Neural mechanisms of oculomotor abnormalities in the infantile strabismus syndrome

Abstract

Infantile strabismus is characterized by numerous visual and oculomotor abnormalities. Recently nonhuman primate models of infantile strabismus have been established, with characteristics that closely match those observed in human patients. This has made it possible to study the neural basis for visual and oculomotor symptoms in infantile strabismus. In this review, we consider the available evidence for neural abnormalities in structures related to oculomotor pathways ranging from visual cortex to oculomotor nuclei. These studies provide compelling evidence that a disturbance of binocular vision during a sensitive period early in life, whatever the cause, results in a cascade of abnormalities through numerous brain areas involved in visual functions and eye movements.

childhood-onset strabismus encompasses a heterogeneous set of disorders of binocular visual development with different clinical features and different ages of onset. While acquired accommodative esotropia and intermittent exotropia are among the most common, the form of childhood strabismus with the earliest age of onset is the infantile strabismus syndrome, so named for the unique constellation of striking oculomotor abnormalities that develop along with the ocular misalignment. Not coincidentally, given the well-defined and unusual features of the syndrome, it is this form of strabismus that has been most studied at a neurophysiological level, thanks to nonhuman primate (NHP) models in which a chronic misalignment of the eyes is induced in infancy. In a recent review, Das (2016) provides a broad overview of issues related to the use of NHP models in the study of strabismus, including the behavioral characteristics, neurophysiology, methods for inducing strabismus in experimental animals, and studies that aim to directly test clinical treatments. The present review will focus primarily on the neurophysiology of infantile strabismus, with the goal of providing a detailed assessment of the known abnormalities across a variety of brain areas. Wherever possible, we attempt to consider how abnormalities in a given area might influence the development of tuning properties of neurons in downstream areas.

Vision

Primate binocular vision.

Optimal visual performance of human and nonhuman primates depends heavily on frontally placed eyes, retinal specializations, and binocular vision. The oculomotor system is responsible for maintaining precise eye alignment during fixation and eye movement to support clear vision. Binocular coordination of eye movements ensures that the image of an object of interest is placed on or near the fovea of each eye. Other equidistant scene elements fall on the Vieth-Müller circle or horopter (Howard and Rogers 2002; Leigh and Zee 2015) and impinge on corresponding points of each retina. This precise active eye alignment function, called “motor fusion,” allows later stages of the visual system to combine information from each eye to produce a percept of single vision, known as “sensory fusion.” Sensory fusion also supports a percept of stereoscopic depth for targets located slightly in front or distant to the horopter (Panum’s fusional area). A target displaced outside Panum’s area is perceived as two distinct objects. Failure of the oculomotor system to develop correct eye alignment and coordinated binocular eye movements is associated with strabismus and visual suppression instead of motor and sensory fusion.

Retinal specialization.

The primate retina is not uniform from center to periphery. Rather, it contains a central “foveal” region where the density of retinal neurons is highest. The fovea is located at the optical axis of the eye subtending ~3° of visual angle (Curcio et al. 1987). The primate visual system uses this feature to emphasize central vision where acuity is highest. The oculomotor system supports central vision by directing the eyes so that objects of interest are imaged on or near the fovea. Extrafoveal visual acuity diminishes steeply with increasing distance from the foveal center.

Retinal projections.

Retinal ganglion cells (RGC) emanating from the eye form different visual pathways supporting different aspects of visual perception, oculomotor control, and homeostatic functions. Table 1 summarizes these pathways and their primary roles in vision and eye movements. Here we focus on those pathways that play a role in supporting clear vision and binocular coordination of eye movements. These pathways deliver visual information related to the contralateral visual hemifield to primary visual cortex (V1) and the superior colliculus (SC). This is achieved by a partial decussation at the optic chiasm such that RGCs from the nasal retina project to the contralateral side, and those of the temporal retina project ipsilaterally, resulting in binocular representation of each visual hemifield in the contralateral hemisphere. Disorders in the pattern of RGC projections (overdecussation) are sometimes associated with strabismus in cases of albinism (Guillery et al. 1984).

Table 1.

Retinal pathways supporting different visual and oculomotor functions

Retinal Projections to	Major Function in Vision	Major Eye Movement
Lateral geniculate nucleusa,b	Principal pathway for visual perception Retinotopic mapping for visual location and perception (magnocellular and parvocellular)h	Volitional saccades, smooth pursuit, vergence, ocular following
Superior colliculusc	Retinotopic map for visual orienting, visual attentionh	Saccades, microsaccades, and smooth pursuit
Accessory optic system terminal nuclei (LTN, DTN, MTN) and pretectal nucleus of the optic tract (NOT)d	Visual-vestibular interaction and optokinesish	Optokinesis and modifying vestibular ocular reflex
Pregeniculate nucleuse	Gating visual stimulih	Saccades
Pulvinarf	Visual attentionh	Fixation and saccades
Hypothalamus (SCN)g	Light entrainment of circadian rhythmsh	None

LTN,  lateral terminal nucleus; DTN,  dorsal terminal nucleus; MTN,  medial terminal nucleus; NOT,  nucleus of the optic tract; SCN,  suprachiasmatic nucleus.

^aSherman (2005);

^bCallaway (2005);

^cWurtz and Albano (1980);

^dGiolli et al. (2006);

^eLivingston and Mustari (2000);

^fItaya and Van Hoesen (1983);

^gHannibal et al. (2014);

^hLeigh and Zee (2015).

The retinal-geniculo-striate pathway plays a dominant role in visual perception. RGCs from each eye deliver visual information to alternating layers of the lateral geniculate nucleus (LGN) for eventual combination in V1. Neurons in the LGN are sensitive to visual information associated with either the left or right eye. Furthermore, different classes of RGCs project to different layers of the LGN (Dacey 2000; Nassi and Callaway 2009). The non-color-coding, large cell (alpha) RGCs project to the ventral magnocellular layers of the LGN. The color-coding medium (beta) RGCs project to the dorsal layers of the LGN and the small (gamma) class RGCs project to intercalated layers of the LGN. These pathways support different aspects of visual-oculomotor function.

Signals from left and right eyes are brought together in V1, mostly outside layer 4, to create binocularly sensitive neurons essential for sensory fusion. A special group of binocularly sensitive neurons in V1, V2, and the middle temporal visual (MT) area extracts information related to the specific amounts of retinal image disparity produced by three-dimensional objects that are located near the point of fixation along the horopter, in Panum’s fusional area (see Primate binocular vision). These neurons are sensitive to the relative disparity of visual target or scene components with respect to locations in front of or distant to the fixation point (Harrold and Grove 2015; Turski 2016).

The retinal-geniculo-striate pathway is essential for high acuity visual function in primates. Furthermore, a large proportion of V1 is dedicated to central vision (e.g., central 5–10°). The peripheral retina plays an important role in visual orienting responses, motion perception, and retinal image stabilization through the optokinetic and ocular following responses. Lesions in the retinal-geniculo-striate pathway produce characteristic deficits in visual perception and visually guided eye movements.

Extrastriate visual areas.

Visual information from V1 is distributed to extrastriate cortical areas where further processing of information is parsed into major components (Gattass et al. 1990; Goodale and Milner 1992). For example, cortical areas in the “dorsal stream” feature visual position, motion direction, and disparity information (DeAngelis 2000; Roe et al. 2007). These areas play a role in saccades, smooth pursuit, and vergence to visual targets. Cortical areas in the “ventral stream” (which involves the temporal lobe) play a dominant role in object vision, face recognition, and color perception (Goodale and Milner 1992). There are extensive cortical connections between distinct visual areas (Felleman and Van Essen 1991; Van Essen 2004). Eye movements are driven by cortical projections to specific brain stem regions that are related to different classes of eye movement control (Büttner-Ennever 2006; Leigh and Zee 2015). Lesions of extrastriate cortical areas are associated with specific visual and eye movement deficits.

Different types of eye movements serve vision.

Primate vision is supported by both reflexive (optokinetic reflex, or OKR, and vestibular ocular reflex, or VOR) and volitional (smooth pursuit, saccades, vergence, fixation) eye movements. The VOR moves the eyes in opposition to head movement, independently of visual input, helping to preserve clear vision during body movement. The OKR provides additional support by moving the eyes in the direction of large field visual motion, such as occurs with visual input during head movements. The VOR and OKR systems therefore attempt to stabilize the visual scene across the retina. The fovea demands special attention in primate vision. Volitional smooth pursuit, saccades, and vergence are designed to place the image of an object of interest on the fovea of both eyes. Fixation is an active process designed to maintain the image of a stationary target on the fovea (Martinez-Conde et al. 2013). Table 2 shows the different eye movement systems, their essential error signals, and common disorders affecting them.

Table 2.

Eye movements, error signals, and associated disorders

Eye Movement Oculomotor Function	Error Signals	Developmental or Acquired Disorders
Vestibular ocular	Acceleration of head	Dizziness and vertigo
Optokinetic	Large-field retinal image motion	Abnormal visual-vestibular calibration
Fixation	Target position	Fixation errors, drifts, saccadic intrusion
Saccades	Target position,velocity	Hypometric, hypermetric or disconjugate saccades
Smooth pursuit	Target position, velocity	Nasotemporal asymmetry, abnormal gain control
Vergence	Retinal disparity, accommodative blur, target proximity, velocity	Abnormal AC/A, convergence insufficiency, defective disparity vergence
Gaze holding	Eye position, visual motion	Latent nystagmus, CN, GEN

AC/A, ratio of accommodative convergence to accommodation; CN, congenital nystagmus; GEN, gaze-evoked nystagmus.

Postnatal Development of Oculomotor and Visual Sensitivity

Primate visual and eye movement functions are immature at birth. Coordinated binocular visual and eye movement experience during an early sensitive period are required for normal maturation and function. This cycle of normal visual-oculomotor experience can be disrupted by prenatal defects in brain circuits and/or the eye per se. Similarly, impaired postnatal experience can lead to miscalibration of neural circuits at multiple levels, from visual cortex to motor neurons (Horton and Hocking 1997; Mustari and Ono 2011; Tychsen and Lisberger 1986; Walton and Mustari 2015).

Human and nonhuman primates (NHPs) follow a similar developmental sequence for visual acuity and eye movement development but with different timing (Boothe and Fulton 2000; Kiorpes 2006). For example, a 1-wk-old monkey has comparable grating visual acuity to that of a 1-mo-old human infant. Near mature acuity values are found at 6 wk and 6 mo for NHPs and human infants, respectively. Therefore, the first 6 wk of life in a NHP comprise a sensitive or critical period for visual acuity development. Other aspects of visual function such as visual motion processing may have extended sensitive periods compared with acuity development (Kiorpes 2016; Kiorpes and Movshon 2004; Kourtzi et al. 2006). Similarly, aspects of the development of oculomotor function may also have extended sensitive periods related to development of motion processing (Kiorpes 2016). Numerous studies have shown that the development of binocular visual sensitivity of neurons in V1 and associated binocular motor coordination depends on early visual experience (Hubel and Wiesel 1970; Kumagami et al. 2000; Nakatsuka et al. 2007; Smith et al. 1997). Early disruption of binocular visual experience leads to strabismus in primates (see Modifying visual experience). The primary site involved is thought to be located in V1.

The degree to which visual and oculomotor function is compromised by abnormal early visual experience depends on the onset time, duration, and type of disruption. As explained above, the first 6 wk of life are most important for the development of normal binocular visual acuity and eye alignment in NHPs (Horton and Hocking 1997; Hubel et al. 1977; LeVay et al. 1980; Tusa et al. 1991; Tychsen 2007). Appropriate visual experience is necessary to develop and refine the spatial, binocular, and disparity sensitivities of striate cortical neurons (Chino et al. 1997; Harwerth et al. 1990; Kiorpes et al. 1998; Kourtzi et al. 2006; Nakatsuka et al. 2007; Zhang et al. 2005). Although there are binocular and disparity sensitive neurons in V1 and V2 present at birth, infant cortical neurons have less mature properties in terms of speed, direction, and disparity sensitivity compared with mature V1 neurons (Blakemore and Van Sluyters 1975; Chino et al. 1997; Hubel and Wiesel 1963; Zheng et al. 2007). It is likely that refinements of disparity sensitivity and eye alignment are codependent. Disruption of this interaction likely underlies strabismus produced by abnormal early sensory experiences (Chino et al. 1997; Kumagami et al. 2000; Tusa et al. 2002).

The cascade of events leading to abnormal visual-oculomotor function could start at different loci in the visual and oculomotor circuitry. For sensory induced strabismus, disruption of V1 binocular sensitivity is certainly involved in loss of sensory and motor fusion. Loss of sensory fusion and maldevelopment of binocular disparity sensitive neurons may deprive critical cortical and brain stem centers of the error signals needed to guide the development of orthotropic eye alignment. Loss of these binocular signals also affects other eye movement systems (Das 2011; Fleuriet et al. 2016; Kiorpes et al. 1996; Mustari and Ono 2011; Tychsen et al. 1996; Walton and Mustari 2015; Wong et al. 2003). Less is known about the role of cortical areas outside V1 in the abnormal visual-oculomotor behavior of strabismic subjects, although we do know that neurons in extrastriate visual areas middle temporal area (MT) and middle superior temporal area (MST) have altered binocular sensitivity in strabismic NHPs (Kiorpes et al. 1996; Mustari and Ono 2011; Mustari et al. 2008).

Developmental Disorders in Oculomotor Function Associated with Abnormal Early Experience

Development of the systems that control eye movements, gaze holding, and eye alignment are all susceptible to alterations of early visual experience (Das and Mustari 2007; Economides et al. 2007; Tusa et al. 1991, 2001; Tychsen 1993; Tychsen and Lisberger 1986). This is because each of these eye movement systems uses visual and eye movement signals to generate appropriate oculomotor commands (Lisberger and Pavelko 1989; Lisberger and Westbrook 1985) and to provide error signals to calibrate oculomotor output (for a review see Leigh and Zee 2015). Altered binocular visual experience, early in life, has a more profound impact on oculomotor performance, eye alignment, and visual capability than similar disruptions with a later onset (Mustari et al. 2001; Quick and Boothe 1989; Tusa et al. 1991, 2001, 2002; Tychsen and Lisberger 1986; Wong et al. 2003). For example, nystagmus resembling latent nystagmus occurs in monkeys following surgical induction of esotropia within the first 2 wk of life, but not if esotropia is induced in a mature animal (Kiorpes et al. 1996; Mustari et al. 2008; Wong et al. 2003). An important clinical syndrome of disordered oculomotor system development is seen in human infants with infantile esotropia (a.k.a. congenital esotropia) (Kiorpes and Movshon 2004; Tychsen 1993; Tychsen et al. 2004).

The Infantile Strabismus Syndrome

The infantile strabismus syndrome is associated with latent nystagmus (LN) and persistence of the immature nasalward bias in monocular smooth pursuit and monocular optokinetic nystagmus (OKN) (Leigh and Zee 2015; Schor 1993; Tychsen 2007; 1993; Tychsen and Lisberger 1986). The nasalward bias in smooth pursuit or OKN consists of near unity gain (eye velocity/visual stimulus velocity) tracking during nasalward visual motion but low gain during temporalward visual motion under monocular viewing conditions. For smooth pursuit, the direction of the nasalward bias is dependent on the eye that is fixating. Although strabismus, LN, and asymmetric monocular pursuit are typically found together in cases of infantile esotropia (Schor et al. 1997), each of these disorders involves different neural sites.

Additional oculomotor abnormalities that are common in the infantile strabismus syndrome are pattern strabismus and dissociated vertical deviation (DVD). In concomitant strabismus, the misalignment of the eyes remains constant across different orbital positions. In contrast, in pattern strabismus the vertical and horizontal misalignments are influenced by the positions of the eyes along the orthogonal axis. This incomitance is called pattern strabismus because plotting the positions of the left and right eyes during pursuit or static fixation often forms either an “A” or “V” pattern, whereas, in those with concomitant strabismus, plots of left and right eye positions produce parallel lines. Pattern strabismus is typically characterized in terms of oblique muscle over- or underaction, but this clinical description does not fully capture the underlying cause. Mounting evidence indicates that the etiology is at least partially supranuclear in origin (Das and Mustari 2007; Ghasia et al. 2015; Joshi and Das 2011; Walton and Mustari 2015; Walton et al. 2013). DVD is a unique phenomenon characteristic of early-onset disruption of binocular visual development, in which occlusion of either eye results in elevation, abduction, and extorsion of the occluded eye (Wright and Strube 2012).

Abnormalities of Eye Muscles and Orbital Tissues

Abnormalities of the extraocular muscles themselves can be seen in strabismus. Indeed, muscle changes are the root cause of strabismus in some forms of acquired strabismus, such as in thyroid ophthalmopathy, but changes in the motor “plant” are also seen in other types of strabismus. In a 2012 study, MRI imaging was used to compare the extraocular muscles in 12 human patients with concomitant esotropia of mixed etiology (excluding paretic and thyroid ophthalmopathy-related strabismus) and 13 normal controls (Schoeff et al. 2013). In the strabismic patients, the cross-sectional area of the medial rectus muscles averaged 39% larger than in normal controls. Medial rectus contractility was also significantly higher for the esotropic patients.

Altick and colleagues (2012), in a study of extraocular muscle samples that were mostly from medial rectus muscles resected in exotropic patients, identified 22 muscle-specific genes that were downregulated in these muscles. The proteins encoded by these genes are believed to play roles in muscle contractility, and the extracellular matrix. Agarwal et al. (2016) also found changes in gene expression in muscles from strabismic individuals. Cheng and colleagues (2003) showed changes in gene expression and extraocular muscle fiber types in monkeys with early-onset monocular deprivation.

There is also evidence that some cases of pattern strabismus are associated with abnormalities of the orbital tissues. A 1995 MRI imaging study showed that all of the recti extraocular muscles pass through fibromuscular pulleys that determine their pulling directions (Demer et al. 1995). A subsequent MRI study showed that these pulleys could be mislocated in some patients with pattern strabismus (Clark et al. 1998). A more recent MRI study showed that rectus muscle pulleys are misplaced in human subjects with pattern exotropia, but not in patients with concomitant exotropia (Hao et al. 2016). Another recent MRI study showed that rectus muscle pulleys can also be misplaced in patients with superior oblique palsy (Suh et al. 2016).

Abnormalities of Visual Processing Associated with Infantile Strabismus

In addition to the oculomotor components of the disorder, infantile strabismus syndrome is associated with abnormalities of visual perception, such as decreased stereoacuity, facultative suppression of the nonfixating eye, which can alternate with alternating fixation, and directional asymmetry of cortical motion visual evoked responses. Esotropic infants may also develop amblyopia in one eye, though many do not (Schor et al. 1997).

Regardless of the initial cause of strabismus, infantile strabismus is always associated with a disruption of binocular vision during a sensitive period early in life. Although most neural pathways are formed before birth, there is considerable development and subsequent pruning of synaptic connectivity during early postnatal life (Cragg 1975; LeVay et al. 1978; Sanes et al. 2006). Single neurons in the visual system depend on experience to develop mature tuning characteristics (Kiorpes and Movshon 2004). Infantile strabismus also leads to reductions in disparity sensitivity and the number of binocularly responsive units in V1 in monkeys (Crawford and von Noorden 1979; Kumagami et al. 2000; Mori et al. 2002). Reductions in binocular visual sensitivity have also been reported for area MT (Kiorpes et al. 1996) and MST (Mustari et al. 2008) of esotropic monkeys.

Disruption of cortical binocularity at the level of, at least, V1 and MT/ MST could trigger a cascade of maldevelopment of neurons throughout the oculomotor system. For example, Tychsen and Lisberger (1986) reported that subjects with early-onset esotropia perceived visual stimuli moving nasalward to be traveling faster than stimuli moving in the opposite direction (during monocular viewing). They also had defective monocular smooth pursuit during attempted pursuit of target moving in a temporalward direction. In contrast, their nasalward smooth pursuit was essentially normal. Tychsen (1999) also found that subjects with early-onset strabismus perceived a flickering grating to be moving in a nasalward direction even though normal observers typically perceived such stimuli as stationary. This aberrant perception might be due, in part, to gaze-holding disorders (LN), but it could also be related, for example, to a bias in the population of visual motion direction sensitive neurons in MT (DeAngelis and Uka 2003; Kiorpes et al. 1996).

Infant monkeys raised with 20-diopter prism goggles that deviate the vertical axis of one eye and the horizontal axis of the fellow eye for the first 3, 12, or 24 wk of life had progressively more severe pursuit disorders (Tychsen 2007). Longer durations of prism experience were associated with more disruption of V1 binocularity (see Modifying visual experience). If the disruption of binocular vision was of short duration (3 wk), visual cortical binocularity was able to recover.

When strabismus occurs early in life, subjects typically do not experience diplopia because vision is suppressed in the eye that is not currently fixating the target of interest (Von Noorden and Campos 2002). When visual acuity is good in both eyes, subjects may alternate fixation. Fixation switching is thought to be driven by suppression of visual information derived from specific areas of the retina (Sireteanu 1982; Steinbach 1981; van Leeuwen et al. 2001). For example, when a saccade target activates the nasal retina of one eye and the temporal retina of the other, both human subjects with intermittent exotropia (Economides et al. 2012, 2014) and monkeys with experimentally induced exotropia (Agaoglu et al. 2014) typically use the eye that received nasal stimulation to acquire the target. This result is in good agreement with maps of suppression scotomata derived from dichoptic presentation (Economides et al. 2014). Similar experiments in esotropes have produced less consistent results, although there seems to be general agreement that suppression affects signals derived from at least some portions of the nasal hemiretina (Agaoglu et al. 2014; Brodsky and Klaehn 2017; Joosse et al. 1997; Pratt-Johnson and Tillson 1984; Sireteanu 1982).

Experimental Induction of Strabismus in Nonhuman Primates

The eyes are frequently slightly misaligned at birth in human and NHPs (Archer et al. 1989; O’Dell and Boothe 1997). Allowing normal binocular visual experience during the first 12 wk of a macaque’s life usually leads to normal eye alignment. Disruption of normal binocular visual experience during this sensitive period can be used to induce strabismus, using one of two broad categories of approaches. In the first, binocular visual experience is disrupted optically without directly altering the muscle plant (atropine, alternating monocular occlusion, prism-rearing). In the second, the intervention is at the level of the eye muscles (eye muscle surgery, botulinum-A neurotoxin). In this section, we will also discuss possible future genetic or molecular approaches.

Modifying visual experience.

Different techniques that disrupt visual input have been shown to be effective for creating experimental strabismus. This set of techniques aim to visually deprive one or both eyes. This can be done either by occluding one or both eyes for an extended period or by alternately covering each eye (opaque contact lens) for 24-h periods (alternating monocular occlusion; AMO). Occluding both eyes for the first 25–40 days of life leads to other deficits in addition to eye misalignment, such as a strong latent nystagmus (Tusa et al. 2002). If the binocular deprivation is extended to 55 days after birth, gaze-evoked and pendular nystagmus also occur. These models are not optimal for studying infantile strabismus, because they are confounded by the effects of bilateral form deprivation amblyopia and sensory nystagmus. Occluding only one eye leads to severe unilateral amblyopia (Cheng et al. 2008). AMO prevents binocular visual fusion during the early sensitive period and results in strabismus. However, AMO rearing is not associated with significant spontaneous nystagmus or strong amblyopia. This is attributable to the fact that the AMO prevents binocular competition while allowing each eye to receive clear vision on alternating days.

In addition to occlusive visual deprivation methods, optical methods using prisms can also induce strabismus. By fitting prism goggles on the infant macaque, binocular coordination related to corresponding points on the retina is prevented. This is because the applied prism power, often a combination of horizontal and vertical prism power, deviates the optical axis of each eye beyond the capacity of the visual system for sensory or motor fusion (Crawford 1996). This technique also has the advantage of a decreased risk of form deprivation amblyopia (Smith et al. 1992), compared with occlusion methods (although unilateral strabismic amblyopia may still develop). After wearing prisms for 12 to 20 days, monkeys can be observed alternating fixation between the eyes. It has been shown that rearing with dissociating prisms for 60 days is enough to severely reduce the number of binocular neurons in V1 (Crawford and von Noorden 1980), and this loss cannot be recovered. All of the techniques described above have been shown to trigger abnormalities in oculomotor structures, as will be discussed below.

Extraocular muscles (surgical methods).

Eye muscle surgery can be used to improve eye alignment early in life to treat strabismus; conversely it can be used to induce strabismus in experimental settings. There are several main surgical approaches used by clinicians to correct eye misalignment, recession, resection, and plication (Chaudhuri and Demer 2014; Von Noorden and Campos 2002). Both recession and resection have been used to create strabismus in animal models. Recession is a technique consisting of disinserting the EOM tendon from the sclera and reattaching it to the sclera at a more posterior location. Resection of the EOM removes some EOM before reinsertion near the normal insertion point. Recession and resection act in opposite ways, with the recession essentially weakening the pull of the muscle on the eye and resection strengthening its pull.

This intervention must be made within the first two weeks of life to disrupt visual binocular fusion during the sensitive period. If the muscle is too thin for sutures, a free tenotomy can be performed in lieu of a recession, avoiding the reattachment of the EOM.

In practice, surgical techniques are less effective than techniques manipulating visual input. The success of these interventions depends critically on the duration of the surgically induced perturbation. The duration depends on the adaptability of the EOMs but also on behavioral compensation, such as adjusting the position of the head. For this reason, the neurophysiologist may prefer to intervene on each eye and combine horizontal and vertical deviations to give the greatest chance of a lasting strabismus.

Botulinum-A injection into a specific EOM provides an alternative method for “weakening” the injected muscle to induce strabismus (Kiorpes 1992; Kiorpes et al. 1996). Botulinum A toxin is a neurotoxic protein that acts by blocking nerve function through inhibition of acetylcholine release.

Growth factors.

Growth factors are molecules capable of stimulating cellular growth and extraocular muscles have receptors to these molecules that could play a role in the maturation of muscle function (Anderson et al. 2006; McLoon and Christiansen 2003; Willoughby et al. 2012). Growth factors could potentially be used to modify skeletal muscles and motor neurons during development without mechanical intervention. For this approach to be successful the EOM must be exposed to the growth factor for a prolonged period of time. Sustained release of growth factors can be achieved by the use of slow-release pellets placed under the conjunctiva on the surface of EOMs (Willoughby et al. 2012; Willoughby et al. 2015a, 2015b). So far, three different molecules have been tested in infant macaques: insulin-like growth factor 1 (IGF-1, Willoughby et al. 2012, 2015a), brain-derived neurotrophic factor (BDNF, Willoughby et al. 2015b), and glial cell-derived neurotrophic factor (GDNF, McLoon et al. 2016). IGF-1 or IGF-2 have been shown to strengthen the EOMs (Anderson et al. 2006; McLoon et al. 2006; McLoon and Christiansen 2003). Effects of IGF-1 on eye alignment depended on the particular strategy used. In one experiment the IGF-1 pellet was placed bilaterally on the medial recti (Willoughby et al. 2012) leading to no misalignment. In another experiment a chronic strabismus was produced by unilateral placement of the IGF-1 pellet on the medial rectus muscle (Willoughby et al. 2015a). On the other hand, unilateral sustained release of BDNF did not induce eye misalignment (Willoughby et al. 2015b). Eye alignment can also be improved by the sustained release of IGF-1 on EOMs in adult NHPs with strabismus (McLoon et al. 2016).

Growth factors have also been used in the visual pathway to alter the sensitive period in rats. It has been shown that BDNF, produced in the retina, influences ocular dominance at the level of the V1 (Mandolesi et al. 2005). The authors showed that the imbalance induced by monocular deprivation was compensated for by injection of BDNF in the deprived eye and injection of antisense oligonucleotides preventing translation of BDNF mRNA in the open eye. Interestingly, a small shift in ocular dominance was also induced by modulation of retinal BDNF levels in animals with normal visual experience. Serotonin seems to be a crucial factor for triggering the cascade of molecular events such as reduced intracortical GABAergic inhibition and increased BDNF expression (Baroncelli et al. 2010). As shown in this study, environmental enrichment leads to enhanced serotoninergic transmission and reactivates ocular dominance plasticity in the adult visual cortex of rats. These effects on ocular dominance plasticity have yet to be tested for potential effects on eye alignment in primates.

Genetic factors.

Modification of genes that play a role in coding for muscle strength, the development of motor or premotor neurons, eye growth, or other aspects of visual development might have future use in animal models of strabismus. One difficulty of course, would be to affect genes without triggering other abnormalities, a true challenge due to pleiotropy. We know for example that the incidence of strabismus is higher in Down syndrome (Yurdakul et al. 2006), but the etiology of nonsyndromic infantile-onset strabismus is typically unknown (Von Noorden and Campos 2002). A few rare forms of strabismus present from infancy have a known genetic basis; the congenital cranial dysinnervation syndromes, for instance, impact the development of brain stem neuronal populations in utero, resulting in loss of brain stem neurons, subnormal or aberrant innervation of extraocular muscles, and missing or fibrotic muscles (Engle 2007). On the other hand, these disorders are not necessarily informative with regard to typical infantile strabismus; in particular, the incomitant strabismus resulting from these forms of strabismus is frequently compatible with normal ocular alignment in certain gaze positions, often resulting in development of normal binocular sensory fusion (Duane syndrome is a classic example).

Relevance of nonhuman primate models to infantile strabismus in humans.

Monkeys with strabismus induced by the above techniques typically display the same clinical features that are common in human patients with infantile-onset strabismus, including horizontal and vertical misalignments (Das et al. 2005), latent nystagmus (Tychsen et al. 1996), A or V patterns (Das and Mustari 2007), alternating fixation (Agaoglu et al. 2014; Das 2009), interocular suppression (Agaoglu et al. 2014), nasalward bias of smooth pursuit (Mustari and Ono 2011; Tychsen 2007), saccade disconjugacy (Fu et al. 2007; Walton et al. 2014), and impaired disparity vergence (Tychsen 2007). Early termination of the disruption of binocular visual experience in NHP permits recovery and avoidance of the visual and oculomotor sequelae of longer disruption, similar to the improved sensorimotor outcomes reported with early surgical correction of infantile esotropia in humans (Wong 2008).

Although humans with infantile strabismus syndrome are far more often esotropic than exotropic, animals with experimentally induced strabismus can be either esotropic or exotropic. NHPs occasionally have strabismus angles in excess of 40° (i.e., monkey XT1 with the left eye viewing in Walton et al. 2014), but misalignments are typically between 7 and 30° (Das et al. 2005; Mustari and Ono 2011; Tychsen 2007; Upadhyaya et al. 2016; Walton et al. 2014), which translates to ~12–52 prism diopters (PD). Castro et al. (2011), examining surgical outcomes in a large sample of human patients with congenital esotropia, reported that over 50% of their 127 subjects had presurgical deviations in the range of ~17–22° (30–40 PD). In most animals there is either an A pattern or a V pattern, although it is usually mild. For a more detailed discussion of animal models of developmental strabismus, see the recent review by Das (2016).

It is clear that NHP models of infantile strabismus have many striking similarities to human infantile-onset strabismus. On the other hand, recently some authors have argued that human infantile strabismus syndrome may be associated with a primary, congenital neural deficit of the disparity-driven vergence system (Kelly et al. 2016). In primate models, abnormalities of this system presumably do not develop until after birth, when researchers intervene to prevent binocular vision. In addition, it should be noted that findings from NHP models of infantile strabismus are not necessarily generalizable to more common forms of childhood strabismus, such as acquired accommodative esotropia and intermittent exotropia.

Understanding Neurophysiology in Normal and Strabismic Primates

While disease, injury, or developmental abnormalities of the extraocular muscles themselves can cause strabismus, it is clear that such peripheral abnormalities are not a prerequisite for strabismus to occur. Furthermore, it is clear that the oculomotor symptoms of strabismus are also associated with disordered neural development. As discussed above, a permanent strabismus can be induced in monkeys by fitting them with prism goggles for the first several months of life (Mustari and Ono 2011; Tychsen 2007; Walton et al. 2014; Wong et al. 2003) or alternating monocular occlusion (Das et al. 2005; Tusa et al. 2002). In addition, when eye muscle abnormalities are found in cases of strabismus (see Abnormalities of Eye Muscles and Orbital Tissues), it is often difficult to determine whether eye muscle abnormalities are the primary cause of strabismus, or an adaptation to an abnormal neural drive (Schoeff et al. 2013). Interestingly, there is evidence that disturbances of early visual experience can alter eye muscle specific myosin expression (Brueckner et al. 1999), indicating a continuous interaction between the visual and oculomotor systems. Finally, as discussed in the remainder of this review, numerous studies have reported neurophysiological abnormalities in monkeys with experimentally induced strabismus.

Oculomotor Neurophysiology

Maintenance of proper eye alignment for the purposes of high-grade stereopsis requires a highly precise coordination of the neural signals sent to extraocular muscles. The most obvious example of this is the vergence system, which is driven by a combination of binocular disparity visual signals, accommodative blur, and target proximity and serves to change the angle formed by the lines of sight of the two eyes. If the vergence system sends inappropriate commands to the extraocular muscles, maintenance of proper eye alignment will be impossible. In the preceding sections, we have reviewed evidence that binocular visual signals are impoverished in infantile strabismus. It is, therefore, not surprising that disparity-driven vergence is severely impaired as well (Kenyon et al. 1980, 1981; Tychsen 2007).

In recent years, a number of studies have suggested that the fundamental organization of oculomotor systems is monocular, even in normal primates (King and Zhou 1995; 2002; Sylvestre and Cullen 2002; Sylvestre et al. 2002; Van Horn et al. 2008; Waitzman et al. 2008; Zhou and King 1998). With normal binocular development, a superimposed system of binocular input results in coordinated eye movements. However, when normal binocular development is disrupted, the monocular nature of oculomotor control becomes more apparent. As reviewed below, there is compelling evidence that the brain sends different commands to the two eyes in monkeys with strabismus induced in infancy (Das and Mustari 2007; Joshi and Das 2011; Walton et al. 2013).

Eye velocity signals from various oculomotor subsystems are mathematically integrated in the interstitial nucleus of Cajal (INC) and the nucleus prepositus hypoglossi (NPH) to produce vertical and horizontal eye position signals, respectively (Belknap and McCrea 1988; Cullen et al. 1993; Dale and Cullen 2013; Fukushima and Kaneko 1995; Kaneko 1999; McCrea and Horn 2006; Sylvestre et al. 2003). For these reasons, abnormalities of oculomotor circuitry have the potential to affect eye alignment and are, therefore, relevant to our understanding of developmental strabismus. The most direct demonstration of this has come from single-unit recordings showing that premotor neurons in the vergence system in monkeys with experimentally induced exotropia modulate their activity in association with changes in horizontal strabismus angle (Das 2011, 2012) and that motoneurons serving horizontal and vertical rectus muscles encode cross-axis changes in eye alignment (vertical movements of the nonfixing eye associated with horizontal movements of the fixing eye, or vice versa) that are characteristic of pattern strabismus (Das and Mustari 2007; Joshi and Das 2011). In the sections that follow, we describe the numerous oculomotor abnormalities that have been reported in humans and monkeys with strabismus, and we review studies that have provided evidence relevant to our understanding of their neural basis.

Optokinetic nystagmus.

The reflexive optokinetic response provides a convenient measure of the maturation state of the oculomotor system (Mustari and Ono 2007). Reflexive optokinetic eye movements are elicited by large-field visual motion in a specific direction. During continuous unidirectional motion, OKN composed of slow following movements of the eyes in the direction of visual motion occurs with oppositely directed quick phases (saccades). Mature OKN in NHPs is composed of an initial rapid rise in eye velocity to ~60% of visual stimulus velocity, followed by a slow buildup in eye velocity over 15–20 s (due to velocity storage mechanisms) until eye velocity matches visual stimulus velocity (Mustari and Ono 2007; Raphan et al. 1977; Tusa et al. 2001).

OKN is immature at birth in primates (Mustari and Ono 2007; Ornitz et al. 1985; Schor 1993; Weissman et al. 1989). This immaturity includes a decreased steady-state OKN with increasing visual stimulus velocity (Distler 1996; Grasse and Cynader 1988; Schor and Narayan 1981). During the first few weeks of life in NHPs, the monocular optokinetic response is asymmetric, with much higher gain during nasalward compared with temporalward visual motion (Naegele and Held 1982; Schor 1993; Tusa et al. 2002). Symmetry and an early rapid rise in eye velocity at the beginning of OKN is reached after 6 wk of age in NHPs (Tusa et al. 2001). This maturation is dependent on the development of binocular sensitivity in V1 and the pretectal nucleus of the optic tract (NOT). Interference with the coordinated use of the eyes early in life (e.g., strabismus, anisometropia, uniocular congenital cataract, monocular deprivation) can prevent maturation of essential cortical-brain stem pathways involved in OKN (Maurer et al. 1983; Schor and Levi 1980; Shawkat et al. 1995; Sparks et al. 1986; Tusa et al. 2002; Westall and Schor 1985).

The NOT has been shown to be essential for horizontal OKN (Mustari and Ono 2007). Maturation of OKN to the mature symmetric pattern is thought to depend on development of V1 and MT pathways that bring direction-selective visual motion information from the ipsilateral eye to the NOT (Distler and Hoffmann 1999; Mustari and Ono 2007). In normally reared monkeys, all NOT neurons are binocularly driven with approximately equal sensitivity to either eye (Hoffmann et al. 1988; Inoue et al. 2000; Mustari and Fuchs 1990; Mustari et al. 2001). In animals with some forms of early-onset visual deprivation, such as binocular or monocular deprivation, a loss of binocular sensitivity of NOT neurons occurs and a corresponding asymmetric OKN with weak or absent temporalward response is observed (Hoffmann et al. 1996; Mustari et al. 2001; Tusa et al. 2002). Hoffmann and colleagues (1996) found only a slight correlation between primary visual cortical binocularity and asymmetry in OKN. Rather, it was loss of binocular neurons in the NOT that correlated best with the degree of asymmetry in the OKN. Similar findings have been reported in naturally strabismic monkeys (Distler 1996).

Smooth pursuit.

Smooth pursuit is a volitional eye movement and is a more complex behavior than OKN; it depends on different cortical areas, such as frontal eye fields (FEF). Numerous studies in normal subjects have shown that signals carried in MT, MST, and FEF pathways feature different aspects of visual and eye motion (Mustari et al. 2009) needed for control of smooth pursuit. Smooth pursuit is immature at birth in primates and evinces a bias for nasalward tracking like that of the immature OKN. Both OKN and smooth pursuit achieve symmetry in gain at about the same postnatal age (>6 wk), likely due to maturation of binocular cortical neurons that project to critical brain stem regions.

Smooth pursuit typically requires visual motion on the retina for initiation (Leigh and Zee 2015). The retino-geniculo-striate system projection provides the main afferent limb for smooth pursuit. In addition, retinal projections to the NOT may play a special role in some forms of smooth pursuit plasticity (Livingston and Mustari 2000; Mustari et al. 2009). Visual motion information for smooth pursuit is carried in dorsal stream pathways involving MT and MST and their connections with the FEF. Each of these regions is essential for different aspects of smooth pursuit, including initiation, maintenance, gain control, and plasticity. These cortical areas have direct projections to brain stem areas [e.g., rostral nucleus reticularis tegmenti pontis (NRTP), dorsolateral pontine nuclei (DLPN), NOT, SC], which in turn project to different regions of the cerebellum (flocculus, paraflocculus, and vermis) to support complementary roles in smooth pursuit (Krauzlis 2004; Leigh and Zee 2015; Ono and Mustari 2009). The central thalamus brings eye position and velocity signals (Tanaka 2005) back to cortical areas involved in smooth pursuit (see Lynch and Tian 2006). This thalamocortical eye velocity feedback could play a role in the control and monitoring of eye movements during ongoing smooth pursuit, gain control, gaze, and other functions.

Like other types of eye movements, smooth pursuit eye movements are abnormal in infantile strabismus. Human patients with this disorder show strong asymmetry in pursuit during monocular viewing. The direction of the asymmetry depends on the viewing eye and switches instantaneously when occlusion is moved to the other eye (Tychsen 1993). Nasalward pursuit is smooth and closely matches target motion. In contrast, temporalward pursuit is interrupted by more frequent saccades, as smooth pursuit gain is insufficient to keep the eye on target (Fig. 1). Normal infants, before the development of binocularity in the cortex, show similar asymmetry during monocular pursuit.

Fig. 1.

Horizontal smooth pursuit and VOR in strabismic monkey, testing during monocular and binocular viewing conditions. Smooth pursuit has a defective gain and saccadic intrusions during temporalward tracking. A: step-ramp tracking paradigm showing eye position (left column) and eye velocity (right column) during successive trails of pursuit. B: sinusoidal smooth pursuit and inset plots showing the horizontal (HE) and vertical (VE) position of each eye during successive cycles of pursuit. A dissociated vertical deviation is evident. C: horizontal VOR during whole body rotation (±5°; 0.50 Hz) in an illuminated environment. Smooth eye movements of the VOR are free of saccadic intrusions during temporalward or nasalward conditions. During binocular viewing alternation of the attending eye is evident. Conventions: right eye, red; left eye, blue; target or head position, gray dashed lines. Figure reproduced from Mustari and Ono (2011) with permission.

Mustari and Ono (2011) investigated the smooth pursuit behavior of macaques with infantile strabismus. Their animal models included both surgical and prism-reared strabismus models (see Postnatal Development of Oculomotor and Visual Sensitivity) and showed that either method produces an animal model that replicates abnormalities seen in human infantile esotropia syndrome. The authors verified that smooth pursuit eye movements per se were altered, but the smooth eye movements associated with the vestibulo-ocular reflex (VOR) remained intact. The presence of a symmetric VOR in these strabismic animals rules out a primary defect at the level of the eye muscles or premotor circuits. This demonstration is represented in Fig. 1.

Interestingly, similar deficits in smooth pursuit are found when unilateral lesions disrupt the pathway through which visual motion information is sent to the cerebellum via the pons. Muscimol injections in the DLPN produce ipsilesional deficits in smooth pursuit similar to the pursuit deficits observed in strabismus (May et al. 1988). Thus occluding one eye of a strabismic subject produces a defect like that associated with unilateral lesion of the cortical-ponto-cerebellar system originating in the contralateral hemisphere. However, in strabismus the pursuit deficit reverses when switching fixation and nearly disappears during binocular viewing.

Mustari and Ono (2011) suggest that the key to understanding this may lie in the loss of cortical binocular cells in infantile strabismus. According to this hypothesis, the right DLPN does not have access to information from right-eye monocular cells. In normal individuals, when the left eye is closed, or not able to see the target, binocular cells provide the necessary input to drive the right DLPN. Since these cells are lost in strabismus, closing the left eye results in insufficient drive for rightward pursuit, as though right DLPN had suffered a lesion, resulting in the observed leftward or nasalward bias in the unoccluded right eye. An important caveat is that it is unclear whether this is due simply to insufficient cortical input to DLPN or whether, like cortex, DLPN also suffers a loss of binocularity during development in infantile strabismus. The extent to which the organization of other brain stem regions are disrupted supports the need for further, direct study of pursuit-related brain stem areas in strabismus.

There have been few single-unit recording studies in MT and MST cortical areas of specially reared monkeys. As reviewed above, strabismus is associated with a deficit in monocular visual motion perception, especially for temporalward motion, which might suggest involvement of cortical area MT. Kiorpes and colleagues (1996) studied smooth pursuit in monkeys with early-onset (<11 days old) artificial strabismus. The strabismic animals developed asymmetric monocular smooth pursuit (nasalward bias) and latent nystagmus. Single unit recording studies in paralyzed and anesthetized preparations of these monkeys revealed a loss of binocularly sensitive neurons in MT. However, the motion sensitivity of MT neurons was indistinguishable from that found in MT neurons of normal monkeys. Therefore, Kiorpes and colleagues (1996) suggested that the nasalward smooth pursuit bias of strabismic monkeys may not depend on a loss of directionally sensitive neurons in MT, but rather on alterations in downstream structures. Other studies in behaving monkeys found a loss of binocularly sensitive neurons in lateral and dorsal divisions of the MST area (MSTl and MSTd, respectively). Both visual and eye movement responses tended to be dominated by one eye, in contrast to the equal sensitivity to both eyes found in corresponding neurons of normal animals (Mustari et al. 2008). Some MSTd neurons responded preferentially for movement of one eye during smooth pursuit. Such neurons have not been reported in normal monkeys. One possibility is that this lack of binocularity, combined with loss of information from the nonviewing eye during occlusion, leaves cerebellar and brain stem pursuit neurons with inadequate drive for temporalward pursuit (Mustari and Ono 2011). A similar suggestion was made by Hasany and colleagues (2008), who hypothesized that cortical pursuit pathways may be innately wired for nasalward pursuit and that binocular connections are crucial to the development of normal temporalward pursuit. They proposed that, in strabismus, the loss of binocular visual signals in cortex leaves the pursuit system unable to produce strong enough commands for temporalward pursuit. Thus, although the neural abnormalities that affect smooth pursuit in strabismus are not fully understood, the few existing studies point to impairment of binocular visual signals in cortex as a likely causative factor.

Figure 1 shows examples of eye movement responses in developmental strabismus cases produced by prism rearing. A clear pattern of nasalward bias during smooth pursuit is present. During monocular viewing conditions, smooth pursuit gain is lower during temporalward motion and saccades are required to keep the eye on target. There is also a DVD present (Fig. 1, inset plots). Importantly, the ability to move the eyes smoothly and symmetrically is preserved during the vestibular-ocular reflex (Fig. 1 right panel). This indicates that the smooth pursuit abnormalities must be neural in origin, and upstream from structures shared with the VOR.

In addition to showing nasotemporal asymmetry, smooth pursuit in strabismic primates is characterized by another type of abnormality, so-called “cross-axis” eye movements. Smooth pursuit movements in one axis (e.g., horizontal) in the fixing eye are accompanied by both horizontal and vertical displacement of the nonfixing eye. Some degree of cross-axis movement in strabismus might be explained on mechanical grounds alone, due to relative ocular torsion or extraocular muscle pulley displacement causing rectus muscles in the nonfixing eye to exert rotational forces on the globe that are not aligned with their normal direction of action or with the direction of pull of the corresponding yoke muscle in the fixing eye. However, such mechanical effects alone do not explain the full range of cross-axis or directionally disconjugate eye movements observed in strabismic individuals. During smooth pursuit, for example, single-unit recordings of burst-tonic motoneurons in strabismic NHPs indicate that cross-axis vertical eye movement in the nonfixing eye during horizontal pursuit by the fixing eye is associated with activity in motoneurons controlling vertical eye muscles (Das and Mustari 2007), despite the fact that no such vertical motoneuron activity is observed when that eye is the one engaged in horizontal pursuit. This raises the possibility that the commonly observed V and A patterns of infantile strabismus may be due in part to mismatched supranuclear input to the two eyes, or uncoupling of the normal system of coordinated, “yoked” control of binocular eye movements.

Cross-axis phenomena are discussed further below in the context of saccadic eye movements, in Saccades.

Vergence.

In normal individuals, vergence eye movements are driven largely by a combination of binocular disparity and accommodative effort (itself driven by blur) (Fincham and Walton 1957; Maddox 1886) (proximity may also play a role, especially under open-loop conditions). The premotor neurons for the vergence system are found in an area dorsolateral to oculomotor nucleus, the supraoculomotor area (SOA). Antidromic stimulation studies have confirmed that neurons in this area project bilaterally to the medial rectus subdivision of the oculomotor nucleus (Zhang et al. 1991, 1992). Single neurons show a tonic discharge related to vergence angle (Mays 1984), bursts of spikes related to vergence velocity (Mays et al. 1986), or both. Some neurons increase their firing rates for convergence, and others for divergence; they are called “near response” and “far response” cells, respectively. In one classic study a display system was used that allowed binocular disparity and accommodative demands to be manipulated independently (Zhang et al. 1992). The authors reported that near response cells carry varying combinations of vergence demand-related and accommodation-related signals.

In infantile strabismus, binocular disparity signals are disrupted, resulting in severe impairments of disparity-driven vergence (Kenyon et al. 1980, 1981). This raises an interesting possibility. As noted above, visual cortical signals appear to be largely monocular in infantile strabismus. If binocular disparity-related visual signals are lost or diminished, perhaps the vergence system comes to be driven, instead, by some combination of monocular visual signals and accommodation. If the neuroanatomical connections of neurons in SOA are intact, any changes in their firing rates should influence the horizontal angle of strabismus. Indeed, recordings from SOA neurons in monkeys with experimentally induced strabismus have shown that the tonic firing rates are correlated with changes in horizontal strabismus angle associated with switching fixation from one eye to the other (Das 2011, 2012).

Outside the laboratory, vergence eye movements are most often made in combination with saccades. The coordination of saccadic and vergence movements is still an area of ongoing research and debate, but we address the research relevant to infantile strabismus in the next section.

Saccades.

When human patients with infantile strabismus make horizontal saccades, the amplitudes often differ for the two eyes (Bucci et al. 2002; Kapoula et al. 1997; Maxwell et al. 1995). The same is true for monkeys with strabismus induced in infancy (Fu et al. 2007). Two recent studies have examined vertical and oblique saccades in monkeys (Walton et al. 2014) and human patients (Ghasia et al. 2015) with pattern strabismus and showed disconjugacy of both amplitude and direction. One of the animals in the NHP study showed directional disconjugacy for saccades in certain directions, but not others, in a manner inconsistent with overall torsion of the globes, muscle pulley displacement, or the predicted effects of oblique muscle overaction. The human study similarly found that cross-axis effects were not correlated with either ocular torsion, or with the severity or predicted field of action of inferior oblique muscle overaction.

Despite these abnormalities, monkeys with experimentally induced strabismus are able to make saccades that accurately bring one eye to the target, with either eye fixating (Fu et al. 2007). This implies that the brain is able to generate different saccadic commands to satisfy a given desired displacement vector, depending on which eye is viewing the target.

Potentially, this could be accomplished using the same processes that control saccadic adaptation in normal primates. Saccadic adaptation is typically studied by shifting the location of the target during the movement (McLaughlin 1967). This introduces an error signal that the brain uses to gradually adjust the amplitudes of subsequent saccades (Wallman and Fuchs 1998). Saccadic adaptation in normal human subjects can be specific to a particular context, such as initial eye position, viewing distance, or head tilt (Chaturvedi and van Gisbergen 1997; Shelhamer et al. 2005; Shelhamer and Clendaniel 2002). In this context it is somewhat surprising that in monkeys with experimentally induced strabismus, saccadic adaptation occurs in both the viewing and nonviewing eyes and is approximately conjugate (Das et al. 2004). Perhaps humans and monkeys with strabismus are able to maintain different adaptive states, depending on which eye is viewing the target.

There are a number of potential explanations for neurally driven saccade disconjugacy in strabismus. First, in the absence of normal binocular disparity signals in visual cortex, the oculomotor system might generate commands that would be appropriate for gaze shifts between targets that differ in both direction and distance. That is, the disconjugate saccades might be caused by a contextually inappropriate activation of a disordered vergence system. This could explain disconjugacy of saccade amplitude, although it would not explain cross-axis effects.

In considering whether amplitude disconjugacy could be explained by inappropriate vergence activation, it is useful to first consider saccade-vergence interactions in normal primates. Saccades are mostly conjugate when they occur between equidistant targets. Conversely, when a change in the depth of fixation occurs in the absence of a change in viewing direction, the eyes rotate equally, but in opposite directions, either diverging or converging. Saccades and vergence eye movements have very different dynamics, with the latter being much slower. When changing fixation between targets that differ in both direction and distance, the resulting movement does not seem to be a simple linear summation between a high-velocity saccade and a slow vergence movement. Instead, vergence velocity is substantially higher during the saccade (Enright 1984; Erkelens et al. 1989b; Maxwell and King 1992; Ono et al. 1978), although analysis is complicated by the presence of high-velocity transient changes in vergence position that accompany saccades, even those between equidistant targets. These so-called vergence transients tend to be idiosyncratic between subjects but are correlated with various metrics of saccade dynamics such as peak conjugate velocity, saccade direction, and target distance (Busettini and Mays 2005a). For the past several decades, the explanation for this phenomenon has been a matter of considerable controversy. According to one idea, the saccadic and vergence systems share one or more elements, producing a nonlinear interaction between the two subsystems (Busettini and Mays 2003, 2005a, 2005b).

Alternatively, the saccadic system might directly program disconjugate saccades when normal primates change fixation between targets that differ in both direction and distance. This latter hypothesis implies that the fundamental organization of the saccadic system is monocular, even in normal primates. This possibility inspired a series of experiments aimed at recording from saccade-related neurons in brain stem while monkeys made disjunctive eye movements between targets that differed in both direction and distance (Sylvestre and Cullen 2002; Van Horn and Cullen 2008; Waitzman et al. 2008; Zhou and King 1998). In each of these studies, the authors concluded that the activity of many individual neurons was better correlated with the movement of one eye.

Potentially, then, horizontal saccade disconjugacy in strabismus might result from normal processes operating under inappropriate circumstances (i.e., when making saccades between equidistant targets). This might even be the case for vertical saccade disconjugacy. Although vertical fusional vergence has a very limited range (Bharadwaj et al. 2007; Houtman et al. 1981; Howard et al. 1997), some adaptation is possible in response to vertical image disparities in normal subjects (Irsch et al. 2013). When normal human subjects wear anisometropic spectacles they can adapt to make vertically disconjugate saccades (Erkelens et al. 1989a; Lemij and Collewijn 1991, 1992). It has been suggested that dissociated vertical deviation in strabismus may be the result of vertical fusional vergence mechanisms, operating in the absence of normal vertical image disparity signals (Van Rijn et al. 1997).

Saccade disconjugacy in strabismus might also be the result of abnormalities within brain stem circuitry specific to saccadic eye movements. This explanation would better account for the combination of both amplitude and directional disconjugacy seen in saccadic eye movements in strabismic individuals. One approach to studying this is to apply microstimulation to brain stem structures that carry signals specifically related to saccades. If the evoked movements are normal, then the artificially imposed command must have bypassed the abnormalities responsible for the unequal saccades. On the other hand, if the evoked movements show a pattern of disconjugacy that cannot be fully accounted for by peripheral abnormalities, then the saccadic system itself must be involved. A good target for such an experiment is the pontine paramedian reticular formation (PPRF), which contains premotor burst neurons encoding instantaneous horizontal saccade velocity (Van Horn et al. 2008). As expected, microstimulation of this structure evoked conjugate horizontal ramp eye movements in normal monkeys. In monkeys with strabismus, however, the evoked movements were often disconjugate, in terms of both velocity and direction (Walton et al. 2013). Importantly, however, the disconjugacy varied widely (Fig. 2). For some sites the evoked movements were nearly conjugate (Fig. 2, E and F); for other sites the direction of the evoked movements differed for the two eyes by more than 45° (Fig. 2, C and D). This high degree of between-site variability is incompatible with the hypothesis that saccade disconjugacy is due solely to effects at the level of eye muscles or the plant. This, of course, does not rule out the possibility that peripheral abnormalities may exist, but it provides some of the strongest and most direct evidence to date that abnormalities exist within the premotor saccadic circuitry itself.

Fig. 2.

Example data from three stimulation sites in PPRF. Blue,  left eye position; red,  right eye position. Gray shaded areas indicate the 200-ms periods of microstimulation. A and B: in the normal monkey the stimulation evokes a ramplike eye movement that is almost perfectly conjugate. C and D: microstimulation of this right PPRF site in an exotropic monkey caused the two eyes to move in very different directions. Even though this structure contains the horizontal premotor burst neurons that directly project to abducens nucleus, note that the left eye’s movement is only upward. E and F: microstimulation of this different site in the right PPRF of the same exotropic monkey evoked movements that were nearly conjugate (see velocity traces).

A recent single-unit recording study has provided additional support for the hypothesis that PPRF develops abnormally in pattern strabismus (Walton and Mustari 2015). The preferred direction of each neuron was estimated by plotting the number of spikes as a function of polar saccade direction and fitting the data with Gaussian functions. In the normal control monkey, all 24 PPRF neurons had preferred directions within 30° of horizontal. In two monkeys with pattern strabismus, however, PPRF neurons on one side of the brain had an abnormally broad distribution of preferred directions (Walton and Mustari 2015). Indeed, 12 of 60 neurons recorded from strabismic animals preferred vertical saccades. An additional 7 of 60 neurons showed abnormally weak bursts, with no correlation between the discharge characteristics and saccade kinematics. To our knowledge, similar “untuned” neurons have never been found in PPRF of normal monkeys.

Thus both stimulation and recording studies support the hypothesis that horizontal and vertical saccadic signals are mixed to an abnormal degree in PPRF in monkeys with experimentally induced pattern strabismus. This suggests a possible neural basis for A and V patterns. According to most models of the saccadic system, premotor horizontal and vertical velocity signals are mathematically integrated in NPH and the INC, respectively. This ensures that the step change in motoneuron firing rates will match the pulse. If, in pattern strabismus, horizontal and vertical eye velocity commands are inappropriately mixed, then the integration of these signals would influence the static eye position.

When considering this as a possible mechanism contributing to A and V patterns, however, several issues must be addressed. First, the neural integrators would have to send different eye position signals to the two eyes. There is evidence consistent with this idea. When normal monkeys make gaze shifts between targets that differ in both direction and distance, the activity of many NPH neurons is better correlated with the position and velocity of one eye than the cyclopean eye (Sylvestre et al. 2003). Thus existing data from normal monkeys supports the plausibility of monocular signals in the neural integrators.

Second, one must consider whether abnormalities in saccade-related brain stem structures are a plausible mechanism, given that A and V patterns are also seen during static fixation and cross-axis movements are observed for other types of eye movements, such as smooth pursuit. One possibility is that the disruption of horizontal and vertical disparity signals in visual cortical areas, during a sensitive period in early postnatal life, results in a “cascade” of abnormalities that disrupts the directional tuning of saccade and pursuit-related neurons in similar ways. This hypothesis implies that neurophysiological abnormalities are widespread within the saccadic system and probably present from early stages of processing. In this case, the directional tuning of saccade-related neurons should be abnormal in numerous brain stem areas. To our knowledge, the aforementioned PPRF study is the only published single-unit recording study targeting a saccade-specific brain area in monkeys with experimentally induced strabismus; additional studies in other saccade-specific areas would help test the validity of this hypothesis.

In addition, important insights might be gained from neurophysiological studies that target areas that play important roles in sensorimotor processing. One such area is the SC, a layered structure in the midbrain that is thought to generate a desired displacement signal. Both visual and saccade-related signals can be found in the intermediate layers, sometimes in the same neuron (Mays and Sparks 1980). Individual neurons exhibit high-frequency bursts of spikes, preferentially for saccades of a specific vector, and have movement fields that are analogous to the receptive fields of visual neurons (Mohler and Wurtz 1976; Sparks et al. 1976; Sparks and Mays 1980). Saccade vectors are encoded by the spatial location of activity within a topographic map, with small-amplitude saccades represented rostrally and large saccades caudally (Robinson 1972; Schiller and Stryker 1972). The medial and lateral portions of SC contain neurons that prefer saccades with upward and downward components, respectively. This vectorial signal is then sent to the horizontal and vertical brain stem saccade generators in PPRF and the rostral interstitial nucleus of the medial longitudinal fasciculus (riMLF), respectively (Keller et al. 2000; Moschovakis et al. 1996). Neurons in these structures encode instantaneous component saccade velocity (Van Horn et al. 2008). There is evidence that this transformation of signals from a spatial code to a temporal one is the result of a gradient of connection weights from SC sites to the horizontal and vertical burst generators (Moschovakis et al. 1998).

Stimulation of the intermediate and deep layers of SC in normal monkeys generates conjugate saccades (Fleuriet et al. 2016). In monkeys with strabismus, however, stimulation produces disconjugate saccades (Economides et al. 2016; Fleuriet et al. 2016; Upadhyaya et al. 2016). Unfortunately, the interpretation of this result is not straightforward. One possibility is that monocular visual signals in cortex might lead to the development of separate topographic saccade maps for the two eyes. In pattern strabismus, these maps might be spatially distorted in different ways. For example, if there was an overrepresentation of downward vectors for one eye and upward vectors for the other eye, then activation of a single anatomical site would cause SC to output two different displacement commands. If saccade-related neurons in SC are primarily binocular in pattern strabismus, then the predicted result would be an intermediate vector, due to vector averaging (Lee et al. 1988; Walton et al. 2005) or vector summation (Goossens and van Opstal 2012; van Opstal and Goossens 2008). However, if the circuits are separate for the two eyes, disconjugate commands at a single anatomical site would not be combined by these mechanisms. This hypothesis predicts that 1) the movement fields of saccade-related neurons in SC will have abnormal sizes and/or shapes and 2) the bursts for individual neurons will be stronger when the “preferred” eye is to be directed to the target.

Another possibility, proposed by Economides et al. (2016) but also discussed in Fleuriet et al. (2016), is that the SC encodes a single desired displacement in pattern strabismus. This hypothesis suggests that disconjugacy occurs because of erroneous signaling within downstream, monocularly organized circuits. According to many modern models of the normal saccadic system, the movement is controlled by a local feedback loop that dynamically compares desired displacement with an estimate of current displacement (Becker and Jürgens 1990; Scudder 1988; van Gisbergen et al. 1985; Walton et al. 2005). When the difference between the two reaches zero, the saccade ends. The addition of an erroneous signal, within this feedback loop, would tend to be compensated by the normal action of this circuit. If, in pattern strabismus, the input to the local feedback loop is a single desired displacement command for both eyes, then the neural basis for directional disconjugacy must either be downstream from the local feedback loop, or the comparator would have to effectively ignore the current displacement signal for the nonviewing eye. A potential advantage of the former possibility is that it would provide a relatively parsimonious explanation for the observation that directional disconjugacy is observed for both saccades and smooth pursuit in pattern strabismus (Das and Mustari 2007; Ghasia et al. 2015; Joshi and Das 2011; Walton et al. 2014). For example, abnormal cross talk between horizontal and vertical pathways might occur at the level of a structure that is shared between different oculomotor subsystems. Numerous studies have reported evidence that NPH and INC serve as integrators for saccades, vergence, smooth pursuit, and the VOR (Belknap and McCrea 1988; Cullen et al. 1993; Dale and Cullen 2013; Fukushima and Kaneko 1995; McCrea and Horn 2006; Sylvestre et al. 2003). If this is correct, then abnormalities at this level could lead to analogous cross-axis disconjugacies across different oculomotor systems. However, Upadhyaya et al. (2016) reported saccade disconjugacies despite small misalignment and small amount of pattern strabismus.

Control systems models typically treat NPH and INC as pure velocity-position integrators but the neurophysiological data indicate that more complex processing occurs in these structures. For example, eye velocity signals (i.e., saccade-related bursts) have been found in both structures (Crawford et al. 1991; Delgado-García et al. 1989; Fukushima et al. 1992; McFarland and Fuchs 1992). This being the case, it is possible that cross talk at the level of the neural integrators might affect the saccade direction for one eye. Notably, INC also carries signals related to ocular torsion (Crawford et al. 1991), which could partially account for the torsional abnormalities observed in some patients with pattern strabismus. These considerations suggest that the neural integrators may be likely candidates for a neural basis of pattern strabismus. Neuroanatomical studies have demonstrated projections from NPH to INC in squirrel monkey (Belknap and McCrea 1988) and from INC to abducens nucleus (Graf et al. 2002; Ugolini et al. 2006).

In normal individuals, there is usually a small postsaccadic drift after the end of a horizontal saccade. Typically, there is both a versional and a convergent component. The latter is believed to counteract a small, transient intrasaccadic divergence that is observed, even when normal monkeys make saccades between distant targets. Numerous studies have reported abnormally large postsaccadic drifts in humans and monkeys with strabismus (Bucci et al. 2009; Fu et al. 2007; Kapoula et al. 1997; Walton et al. 2014). In general, this phenomenon could result from the effect of a mismatch between the saccade-related step and the tonic eye position signal on motoneurons. The presence of abnormally large drifts in strabismus, therefore, suggests the possibility of faulty integration of eye velocity commands. Alternatively, the asymmetrical drift may occur because a disconjugate saccade produces a horizontal angle of strabismus that is inconsistent with what has been requested by a (disordered) vergence system. If this is the case then normal integrators might simply be responding to disordered inputs. A recent SC stimulation study has provided support for this latter hypothesis. In normal monkeys, prolonged stimulation of SC leads to a “staircase,” a series of saccades separated by periods without eye movement. In monkeys with strabismus, the same procedure evokes a staircase of disconjugate saccades, separated by periods of slow, disconjugate movements that were greater than those observed after visually guided saccades (Upadhyaya et al. 2016). The authors proposed that these slow, disconjugate movements were the result of activation of vergence-related neurons that have recently been reported in rostral SC (Van Horn et al. 2013).

Future Directions

A common theme that runs through much of this review is that the loss of binocular vision during a sensitive period early in life likely results in a cascade of abnormalities, from visual cortical areas through various oculomotor subsystems (saccades, vergence, smooth pursuit, etc.). Even if the initial abnormality involves the orbital tissues or eye muscles, the result will be a disturbance of the input that visual neurons need to develop normal tuning properties. Of course, neurons that carry signals related to vision and/or eye movements are part of complex processing streams involving numerous brain areas. From a developmental perspective, it seems rather improbable that serious tuning abnormalities would arise in one or two nodes within an otherwise normal circuit. Disordered tuning in one brain area is likely to influence the signals carried by downstream neurons. In this way, abnormalities might propagate through the circuit in very complex ways. Verifying this general idea will require single-unit recording studies targeting many visual and oculomotor areas in primates with strabismus that is present in infancy. Much can be learned by identifying the precise tuning abnormalities across different nodes in known circuits.

As noted early in this review, pattern strabismus is often clinically described in terms of oblique muscle dysfunction. MRI studies of rectus muscle pulleys (Hao et al. 2016; Suh et al. 2016) demonstrate that other peripheral factors can contribute to incomitance. However, the neurophysiological evidence discussed above indicates that pattern strabismus is also associated with fundamental abnormalities in the neural signals controlling eye movement. For example, several authors have suggested that pattern strabismus is associated with abnormal cross talk between horizontal and vertical pathways in brain stem (Das 2012, 2016; Ghasia et al. 2015; Walton and Mustari 2015; Walton et al. 2013). As discussed above, the circuits controlling smooth pursuit and saccades are largely separate, yet cross-axis movements are observed in both. The currently available data are insufficient to determine whether the hypothesized cross talk between horizontal and vertical pathways occurs at a level shared by both pursuit and saccadic systems (such as motoneurons and/or the neural integrators in NPH and INC; Fig. 3C) or whether a more general breakdown of directional tuning affects both systems in analogous ways. The latter hypothesis predicts that directional tuning should be abnormal in numerous brain areas involved in saccades and smooth pursuit. The former hypothesis predicts that faulty directional tuning should be more severe in shared structures (such as NPH and INC) than in upstream saccade-related and pursuit-related areas.

Fig. 3.

Schematic representation of brain stem circuitry activated in association with a saccade made in response to a rightward target step. A: normal monkey. Blue and red lines and shapes represent horizontal and vertical pathways, respectively. In the intermediate layers of SC a region near the center of the topographic map is activated (yellow star, situated in the blue region). Downstream, the horizontal pathway is activated but vertically tuned saccadic burst neurons in riMLF are quiescent. Vertical eye position cells in interstitial nucleus of Cajal (INC) would be active, but would not modulate. Arrows labeled “1” (Belknap and McCrea 1988) and “2” (Graf et al. 2002; Ugolini et al. 2006) indicate connectivity between horizontal and vertical pathways; the dashed lines are shown in gray to indicate the possibility that this connection may have little influence in normal primates. Similarly the line from the vergence box is dashed and gray to represent the fact that this pathway has no influence for a saccade between equidistant targets. B: hypothesized scheme for a monkey with pattern strabismus. Spatial distortions of visual representations result in a distortion of the topographic saccade maps in SC. Although the same anatomical site is activated as in A, the topographic map of saccade vectors is distorted, such that the area encoding upward vectors is abnormally large. As a result, SC outputs a command for an oblique saccade. In this scheme the distortion is assumed to differ for the two eyes, with conjugate saccadic adaptation (Das et al. 2004) ensuring that the viewing eye is accurately brought to the target. The assumed distortion of the collicular map would affect the decomposition of vectorial commands into component velocity signals. As a result, monocularly tuned premotor saccadic burst neurons in PPRF 3 (Walton and Mustari 2015) and riMLF have abnormal preferred directions. The breakdown of directional tuning propagates from node to node, ultimately affecting static eye position signals for motoneurons, which then encode the monocular cross-axis movements “4” (Das and Mustari 2007; Joshi and Das 2011). Other oculomotor subsystems (such as pursuit) are assumed to develop separate (but analogous) abnormalities that result in cross-axis movements. The vergence drive is marked with a question mark to represent uncertainty about whether vergence neurons show abnormal modulation associated with saccades between equidistant targets in infantile strabismus. C, left: in this alternative scheme, SC is assumed to be effectively normal. The cross talk between neural integrators and motoneurons for the horizontal and vertical pathways is assumed to be abnormally strong (indicated by colored, solid lines). Right: because NPH and INC are shared by multiple oculomotor subsystems, this scheme assumes that the neural basis of pattern strabismus would be the same for different types of eye movement. III, oculomotor nucleus; IV, abducens nucleus.

It is possible to conceive of specific scenarios by which abnormal signaling could develop across different nodes within known circuits. For example, under Abnormalities of Visual Processing Associated with Infantile Strabismus we reviewed evidence that exotropia and esotropia are associated with suppression of visual signals arising from the temporal and nasal hemiretinae, respectively. Visual and saccade-related signals are mixed, often in the same neurons, in SC (Mays and Sparks 1980). It is, therefore, not unreasonable to hypothesize that distorted representations of visual space in cortex might ultimately lead to analogous distortions of topographic maps of saccade vectors in SC (Fig. 3B). This, in turn, would be expected to affect the spatial-temporal transformation that determines the firing rates of premotor burst neurons in PPRF and riMLF that encode saccade velocity. The mathematical integration of these disordered velocity signals in NPH and INC would affect the eye position signals carried by motoneurons. Currently, this scenario is highly speculative but it is testable, since it makes rather specific predictions about the abnormalities expected in various brain areas in monkeys with pattern strabismus (distorted movement fields in SC neurons, abnormal preferred directions for riMLF neurons, inappropriate cross axis position signals in NPH and INC neurons, etc.).

Another key area of interest for future research concerns visuomotor transformations in strabismus. In numerous areas of the brain, there exist neurons that carry both visual and oculomotor signals. Neurophysiological studies targeting areas such as FEF, SC, MST, and the lateral intraparietal area might provide key insights into how abnormal development of signaling in visual cortical areas leads to oculomotor abnormalities.

With respect to the vergence system, it is easy to see how a loss of binocular responses in visual cortical areas could disrupt the tuning of neurons that would normally encode disparity vergence. Nonetheless, the details of how these abnormal signals develop across cortical and brain stem vergence circuits remain to be determined. In our discussion of vergence eye movements in strabismus we considered the hypothesis that, in developmental strabismus, neurons that would normally encode disparity vergence are driven, instead, by monocular visual signals and/or accommodation. This possibility could be tested by single-unit recording studies targeting vergence-related areas such as FEF, NRTP, caudal fastigial nucleus, posterior interposed nucleus, and SOA.

One highly characteristic oculomotor feature of the infantile strabismus syndrome that has not been addressed in any depth in this review is the phenomenon of DVD, an unusual and striking slow elevation of the nonfixing eye that occurs under occlusion and to some extent spontaneously. Although numerous authors have published hypotheses relating to the origin of DVD, it has received little to no attention from neurophysiologists, despite representing a significant treatment challenge for clinicians. Further research into the neural basis of DVD would provide a welcome complement to the existing body of work related to the neural basis of the other characteristic oculomotor features of infantile strabismus.

Finally, we recognize that while progress has been made in understanding the neural basis of the oculomotor abnormalities in the infantile strabismus syndrome, little is known about the neurophysiology of other, more common forms of childhood-onset strabismus. Acquired esotropia (usually accommodative or partially accommodative esotropia) is far more common than infantile esotropia and is typically acquired well after the development of normal binocular alignment and sensory and motor fusion. Although a key etiological factor in this type of strabismus is the excess accommodative demand required to compensate for hyperopic refractive error, resulting in excessive accommodative convergence, it is not known why not all children with equivalent levels of hyperopia develop esotropia, or why refractive correction is not always sufficient to control the misalignment once strabismus has developed. Intermittent exotropia is another very common form of childhood strabismus, in which periods of normal eye alignment and normal stereopsis alternate with episodes of large-angle strabismus, with corresponding sensory adaptations including suppression of the deviating eye. Whereas abnormal binocularity appears to be at the root of the neural abnormalities in the infantile strabismus syndrome, normal binocularity precedes or coexists with strabismus in both accommodative esotropia and intermittent exotropia. One would therefore predict different patterns of neurophysiological abnormalities in these other forms of childhood strabismus.

General Conclusions

From the available literature, it is clear that efforts to understand the etiology of strabismus must investigate the disorder at many different levels. The widespread deficits seen throughout the visual and oculomotor systems associated with strabismus highlight the highly balanced nature of the normal visuo-oculomotor system. Early disruption in visual signals during the sensitive period can have a widespread impact on the development of the oculomotor system, just as an early disruption in the oculomotor system can produce deficits in the visual sensory system. Normal binocular development involves the simultaneous consolidation of a binocular sensory system and an oculomotor system that supports bifoveal fixation, neither of which can exist in the absence of the other. How the brain is usually able to achieve this starting from an immature state, and why it sometimes does not, remains unknown. Early surgery for infantile strabismus may lead to better sensorimotor outcomes, yet excessively early surgery is not warranted for the high proportion of infants who show some degree of ocular misalignment in the early neonatal period. It may be hoped that better understanding of the neurophysiology underlying normal and abnormal binocular visual development might someday lead to better outcomes from treatment interventions for infantile strabismus.

GRANTS

This work was supported by EY024848 (M. M. G. Walton), EY06069 (M. J. Mustari), ORIP P51OD010425, and Research to Prevent Blindness.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.M.W., J.F., and M.J.M. conceived and designed research; M.M.W. and J.F. performed experiments; M.M.W., J.F., and M.J.M. analyzed data; M.M.W., A.C.P., J.F., M.J.M., and K.T.-H. interpreted results of experiments; M.M.W. and M.J.M. prepared figures; M.M.W., A.C.P., J.F., and M.J.M. drafted manuscript; M.M.W., A.C.P., J.F., and K.T.-H. edited and revised manuscript; M.M.W., A.C.P., J.F., M.J.M., and K.T.-H. approved final version of manuscript.