Brainstem Neural Circuits Triggering Vertical Saccades and Fixation

Abstract

Neurons in the nucleus raphe interpositus have tonic activity that suppresses saccadic burst neurons (BNs) during eye fixations, and that is inhibited before and during saccades in all directions (omnipause neurons, OPNs). We have previously demonstrated via intracellular recording and anatomical staining in anesthetized cats of both sexes that OPNs are inhibited by BNs in the medullary reticular formation (horizontal inhibitory BNs, IBNs). These horizontal IBNs receive monosynaptic input from the caudal horizontal saccade area of the superior colliculus (SC), and then produce monosynaptic inhibition in OPNs, providing a mechanism to trigger saccades. However, it is well known that the neural circuits driving horizontal components of saccades are independent from the circuits driving vertical components. Thus, our previous results are unable to explain how purely vertical saccades are triggered. Here, we again apply intracellular recording to show that a disynaptic vertical IBN circuit exists, analogous to the horizontal circuit. Specifically, we show that stimulation of the SC rostral vertical saccade area produces disynaptic inhibition in OPNs, which is not abolished by midline section between the horizontal IBNs. This excludes the possibility that horizontal IBNs could be responsible for the OPN inhibition during vertical saccades. We then show that vertical IBNs in the interstitial nucleus of Cajal, which receive monosynaptic input from rostral SC, are responsible for the disynaptic inhibition of OPNs. These results indicate that a similarly functioning SC–IBN–OPN circuit exists for both the horizontal and vertical oculomotor pathways. These two IBN-mediated circuits are capable of triggering saccades in any direction.

Significance Statement Saccades shift gaze to objects of interest, moving their image to the central retina, where it is maintained for detailed examination (fixation). During fixation, high gain saccade burst neurons (BNs) are tonically inhibited by omnipause neurons (OPNs). Our previous study showed that medullary horizontal inhibitory BNs (IBNs) activated from the caudal superior colliculus (SC) inhibit tonically active OPNs in order to initiate horizontal saccades. The present study addresses the source of OPN inhibition for vertical saccades. We find that OPNs monosynaptically inhibit vertical IBNs in the interstitial nucleus of Cajal during fixation. Those same vertical IBNs are activated by the rostral SC, and inhibit OPN activity to initiate vertical saccades.

Introduction

Saccadic eye movements redirect gaze to an object of interest in the visual field and bring the object's image onto the central retina for visual fixation. During such fixations, all eye movements to other targets are suppressed. There are two types of saccade-related neurons in the superior colliculus (SC) and brainstem; burst neurons (BNs) that drive saccades and tonically active neurons that maintain fixation (Peck, 1989; Wurtz and Goldberg, 1989).

Neurons in nucleus raphe interpositus (RIP) (Büttner-Ennever et al., 1988) show tonic activity during fixation in contrast to saccadic BNs in caudal SC or brainstem, but cease firing before and during saccades in all directions (omnipause neurons, OPNs) (Cohen and Henn, 1972; Luschei and Fuchs, 1972; Evinger et al., 1977, 1982; Keller, 1977; King et al., 1978, 1980; Kaneko and Fuchs, 1982). Stimulation of the OPN region stops saccades (Keller, 1974). Based on this finding, Robinson (1973, 1975, 1976) proposed a computational model of saccades which employs a “trigger” input to inhibit OPN activity and permit saccade initiation. For this to work, two parallel ,nd signals from a higher brain center are necessary for saccades: (1) a saccade driving signal to activate oculomotor BNs for saccade generation, and determine the metrics (i.e., direction, kinematics, length) of the saccade, and (2) a saccade triggering signal to suppress OPNs and gate all saccades. RIP OPNs have been assumed to be the primary mechanism for gating saccade initiation, but the neural substrate carrying the signal inhibiting OPNs and initiating saccades had remained elusive.

Recently, we located the neural substrate for the OPN trigger mechanism in the well-studied horizontal saccade pathway (Takahashi et al., 2005b, 2022a). Specifically, we showed via intracellular recording that stimulation of the rostral SC evokes monosynaptic excitation in OPNs and that stimulation of caudal SC evokes disynaptic inhibition in OPNs, with the inhibition mediated by horizontal inhibitory BNs (IBNs) in the paramedian medullary reticular formation (PMRF) (Hikosaka and Kawakami, 1977). Intracellular HRP staining of horizontal IBNs confirmed that the IBNs terminate in the OPN region with extensive axon collaterals (Takahashi et al., 2022a). We thus showed that the same pathway carries both signals driving BNs to provide saccade metrics and signals inhibiting OPNs (via IBNs) to trigger saccade initiation.

While it is generally accepted that separate neural pathways control horizontal and vertical components of saccades, most circuit mechanisms have only been mapped for the horizontal saccade system. Due to their complex bilateral interactions and role in torsion, the details for the vertical saccade systems have remained elusive. However, our recent work has led to clarifications of the vertical saccade system. Specifically, we have determined via both electrophysiological and anatomical techniques that the SC exerts disynaptic excitation of ipsilateral vertical motoneurons (MNs) via Forel's field H (FFH), known as rostral interstitial nucleus of the medial longitudinal fasciculus (riMLF) in primate (Büttner-Ennever and Büttner, 1978), and disynaptic inhibition of contralateral vertical MNs via GABAergic neurons in the interstitial nucleus of Cajal (INC) (Sugiuchi et al., 2013; Takahashi and Shinoda, 2018). Having determined that INC contains the IBNs for vertical saccades, an obvious question is whether there exists a vertical IBN–OPN circuit analogous to the horizontal circuit we previously established as responsible for the trigger signal for horizontal saccades.

In the present study, we dissect the neural circuits from the SC that trigger vertical saccade initiation and then suppress vertical saccades during fixation. As previously, we apply intracellular recording techniques combined with electrical stimulation in anesthetized cats. We find that vertical IBNs in the INC are present that can shut down OPN activity and initiate vertical saccades, implementing a conceptually identical mechanism to the horizontal trigger circuit we previously described. In other words, both vertical IBNs (in INC) and horizontal IBNs (in PMRF) receive input from SC and converge onto OPNs, explaining how saccades in any direction are initiated and guided to their targets.

Materials and Methods

Fifteen cats weighing 3.0–4.5 kg (9 males and 6 females) were used for the experiments described in this report. Data is included from four of the animals that were previously used in Sugiuchi et al. (2013) and five animals that were previously used in Takahashi et al. (2022a). All animal experimentation was performed in accordance with “Policies on the Use of Animals and Humans in Neuroscience Research” revised and approved by the Society for Neuroscience in 1995, and “Guiding Principles for the Care and Use of Animals in the Field of Physiological Sciences” (The Physiological Society of Japan, revised in 2001). The experimental protocol was approved by the Animal Care Committee of Tokyo Medical and Dental University. The animals were initially anesthetized with ketamine hydrochloride (Ketalar, Parke-Davis: 25 mg/kg, i.m.) followed by α-chloralose (40–45 mg/kg, i.v.). This was supplemented by additional doses of 10–25 mg/kg, i.v. or pentobarbital sodium (10–15 mg/kg, i.v.), when necessary, throughout the remainder of the experiment. During recording, the animals were artificially ventilated with the end-tidal CO₂ held at 35–40 mmHg. Their heart rate was constantly monitored by electrocardiogram, and the body temperature was maintained between 37.0 and 39.0°C by a heating pad. The animals were mounted on a stereotaxic frame. For intracellular recording, the experimental setup is essentially identical to that of previous reports (Sugiuchi et al., 2005, 2013; Takahashi et al., 2007, 2022a). Briefly, the bone over the parietal and occipital cortex was removed, and the cerebral cortex was removed by aspiration bilaterally to allow the introduction of stimulating electrodes into the superior colliculi (SCs) on both sides under direct visual observation. Four concentric bipolar stimulating electrodes were arranged at 0.8–1.0 mm intervals rostrocaudally along the presumed horizontal meridian (Sugiuchi et al., 2005; Takahashi et al., 2022a) or mediolaterally along the presumed vertical meridian of the SC motor map (McIlwain, 1986) on either side (Sugiuchi et al., 2013), with their tips positioned in the intermediate or deep layer (1.5–2.0 mm deep from the surface) of the SC (Kawamura and Hashikawa, 1978; Moschovakis and Karabelas, 1985; Izawa et al., 1999). Two concentric bipolar stimulating electrodes were placed in the FFH (A, 7.0–7.5; L, 1.5–2.5) (Nakao and Shiraishi, 1985), and two were placed in the INC (A, 5.0–5.5; L, 0.8–1.2) contralateral to the recording site (Sugiuchi et al., 2013), with the aid of a stereotaxic atlas of the brainstem (Snider and Niemer, 1961). Their dorsoventral positions were fixed at the depths where maximal field potentials were evoked by electrical stimulation of the ipsilateral SC.

Intracellular recording was made from OPNs in the RIP, and from putative BNs in INC or FFH. Glass microelectrodes for intracellular recording were filled with 0.4 M KCl or 2 M potassium-citrate and had a resistance of 15–25 MΩ. The location of the abducens nucleus was identified by mapping type II responses to horizontal rotation of the turntable on which the stereotaxic apparatus rested (Shinoda and Yoshida, 1974) to provide a landmark for locating the OPN regions. OPNs were penetrated in the medial part of the pons from the middle level of the abducens nucleus to 2.5 mm rostral to it, at a depth of 1.0–4.0 mm deep from the floor of the fourth ventricle, and within 0.5 mm from the midline (Keller, 1977; King et al., 1978, 1980; Nakao et al., 1980; Curthoys et al., 1981; Evinger et al., 1982; Langer and Kaneko, 1983; Ohgaki et al., 1987, 1989; Takahashi et al., 2022a). For stimulation of OPNs and IBNs, an array of two concentric electrodes arranged at 1 mm intervals in a dorsoventral direction were placed in both the OPN region and PMRF (Izawa et al., 1999; Sugiuchi et al., 2005). We delivered negative pulses of 0.2 ms in duration at <500 µA (usually <300 µA) for stimulation of the SC, the INC, the FFH and OPNs, and at <200 µA (usually <100 µA) for microstimulation of the PMRF. Ranck (1975) estimated the effective current spread of 1.0–1.5 mm around an electrode tip for monopolar stimulation at 500 µA (200 µs duration pulse) in the mammalian CNS. The effective current spread should be much less than the estimated values by Ranck (1975), because we used bipolar stimulation with a concentric electrode (Shinoda et al., 1977). Sasaki et al. (1970, 1972) estimated that a 500 µA stimulus with a concentric bipolar electrode would not activate fibers or cells beyond 1.0 mm from an electrode tip in the cat. The positions of the stimulating electrodes in the SC and brainstem were marked by electrolytic lesion, passing negative currents of 20 µA for 20 s after each experiment, and were histologically confirmed in sections stained with thionine.

To identify recording sites histologically, a glass microelectrode used for intracellular recording was left in the OPN region or INC region for a marker at the end of each experiment. The position of a broken tip of the glass microelectrode was identified histologically on serial sections and used for reference to reconstruct the other electrode tracks.

All quantitative data are presented as the mean ± SD; an unpaired t test was used to characterize the statistical significance when comparing different mean values. The significance level was set at p < 0.05.

Results

Identification of INC IBNs

The INC is known to consist of intermingled neurons of many different groups and functions. In addition, among eye movement-related neurons, there are BNs, putative integrators, and burst-tonic cells (Hess 1954; Hassler, 1972; Büttner-Ennever et al., 1982, 1988; Fukushima, 1987, 1991; Hepp et al., 1989; Fukushima et al., 1990a,b, 1991, 1992, 1995; Chimoto et al., 1992; Henn, 1992; Crawford and Vilis, 1993; Fukushima and Kaneko, 1995; Helmchen et al., 1996; 1998; Chen and May, 2007). Under normal (awake) conditions, IBNs are identified by their bursts of action potentials during saccades or the quick phase of nystagmus. However, our animals are anesthetized and do not make eye movements. Additionally, unlike horizontal centers, which are lateralized, the INC on both sides of the brain contains upward-tuned and downward-tuned cells of various types, which makes it comparatively difficult to investigate the IBN circuits for a single saccade direction. In the upward vertical system, excitatory BNs located in the FFH innervate a yoked pair of elevator MNs (superior rectus and inferior oblique MN pools) on the ipsilateral side, and inhibitory BNs in the INC innervate a yoked pair of depressor MN pools (inferior rectus and superior oblique MNs) on the opposite side. In the downward vertical system, excitatory BNs located in the FFH innervate a yoked pair of depressor MN pools (inferior rectus and superior oblique MNs) on the ipsilateral side and inhibitory BNs in the INC innervate a yoked pair of elevator MN pools (superior rectus and inferior oblique MNs) on the contralateral side (Sugiuchi et al., 2013). The upward saccade system on one side and the downward saccade system on the opposite side mutually inhibit each other via IBNs in the INC (Takahashi and Shinoda, 2018). Accordingly, there are upward and downward IBNs in each INC. Because of this, it is necessary for the present study to pre-classify the INC cells to insure we only include IBNs and classify them as upward or downward neurons in our analysis. In addition to targeting MNs, many INC neurons project ipsilaterally in the brainstem. The interstitiospinal tract originates from neurons in the INC and its vicinity, and descends in the dorsal portion of the ipsilateral MLF in the brainstem (Verhaart, 1964). INC neurons projecting to the inferior olive and prepositus hypoglossi also run ipsilaterally (Voogd et al., 2013). Other INC neurons cross the midline at the same level as their cell bodies, and project to the contralateral INC (Onodera, 1984). This finding is consistent with the electrophysiological finding that INC IBNs should be activated antidromically from the contralateral INC, and also receive strong monosynaptic inhibition from the same INC site (Sugiuchi et al., 2013). This mutual inhibition between the lNC IBNs exists between the upward vertical system on one side and the downward vertical system on the opposite side (Takahashi and Shinoda, 2018). Furthermore, though it is not easy to delineate the exact lateral border of INC, we know the location of INC IBNs terminating on contralateral trochlear MNs in our previous study, because their location was determined by using a transneuronal labeling method after the injection of wheat germ agglutinin-horseradish peroxidase into the trochlear nerve (see Fig. 1, Sugiuchi et al., 2013). Therefore, we used the following and reliable identification criteria to identify IBNs in INC. Thus, we systematically searched for and analyzed only inhibitory last-order premotor neurons in INC which satisfied a set of criteria associated with INC IBNs (Izawa et al., 2007; Sugiuchi et al., 2013; Takahashi and Shinoda, 2018). These criteria are: (1) neurons are located in the INC; (2) neurons are activated antidromically by stimulation of the contralateral trochlear nucleus and/or oculomotor nucleus, since IBNs cross the midline through the posterior commissure and inhibit vertical ocular MNs on the contralateral side, whereas EBNs in the FFH excite vertical MNs on the ipsilateral side; (3) neurons receive monosynaptic inhibition from the contralateral INC; (4) neurons are antidromically activated from the contralateral INC and/or FFH, since the upward torsional saccade system on one side and the downward torsional saccade system on the opposite side mutually inhibit each other through INC IBNs; (5) neurons receive monosynaptic excitation from the ipsilateral SC, and disynaptic inhibition from the contralateral SC. All lateralities in this article are described with reference to the recording site, if not stated otherwise.

Figure 1.

Identification of a last order-premotor neuron in the INC that projects to the vertical extraocular MN pool. A, Schematic diagram showing the arrangement of stimulating electrodes in the midbrain (dorsal view). Intracellular recording was made from a neuron in the left INC. B–E, Antidromic spikes were evoked at a latency of 0.7 ms by stimulation of contralateral III nucleus at 380 μA (B), but no spikes were evoked from the contralateral IV nucleus at 500 μA (C). The same neuron was antidromically activated from the contralateral Forel's field H (FFH) at a latency of 0.8 ms (D). In Ea, spikes were evoked from the contralateral INC at a latency of 1.0 ms in an all or none manner, without latency jittering at threshold stimuli (110 μA), although the latency of these evoked spikes was the longest among the evoked antidromic spikes in this neuron. These spikes responded to double pulse stimuli with two spikes at an interval of 0.5 ms (Eb), but not at 0.4 ms interval (Ec). In such a way, we identified the evoked spikes as antidromic, when PSPs were not observed or spikes were recorded extracellularly. Calibration in B also applies to C–E. INC, Interstitial nucleus of Cajal; MN, motoneuron; SC, superior colliculus; III, oculomotor nucleus; IV, trochlear nucleus; FFH, Forel's field H.

Projection patterns to vertical MN pools classify INC neurons (“upward” or “downward”)

We include in our analysis only last-order premotor INC neurons that were activated antidromically from the III and/or IV nucleus (Fig. 1A). The pattern of projection to the two nuclei allows us to classify each INC neuron as either an upward-tuned cell or a downward-tuned cell. We are able to do this because in a previous study, we showed that individual INC IBNs innervate yoked pairs of vertical ocular MN groups on the contralateral side via axon collaterals. Specifically, upward IBNs innervate yoked pools of inferior rectus MNs in III nucleus and superior oblique MNs in IV nucleus, whereas downward IBNs innervate yoked pools of superior rectus and inferior oblique MNs in III nucleus (Sugiuchi et al., 2013). Based on these different patterns of innervation, we separate our INC cells into two groups. One group of neurons (downward ipsitorsional INC inhibitory last-order premotor neurons) projects to the contralateral III nucleus (Fig. 1B), but not IV nucleus (Fig. 1C). The other group (upward ipsitorsional INC inhibitory last-order premotor neurons) projects to the contralateral IV nucleus (Fig. 4). To understand this classification scheme, it is necessary to recall that INC inhibitory neurons project to MNs contralaterally (Fig. 1B) and to the contralateral FFH (Fig. 1D) and INC (Fig. 1E), in contrast to the excitatory neurons in the FFH, which project to ipsilateral MN nuclei (Fig. 6B,C; Takahashi and Shinoda, 2018). In this study, 105 INC IBNs and 15 EBNs in the FFH, from which intracellular potentials were recorded, were used for later analysis.

Figure 4.

Inhibition of an upward INC IBN by OPNs. A, Experimental setup. B,C, IPSPs were evoked from the contralateral OPN region (B), but not from the IBN region (C). Ba, Single pulse stimuli evoked large IPSPs at 1.3 ms, which was preceded by small EPSPs at 1.0 ms. Bb, Double pulse stimuli evoked larger IPSPs with shorter latencies. D,E, Antidromic spikes were evoked from the contralateral IV nucleus (D) and INC (E). F,G, The largest EPSPs were evoked from the medial part (site SC1) of the ipsilateral SC (F), and EPSPs followed by the largest IPSPs after stimulation from the lateral part (site SC 8) of the contralateral SC (G), indicating this INC neuron is an upward INC IBN inhibiting downward vertical MNs.

Figure 6.

Inhibition of a downward FFH EBN by OPNs. A, Experimental setup. B,C, Antidromic spikes were evoked at 270 μA by stimulation of the ipsilateral IV nucleus (B), but not from the contralateral IV nucleus at 500 μA (C). D, OPN-evoked hyperpolarizing IPSPs (black traces) that were reversed to depolarizing potentials (red traces) after spontaneous diffusion of Cl⁻ into the penetrated cell. E, Large monosynaptic IPSPs were evoked from the contralateral INC, but antidromic spikes could not be evoked at 500 μA. F,G, Synaptic inputs from the ipsilateral (F) and contralateral SC (G). EPSPs were largest from Lt SC4 and IPSPs were largest from Rt SC5. Calibration in C also applies to B and D–G.

SC projects to INC IBNs in a characteristic topographic pattern

Our goal was to show that the vertical SC–IBN–OPN saccade circuit is identical to the horizontal one, including the IBN–OPN triggering mechanism. The first step was to show that the vertical IBNs (i.e., the inhibitory last-order premotor neurons described above) are driven by SC saccade BNs, and mediate the saccade signal. To show this, we investigated the nature of the synaptic inputs from the SCs onto our putative INC IBNs. In the SC, a large number of cells which project to BN-containing vertical brainstem reticular areas (tectoreticular neurons, TRNs) are located along the vertical meridian running mediolaterally in the rostral SC, whereas a large number of cells which project to horizontal BN areas are located along the horizontal meridian running rostrocaudally in the SC (Takahashi et al., 2007, 2010). Knowing this, we placed stimulating electrodes in the rostral SC along a mediolateral axis, and recorded intracellular potentials from the INC premotor neurons. Figure 2 shows an example of signals recorded in an INC inhibitory premotor neuron after electrical stimulation of the SCs. Ipsilateral SC stimulation usually evoked large depolarization at monosynaptic latency, followed by a small hyperpolarization (Fig. 2B, black traces). In contrast, contralateral SC stimulation evoked small monosynaptic depolarization, which was always followed by larger hyperpolarization (Fig. 2C, black traces). To identify the nature of these evoked postsynaptic potentials (PSPs), we performed additional trials during which we passed depolarizing or hyperpolarizing currents through the recording electrode, or injected Cl⁻ into the penetrated cells (Coombs et al., 1955a,b; Eccles, 1964). Injection of Cl⁻ into the penetrated INC cell caused a reversal of the later hyperpolarizing potentials (to depolarizations), but did not change the polarity of the early (monosynaptic) depolarizations (Fig. 2B,C, red traces). This indicates that the early depolarizations are indeed monosynaptic excitatory postsynaptic potentials (EPSPs), which were evoked by the SC stimulation. It also indicates that the later hyperpolarizations are in fact inhibitory postsynaptic potentials (IPSPs), which were produced by chloride channels (Coombs et al., 1955b; Eccles, 1964). When K-citrate electrodes were used, depolarizing responses became larger and hyperpolarizing responses became smaller or reversed to depolarizing potentials during passage of hyperpolarizing currents, whereas during passage of depolarizing currents, depolarizing responses became smaller and hyperpolarizing responses became larger, indicating that the depolarizing responses were EPSPs and the hyperpolarizing responses were IPSPs (Eccles, 1964).

Figure 2.

Properties of synaptic inputs from both SCs and the contralateral INC to an INC IBN. A, Schematic diagram showing an experimental setup. Three stimulating electrodes were placed mediolaterally in the rostral SC along the presumed vertical meridian of the SC motor map on either side. B,C, Synaptic inputs from the ipsilateral (B) and contralateral SC (C). Upper traces (black traces), control intracellular potentials; middle traces (red traces), after injection of Cl⁻ ions into the penetrated cell; lower traces (blue traces), juxtacellular field potentials. D, Synaptic inputs from the contralateral INC. Hyperpolarizing potentials that followed the earlier depolarizing potentials in B–D were reversed to depolarizing potentials without reversal of the early depolarizations after iontophoretic injection of Cl⁻ into the penetrated cell, indicating that the early depolarizations are EPSPs and the later hyperpolarizations are IPSPs. The same color scheme is used to show synaptic inputs in the following figures. Note excitation in C due to activation of commissural collaterals of upward TRNs in the left SC and inhibition in B due to activation of downward TRNs in the right SC based on our previous study in A [Takahashi et al. (2007), their Fig. 14]. Calibration in D also applies to B,C.

We argue that this approach of analyzing monosynaptic inputs from SC to BNs is a more straight forward method to unravel the complex nature of oculomotor circuits than trying to analyze the complex interplay of parallel disynaptic inputs in ocular MNs (Sugiuchi et al., 2013). This is due in part to the additional complexity arising from commissural axon collaterals of TRNs (Sugiuchi et al., 2013), which cause simultaneous bilateral signals to arrive in MNs via multiple pathways through BNs. In the case of SC stimulation, downstream PSPs must be interpreted with care due to the fact that SC stimulation not only evokes action potentials in TRN axons emanating from the stimulated SC sites, but also causes antidromically propagated action potentials in commissural axons of contralateral TRNs. These antidromically propagating action potentials flow backwards along the commissural branches activating all axon branches of the contralateral TRNs. From the point of downstream cells, these originally antidromically propagating (axon reflex) action potentials are indistinguishable from action potentials initiated by the contralateral TRNs.

While the commissural connections between the SCs were historically thought to be exclusively inhibitory (Appell and Behan, 1990), we have since identified excitatory commissural connections between SCs both electrophysiologically and morphologically (Takahashi et al., 2005a, 2007, 2010). Figure 3 shows an example of such a TRN (unfilled cell in right SC) with a commissural collateral. We recorded from an axon passing through the OPN region (Fig. 3A). The penetrated axon showed spikes after stimulation of contralateral FFH (presumably via stimulation of an axon reflex through the TRN's axonal branches) at a latency of 0.6 ms (Fig. 3B), indicating that they were most likely activated directly. Single stimuli of the contralateral caudal SC (Fig. 3D8) did not evoke spikes at 500 µA, but evoked spikes with fluctuating latencies of about 1.6 ms from the second stimuli. These spikes were considered to be synaptically evoked spikes. Stimulation of site 6 of the contralateral SC evoked synaptically activated spikes with fluctuating latencies of about 1.7 ms at threshold (125 µA) (Fig. 3D6b), and increased stimuli at 500 µA evoked spikes with a fixed short latency (1.1 ms) (Fig. 3D6a), Therefore, the shorter-latency spikes were activated directly and the longer-latency spikes were synaptically activated spikes. Spikes evoked from the most rostral site in the contralateral SC were complex (Fig. 3D5). Single stimuli evoked spikes in an all or none manner with fluctuating latencies of about 1.7 ms at threshold (530 µA) (Fig. 3D5a). Double stimuli at 500 µA evoked fixed short-latency spikes (1.0 ms) from the second stimuli (Fig. 3D5b). These spikes were most likely evoked directly from the second stimuli, because they were activated at a fixed short latency from the second stimuli at subthreshold. This temporal facilitation observed in activating direct spikes could often occur when a cell body was activated (Jankowska et al., 1975; Shinoda et al., 1976, 1982, 1987). In Figure 3D5b, the intensity of the first stimuli was fixed at a subthreshold level (500 µA) for generating spikes. The first stimuli depolarized the membrane of a cell body by direct current and also EPSPs evoked in the cell body by activating presynaptic axons terminating on the cell. The depolarization caused by the first stimuli alone was not large enough to generate direct spikes. However, with the second stimulus, the membrane potential reached the threshold for generating direct spikes, attributed to the summation of the depolarization due to the second stimuli with the depolarization and EPSPs evoked by the first stimuli. Similarly, spikes evoked from site 7 (Fig. 3D7) were considered to be evoked directly. Because these findings suggest that the penetrated axon arose from a cell body in the immediate vicinity of the stimulation site, this axon was identified as a tectoreticular axon arising from a collicular neuron near the site around #6 electrode (Fig. 3A). In addition, this collicular neuron was activated synaptically from site 1 in the ipsilateral SC (Fig. 3C1) (not drawn in Fig. 3A) and directly at a fixed short latency (1.0 ms) from its site 2 (Fig. 3A,C2).

Figure 3.

Intraaxonal spikes recorded from a TRN axon in the left OPN region following stimulation of either SC. A, Experimental setup. Four stimulating electrodes were placed along the presumed horizontal meridian of the SC motor map on either side. B, Antidromic spikes were evoked from the right FFH at 190 μA, but not from the left FFH at 500 μA. C,D, Direct spikes were evoked from the contralateral rostral SC (D5b, 6a, 7) and also from the ipisilateral rostral SC (C2). Indirectly activated spikes were evoked from the rostral SC (C1), and the contralateral SC (D5a, 6b, 8). This TRN in the right SC had a descending axon passing through the OPN region, and an ascending axon to the ipsilateral FFH, and also had a commissural collateral to the contralateral rostral SC (For details about how to identify evoked spikes as either direct or indirect spikes, see Shinoda et al., 1976, 1977, 1986; Takahashi et al., 2005a,b, 2007). PPRF, paramedian pontine reticular formation; OPN, omnipause neuron; Abd Nucl, abducens nucleus; PMRF, paramedian medurally reticular formation; Lt, left; Rt, right.

The existence of this kind of commissural collateral causes stimulation of the SC to activate not only TRNs on the same side, but also causes an action potential to propagate along the axons of contralateral TRNs via axon reflex, as described above. This explains why we observe EPSPs followed by IPSPs in vertical ocular MNs and INC IBNs after SC stimulation on either side. This point was confirmed by a midline section between the SCs that eliminated this inadvertent effect of our stimulation method (Sugiuchi et al., 2013). After we subtract away the signals caused by the antidromic stimulation of the commissural collaterals of the TRNs in the opposite SC (axon reflex), it is clear that INC neurons receive monosynaptic excitation from the ipsilateral SC and disynaptic inhibition from the contralateral SC (Sugiuchi et al., 2013). Excitatory commissural connections are mirror symmetric, and exist between the medial sites (upward saccade-representing areas) or between the lateral sites (downward saccade-representing areas) of both rostral SCs (for details about commissural collaterals of TRNs, see Takahashi et al., 2007, 2010). The INC neuron in Figure 2 produces an EPSP and then IPSP regardless of whether we stimulated SC ipsilaterally or contralaterally. Due to the existence of commissural axon collaterals of TRNs, evoked PSPs consisted of EPSPs followed by IPSPs from both SCs. However, in Figure 2B, monosynaptic EPSPs were the shortest in latency and the largest in amplitude, and orthodromic spikes were most constantly evoked from site 1 among ipsilateral SC stimulation, whereas in Figure 2C, disynaptic IPSPs were the largest and their slopes of the falling phase were the sharpest from site 6 among contralateral SC stimulation. Therefore, taking into consideration the existence of commissural excitatory collaterals of TRNs, we infer that the INC neuron in Figure 2 receives the strongest monosynaptic excitatory input from the ipsilateral rostrolateral SC (stimulation location 3) and the strongest inhibitory input from the contralateral rostromedial SC (stimulation location 4). This pattern of SC inputs implies that the cell is an INC IBN for downward saccades. As explained below, we then checked the output pattern of projection of an INC cell to the vertical MN nuclei to confirm this interpretation based on input patterns from the SCs.

We grouped INC IBN neurons into two groups: one group projected to the III nucleus (Fig. 1B), but not the IV nucleus on the contralateral side (Fig. 1C), and the other group projected to the contralateral IV nucleus (Fig. 4D). This latter group received larger excitation from the medial part of the ipsilateral rostral SC where upward saccades are represented, and a much smaller excitation followed by larger inhibition from the lateral part of the contralateral rostral SC where downward saccades are represented. Therefore, this latter group represents the part of the upward saccade system that inhibits the downward saccade system. The former group received larger monosynaptic excitation from the lateral part of the ipsilateral rostral SC and small excitation followed by larger inhibition from the medial part of the contralateral rostral SC. This former group, therefore, belongs to the downward saccade system, which inhibits the upward system. We recorded from more than 150 INC neurons. Among them, 65 INC neurons were classified as putative upward-related IBNs. The remaining 40 INC neurons were classified as putative downward-related IBNs. Other INC neurons could not be classified as either upward- or downward-IBNs.

OPNs inhibit INC IBNs

While our primary goal is to show the existence of the vertical saccade SC–IBN–OPN circuit, OPNs are also known to send axons to the FFH, which contains vertical EBNs (Langer and Kaneko, 1983; Nakao et al., 1988; Ohgaki et al., 1989; Takahashi et al., 2022a). The tonic inhibition provided by this OPN projection onto EBNs prevents eye movements during eye fixations. IBNs are equally involved in saccade generation, inhibiting MNs and quickly relaxing antagonistic muscles, and thus OPN should project to and inhibit IBNs as well. To determine whether OPNs inhibit INC IBNs, we recorded intracellularly from INC IBNs while stimulating the OPN region. The INC neuron in Figure 4 was antidromically activated from both the contralateral IV nucleus (Fig. 4D) and the contralateral INC (Fig. 4E). EPSPs evoked from the ipsilateral rostral SC were the largest when stimulating medially (Fig. 4F, SC1), and IPSPs from the contralateral rostral SC were the largest when stimulating laterally (Fig. 4G, SC8). Therefore, this INC neuron is an upward IBN. Stimulating the contralateral OPN region caused the INC neuron to show larger IPSPs at 1.3 ms, which was preceded by small EPSPs at a latency of 1.0 ms (Fig. 4B). These IPSPs had latencies ranging from 1.2 to 1.8 ms (mean ± SD, 1.5 ± 0.2 ms, N = 21). Latencies of antidromic intrasomatic spikes in OPNs were evoked by stimulation of the INC ranged from 0.6 to 1.7 ms (1.0 ± 0.3 ms, N = 24) (Fig. 8). These values were further supplemented by latencies of antidromic extracellular spikes of OPNs evoked by INC stimulation (1.1 ± 0.4 ms, N = 21). The values indicate that the latencies of the IPSPs are 0.5 ms longer than the latencies of the antidromic spikes in OPNs (p < 0.001, t-test). Although the latency difference is slightly longer for synaptic transmission (usually 0.3–0.4 ms, Eccles, 1964), these IPSPs were regarded as monosynaptic from the OPN region, because stimulation of the INC likely activates stem axons rather than axon collaterals near their terminals. The EPSPs were most likely caused by stimulation of passing TRN axons on their way to PMRF (Fig. 4C), because individual TRNs project to both the midbrain (vertical) BN areas and the horizontal BN areas via the ascending and descending axon collaterals they produce (as exemplified in Fig. 3) (Takahashi et al., 2022b).

Figure 8.

Inhibition of an OPN by the INC and mutual interaction between INC IBNs and OPNs. A, Experimental setup. B,C, Antidromic spikes were evoked in a right OPN from the left FFH (B) and INC (C) at 500 μA. D, Stimulus intensities of the left INC stimulation were adjusted to threshold intensity for the antidromic spikes (260 μA), so that monosynaptic reversed IPSPs (red traces) appeared in about half of the traces (compare the red traces in D and E). E, Large monosynaptic IPSPs were evoked from the same bipolar electrode in the left INC, but with the reversed polarity. F,G, Monosynaptic IPSPs were evoked from the Rt INC (F) and FFH (G).

An example of a downward INC IBN is shown in Figure 5. Antidromic spikes could be observed after stimulation of contralateral III nucleus (Fig. 5D), but not contralateral IV nucleus (Fig. 5E). In addition, the lateral site (Fig. 5G4) produced the largest excitation among ipsilateral rostral SC sites (Fig. 5G), whereas the inhibition was larger as the more medial sites were stimulated among contralateral rostral SC stimulation sites (Fig. 5H), classifying this neuron as a downward IBN. Single pulse stimulation of the OPN region evoked early EPSPs (likely caused by antidromic stimulation of TRN fibers, as described above) followed by IPSPs (Fig. 5Ba), and double-pulse stimulation increased the amplitude of the IPSPs (Fig. 5Bb). Since stimulation of two sites in the horizontal IBN region (PMRF) evoked only much weaker IPSPs in the same neuron (Fig. 5Ca,b), it was concluded that the IPSPs observed in INC IBN must be due to OPN input.

Figure 5.

Inhibition of a downward INC IBN by OPNs. A, Experimental setup. B,C, Synaptic inputs evoked from the contralateral OPN (B) and IBN regions (C). Large IPSPs at 1.2 ms that followed early EPSPs at 0.8 ms appeared only from the OPN region (B), but not from the IBN regions (C). Single pulse (Ba) and double pulse stimulation (Bb). This INC neuron was antidromically activated from the III nucleus (D), but not from the IV nucleus on the contralateral side (E). F, Large monosynaptic IPSPs were evoked from the contralateral INC. G,H, Typical synaptic inputs from the ipsilateral (G) and contralateral SC (H) for this downward INC IBN. The largest EPSPs were evoked from the lateral site 4 of the ipsilateral SC (G) and the largest IPSPs from the medial site 6 of the contralateral SC (H). Calibration in F also applies to B–E and G,H. hIBN: horizontal IBN.

In addition to their inhibition of INC vertical IBNs, OPNs also inhibit (vertical) EBNs in the FFH (as mentioned above). Figure 6 shows an example of such an EBN. During the search for INC IBNs, some neurons penetrated rostral to the INC were not antidromically activated from the contralateral IV nucleus (Fig. 6C), but instead from the ipsilateral IV nucleus (Fig. 6B) (N = 15). The example cell was monosynaptically inhibited by stimulation of the contralateral INC (Fig. 6E). Furthermore, in the example cell, EPSPs caused by ipsilateral SC stimulation were larger from the lateral rostral SC site (Fig. 6F, SC4) and IPSPs due to contralateral SC stimulation were stronger from the medial SC site (Fig. 6G, SC5), indicating that this neuron is an EBN driving downward saccades. Such EBNs are likely located in the FFH, and were observed to be monosynaptically inhibited by OPN stimulation in four cases (Fig. 6D).

Individual INC IBNs send axon collaterals to the OPN region

Previously, INC IBNs have been shown to project not only to their target vertical ocular motor nuclei, but also to the INC and FFH on the opposite side of the brain via axon collaterals. In the horizontal saccade system, IBNs terminate not only on OPNs, but also on contralateral abducens MNs, as well as IBNs and EBNs located contralaterally (Takahashi et al., 2022a). If the vertical IBNs follow the same pattern of projection, then INC IBNs should also terminate on OPNs. To find axonal projection of INC IBNs to the OPN region, we recorded intracellularly from INC neurons identified as IBNs while stimulating the contralateral OPN region (Fig. 7). The penetrated neuron was directly activated from the contralateral INC (Fig. 7C), contralateral FFH (Fig. 7D) and the contralateral IV nucleus (Fig. 7E), as expected of an upward saccade IBN. Spikes were synaptically evoked from the ipsilateral SC, and their latencies were shorter and more fixed than those from the lateral SC, indicating that the spikes were more effectively activated from its medial part (Fig. 7F1,F2), but could not be evoked from the contralateral SC, except its medial part (most likely, due to accidental stimulation of intracollicular commissural collaterals of left TRNs) (Takahashi et al., 2007, 2010). Since we were also able to antidromically activate this same INC neuron at a latency of 0.9 ms via stimulation in the OPN region (Fig. 7B), we can conclude that vertical INC IBNs indeed project to the OPN region with their descending axon collaterals. Latencies of antidromic spikes evoked in INC IBNs from stimulation of contralateral OPN region ranged from 0.4 to 1.2 ms (0.8 ± 0.2 ms, N = 26).

Figure 7.

Projection of a single upward INC IBN to the OPN region with its descending axon collateral. A, Experimental setup. Intracellular recording was made from a neuron in the left INC. B–E, Antidromic spikes were evoked from the OPN region (B), INC (C), FFH (D), and IV nucleus (E) on the contralateral side. F,G, EPSPs with spikes were evoked from the ipsilateral SC (the strongest input from SC1) (F). Disynaptic IPSPs were evoked from the contralateral SC except SC5 (G). Calibration in D also applies to B,C and E–G.

INC IBNs inhibit OPNs

Having shown that putative IBN cells in INC send axons to the OPN region, our next step was to show that these axons make synapses on them and functionally inhibit the OPNs. To do so, we recorded intracellularly from OPNs while stimulating the contralateral INC (Fig. 8). However, our anesthetized preparation prevented us from identifying OPNs via their typical behavioral definition (cessation of tonic firing that is typical of OPNs during saccades or the quick phases of nystagmus). Instead, we identified OPNs using different criteria. We systematically searched for neurons in RIP, identifying them as OPNs when they satisfied the following criteria (Takahashi et al., 2022a): (1) neurons were located in the OPN region mentioned in Materials and Methods; (2) neurons were activated antidromically by stimulation of the FFH, INC and/or the horizontal IBN region on either side of midline; (3) neurons received monosynaptic excitation from the most rostral SC (either ipsilateral or contralateral) due to a unique pathway of a subset of rostral SC cells to OPNs (Raybourn and Keller, 1977; Chimoto et al., 1992; Paré and Guitton, 1994; Büttner-Ennever et al., 1999; Yoshida et al., 2001; Takahashi et al., 2005b); and (4) neurons received disynaptic inhibition from the caudal SC (Takahashi et al., 2022a).

Figure 8 shows a neuron penetrated in the right OPN region. This neuron was antidromically activated from the left FFH (Fig. 8B) and INC (Fig. 8C) at 500 µA. Reducing the left INC stimulus intensity to 260 µA evoked spikes at a latency of 0.9 ms in an all-or-none manner without EPSPs (Fig. 8D, black traces vs red traces). Reversing the polarity of the concentric bipolar stimulating electrode in the left INC caused large hyperpolarizing potentials at a latency of 1.2 ms at 500 µA (Fig. 8E, black traces), which we interpret as including orthodromic (synaptically mediated) signals from INC IBNs. Spontaneous diffusion of Cl⁻ ions into the OPN cell reversed the hyperpolarizing potentials in a depolarizing direction (Fig. 8E, red traces, and also Fig. 8D, red traces), indicating that these hyperpolarizing potentials were IPSPs (Eccles, 1964). This result confirms that INC IBNs inhibit OPNs. Stimulation of the right INC (Fig. 8F) and FFH (Fig. 8G) evoked inhibition in the same neuron. However, these were most likely induced via reflex activation of individual INC IBNs on the opposite side (see the branching pattern of an INC IBN in Fig. 8A). Latencies of INC-evoked IPSPs in OPNs ranged from 0.9 to 1.6 ms (1.2 ± 0.2 ms, N = 15). These latencies are 0.4 ms longer on average than the antidromic spike latencies of the INC IBNs (p < 0.001, t test), and regarded as monosynaptic from the INC IBNs.

INC IBNs suppress OPN activity to trigger vertical saccades

The above electrophysiological investigation clearly shows that INC IBNs directly terminate on contralateral OPNs. Our previous study in the horizontal saccade system showed that horizontal IBNs mediate disynaptic inhibition from the SC to OPNs (Takahashi et al., 2022a). This disynaptic inhibition of OPNs was stronger from more caudal SC stimulation, where larger horizontal saccades are represented (McIlwain, 1986). To determine whether a similar retinotopic topology exists for the vertical INC IBNs, we examined whether INC IBNs that project to the OPN region also receive stronger input from rostral SC (representing the visual vertical meridian and eye movements along it) than caudal SC. Figure 9 shows an example of an INC IBN that projects to the OPN region. This neuron was identified as an upward INC IBN, since it was antidromically activated from the contralateral IV nucleus (Fig. 9B) and additionally from the contralateral INC (Fig. 9C) and FFH (Fig. 9D). This neuron was also antidromically activated from the OPN region at fixed latencies of 0.4 ms at 60 µA (Fig. 9E), but not activated from the horizontal IBN region at 300 µA (Fig. 9F), suggesting that this neuron did not project further caudally than the OPN region. Single stimuli of all sites in the ipsilateral SC evoked EPSPs with spikes, and the latency of the spikes was shorter from the medial sites (Fig. 9G1,2) than from the lateral site (Fig. 9G4). On the other hand, single stimuli of the contralateral SC could not evoke spikes, and double stimuli evoked spikes with fluctuating latencies (Fig. 9H5–7). These spikes were evoked from the medial SC at shorter latencies of 1.2 ms (Fig. 9H5,6) than from the lateral SC (Fig. 9H7), suggesting that the spikes evoked from the contralateral SC might be due to activation of commissural collaterals of left TRNs (shown as a putative commissural axon collateral of a left medial TRN in Fig. 9A). Thus, this INC IBN likely mediates vertical saccade signals from the ipsilateral SC to OPNs.

Figure 9.

Bilateral input pattern from the SCs to an upward INC IBN projecting to the contralateral OPN region. This INC neuron was antidromically activated from the contralateral IV nucleus (B), INC (C) and FFH (D). E,F, Antidromic spikes were evoked from the contralateral OPN region (E), but not from the horizontal IBN region (F). G,H, Inputs from the ipsilateral (G) and contralateral SC (H) to the same INC IBN at 500 µA. The strongest excitation was from site 2 in the ipsilateral SC (G). The spikes evoked from SC 5–7 might be due to activation of axon collaterals of left TRNs (see a putative commissural collateral of the left SC neuron drawn in A based on Takahashi et al., 2010). The lack of spikes in H8 was either due to absence of excitatory input or due to active suppression via the INC.

Stepwise, the latencies of the excitation from SC stimulation to INC IBNs ranged from 0.7 to 1.3 ms (mean ± SD, 1.1 ± 0.2 ms, N = 31), and those of antidromic spikes in INC evoked from the OPN region ranged from 0.4 to 1.2 ms (0.8 ± 0.2 ms, N = 26). Summing from the stepwise latencies, we find a range of 1.1 to 2.5 ms, which lines up reasonably with the disynaptic latencies of SC-evoked IPSPs in OPNs (Takahashi et al., 2022a) when synaptic transmission delay and membrane time constants are taken into account.

To further confirm that the vertical saccade signals from the SC suppress OPNs via INC IBNs, we made a midline section between the two horizontal IBN regions in order to eliminate inhibition via horizontal IBNs. In our previous study, we investigated the distribution of horizontal IBNs labeled after injection of wheat germ-agglutinine-horseradish into the abducens nerve (Fig. 10E; Sugiuchi et al., 2005). Based on this distribution, we made a midline section with a fine blade (Fig. 10F). Figure 10B,C shows the effects of SC stimulation after the midline section (recording intracellularly in a putative OPN cell). First, as a control, we confirmed that disynaptic IPSPs were always evoked via horizontal IBNs from stimulation of the ipsilateral caudal SC before midline section (Sugiuchi et al., 2005). After the midline section, these IPSPs could not be evoked in three horizontal IBNs examined, confirming that the midline section was wide enough to eliminate the projection of horizontal IBNs to the contralateral side. Then, in the same preparation, no short-latency PSPs were evoked from the caudal SCs (Fig. 10C3,4,D7,8), but disynaptic IPSPs with small preceding monosynaptic EPSPs were evoked from either of the rostral SCs (Fig. 10C1,D5). The disynaptic IPSPs evoked by rostral SC stimulation, could be mediated by either horizontal IBNs driven by TRNs in the SC representing small amplitude-horizontal saccades or vertical IBNs driven by TRNs in the SC representing vertical saccades, but our midline section has eliminated the possibility that disynaptic inhibition is due to horizontal IBNs. Thus, we could conclude that the vertical saccade driving signals from the rostral SC inhibited OPNs via INC IBNs (N = 3).

Figure 10.

Effects of a midline section between the horizontal IBN regions on SC-evoked PSPs in an OPN. A, Experimental setup. Thick midline bar between the bilateral horizontal IBN regions indicates the rostrocaudal extent of the midline section shown in (F). B, Partial antidromic spikes were evoked at the initial segment by stimulation of the right FFH at 290 μA. C,D, Synaptic inputs from the left (C) and right SC (D) to an OPN. Short-latency PSPs could not be evoked from the caudal SC sites (C3,4 and D7,8), but disynaptic IPSPs with preceding monosynaptic EPSPs were evoked from the rostral SC sites (C1 and D5). The most rostral SC stimulation sites activated “fixation neurons” exciting this OPN (Gandhi and Keller, 1997, 1999) and vertical saccade neurons simultaneously, so that PSPs in OPNs became monosynaptic excitation followed by disynaptic inhibition. This result confirmed that vertical saccade driving signals from the rostral SC inhibit OPNs via INC IBNs. Calibration in D also applies to B and C. E, Lateral view of the brainstem showing the distribution of horizontal IBNs in the PMRF of the brainstem that were labeled transneuronally after injection of wheat germ agglutinin–horseradish peroxidase into the contralateral abducens nerve in the cat. Labeled neurons are projected onto a parasagittal plane 1.2 mm from the midline. Broken line indicates the dorsal surface of the fourth ventricle in the midline [modified from Fig. 1B in Sugiuchi et al. (2005)]. F, Lateral view of a midline section between the horizontal IBN regions reconstructed from 100-µm-thick coronal sections of the brainstem stained with thionine. This lesion covered the distribution area of the labeled IBNs in E, and sectioned their main crossing axons, because the main axons of horizontal IBNs cross the midline at almost the same level as their cell bodies (Takahashi et al., 2022a).

On the other hand, the monosynaptic EPSPs are thought to be caused by “fixation neurons” in the most rostral SC (King et al., 1978, 1980; Peck, 1989; Gandhi and Keller, 1997, 1999; Takahashi et al., 2022a). However, the functional role of the rostral pole is controversial. It has been proposed that the SC contains a rostrally located fixation neurons (Munoz and Guitton, 1989, 1991; Munoz and Wurtz, 1993a,b). In an alternative view, the rostral SC of monkeys has been found to contain neurons with movement fields for microsaccades that are otherwise similar to movement fields in the more caudal SC (Hafed et al., 2009; Hafed and Krauzlis, 2012). In the present study, the midline section data indicate that TRNs in the most rostral and caudal sites have differential connections with OPNs, and likely have different functional roles.

INC IBNs and OPNs mutually inhibit one another

Figure 8D,E suggest that OPNs and INC IBNs form mutually inhibitory connections. We have confirmed this possibility in an OPN (Fig. 11). This OPN was antidromically activated from the contralateral FFH (Fig. 11B) and INC (Fig. 11C) at 500 µA. In many penetrated OPNs, the spike-generating mechanism deteriorated very rapidly after electrode penetration, possibly due to small soma size, as noted in previous reports of intracellular OPN recordings (Ohgaki et al., 1987; Yoshida et al., 2001; Takahashi et al., 2022a). As a consequence, often only small spike-like depolarizations remained in response to stimulation of the FFH and the IBN region (Fig. 11B,C). These spike-like sharp depolarizations represent antidromic spikes arriving at the soma, due to the fact that they appear in an all-or-none manner and at the same thresholds which had activated full antidromic spikes in the same cells before their deterioration (Fig. 11C). Therefore, the small spikes are action potentials generated at an initial segment (Eccles, 1964). Figure 11D shows the result when the stimulation strength was adjusted, such that the antidromic spikes appeared in about half of stimulation trials (185 µA). In those trials where the INC IBN stimulation failed to elicit antidromic spikes, we could clearly observe inhibition at a short latency of 1.0 ms, which was followed by larger IPSPs with longer latencies. These short latency IPSPs were small, because the threshold for antidromic activation of this INC IBN was very low, so that it was impossible to increase the stimulus intensity in order to get larger short latency IPSPs and because increased stimulus strength evoked antidromic spikes, which masked the appearance of the short-latency IPSPs. Thus, this OPN is inhibited by INC IBNs, and in turn, projects back to and inhibits contralateral INC IBNs. We frequently observed this type of reciprocal inhibition between OPNs on one side and INC IBNs on the opposite side (Fig. 8). This mutual negative feedback loop is comparable to a positive feedback loop in function (since reduced IBN inhibition by OPNs would increase IBN firing due to the excitatory SC input, producing further suppression of OPNs). This mechanism could quickly suppress OPNs for initiating vertical saccades.

Figure 11.

Mutual inhibitory connections between OPNs and INC IBNs. A, Experimental setup. Intracellular recording from a left OPN. B–D, Antidromic partial and full spikes evoked from the right FFH (B) and the right INC (C). D, Stimulation of the right INC IBN region at around 185 μA evoked antidromic spikes or IPSPs in about half of the traces, indicating that this OPN receives inhibition from the INC IBNs, and in turn projects back to and inhibits INC IBNs. Small IPSPs were evoked at 1.0 ms and they were followed by longer latency-IPSPs that might be evoked by stimulation of presynaptic terminal fibers contacting INC neurons. E, Monosynaptic IPSPs were evoked from the ipsilateral INC, most likely via reflex activation of axons from right INC IBNs. Calibration in C also applies to B, D and E.

Discussion

The present study provides evidence that INC IBNs monosynaptically inhibit OPNs, and such OPNs, in turn, monosynaptically inhibit INC IBNs and FFH EBNs. Thus, INC IBNs that receive saccade driving signals from the rostral SC likely inhibit OPNs at the onset of vertical saccades to allow their initiation and further inhibit OPNs during saccades for their maintenance (latch function). This circuit for vertical saccades is very similar to the circuit for horizontal saccades (Takahashi et al., 2022a), in which caudal SCs inhibit OPNs via PMRF IBNs before horizontal saccades. Since the original model proposed by Robinson (1975), almost all computational models have considered OPNs to play a critical role in initiating saccades. This concept is well proven for the horizontal system, but less so for the vertical one (Takahashi and Shinoda, 2018). Furthermore, these models were uncertain of the identity of the inhibitory neurons that suppress OPNs at the onset of saccades. We have previously demonstrated that instead of an independent saccade triggering pathway, the same horizontal IBNs that carry saccade driving signals from the caudal SC inhibit OPNs to help triggering horizontal saccades (Takahashi et al., 2022a). In the present study, we have demonstrated that the same basic concept holds for the vertical saccade system, and that INC IBNs are the counterparts to the horizontal IBNs in the PMRF.

Some INC neurons are known to project caudally to the brainstem and spinal cord (Isa and Sasaki, 1992), but this is the first direct demonstration that INC IBNs inhibit OPNs with their descending collaterals. Although it is known that OPNs project to midbrain areas (Langer and Kaneko, 1983; Nakao et al., 1988; Ohgaki et al., 1989), the present study clearly reveals that OPNs monosynaptically inhibit vertical FFH EBNs and INC IBNs. Mutual inhibitory connections between INC IBNs and OPNs were found in the vertical saccade system. These are analogous to mutual inhibitory connections between PMRF IBNs and OPNs in the horizontal saccade system.

If INC IBNs are the source responsible for inhibition of OPNs just before and during vertical saccades, there should be some IBNs that show burst activity before the cessation of OPN tonic activity. Then, OPNs will be quickly suppressed by the positive feedback connection between IBNs and OPNs. In fact, two types of IBNs, short-lead and long-lead IBNs (LLIBNs), have been reported during horizontal saccades of awake monkeys (Scudder et al., 1988; Cullen and Guitton, 1997). LLIBNs start firing before the onset of saccade velocity by 21 ± 7.9 ms (N = 12/28) (Cullen and Guitton, 1997). The existence of such a subset of IBNs that start firing earlier than the cessation of OPN activity supports our interpretation that horizontal IBNs cause a decrease in tonic activity of OPNs before saccade initiation (Takahashi et al., 2022a). In vertical saccades in awake monkeys, latencies of the first spike of INC BNs relative to vertical saccade onset were reported to have a mean of 10.4 ms (range 5–23 ms) by Helmchen et al. (1996). Similarly, 15% of these neurons had latencies that preceded the saccade by more than 15 ms (16–23 ms) and would consequently be classified as long-lead BNs (Hepp et al., 1989). This suggests the existence of such a subset of INC IBNs in vertical saccades and so supports our interpretation.

The present study shows that both horizontal and vertical IBNs converge onto OPNs. This finding explains the omnipause nature of OPN activity in all directions, and also suggests a source for the origin of GABAergic terminals on OPNs. OPNs are glycinergic and have GABAergic and glycinergic terminals on them (Horn et al., 1994). PMRF IBNs are glycinergic, terminating on abducens MNs (Spencer et al., 1989). These same horizontal IBNs can terminate on OPNs via glycinergic collaterals (Takahashi et al., 2022a). Vertical INC IBNs are GABAergic (Horn et al., 2003), and a portion of these cells terminate on trochlear MNs (Sugiuchi et al., 2013). The present study showed that individual INC IBNs directly terminate on OPNs and trochlear MNs, indicating that terminals of INC IBNs are among GABAergic terminals on OPNs. On the other hand, one of the origins of GABAergic terminals on OPNs was central mesencephalic reticular formation (Wang et al., 2013). In a pharmacological study, strychnine (glycine antagonist) and bicuculine (GABA antagonist) were iontophoretically applied onto an OPN (Kanda et al., 2007). The former increased spike activity during fixation, and greatly decreased the pause duration of the OPN, although some pause still was present. This remaining pause might be caused by GABA, but bicuculine application did not shorten the pause duration. Since oblique saccades were included in their study, and in light of the present findings, further studies are required to determine a possible contribution of GABA to initiation of vertical saccades.

The mutual inhibitory loop between IBNs and OPNs is common for both horizontal and vertical saccade systems (Takahashi et al., 2022a), and can work in such a way that increased IBN activity may inhibit OPN tonic activity, which in turn disinhibits IBN activity. This mutual inhibitory negative loop, effectively equivalent to a positive feedback loop between IBNs and OPNs, will cause depolarization in IBNs from the hyperpolarized state maintained by tonic OPN inhibition during fixation. In addition to excitatory saccadic driving signals from the SC, such abrupt disinhibition due to decreased OPN inhibitory activity may accelerate activation of low-threshold Ca²⁺ channels for spike burst in IBNs and EBNs (Serafin et al., 1990; Miura and Optican, 2006). This will contribute to an increase in saccade peak velocity. On the other hand, during saccade termination, decreased IBN activity may quickly increase OPN activity due to the same double negative loop, resulting in rapid termination of burst activity in IBNs and EBNs. The time course of the decay phase of the membrane hyperpolarizing potentials recorded from OPNs is strikingly similar to the falling phase of saccade velocity (Yoshida et al., 1999). This observation is consistent with the present finding that OPNs are monosynaptically inhibited by IBNs, which encode saccade velocity signals. This interpretation fits with a previous lesion study in the monkey where ibotenic acid injections in the OPN region increased saccade duration and decreased peak velocity (Kaneko, 1996). Suppression of high gain BN activity by OPN tonic activity plays an essential role in maintaining stable fixation for acquiring visual information. The timing of saccade termination is most important for accurate acquisition of new visual information. IBN suppression of OPNs reduces at the termination of saccades, and then OPNs resume spiking, so that IBNs further reduce spike activity and result in rapid saccade termination. Therefore, double negative loops between IBNs and OPNs have crucial roles in both rapid saccade initiation and termination.

A human report also provides insight into our saccade generation model. The only neurological symptom found in this patient was slow horizontal saccades (Uemura et al., 1987). The quick phases of caloric and optokinetic nystagmus were also abnormally slow, but vertical saccades appeared normal. Later careful neuropathological analysis revealed fibrous gliosis of the midline in the pons and medulla without loss of cells in the EBN and OPN areas (Büttner-Ennever et al., 2008). The midline gliosis was most pronounced at the level of the caudal pons, where horizontal IBN axons decussate across the midline (Sugiuchi et al., 2005). It seems likely that this midline gliosis could account for slowing of horizontal saccades, as expected in the midline lesion between the horizontal IBN regions in our experiment. The rostral mesencephalon was not affected by the scar, which might explain the preservation of vertical saccades due to the preserved reciprocal connections between INC IBNs and OPNs.

Figure 12 summarizes our current understanding of the neural circuit in which TRNs in the rostral SCs trigger vertical saccades (and how the downstream circuits suppress saccades during fixation via OPN projections). “Fixation TRNs” in the rostral pole tonically excite OPNs causing them to suppress high gain activity of FFH EBNs and INC IBNs during fixation (data not shown in Fig. 12, see Takahashi et al., 2022a). In contrast, vertical INC IBNs excited by more medially or laterally situated rostral TRNs inhibit OPNs to initiate vertical saccades and continue inhibiting OPNs for saccade duration. The reciprocal inhibition via lNC IBNs exists between the upward vertical system on one side and the downward vertical system on the opposite side (red and blue neurons, respectively). For pure vertical saccades, TRNs in symmetric sites of the two SCs must be simultaneously activated via excitatory commissural connections (Takahashi et al., 2007; Takahashi and Shinoda, 2018). Therefore, for upward saccades, TRNs in the rostromedial SCs activate upward FFH EBNs and INC IBNs on both sides, and such INC IBNs further inhibit OPNs for saccade initiation and, at the same time, inhibit downward EBNs and IBNs bilaterally for reciprocal inhibition. Similarly, for downward saccades, TRNs in the rostrolateral SCs activate downward FFH EBNs and INC IBNs on both sides, and such INC IBNs further inhibit OPNs and upward EBNs and IBNs bilaterally.

Figure 12.

Schematic summary diagram showing the neural circuit for triggering and suppressing vertical saccades via INC IBNs and OPNs. Red neurons (u-EBN, u-IBN, SR and IO, u-SC) indicate neurons in the upward saccade system with upward ipsitorsional on-direction (thick red arrows) [Takahashi et al. (2007), their Fig. 14A]. Blue neurons (d-EBN, d-IBN, IR and SO, d-SC) indicate neurons in the downward saccade system with downward ipsitorsional on-direction (thick blue arrows) [Takahashi et al. (2007), their Fig. 14B]. The proposed SC–INC IBN–OPN pathway is highlighted with heavier red lines. Open neurons, excitatory neurons; filled neurons, inhibitory neurons. Red filled lines indicate upward saccade driving pathways, and blue dashed lines indicate downward saccade driving pathways that are suppressed by INC IBNs in the contralateral upward saccade system during upward saccades (Takahashi and Shinoda, 2018). Note that for pure vertical upward saccades, TRNs in the upward ipsitorsional saccade areas in the rostromedial SCs are simultaneously activated due to excitatory SC commissural connections in order to cancel torsional components. IR, inferior rectus; SR, superior rectus; IO, inferior oblique; SO, superior oblique; PPRF, paramedian pontine reticular formation; OPN, omnipause neuron; Abd nucl, abducens nucleus; PMRF, paramedian medullary reticular formation.