Excitatory and inhibitory synapses to octopus cells form two domains.

(A) Illustration of spiral ganglion neuron (SGN) central axons branching within a parasagittal section of the mouse cochlear nucleus complex (CNC). SGN somas in the cochlea are tono-topically organized according to frequency. Axons remain organized throughout the ventral (VCN) and dorsal (DCN) divisions of the CNC. Octopus cells (inset) are found in the octopus cell area (OCA) of the VCN. (B) Excitatory SGN puncta labeled with a VGLUT1 antibody (left) and glycinergic puncta labeled with Glyt2Cre-dependent syp/tdT (Glyt2-syp/tdT; right) in a par-asagittal section of the CNC. The teardrop-shaped OCA is not devoid of inhibitory inputs, although less prominent than in the surrounding CNC. (C) A Thy1 sparsely labeled octopus cell with excitatory SGN (VGLUT1) and inhibitory (Glyt2-syp/tdT) puncta. Micrographs of 3µm confocal z-stacks show puncta on the medial surface of a soma (top) and a dendrite (bottom). (D) Representative reconstruction of excitatory SGN puncta labeled with Foxg1Cre-dependent syp/tdT (Foxg1-syp/tdT; blue) onto a Thy1+ octopus cell. (E) Representative reconstruction of inhibitory puncta labeled with Glyt2-syp/tdT (green) onto a Thy1 sparsely labeled octopus cell (white). (F) Puncta density for excitatory SGN (Foxg1-syp/tdT; black: 10.7 ± 3.0, n = 8 cells, 4 mice) and glycinergic puncta (Glyt2-syp/tdT; green: 4.2 ± 0.8, n = 8 cells, 3 mice) on octopus cell dendrites. Data are presented as mean ± SD. (G) Puncta density on somas for excitatory SGN (black: 13.3 ± 2.2, n = 8 cells, 4 mice) and glycinergic puncta (green: 1.8 ± 0.1, n = 8 cells, 3 mice) and the density along the length of dendrites. Data are presented as mean ± SEM. (H) Top: Illustration of an octopus cell and the ratio between excitatory SGN puncta and glycinergic puncta. Inset: Illustration of an octopus cell and the relative innervation densities of excitatory SGNs (blue) and inhibitory puncta (green).

Type Ia SGNs are the primary excitatory contributors to octopus cells.

(A) Ia (yellow), Ib (orange), and Ic (magenta) SGN axons innervate the CNC. SGNs have a continuum of properties organized along the pillar-modiolar axis of inner hair cells (IHCs) and the habenula. Ntng1-expressing Ib/c fibers are positioned on the modiolar side (closest to the ganglion). Strongly calretinin immunopositive Ia fibers are on the other side (closest to the pillar cells). This organization correlates with spontaneous rates (SR) and thresholds (thresh.) measured in vivo. Somas of all SGN subtypes are found at all tonotopic locations. (B) Calretinin (CR) immunolabeling distinguishes SGN subtypes. Ia/b somas label with high (CR++) and medium (CR+) levels of CR, respectively. Ic somas label with low to undetectable levels of CR (CR-). Ntng1Cre-mediated expression of tdT (Ntng1-tdT) labels Ib/c SGNs. (C) Ias (tdT-, CR++) make up 39.9 ± 2.6% of the SGN population. Ib/cs (tdT+) make up 60.1 ± 2.6% of the SGN population. Ibs (tdT+CR+) make up 28.5 ± 12.2% of the SGN population (n = 1599 neurons, 4 mice). Data are mean ± SD; individual points represent percent coverage per animal, lines connect measurements from the same animal. Dotted lines indicate percentages from 1: Petitpré et al. 2018, 2: Shrestha et al. 2018, and 3: Sun et al. 2018. (D) Reconstructed Ib/c puncta labeled with Ntng1Cre-dependent syp/tdT (Ntng1-syp/tdT; magenta) onto a Thy1+ octopus cell. (E) Puncta density for all SGNs (Foxg1-syp/tdT; black: data from Fig. 1F), Ias (Foxg1 -Ntng1; yellow: 6.6 ± 1.0), and Ib/cs (Ntng1-syp/tdT; magenta: 4.1 ± 1.0, n = 9 cells, 5 mice) along the total dendritic length. Ia density was calculated by subtracting Ib/c density from total SGN density; lines connect measurements from the same reconstruction. Data are mean ± SD. (F) Puncta density on somas for all SGNs (Foxg1-syp/tdT; black: data from Fig. 1E), Ias (Foxg1 -Ntng1; yellow: 8.4 ± 2.3), and Ib/cs (Ntng1-syp/tdT; magenta: 4.9 ± 1.2, n = 9 cells, 5 mice) and the density along the length of dendrites. Data are mean ± SEM. (G) Relative innervation densities of Ias (yellow), Ib/cs (magenta), and inhibitory puncta (green).

SGN subtype inputs to octopus cells do not differ in short term plasticity.

(A) Illustration of the experimental paradigm and representative EPSPs recorded during in vitro whole-cell current clamp recordings of octopus cells. SGN stimulation method included electrical stimulation or full-field, light-evoked activation of Foxg-ChR2 or Ntng1-ChR2 SGNs. TTL trigger pulses are shown in gray. (B) Paired pulse ratios for electrically stimulated SGNs (open circles: n = 5 cells, 3 mice), ChR2 stimulated SGNs (Foxg1-ChR2; black: n = 8 cells, 5 mice), and ChR2 stimulated Ib/c SGNs (Ntng1-ChR2; magenta: n = 7 cells, 6 mice) at three interstimulus intervals. With electrical stimulation, SGN inputs to octopus cells were stable and exhibited slight facilitation at 50 Hz (20ms interstimulus interval). ChR2 stimulation caused paired pulse depression not seen with electrical stimulation. Data are presented as mean ± SD. Each data point represents the average paired pulse ratio for a cell. p values from ANOVA and subsequent Tukey HSD test are reported for comparisons between methods of SGN activation (electrical and ChR2) and SGN subpopulation composition within method of activation (SGN-ChR2 and Ib/c-ChR2). Welch’s ANOVA was used for comparisons at 20ms interstimulus interval (50Hz) as data in this condition did not meet the homogeneity of variance assumption. (C) Pulse ratios at 50Hz for electrically stimulated SGNs with physiological 1.4mM Ca2+ ACSF (open circles: n = 5 cells, 3 mice), electrically stimulated SGNs with 2.4mM Ca2+ ACSF (grey: n = 3 cells, 2 mice) and ChR2 stimulated SGNs with physiological ACSF (Foxg1-ChR2; black: n = 8 cells, 5 mice). Data are presented as mean ± SD. p < 0.001 from ANOVA and subsequent Tukey HSD test for all comparisons between methods of SGN activation (electrical and ChR2). There were no statistically significant differences for all comparisons under 1.4mM and 2.4mM Ca2+ (p > 0.100, ANOVA).

Octopus cells receive glycinergic inhibitory post synaptic potentials.

(A) Voltage responses to a -200pA current injection. This representative neuron hyperpolarized 0.7mV (black) in control conditions. After bath application of 100µM 4-Aminopyridine (4-AP) and 50µM ZD 7288 (ZD), hyperpolarizing responses to the same -200pA current injection increased to 8.8mV at steady state (green). (B) Postsynaptic responses to ChR2 stimulation of glycinergic terminals (Glyt2-ChR2) with a 5Hz train (gray) of 1ms full-field blue light pulses before (black) and after bath application of 100µM 4-AP, 50µM ZD, and 15µM NBQX (green: n = 9 cells, 8 mice). Increased input resistance reveals inhibitory potentials that are difficult to detect. (C) Postsynaptic responses to Glyt2-ChR2 stimulation after bath application of 100µM 4-AP, 50µM ZD, and 15µM NBQX (green), with further sequential addition of 20µM picrotoxin (PTX, pink), 100µM cyclothiazide (CTZ, blue), and 500nM strychnine (STN, orange; n = 6 cells, 5 mice). (D) Change in membrane voltage in response to hyperpolarizing somatic current steps in a morphologically and biophysically realistic model of octopus cells before (black) and after removal of voltage-gated potassium (Kv) and hyperpolarization-activated cyclic nucleotide-gated (HCN) channels (blue). As in in vitro current-clamp recordings, removing Kv and HCN channels increased the magnitude of voltage responses (ΔVm) to hyperpolarizing current. (E) IPSPs measured at the soma of a modeled octopus cell before (black) and after removal of Kv and HCN channels (blue). As in in vitro current-clamp recordings, this allows for somatic IPSP detection. (F) IPSP magnitude in experimental data (green) and the model (blues) as a function of input resistance. In somatic measurements, IPSP size increases with input resistance. Modeled IPSPs are shown for two conductance levels (1nS, dark blue; 100nS light blue). (G) Fold change in magnitude of soma-measured IPSPs (blue) or dendrite-measured IPSCs at proximal (dark orange) and distal (light orange) dendritic locations after removal of Kv and HCN channels. Kv and HCN block increases the magnitude of soma-measured IPSPs. The size of dendritic IPSCs are not changed with Kv and HCN or synapse location.

Coincident excitation and inhibition on octopus cell dendrites advances EPSP peak times.

(A) The impact of distance between excitatory and inhibitory synapses was measured in a computational model of octopus cells. Inhibitory synapses were placed either proximally (on-path) or distally (off-path) to excitation. Excitatory synapses were placed at varying locations along the dendritic arbor to change the anatomical distance (Δd) where d=0 is the location of inhibition (GGly) and d=1 is the condition where excitation (GAMPA) and inhibition are maximally separated. EPSPs were measured at the soma in all conditions (green). (B-E) Quantification of the percent change in soma-measured EPSP magnitude and the shift in EPSP peak timing in models of (B-C) on-path and (D-E) off-path inhibition. Example traces show EPSPs with (green) and without (black) inhibition at d=0 and d=1. Distal dendrites (d=1) have higher local input resistance and lower IPSP attenuation due to the sealed end. Inset scale bars are 1mV, 200ms. (F-H) Coincident stimulation of excitation and inhibition changes EPSP shape. (F) Representative responses to independent stimulation of excitatory SGNs (black), independent stimulation of inhibitory inputs (light blue), coincident stimulation of both excitation and inhibition (green), and independent stimulation of excitatory SGNs with the addition of 1µM strychnine (STN). Quantification of (G) the percent change in EPSP height and (H) the shift in EPSP peak timing during coincident Glyt2-ChR2 activation of inhibitory inputs (green: n = 8 cells, 6 mice), bath application of 25µM glycine (dark blue: n = 4 cells, 3 mice), and bath application of 1µM STN (orange: n = 5 cells, 4 mice). Activation of glycinergic receptors during excitation decreases EPSP heights and advances EPSP peaks. Blocking of tonically active glycine receptors slows and delays EPSPs. Data are presented as mean ± SD. Markers represent the average quantification for a cell.

Proposed model of flexible dendritic timing for precise somatic coincidence detection.

(A-B) Co-occurring frequencies in broadband onsets and frequency modulations occur on different time scales. (Top) Spectrograms illustrate frequency strength over time. (Bottom) Waveform illustrate total intensity over time. Both of these stimuli drive firing in octopus cells despite the limitations imposed by a ∼1ms window for temporal integration. (C) Excitation must summate at the soma of octopus cells within a narrow time window (∼1ms) to achieve a depolarization rate rapid enough to trigger action potentials. Summation of excitation alone accounts for responses to sound onsets (inset, top). Stimuli that require summation over a longer time period, such as frequency modulated sweeps, require synaptic inhibition to modulate the timing of excitation on octopus cell dendrites before they reach the soma for coincidence detection computations (inset, bottom). We propose a mechanism for preferential processing of a subset of excitatory inputs where selective temporal advancement of a subset of EPSPs could expand the effective window for coincidence detection at the soma.

Summary and description of experimental genotypes presented in figures.

Ntng1Cre has high specificity for Ib/c SGNs.

(A) Calretinin immunopositive (CR+) Ia/b fibers (CR, yellow) preferentially innervate the pillar side of inner hair cells (IHCs). Ib/c fibers with Ntng1Cre-mediated expression of tdTomato (Ntng1-tdT, magenta) preferentially innervate the modiolar side of IHCs. IHCs also immunolabel for CR. (B) CR+ Ia/b fibers (yellow) pass through the pillar side of the habenula while Ntng1-tdT+ Ib/c fibers (magenta) pass through the modiolar side. Arrowhead highlights a Ntng1-tdT fiber passing through the pillar side of the habenula but ultimately terminating on the modiolar side of the IHC. (C) Normalized position of CR+ Ia/b (yellow) and Ntng1-tdT+ Ib/c (magenta) fibers along the pillar to modiolar axis of the habenula (n = 124 fibers; 5 mice). (D) In the central nervous system, Ntng1-tdT is present throughout the brain, but is restricted to SGN axons in the ventral cochlear nucleus where the octopus cell area (OCA) is found. (E) In the OCA, CR immunolabel is present in Ia/b SGN axons and puncta. As in the ganglion, CR co-labels with some Ib/c puncta (Ntng1-syp/tdT).

Myo15iCre sparsely labels Ic SGNs.

(A) Cochlear sections with calretinin (CR) immunolabeling of hair cells and type Ia/b SGNs and Myo15iCre-mediated expression of tdTomato (Myo15-tdT) in hair cells and some type Ic SGNs. (B) CR+ Ia/b fibers (yellow) preferentially innervate the pillar side of IHCs. Sparse Ic fibers with Myo15-tdT (magenta) preferentially innervate the modiolar side of IHCs. IHCs label with both tdT and CR. (C) CR+ Ia/b fibers (yellow) pass through the pillar side of the habenula while sparsely labeled Myo15-tdT+ Ic fibers (magenta) pass through the modiolar side. (D) Normalized position of CR+ Ia/b (yellow) and Myo15-tdT+ Ic (magenta) fibers along the pillar to modiolar axis of the habenula (n = 90 fibers; 4 mice). (E) 65µm cochlear section containing SGN somas. SGNs have variable levels of calretinin (CR) immunolabeling corresponding to the three molecular subtypes. Ia/b SGN somas label with high and medium levels of CR, respectively. Ic somas label with very low levels of CR. Myo15-tdT is sparsely found in Ic SGNs. All SGN somas are labeled with Foxg1Flp-mediated expression of EYFP (Foxg1-EYFP). (F) tdT+CR-SGNs make up 4.7 ± 2.3% of the SGN population (n = 2150 neurons, 5 mice), indicating sparse reporter expression. Data are presented as mean ± SD; individual data points signify percent coverage per animal. Dotted lines are estimated percentages for type Ic SGNs from 1: Petitpré et al., 2018, 2: Shrestha et al., 2018, and 3: Sun et al., 2018. (G) In the OCA, VGlut1 immunolabel neatly tiles around octopus cells with sparse Myo15-syp/tdT puncta. (H) Density of all SGNs (black: data from Fig. 1D), Ib/c SGNs (magenta: data from Fig. 2D), and sparse Ic inputs (open magenta circles: 1.1 ± 0.6, n = 6 cells, 3 mice). Data are presented as mean ± SD. Markers represent the total puncta density computed per reconstructed octopus cell. (I) Puncta density per 100µm2 of soma surface area (all SGN, black: data from Fig. 1E; sparse I/c inputs, magenta open circles: 0.7 ± 0.4, n = 6 cells, 3 mice) and density along the length of the dendritic tree, relative to the soma. Data are presented as mean ± SEM.

Dendritic and synaptic reconstructions of octopus cells.

(A) Total length of reconstructed dendritic arbors for 31 octopus cells. (B) Total surface area of reconstructed dendritic arbors for 31 octopus cells. (C) Octopus cell reconstructions were normalized to the longest reconstructed dendrite. Total length of reconstructed dendrites correlated with the longest branch per neuron. (D) Total length of reconstructed dendrites correlated with total dendritic surface area. (E-G) Longest branch length, total dendrite length, and total surface area compared to estimated position of the octopus cell soma in the tonotopic organization of the OCA. (H-J) Total number of reconstructed SGN puncta (Foxg1-syp/tdT, black), inhibitory puncta (Glyt2-syp/tdT, green), Ib/c SGN puncta (Ntng1-syp/tdT, magenta), and sparse Ic SGN puncta (Myo15-syp/tdT, open magenta circles) compared to the longest branch length, total dendrite length, and total dendrite surface area. (K-M) Density of reconstructed SGN puncta (Foxg1-syp/tdT, black), inhibitory puncta (Glyt2-syp/tdT, green), Ib/c SGN puncta (Ntng1-syp/tdT, magenta), and sparse Ic SGN puncta (Myo15-syp/tdT, open magenta circles) compared to longest branch length, total dendrite length, and total dendrite surface area.

Optimizing active and passive properties of an octopus cell model.

(A) Subthreshold somatic voltage response to a hyperpolarizing current injection in an in vitro whole-cell current clamp recording of an octopus cell. (B) Somatic voltage responses from a morphologically realistic octopus cell model for various scaling factors of maximal conductances of active ion-channels. (C) Comparison of somatic hyperpolarizations in a model reproducing experimental data during control (black) and Kv and HCN block conditions (blue). (D) Illustration of injection and recording locations for panels E-G in a morphologically realistic octopus cell model. (E) IV curves (change in somatic membrane potential as function of current injection magnitude) from a representative experimental octopus cell. Dotted lines plot linear fits of the experimental data. Rin dotted line is the slope of the fit. (F-G) Impact of leak conductance (ḡleak) and scaling factor (scl) on passive properties of the model. Input resistance of the octopus neuron model as a function of membrane resistivity (Rm) with ḡleak=0 (F) and ḡleak=1.67 pS (g) for various scaling factor values indicated in different colors. (H) Illustration of injection and recording locations for panels I-K in a morphologically realistic octopus cell model. (I) Inhibitory post synaptic currents (IPSCs) recorded from glycinergic synapses in proximal (blue) and distal (orange) stimulation during control (solid) and Kv and HCN block conditions (dotted). (J) Peak IPSC magnitude as function of glycine conductance in proximal (blue) and distal (orange) stimulation during control (solid) and Kv and HCN block conditions (dotted). (K) Transfer impedance as function of frequency from proximal dendrites to soma (blue) and distal dendrites to soma (orange) during control (solid) and Kv and HCN block conditions (dotted).

Impact of inhibitory synaptic location and distance between excitatory and inhibitory synapse on somatic EPSP amplitude and timing.

(A-B) Illustration of injection and recording locations for off-path (A) and on-path (B) inhibition paradigms and the normalized relative distance (d) between excitatory synapses. The impact of on-path and off-path inhibition in the dendrites is primarily determined by the local potential change by EPSP and the attenuation or the length constant (λ) of the IPSP towards the excitatory synaptic location. The exponential decay of membrane voltage is asymmetric, with lower λ for the open end and higher λ for sealed end propagation. Distal parts of the dendrites have higher local input resistance and lower attenuation of IPSP due to the sealed end. (C-D) Percentage change in somatic EPSP height with dendritic glycinergic inhibition as function of normalized distance between excitatory and inhibitory synapses in off-path (C) and on-path (D) inhibition for various E/I ratio with excitatory AMPA conductance (GAMPA) set at 5nS. Average shown in black with SEM in shaded region. (E-F) Somatic EPSP peak time shift with dendritic glycinergic inhibition as function of normalized distance between excitatory and inhibitory synapses in off-path (E) and on-path (F) inhibition for various E/I ratio with excitatory AMPA conductance (GAMPA) set at 2nS. Average shown in black with SEM in shaded region.