Organization of vestibular circuits for postural control in zebrafish

Abstract

Most animals begin controlling their posture, or orientation with respect to gravity, at an early stage in life. Posture is vital for locomotor function. Even animals like fish, which are capable of swimming upside-down, must actively control their orientation to coordinate behaviors such as capturing prey near the water’s surface. Here we review recent research from multiple laboratories investigating the organization and function of the vestibular circuits underlying postural control in zebrafish. Some findings in zebrafish strongly align with prior observations in mammals, reinforcing our understanding of homologies between systems. In other instances, the unique transparency and accessibility of zebrafish has enabled new analyses of several neural circuit components that remain challenging to study in mammalian systems. These new results are also explored in depth in a recent review [1].

Development of postural behaviors

Postural control in fish consists of two elements: roll (left-right stabilization) and pitch (rostrocaudal stabilization). Young zebrafish first display postural behaviors at 3-4 days post fertilization (dpf), coinciding with the earliest vestibulo-ocular reflexes [2-5]. Postural control is vital for survival during development, as it is required for feeding and swim bladder inflation. Pitch, which can be voluntarily adjusted as fish navigate the water column, undergoes significant refinement during the first three weeks of life. This improvement stems from a combination of enhanced swim bout timing to counterbalance the naturally heavier head and increased reliance on coordination between fins and body [5,6]. Similar to many vertebrates, zebrafish demonstrate progressively precise postural control across development. Unlike in mammals, however, this development takes place externally, providing convenient experimental access to both the behavior and its underlying neural circuits (reviewed in [7]).

Distinctions between classes of gravity-sensing hair cells

The sense of orientation relative to gravity originates from the vestibular otoliths, located in the inner ear. These are distinct from the semi-circular canals, which sense rotational movement but cannot transduce gravity sensation. Otolithic hair cells serve as motion sensors to detect head movement and tilt, with gravity sensation in zebrafish arising specifically from the utricle (the zebrafish saccule is specialized for hearing and dispensable for postural control [4,8]). The ciliary orientation of these hair cells determines their directional tuning, while variations in ciliary heights and intrinsic properties establish their temporal dynamics. Hair cells selective for higher frequencies of head movement are located in a central zone of the macula, known as the striolar zone, while hair cells that preferentially encode lower frequencies populate the remaining area, namely the extrastriolar zone [9] (Fig. 1).

Figure 1. Convergent characterization of utricular hair cells.

Three recent studies found similar patterns of utricular hair cells at the level of molecular expression [12], ultrastructural morphology [14], and sensory tuning of hair cells [15]. Striolar (dark blue) and extrastriolar (light blue) zones display distinct features. (A) Cabp2b and cabp1b are specifically expressed in different regions that define striolar and extrastriolar hair cells, respectively. (B) In striolar hair cells, the kinocilia and longest stereocilia exhibit similar lengths, whereas the kinocilia in extrastriolar hair cells are significantly longer than their stereocilia. Hair cells are topographically organized by their orientations, divided by the line of polarity reversal (dashed). (C) In functional imaging, striolar hair cells demonstrate greater sensitivity to linear vibration, while extrastriolar hair cells exhibit higher sensitivity to static tilts. Hair cell directional tuning (arrows representing direction of highest sensitivity) is consistent between ultrastructural [14] and functional imaging [15] approaches. The striolar zone was defined somewhat more broadly in [15] based on expression of calcium binding protein S100s, but it remains to be determined how this corresponds to other metrics of striolar identity.

In mammals, striolar and extrastriolar cell types are morphologically and genetically distinct, with each population expressing selective proteins [9]. Previous work in zebrafish has demonstrated ultrastructural features defining striolar and extrastriolar hair cells [10], but the field has lacked systematic analysis of molecular markers to distinguish these populations genetically. Recent advances in single-cell RNA sequencing of the zebrafish inner ear [11] have now been applied to reveal distinct molecular profiles within these sensory populations in zebrafish, leading to the categorization of two major utricular hair cell groups based on gene expression [12] (Fig. 1a). Striolar and extrastriolar hair cells express calcium binding protein genes cabp2b and cabp1b, respectively, resembling the selective expression of calretinin in some mammalian striolar hair cells [13]. These two zones correspond well to both ultrastructural and functional analyses of zebrafish hair cells in two recent significant studies. At the anatomical level, a comprehensive reconstruction of utricular hair cells showed that larval zebrafish striolar hair cells display taller stereocilia relative to their kinocilium length, consistent with observations in both mammals and adult zebrafish [14] (Fig. 1b). Correspondingly, the first direct in vivo calcium imaging of vestibular-evoked activity in utricular hair cells has shown that striolar hair cells, as identified by calcium binding protein S100s expression, are preferentially activated by high-frequency vibrational stimuli, while extrastriolar hair cells are preferentially activated by static tilts [15] (Fig. 1c). Collectively, these three results reinforce each other and suggest future opportunities for genetic manipulations to study the development and behavioral contributions of striolar and extrastriolar hair cells.

Gravity-sensing afferents are topographically organized

Hair cells relay sensory information to the brain via vestibular afferents, which receive glutamatergic synaptic input from hair cells at ribbon synapses. Vestibular afferents are classified based on both their spatial and temporal encoding properties: spatially, they exhibit directional tuning which aligns with the input from hair cells; temporally, they exhibit regular or irregular firing patterns. Regular afferents predominantly receive input from extrastriolar hair cells, whereas irregular afferents predominantly receive input from striolar hair cells, enabling them to preferentially encode head motion in low and high frequency domains, respectively [9]. Physiological recordings from afferents have produced extensive knowledge about these properties, but surprisingly their organization within the vestibular ganglion has remained largely unknown. This absence of known topography contrasts with the tonotopy of the neighboring auditory ganglion, and is due to technical difficulties in tracing afferents’ twisting paths from different vestibular end-organs.

The young zebrafish provides an ideal model to address afferent topography because its compact size and transparency make it amenable to analysis by both light microscopy and serial-section electron microscopy (ssEM) [16]. Two recent papers with complementary approaches showed that the utricular afferent ganglion is approximately organized by both directional tuning and temporal tuning along two anatomical axes [14,15] (Fig. 2). In the rostral-caudal axis, both anatomical and functional analyses found that utricular afferents exhibited graded tuning from rostral to caudal tilt, largely preserving the rostral-caudal organization of the hair cells. In contrast, the mediolateral axis exhibited different topography in the utricular ganglion compared to the hair cell macula. Anatomically, afferents receiving input from the striolar zone were located at the ganglion’s lateral edge, while afferents receiving input from the extrastriolar zone form a shell on the medial aspect [14] (Fig. 2a). In vivo, responses to high-frequency vibrational stimuli were present exclusively in laterally positioned afferents [15] (Fig. 2b). Many afferents with cell bodies located medially were entirely unresponsive to either tilt or vibrational stimuli.

Figure 2. Transformation of topography from hair cells to afferents.

Vestibular afferents are organized by directional tuning in the rostral-caudal axis and temporal dynamics in the medial-lateral axis, both of which are formed by patterned innervation to the utricular macula. (A) Individual afferent innervation pattern revealed by ultrastructural analysis [14] indicates that afferents innervating the striolar zone in the utricular macula are located on the lateral side of the ganglion, are early myelinated, and receive more hair cell synaptic inputs; whereas those innervating the extrastriolar zone are located more medially, are late myelinated, and receive fewer hair cell synaptic inputs. These afferents also connect rostrally located afferent somata with rostrally located hair cells in the utricular macula (white arrows), and caudally located somata with caudally located hair cells (black arrows), forming a rostrocaudal topography. However, contralateral-tilt afferents (red arrows) are intermingled in the ganglion. (B) Functional calcium imaging [15] demonstrated that about half of utricular afferents, primarily located on the lateral side of the ganglion, are responsive to vibration and static tilts. Rostrally and caudally located afferents encode rostral (white arrows) and caudal (black arrows) tilts respectively. In this study, no afferents exhibited consistent responses to contralateral tilt. Vibration activated a small subpopulation of afferent neurons without a clear spatial pattern. It seems likely that the lower myelination and synapse number in the medially located afferents accounts for their lack of response in functional assays.

These results may be due to the smaller number of hair cell synaptic contacts that medial afferents receive at these ages [14], or they may indicate that afferent axons cannot conduct action potentials from the periphery to the soma prior to myelination. Further study of afferent myelin development may illuminate this question [17].

Intriguingly, this organization of temporal dynamics from lateral to medial mirrors the developmental sequence of the ganglion [18,19]. Laterally located afferents were early-myelinated and formed more synaptic contacts with hair cells, whereas medially located afferents were less myelinated with fewer synaptic contacts with hair cells (Fig. 2a). These findings suggest that development may pattern vestibular topography, with early-developing striolar hair cells preferentially contacting early-developing afferents to encode high-frequency stimuli, and later-developing hair cells and afferents encoding lower-frequency stimuli. Future studies of other vestibular afferents, such as those arising from the semi-circular canals, may indicate whether this developmental topography is widespread. It remains an open question whether there is a causal relationship between developmental sequence and vestibular topography.

In addition, the functional analyses help to address a longstanding question about how the vestibular system distinguishes between head tilt and translation, which are both inertial stimuli and thus theoretically indistinguishable. In adult primate cerebellar neurons, tilt and translation can be distinguished by integrating sensory signals from the semicircular canals, which encode angular movements [20]. However, despite the fact that semicircular canals are not yet functional in young fish [21-23], linear movements can generate different compensatory behaviors than those elicited by body tilts [14]. One possibility is that otolith hair cells with different temporal dynamics could be used to distinguish between linear and rotational self-motion. During natural movements in mice, head tilts typically occur in the low frequency domain, whereas head translations commonly occur at high frequencies [24]. Similarly, fish experience high frequencies during head translations [25]. As a consequence, striolar afferents with their rapid dynamics should be more effective at transmitting high frequency movements, which are typically translational, whereas extrastriolar afferents should transmit more information about head tilts, because their slower dynamics are aligned with the lower frequency head tilts [26]. The new functional analyses presented by Tanimoto and colleagues support this idea by showing that although all hair cells respond to tilt, extrastriolar hair cells respond better to static tilt, whereas striolar hair cells respond best to high-frequency stimuli.

Central circuits governing posture

From the utricular ganglion, vestibular afferents travel centrally and form synaptic connections with several classes of brainstem neurons, including vestibulo-ocular and vestibulospinal neurons. In zebrafish, utricular afferents also form synapses onto the Mauthner cells, a pair of specialized reticulospinal neurons responsible for startle-evoked escape behaviors. In addition, vestibular information is distributed widely to other brainstem and mid-brain targets [27,28]. only some of which have been characterized for their vestibular computations.

The vestibulo-ocular pathways for stabilizing vision are well-established, but knowledge about the vestibular pathways controlling posture is still limited. Young fish rely on the utricle for gravity sensing: animals without utricles frequently swim sideways or upside-down [4]. Vestibulospinal neurons have been considered primary candidates for controlling posture because they receive sensory inputs from the utricular afferents and project down the spinal cord to motor targets [29-31]. Vestibulospinal neurons develop early and encode head movements in the roll and pitch axes [32,33]. However, bilateral ablation of vestibulospinal neurons caused only modest deficits in postural control, namely a slight increase in abnormal pitch angles, suggesting that alternate pathways may be important [33].

Recently, Sugioka and colleagues showed that vestibular information takes another route to produce compensatory rolling behaviors [34] (Fig. 3). Utricular afferents excite tangential nucleus neurons, whose axons cross the midline and ascend towards the oculomotor nuclei [2]. Though primarily associated with oculomotor control, tangential neurons form en passant synaptic connections with neurons of the midbrain nucleus of the medial longitudinal fasciculus (nucMLF). Consequently, nucMLF neurons are activated by head tilts towards the contralateral side, and carry these vestibular signals via descending axons to the spinal cord, where they drive motor circuits controlling the posterior hypaxial muscle near the swim bladder [35]. Contraction of this muscle bends the fish’s rostral body towards the ear-up side, repositioning the animal’s center of gravity lateral to its center of buoyancy and creating a moment of force that counteracts the roll. Ablation of any component in this pathway impairs posture control in the roll axis. The pathway seems highly homologous to that of the mammalian interstitial nucleus of Cajal, which directly influences neck motor neurons for control of head orientation [36].

Figure 3: Two vestibular pathways for postural control in zebrafish.

Vestibular afferents relay signals from the inner ear to the hindbrain (blue), forming two neural pathways that are implicated in postural control. Both vestibulospinal (VS) neurons (dark brown) and tangential (Tan) neurons (light brown) receive direct afferent inputs about ipsilateral tilt. VS neurons project to unknown targets in the spinal cord, whereas Tan neurons project to the contralateral nucleus of the medial longitudinal fasciculus (nucMLF) in the midbrain. Descending nucMLF neurons can activate the posterior hypaxial muscles (PHM) to bend the body and create a torque in the roll axis for body balance [34,35]. The targets of the VS neurons and their role for postural control are unclear. Moreover, the brainstem targets of afferents encoding contralateral tilt and the corresponding functions remain largely unknown (red), although some appear to serve computations of head movement triggering escape [14].

A key innovative feature of this work [34] is its use of a head-free system for studying postural control. Prior work demonstrated that utricular displacement and nucMLF activity could elicit rostral body bends [35,37], but those experiments were conducted in head-fixed animals, making it difficult to evaluate whether this bend was used for steering in the yaw axis or for postural control. The new results suggest that postural control in the roll axis is at least one major component of the function, though it does not exclude the possibility that this circuit could also be used for steering.

Formation and function of contralateral tilt pathways

Several recent studies have examined the organization and function of utricular pathways arising from the lateral edge of the macula. The orientation of these lateral hair cells, which report contralateral tilt, is established during development by Emx signaling pathways [38,39]. These hair cells signal via a distinct set of afferents whose cell bodies are intermingled with afferents encoding ipsilateral roll at the ganglion. Their axons form a segregated, lateralized pathway in the young zebrafish brainstem [14,40]. In mice, afferents from the lateral macula project directly to the cerebellum [41,42], but this has not yet been explored in fish. Some of these afferents from the lateral macula form synapses with commissural or other brainstem interneurons in zebrafish [40]. Our recent work showed that commissural neurons contacted by these afferents encoding contralateral tilt project to the escape circuit, specifically the Mauthner cell on the opposite side of the brainstem. This circuit suggests that Mauthner cells might make a directional computation about head movement: each Mauthner cell receives information about ipsilateral tilt directly from utricular afferents, and information about contralateral tilt relayed through commissural neurons. Indeed, in behavioral experiments, zebrafish challenged with a fast lateral displacement made an escape response toward the direction of displacement [14]. Therefore, the lateral macula pathway may support particularly high-acceleration movements, consistent with its responsiveness to vibrational stimuli [15].

Conclusion and future directions

These recent findings on a novel circuit mechanism for postural control have opened up new avenues for exploration. Several interesting questions still remain to be answered. First, it is unknown whether either the nucMLF or the posterior hypaxial muscle are involved in control of pitch, which is frequently adjusted as fish dive or surface. One possibility is that nucMLF descending neurons may develop synaptic connections with the fin motor circuits for later refinement of pitch [6]. Second, it remains unclear how the rest of the body, particularly the axial musculature, participates in control of roll. Previous work demonstrated differential recruitment of motor neurons governing dorsal and ventral musculature on the left and right sides of the body during fictive postural correction, dependent on the utricular signaling [43]. It is yet to be determined whether this trunk musculature would also be recruited by a nucMLF-driven circuit, or whether the trunk and posterior hypaxial muscle are governed separately. A new methodological approach, using ferrofluid injection into the inner ear for controlled manipulation of the otoliths, may provide opportunities to parse these postural circuits in head-fixed animals [44].

Finally, the function of VS neurons in fish requires further characterization. In mice, the homologous neurons in the lateral vestibular nucleus are thought to be important for posture based on lesion studies [45], but because vestibular afferents pass through the LVN, lesions are likely to have non-specific effects. More targeted ablation approaches have demonstrated that LVN neurons are not essential for basic locomotor performance but important for compensatory hindlimb movements during unexpected lateral displacement [46,47]. One interesting possibility is that different pathways could be employed for maintenance of static posture versus active balance during fast locomotion. The tangential pathway could serve fine postural control, which occurs via tonic mechanisms, while the VS pathway could serve as a rapid response pathway to counteract unexpected disturbances or the effects of rapid body bends during fast locomotion. This idea is consistent with the observation that VS ablation has a modest effect in fish [33], presumably because the tangential pathway is able to compensate during spontaneous, low-effort swimming. One counterargument is that in mouse, the VS pathway preferentially contacts slow motor neurons [48]. Mechanistically, the VS pathway might be preferentially activated by neuromodulators during high-threat periods, as supported by recent experiments in mice [49].

The zebrafish model has proven to be a valuable tool for elucidating the organization and function of the peripheral vestibular circuitry, as well as the central processing of vestibular information. These findings offer new insights into the development, organization, and functional role of striolar and extrastriolar hair cells and their associated afferents. They also illuminate alternative pathways for postural control. Further investigation into the remaining questions and the application of novel experimental approaches promise to deepen our understanding of how vestibular signals are integrated into motor systems under a wide variety of contexts, such as computation of heading direction or body stabilization [50,51]. These findings will not only expand our knowledge of the vestibular system in zebrafish but also have broader implications for understanding the function and organization of the vestibular system in other vertebrates.

Highlights:

• We review recent work on neural circuits controlling posture in zebrafish

• Gravity-sensing hair cells express distinct genetic markers for identity

• Hair cell polarity is governed by a molecular pathway involving Emx2

• Vestibular peripheral circuits are organized topographically

• A brainstem / midbrain pathway acts on rostral trunk muscles to control roll