP-­‐Mg,  0.3  GTP-­‐Na,  5  Na2-­‐phosphocreatine,  and  10  HEPES.  Synaptic  currents  were  recorded  with  an  Axopatch  1D  amplifier  (Axon  Instruments).  With  the  internal  and  external   solution   and   a   holding   potential   of   -­‐70   mV   (corrected   for   -­‐5   mV   liquid  junction   potential),   ECl   was   -­‐20   mV.   For   the   current   clamp   recordings   shown   in  Figure   S2,   Cs-­‐methanesulfonate   was   replaced   with   K-­‐gluconate.   The   series  resistance   (5-­‐20   MΩ)   was   compensated   by   60-­‐80%   (Axopatch   1D,   Axon  Instruments)   and   was   continuously   monitored   throughout   the   recordings.   Data  were  filtered  at  2  kHz  (Axopatch  1D)  and  acquired  at  10  kHz  using  custom-­‐written  Labview  data  acquisition  software.    

The  spatial  distribution  of  presynaptic  MNTB  neurons  was  determined  using  focal   photolysis   of   p-­‐hydroxyphenacyl-­‐glutamate   (Givens   et   al.,   1997)   or   CNB-­‐glu  (“caged   glutamate”,   150-­‐200   µM;   Invitrogen)   using   a   fiber   optic-­‐based   system   as  described  previously  (Kandler  et  al.,  2013;  Kim  and  Kandler,  2003).  The  optical  fiber  for   delivering   UV   light   had   a   20-­‐µm-­‐diameter   light-­‐conducting   core   (Polymicro  Technologies  Inc,  Phoenix,  AZ)  that  produced  small  diameter  UV  spots  (~25  µm)  on  the   surface   of   the   slice.   The   duration   of   UV-­‐light   pulses   was   20   ms   for   the   P1-­‐3  group  and  100  ms  for  the  P10-­‐12  group.  Pulse  durations  were  chosen  such  that  the  spike-­‐eliciting   distance   from   an   MNTB   neuron   was   less   than   50   µm   in   both   age  groups  (see  below  and  Figure  S2).    

In   each   slice,   50   to   80   locations   (spaced  ~50  µm  apart)   in   and   around   the  MNTB  were  stimulated.  Uncaging  sites  were  marked  as  the  position  of  the  light  spot  on  the  slice  on  digitized  video  frames  overlaid  with  a  50  µm  x  50  µm  grid  (Figure  2).  Squares   containing   successful   stimulation   sites  were  used   to   calculate   input   areas  (Noh  et  al.,  2010).  To  normalize   input  areas   to  MNTB  cross-­‐sectional  areas,  MNTB  boundaries  were  determined  by  three  investigators  (two  of  them  were  blind  to  the  data  and  animal  group)  using  images  taken  from  the  slices  during  the  experiments  and   the   mean   boundary   constructed   (Adobe   Illustrator).   On   rare   occasions   in  neonatal  animals  of  both  genotypes,  responses  could  be  elicited   from  sites  slightly  dorsal   to   the   MNTB   boundaries   and   these   locations   were   excluded   from   our  analysis.  The   spatial   resolution  of  uncaging  was  measured  by   recording   the  direct  responses   of   MNTB   neurons   to   uncaging   glutamate   in   current   clamp   (Kim   and  

  11  

Kandler,  2003).  With  the  same  uncaging  parameters  as  for  mapping,  only  uncaging  directly  at  the  cell  body  (indicated  by  electrode  tip)  elicited  spikes  (Figure  S2).    

For  minimal   and  maximal   stimulation,  MNTB  axons  were   stimulated   at   the  lateral  boundary  of  the  MNTB  with  patch  pipettes  (1  MΩ)  filled  with  external  bath  solution.   Constant   current   pulses   (0.2   ms)   were   delivered   using   a   stimulation  isolation  unit   (IsoFlex,  AMPI)   at   a   rate  of  0.2  Hz.   In  minimal   stimulation,   stimulus  intensity  was  gradually  varied  around  the  threshold,  such  that  we  could  record  the  smallest   responses   that   emerge   from   failures.   The   first   plateau   of   successful  responses   in   the   input-­‐output   relationship  was   taken   as   the   single-­‐fiber   response  amplitude.  Using  minimal  stimulation  allowed  us   to  detect  small  amplitude  single-­‐fiber   responses   that   would   not   be   resolved   in   input-­‐output   relations   over   a   wide  range   of   stimulus   intensity.   For   maximal   responses,   stimulus   intensity   was  increased   up   to   1,000   µA.   PSC   amplitudes   typically   saturated   around   400   µA,  indicating  our  range  of  stimulus  intensity  was  adequate  for  activating  most  available  inputs.  Only  1  or  2  LSO  neurons  were  recorded  per  MNTB-­‐LSO  pathway  to  minimize  potential  bias  introduced  by  slicing.  The  experimenter  was  blind  to  the  genotype  of  the  animals  until  the  analysis  was  finished.    

The   amplitudes   of   individual   MNTB   inputs   varied   widely   within   each   LSO  neuron.   Therefore,   unlike   in   the   case   where   converging   inputs   have   similar  amplitude,   simply   dividing   the   maximal   amplitude   by   the   minimal   within   each  neuron   may   not   yield   accurate   estimates   of   the   number   of   converging   inputs.   To  better  estimate  the  convergence  and  its  uncertainty,  we  used  bootstrap  resampling  of   maximal   and   minimal   PSC   amplitudes.   PSC   amplitude   values   were   randomly  drawn  from  the  experimental  data  set  (with  replacement)  to  form  a  resample  of  the  same  sample  size.  A  convergence  ratio  was  computed  by  dividing  the  mean  maximal  amplitude  by   the  mean  minimal   amplitude.  Resampling  was   repeated  1000   times,  yielding  1000  convergence  values.  The  mean  and  standard  deviation  of  these  1000  values   represent  bootstrap  estimates   (Figure   3F).   Statistical   significance  between  WT  and  α9  KO  mice  was  determined  using  a  permutation  test.  The  amplitude  values  from  the  two  groups  were  first  combined  and  then  randomly  re-­‐assigned  to  either  group   to   form   permuted   samples.   This   permutation   was   repeated   1000   times,  yielding   a   distribution   of   the   difference   values   in   convergence   between   the   two  permuted   samples.   The   difference   values   from   our   experimental   data   were   then  compared  against  this  distribution  to  calculate  the  p  values  (Figure  3F).    Noise-­‐rearing  protocol  

Noise-­‐reared   (NR)  mice  were   reared   from  P8  until   P19-­‐21   in  pulsed  white  noise.    Litters  were  culled  to  six  pups  and  home  cages  containing  the  pups  and  their  mother  were  placed   in  a  sound-­‐attenuating  chamber   (Coulbourn   Instruments,  PA)  lined   with   anechoic   foam   on   a   12-­‐hour   light/dark   cycle   with   food   and   water   ad  libidum.     White   noise   pulses   at   70   decibels   sound   pressure   level   (dB   SPL)   were  generated   (Coulbourn   tone/noise   generator   A69-­‐20,   Coulbourn   Instruments,   PA)  and   delivered   to   the   chamber   by   an   isodynamic   tweeter   (RT2H-­‐A,   HiVi   Research,  CA)   mounted   30   cm   above   the   cage   floor.     The   loudspeaker   had   a   relatively   flat  frequency   response   over  much   of   the   normal   hearing   range   of   the  mouse   (~2-­‐50  

  12  

kHz).     In   order   to   minimize   adaptation   effects,   white   noise   pulse   length   was  pseudorandomly  set  to  700-­‐1400  ms  and  pulses  were  pseudorandomly  presented  at  a   rate   of   ~36   per   minute.     Acoustic   stimuli   were   calibrated   with   a   ¼”   diameter  microphone  (4939,  Brüel  &  Kjær,  Denmark)  placed  at  the  level  of  the  animals  within  the  housing.    Microphone  output  was  amplified    (Nexus  2690-­‐A-­‐OS1,  Brüel  &  Kjær,  Denmark)   and   analyzed   using   SoundCheck   (Listen,   MA).   NR   pups   showed   no  behavioral   signs   of   distress   and   at   time   of   sacrifice  were   similar   in  weight   to  WT  pups,  indicating  normal  maternal  care.    Of  the  17  total  axons  reconstructed  from  13  NR  mice,   3   axons   exhibited   two  distinctly   separate   areas  of   boutons   in   the  LSO,   a  termination  pattern  that  was  not  observed  in  WT  mice.    Including  these  3  axons  in  the  NR  population  did  not  change  the  results  of  statistical  comparisons  between  NR  and  WT  mice  at  P19-­‐21  (all  p-­‐values  >  0.05),  but  because  they  were  unique  to  the  NR   and   never   observed   in   the   WT   case,   they   were   excluded   from   the   population  data  presented  here.      Single  axon  and  bouton  analysis  Individual  MNTB  neurons  were  filled  with  patch  pipettes  containing  0.1  mM  Alexa  Fluor   568   hydrazide   (Invitrogen,   Carlsbad,   CA)   and   0.5   %   biocytin.   After   filling,  slices  were  incubated  for  at  least  60  min  in  an  interface-­‐type  chamber  before  being  fixed  in  4  %  paraformaldehyde  (1-­‐7  days)  and  processed  for  biocytin  visualization.  Axonal  arbors  and  boutons  within  the  LSO  were  reconstructed  with  60X  and  100X  oil  objectives  using  a  Neurolucida  system  (MBF  Bioscience,  Williston,  VT)  interfaced  with   a   Zeiss   Imager  M1.   Putative   synaptic   boutons  were   identified  based  on   their  characteristic   round   shape   and   their   diameter   greater   than   2x   the   width   of   the  parent  axon.  Cells  with  cut  axon  branches  within   the  LSO  (identified  by  a  bulbous  ending  at  the  slice  surface)  were  discarded  from  analysis.       For   analysis   (Neurolucida   Explorer),   3-­‐dimensional   reconstructions   were  flattened  rostro-­‐caudally  into  a  2-­‐dimensional,  medio-­‐laterally-­‐  and  dorso-­‐ventrally-­‐defined   Cartesian   coordinate   system.   To   determine   bouton   area   in   an   objective  manner,   an   ellipse   was   fit   to   the   coordinates   of   LSO   boutons   that   maximized   the  density  of  boutons  within  the  confines  of  the  ellipse.  This  approach  is  less  likely  to  be   influenced   by   the   presence   of   single   or   small   numbers   of   boutons   located   far  outside  the  main  region  of  termination.  Analysis  using  bouton  areas  determined  by  conventional  approaches  (connecting  peripheral  boutons)  yielded  qualitatively  and  statistically   similar   results.   For   ellipse   fitting,   custom-­‐written   LabView   programs  were  used   to   extract   the  Cartesian   coordinates   of   boutons   located  within   the  LSO  from  Neurolucida  ASC  files  and  a  minimum-­‐volume  enclosing  ellipse  was  then  fit  to  these   coordinates  using   custom-­‐written  MATLAB   (Mathworks,  Natick,  MA)   scripts  based  on   the  Khachiyan  Algorithm.    Due   to   the  S-­‐shaped   curvature  of   the  LSO  we  used  the  length  of  the  short  axis  of  the  ellipse  as  a  measure  of  tonotopic  spread.  The  angle   of   the   ellipse   short   axis   to   the   horizontal  medio-­‐lateral   axis  was  used   as   an  estimate  of   the  orientation  of   the  LSO   frequency  axis   at   the   location  of   the  axonal  boutons   and  when  measuring   LSO   length.   There   was   no   difference   of   the   ellipse-­‐horizontal  angles  between  genotypes  at  any  age  (P  >  0.05,  Students  t-­‐test).       To   create   density   heat   maps   (e.g.,   Figure   5A),   the   boutons   of   each  reconstructed  axon  were  overlaid  with  a  2-­‐dimensional  grid  comprised  of  56.8  μm2  

  13  

(at  P2-­‐4)  or  100  μm2  (at  P12-­‐14  and  P19-­‐21)  bins  within  Neurolucida.  The  chosen  bin  size  encompassed  a  constant  proportion  of  LSO  area  (~0.08  %)  across  all  ages.  The  origin  of  the  grid  was  positioned  at  the  bouton  centroid  (the  center  of  mass)  of  each  axon,  aligning  each  axon  independent  of  the  location  of  its  terminations  within  the  LSO.  Bouton  numbers  per  bin  were  calculated  for  each  axon  and  then  averaged  across  axons.  This  average  bouton  density  for  each  age  group  was  plotted  as  a  heat  map  using  the  ‘contourf’  function  in  MATLAB.  

To generate histograms of bouton density along the LSO frequency axis (e.g., Figure 5B), each axon’s bouton field was rotated around the centroid so that the horizontal axis of the coordinate system now corresponded to the putative LSO frequency axis. The frequency axis was divided into bins of 2 % of the length of the LSO. For each axon, counts of the number of boutons per bin were computed. Data for all of the axons in each age group were then averaged.

In  Figure   7B,   to   demonstrate   the   “sinking   iceberg”   effect,   a   constant   (21.4  boutons)  was  subtracted  from  the  P12-­‐14  histogram  to  match  the  peak  of  the  P19-­‐21  histogram.  The  correlation  coefficient  between  the  downshifted  and  the  P19-­‐21  distribution  was   calculated   using   only   the   bins  with   non-­‐zero   bouton   counts,   and  using  a  finer  binning  (1%  LSO  length)  to  have  enough  bins.  

Throughout  the  paper,  results  are  presented  as  mean  ±  s.e.m,  and  Student’s  t-­‐test  was  used  unless  stated  otherwise.    

 In  vitro  tracing    Following  decapitation  and  removal  of  the  brain  a  small  crystal  of  biocytin  (Sigma-­‐Aldrich)  was   inserted   into   the  cochlear  nucleus  under  visual   control   (Kandler  and  Friauf,   1993).   Brains   were   submerged   in   oxygenated   ACSF   for   3-­‐4   hours   and  immersion  fixed  for  2-­‐5  days  in  4%  PFA.  Brains  were  cryoprotected  in  30%  sucrose,  cut   into  50  µm  thick   coronal   sections  on  a   freezing  microtome,   and  processed   for  biocytin   visualization   following   standard   protocols   (Vectastain   ABC   kit).   Calyceal  axons   were   labeled   in   300   µm   thick   living   brainstem   slices   by   inserting   a   small  crystal  of  rhodamin  dextrane  (Sigma-­‐Aldrich)  into  the  acoustic  decussation  between  both  MNTBs.  Following  30-­‐60  minutes  of  incubation,  slices  were  transferred  into  a  submerged   type  chamber  mounted   to  an  upright  microscope   (Examiner  Z1,  Zeiss)    and   individual   calyces  were   imaged  with   a   2   photon  microscope   (VIVO   2-­‐photon,  Intelligent  Imaging  Innovations)  with  excitation  at  810  nm  and  z-­‐axis  spacing  of  0.5  µm.      

  14  

                   

References    Guinan,   J.J.,   and   Li,   R.Y.   (1990).   Signal   processing   in   brainstem   auditory   neurons  which  receive  giant  endings  (calyces  of  Held)  in  the  medial  nucleus  of  the  trapezoid  body  of  the  cat.  Hear  Res  49,  321-­‐334.  

Sonntag,   M.,   Englitz,   B.,   Kopp-­‐Scheinpflug,   C.,   and   Rubsamen,   R.   (2009).   Early  postnatal  development  of  spontaneous  and  acoustically  evoked  discharge  activity  of  principal  cells  of  the  medial  nucleus  of  the  trapezoid  body:  an  in  vivo  study  in  mice.  J  Neurosci  29,  9510-­‐9520.  

Kim,   G.,   and   Kandler,   K.   (2003).   Elimination   and   strengthening   of  glycinergic/GABAergic   connections   during   tonotopic  map   formation.  Nat  Neurosci  6,  282-­‐290.  Givens,   R.S.,   Jung,   A.,   Park,   C.H.,   Weber,   J.,   and   Bartlett,   W.   (1997).   New  photoactivated   protecting   groups.   7.   p-­‐hydroxyphenacyl   -­‐   a   phototrigger   for  excitatory  amino  acids  and  peptides.  J  Am  Chem  Soc  119,  8369-­‐8370.  Kandler,   K.,   Nguyen,   T.,   Noh,   J.,   and   Givens,   R.S.   (2013).   An   optical   fiber-­‐based  uncaging  system.  Cold  Spring  Harbor  protocols  2013,  118-­‐121.  

Noh,  J.,  Seal,  R.P.,  Garver,   J.A.,  Edwards,  R.H.,  and  Kandler,  K.  (2010).  Glutamate  co-­‐release  at  GABA/glycinergic  synapses   is  crucial   for  the  refinement  of  an   inhibitory  map.  Nat  Neurosci  13,  232-­‐238.  Kandler,   K.,   and   Friauf,   E.   (1993).     Pre-­‐   and   postnatal   development   of   efferent  connections  of  the  cochlear  nucleus  in  the  rat.  J  Comp  Neurol.  328,  161-­‐84.  

 

  15