Gaze-Stabilizing Central Vestibular Neurons Project Asymmetrically to Extraocular Motoneuron Pools

Abstract

Within reflex circuits, specific anatomical projections allow central neurons to relay sensations to effectors that generate movements. A major challenge is to relate anatomical features of central neural populations, such as asymmetric connectivity, to the computations the populations perform. To address this problem, we mapped the anatomy, modeled the function, and discovered a new behavioral role for a genetically defined population of central vestibular neurons in rhombomeres 5-7 of larval zebrafish. First, we found that neurons within this central population project preferentially to motoneurons that move the eyes downward. Concordantly, when the entire population of asymmetrically projecting neurons was stimulated collectively, only downward eye rotations were observed, demonstrating a functional correlate of the anatomical bias. When these neurons are ablated, fish failed to rotate their eyes following either nose-up or nose-down body tilts. This asymmetrically projecting central population thus participates in both upward and downward gaze stabilization. In addition to projecting to motoneurons, central vestibular neurons also receive direct sensory input from peripheral afferents. To infer whether asymmetric projections can facilitate sensory encoding or motor output, we modeled differentially projecting sets of central vestibular neurons. Whereas motor command strength was independent of projection allocation, asymmetric projections enabled more accurate representation of nose-up stimuli. The model shows how asymmetric connectivity could enhance the representation of imbalance during nose-up postures while preserving gaze stabilization performance. Finally, we found that central vestibular neurons were necessary for a vital behavior requiring maintenance of a nose-up posture: swim bladder inflation. These observations suggest that asymmetric connectivity in the vestibular system facilitates representation of ethologically relevant stimuli without compromising reflexive behavior.SIGNIFICANCE STATEMENT Interneuron populations use specific anatomical projections to transform sensations into reflexive actions. Here we examined how the anatomical composition of a genetically defined population of balance interneurons in the larval zebrafish relates to the computations it performs. First, we found that the population of interneurons that stabilize gaze preferentially project to motoneurons that move the eyes downward. Next, we discovered through modeling that such projection patterns can enhance the encoding of nose-up sensations without compromising gaze stabilization. Finally, we found that loss of these interneurons impairs a vital behavior, swim bladder inflation, that relies on maintaining a nose-up posture. These observations suggest that anatomical specialization permits neural circuits to represent relevant features of the environment without compromising behavior.

Keywords: anatomy; asymmetry; motoneuron; reflex; vestibular; zebrafish.

Figure 1.

Vestibular nucleus neurons labeled in Tg(−6.7FRhcrtR:gal4VP16). A, The expression pattern of Tg(−6.7FRhcrtR:gal4VP16); Tg(UAS-E1b:Kaede)s1999t (purple) is shown as a horizontal MIP, with one vestibular neuron, colabeled by focal electroporation of gap43-EGFP (white). Arrows point to the tangential (_TVN) and medial vestibular nuclei (_MVN) and the MLF. Inset, Schematic of a dorsal view of a larval zebrafish. Magenta rectangle represents the location of the image. Scale bar, 50 μm. Horizontal (B) and sagittal (C) MIP of vestibular neurons in Tg(−6.7FRhcrtR:gal4VP16);Tg(UAS-KillerRed) (purple);Tg(isl1:GFP) (green, image γ = 0.5) showing cranial motoneuron somata from nIII/nIV, nV, and nVII (green text). Arrows indicate neurons in the vestibular nuclei (VN) and the MLF. Scale bar, 50 μm. D–F, Close-up of white boxed region in C, showing major branch patterns of vestibular neuron axon fascicle (purple) relative to extraocular motoneurons (green). D, Motoneurons from Tg(isl1:GFP) (green) in nIV (magenta arrow), superior rectus motoneurons of nIII (cyan arrow), and the midbrain/hindbrain boundary (white dotted line). E, Branches of the vestibular neuron axon fascicle (purple), emerging from the MLF (white arrow) in Tg(−6.7FRhcrtR:gal4VP16);Tg(UAS-KillerRed), projecting to nIV (magenta arrow) and nIII (cyan arrow). F, Merge of D and E. Scale bar, 20 μm. G–I, Broad and close-up views of vestibular neuron axonal projection (purple) to nIII cell bodies (green), taken at the horizontal plane delineated by the cyan dotted line in F, SR motoneurons (nIII) encircled in cyan. G, Cyan arrows localize close-ups in H and I. Scale bar, 10 μm. J–L, Broad and close-up view of vestibular neuron axonal projection (purple) to nIV cell bodies (green), taken at the horizontal plane delineated by the magenta dotted line in F, SO motoneurons (nIV, green) encircled in magenta. J, Arrows point to close-up in K and L. Scale bar, 10 μm.

Figure 2.

Projections from singly labeled vestibular nucleus neurons. A, Horizontal MIP of a single vestibular neuron labeled with UAS-ChR2(H134R)-EYFP (purple) in Tg(−6.7FRhcrtR:gal4VP16);Tg(isl:GFP) (green). γ = 0.5 to highlight the sparse label. Scale bar, 100 μm. Pink triangle represents the data in Figure 7D. Twenty-three of 27 neurons studied projected similarly. B, Sagittal MIP of the neuron in A highlighting nIII (cyan arrow), nIV (magenta arrow), and projection to nIV (white arrow). Scale bar, 20 μm. C, Horizontal MIP of nIV (green cell bodies in dotted magenta outline) from A. Vestibular neuron projection (purple, white arrow). Scale bar, 10 μm. D, Horizontal MIP of nIII (green cell bodies in dotted cyan outline) with no proximal vestibular neuron projection (purple). E, Sagittal MIP of a single axon expressing 14×UAS-E1b:hChR2(H134R)-EYFP (purple) in Tg(−6.7FRhcrtR:gal4VP16);Tg(isl1:GFP) (green); Tg(atoh7:gap43-RFP) (cyan) fish. Expression of bright GFP bleeds into the purple channel, making the cell bodies white. nIV (magenta arrow), nIII (cyan arrow), and the vestibular neuron projection to SR motoneurons in nIII (white arrow). Scale bar, 20 μm. Four of 27 neurons projected similarly, exclusively to nIII. F, Horizontal MIP of nIII (cells in blue outline) from E, purple projections from vestibular neuron (white arrow). Scale bar, 10 μm. G, Horizontal MIP of nIV (cells in magenta outline) from E with no purple vestibular neuron projection. Scale bar, 10 μm.

Figure 3.

Tracings of two vestibular nucleus neurons from a single fish at two developmental time points. A, Horizontal (top) and sagittal (bottom) projections of two traced neurons taken from the same fish imaged at 5 dpf. Magenta trace represents the characteristic projection to the nIV motoneuron pool (magenta arrows), whereas the green neuron does not. B, Same two neurons traced in the same fish, at 11 dpf. The same projection to nIV is visible in the magenta tracing (magenta arrow). Scale bars, 100 μm.

Figure 4.

Vestibular nucleus neurons show synaptophysin-positive puncta on their motoneuron targets. A, Sagittal MIP of a labeled SO motoneuron (magenta arrow) in green and the purple synaptic puncta labeled in Tg(−6.7FRhcrtR:gal4VP16); Tg(5×UAS:sypb-GCaMP3). Dotted lines indicate the planes in B, C. Scale bar, 20 μm. B, C, Close-up slice of the motoneuron somata in A with puncta (magenta arrow). Scale bar, 10 μm. D, Close-up of a retrogradely labeled SR motoneuron soma (green) with visible purple puncta (cyan arrow). Scale bar, 10 μm. E, Close-up of the dendrites of SR motoneurons (green) with visible purple puncta (cyan arrow). Scale bar, 10 μm.

Figure 5.

Vestibular nucleus neurons labeled in Tg(−6.7FRhcrtR:gal4VP16) are necessary for both nose-up and nose-down gaze stabilization. A, Horizontal MIP of vestibular and control neurons (nVII) in rhombomeres 4–8 in Tg(−6.7FRhcrtR:gal4VP16); Tg(UAS-E1b:Kaede)^s1999t; Tg(isl1:GFP) fish before and after targeted photo-ablation of vestibular neuron cell bodies. γ = 0.5 highlights dim signal. Colors represent depth over ∼150 μm. White arrows indicate the general region of targeted cell bodies in either the vestibular nuclei (top row) or the facial nucleus (nVII). Scale bar, 150 μm. For anatomical localization, compare with the right side of Figure 1B. B, Vestibulo-ocular reflex gain preablation and postablation of vestibular neurons. C, Vestibulo-ocular reflex gain preablation and postablation of facial nucleus neurons. D, Vestibulo-ocular reflex gain wild-type siblings (WT) and fish with pharmacogenetic (nitroreductase, “nfsb”) and optogenetic ablation (Killer-Red [KR]) of neurons in Tg(−6.7FRhcrtR:gal4VP16).

Figure 6.

The simplified neural circuit underlying the ocular response to pitch and roll tilts. Cyan represents nose-down. Magenta represents nose-up channels. A, Wiring diagram of one hemisphere of the excitatory vestibulo-ocular circuit showing utricular hair cells (cyan/magenta), stato-acoustic ganglion (SAG), central vestibular neurons (VN, cyan and magenta), extraocular motoneuron pools in nIII (SR, IR, IO) and nIV (SO). B, During a roll tilt to the fish's left, the left utricle hair cells (cyan/magenta) are activated, causing cocontraction of superior (SO/SR) eye muscles ipsilateral to the activated utricle, and inferior (IO/IR) muscles contralateral to the activated utricle. C, Utricular hair cells sensitive to nose-up pitch tilts (magenta) ultimately activate only vestibular neurons that project to both nIII and nIV, activating SO (contralateral) and IR (ipsilateral). D, Utricular hair cells sensitive to nose-down pitch tilts (cyan) ultimately activate vestibular neurons that project to exclusively to nIII, activating SR (contralateral) and IO (ipsilateral).

Figure 7.

Activating vestibular nucleus neurons generates downward eye rotations. A, Peak eye rotation as a function of blue light duration. Positive values indicate eyes-down rotations (magenta arrow). Negative values indicate eyes-up (cyan arrow). Black represents ChR2+ fish. Gray represents ChR2− siblings. Points are median ± median absolute deviation. B, Evoked eye rotation in time. Gray lines indicate individual fish. Black lines indicate the median of prelesion data. Red lines indicate the same fish after photoablation of ChR2+ vestibular neurons. Blue represents stimulus (100 ms). C, Gray lines indicate the average responses from individual fish with pan-neuronal expression. Black represents the median across fish. Blue represents stimulus (100 ms). The trace with a downward lobe indicates a nontorsional component; video of this fish is shown as Figures 7-1 and 7-2. D, Evoked ocular rotations from sparsely labeled fish as a function of ChR2+ expression (MLF fluorescence). Black dots represent fish with discriminable vestibular neurons. Green dots represent fish without discriminable vestibular neurons. Pink triangle corresponds to the fish in Figure 2A.

Figure 8.

A, Model schematic. B, One simulation of the model for two different population sizes, 180 neurons (magenta) and 30 neurons (cyan). First column represents the vestibular neuron activity as a spike raster plot and the input function (black). Second column represents the motoneuron spikes. For display, half the generated spikes are shown in each raster. C, The “Output strength” (average firing rate) of the postsynaptic neurons as a function of the population size (rows) and number of inputs per motoneuron (columns). D, The “Encoding fidelity” (variance explained, R) in the input rate function by the summed postsynaptic output.

Figure 9.

Early ablations of vestibular neurons leave fish unable to inflate their swim bladders. A, Tg(−6.7FRhcrtR:gal4VP16); Tg(14×UAS-E1b:hChR2(H134R)-EYFP); mitfa^−/− fish swimming in a cuvette in the dark at 144 hpf. Red arrows point to swim bladders. B, Sibling fish where the vestibular neurons in these fish were photoablated at 72 hpf, before swim bladder inflation. Note the absence of a swim bladder, evaluated here at 144 hpf. Images are taken from Figures 9-1 and 9-2.