DESCENDING PROJECTIONS TO THE AUDITORY MIDBRAIN: EVOLUTIONARY CONSIDERATIONS

Abstract

The mammalian inferior colliculus (IC) is massively innervated by multiple descending projection systems. In addition to a large projection from the auditory cortex (AC) primarily targeting the non-lemniscal portions of the IC, there are less well-characterized projections from non-auditory regions of the cortex, amygdala, posterior thalamus and the brachium of the IC. By comparison, the frog auditory midbrain, known as the torus semicircularis, is a large auditory integration center that also receives descending input, but primarily from the posterior thalamus and without a projection from a putative cortical homolog: the dorsal pallium. Although descending projections have been implicated in many types of behaviors, a unified understanding of their function has not yet emerged. Here, we take a comparative approach to understanding the various top-down modulators of the IC to gain insights into their functions. One key question that we identify is whether thalamotectal projections in mammals and amphibians are homologous and whether they interact with evolutionarily more newly-derived projections from the cerebral cortex. We also consider the behavioral significance of these descending pathways, given anurans’ ability to navigate complex acoustic landscapes without the benefit of a corticocollicular projection. Finally, we suggest experimental approaches to answer these questions.

Introduction

The inferior colliculus (IC) is among the largest auditory nuclei in the vertebrate brain and it is virtually an obligatory synaptic station for ascending input to the medial geniculate body (MGB) ((Aitkin and Phillips 1984) see (Malmierca et al. 2002; Schofield et al. 2014) for exceptions). The large size of the IC and its many connections suggest that it has critical roles in both the ascending and descending limbs of the auditory system. In amphibians, its homolog, the torus semicircularis (TS), is a massive prominence in the caudal midbrain that also receives many converging ascending inputs (Foster and Hall 1978). The importance of the IC is highlighted by the relatively reduced size of the auditory portions of the thalamus in both non-amniote species such as frogs (Hall and Feng 1987) and non-mammalian amniotes such as alligators (Pritz 1974a, b). In addition to the important role of the IC as a nexus of the ascending auditory system, previous studies have demonstrated the importance of cortical descending projections to the IC in auditory behaviors in mammals (Bajo et al. 2010; Xiong et al. 2015). There is also a large body of evidence showing that neuronal responses to sounds in IC are altered by focal electrical stimulation or inactivation of the auditory cortex (AC). Non-auditory areas of the cortex also project to the IC, though their role in modulating IC responses has not been fully explored (Olthof et al. 2019; Lesicko et al. 2016; Yang et al. 2020). As reviewed below, AC stimulation shifts the tuning properties of IC neurons toward those of the stimulated neurons for preferred frequency (Jen et al. 1998; Jen and Zhou 2003; Zhou and Jen 2007; Ma and Suga 2001b; Yan et al. 2005; Yan and Suga 1998), amplitude (Jen and Zhou 2003; Yan et al. 2005; Zhou and Jen 2007), azimuth (Zhou and Jen 2007; Zhou and Jen 2005), and duration (Ma and Suga 2001b). While in mammals the neocortex is the major source of descending control of the auditory midbrain (Suga 2008; Stebbings et al. 2014; Bajo and King 2013), in anurans this role appears to be played by the thalamus (Feng and Lin 1991). A potentially homologous pathway from the thalamus and nuclei of the brachium of the IC (BIC) to the IC is found in mammals (Adams 1980; Winer et al. 2002; Winer 2005; Kuwabara 2012; Kuwabara and Zook 2000; Senatorov and Hu 2002; Patel et al. 2017; Ito et al. 2019), suggesting that mammals carry both an evolutionarily-ancient pathway from subcortical structures to modulate the IC, as well as an evolutionarily newer pathway from the cerebral cortex. Here, we review several anatomical and physiological features of the various descending projection systems to the IC and attempt to draw parallels across these systems seen in mammalian and amphibian species.

Auditory systems of amphibians and mammals

Amphibian and mammalian auditory systems differ in a number of ways and also have a number of similarities. Starting at sound transduction, acoustic signals in amphibians are transduced by two organs – the basilar papilla and the amphibian papilla, which are tonotopically organized, but do not have a basilar membrane, and show tuning based on electrical and mechanical properties (Schoffelen et al. 2008). In contrast, the mammalian cochlea is a coiled tonotopic structure, containing hair cells situated on a flexible basilar membrane whose mechanical properties determine the frequency tuning of constituent hair cells. In the central auditory system, amphibians and mammals share many likely homologous structures in the brainstem, but beyond the midbrain, the structural differences between the amphibian and mammalian auditory systems become more evident. For example, in anurans, the main initial processing nucleus is known as the dorsal medial nucleus (DMN) and is likely homologous to the mammalian cochlear nucleus (Feng 1986a). In fact, Feng and Lin (1996) proposed that the portfolio of cell types seen in the frog DMN resembles the variety of cell types seen in the cochlear nucleus (Feng and Lin 1996). In addition, rather than having a tripartite structure as is observed in the mammalian cochlear nucleus, DMN comprises a single tonotopic region (Fuzessery and Feng 1981; Will et al. 1985). Similar to the cochlear nuclei in mammals, DMN sends multiple outputs, including projections to the bilateral superior olivary nuclei, projections to the nuclei of the lateral lemniscus, and a major contralateral projection to the TS (Wilczynski and Endepols 2007). Interestingly, a descending projection pathway that has been described from the mammalian IC to the superior olivary and cochlear nuclei (Malmierca et al. 1996; Schofield and Cant 1999; Caicedo and Herbert 1993; Vetter et al. 1993; Milinkeviciute et al. 2017) appears to have a parallel in the frog auditory system (Feng 1986b; Luksch and Walkowiak 1998; Matesz and Kulik 1996; Kulik et al. 1994). In addition, both anurans and mammals appear to have a small projection from the auditory brainstem that bypasses the midbrain and directly innervates the thalamus (Malmierca et al. 2002; Schofield et al. 2014; Feng 1986b).

The TS in anurans, similar to mammals, is a major processing node in the auditory system, and receives convergent ascending input from the DMN, superior olivary nucleus and nuclei of lateral lemniscus. The TS shows tonotopy, and contains three main subnuclei: the principal, laminar and magnocellular nuclei. The mammalian IC also has three main nuclei: the central nucleus, the lateral cortex (sometimes called external cortex or lateral nucleus) and a dorsal cortex. The correspondence between these nuclei across anurans and mammals is not entirely clear, though the principal nucleus of the TS and central nucleus of IC share the property of receiving a major ascending input from the contralateral DMN and cochlear nucleus, respectively (Wilczynski 1981; Kulik et al. 1994). Similar to the non-lemniscal nuclei of IC (Lesicko et al. 2016; Jane and Schroeder 1971), the TS also receives ascending somatosensory input, primarily targeted to regions outside of its principal nucleus (Muñoz et al. 1997). As will be described below, TS also receives descending input from the thalamus, primarily to the laminar and magnocellular nuclei, similar to the thalamic input to the non-lemniscal nuclei of the mammalian IC.

Amphibians and mammals differ significantly in terms of their auditory forebrain organization. The main outputs from the mammalian IC are to the auditory thalamus (part of the dorsal thalamus, which we conventionally refer to here as thalamus). Projections to the MGB derive from all IC subnuclei, but the largest input is derived from the central nucleus. In anurans, the principal nucleus of the TS projects to the central and posterior thalamic nuclei. In general, the amphibian thalamus diverges quite substantially from the mammalian thalamus in terms of its structure and physiology. The mammalian thalamus contains multiple subnuclei, with the canonical principal cell type being thalamocortical neurons. These neurons send focused projections to the cerebral cortex, show low-threshold bursting behavior mediated by T-type calcium channels and receive a reciprocal connection from the cerebral cortex (Sherman and Guillery 2002). A subset of thalamic neurons, typically found in midline or intralaminar nuclei, has a major projection to the striatum, has a less focused projection to the cortex, and contains neurons that show variable bursting behavior (Van der Werf et al. 2002; Lacey et al. 2007; Smith et al. 2006). In contrast, the amphibian dorsal thalamus is divided into anterior, central, anterior lateral and posterior lateral regions. The relevant nuclei for the current discussion are the central and posterior regions of the thalamus which receive input from the TS, which are considered part of the collothalamus (Butler 2008). Neurons in these regions do not burst (Yang et al. 2012) and do not have a cortical (i.e., dorsal pallium) projection – their output is relatively diffuse and targets the striatum. It is notable that many of the canonical features of mammalian thalamic organization – reciprocal thalamocortical/corticothalamic projections and bursting – are not present in amphibians. It is not until the appearance of amniotes 350 million years ago that we observe point-to-point and reciprocal thalamic-dorsal pallium interactions resembling the mammalian thalamocortical system (Ulinski 1986).

In anurans, thalamic neurons project primarily to the striatum rather than to a cortex or cortex homolog. The notion that amphibians contain a homolog of the mammalian neocortex has been debated (Striedter and Northcutt 2019). In amniotes, such as reptiles, the dorsal pallium may represent a homolog of the mammalian cerebral cortex (Adams 1980; Northcutt and Kaas 1995). It has been argued that the region in amphibians that has been called dorsal pallium is really a transition zone between medial and ventral pallium and does not share the same connections as dorsal pallium in amniotes (Striedter and Northcutt 2019). The dorsal pallium in amphibians does not display layering, unlike that seen in reptiles (Tosches et al. 2018; Laurent et al. 2016), and is primarily connected with other telencephalic structures, with only a minor diencephalic projection (Roth et al. 2007). Thus, although clear parallels exist when comparing amphibian and mammalian central auditory systems at the level of the upper and lower brainstem, they diverge quite dramatically in the forebrain. However, in both classes of organisms the auditory midbrain is structurally heterogeneous and appears to occupy a central role in auditory processing and receives massive convergent ascending and descending projections (see Fig. 1 for schematic representations of the amphibian and mammalian auditory systems). Below, we review what is known about the descending projections to the midbrain in both sets of species to understand more about their evolutionary relationship, and ultimately their roles in sensory processing. For mammals, we consider both a highly specialized and heavily-studied group of animals – echolocating bats – as well as species that are less specialized for hearing.

Fig. 1:

Diagrams of the major structures and connections in the anuran and mammalian central auditory systems. Where projections are bilateral, only the stronger of the two is drawn, for purposes of clarity. AC = auditory cortex, AMY = amygdala, AP = amphibian papilla, BP = basilar papilla, DMN = dorsomedial nucleus, IC = inferior colliculus (C=central nucleus, D=dorsal cortex, L=lateral cortex), MGB = medial geniculate body (D=dorsal, V=ventral, M=medial), NLL = nuclei of lateral lemniscus (V=ventral, D=dorsal), SON = superior olivary nuclei (LSO = lateral superior olive, MSO=medial superior olive, MNTB = medial nucleus of the trapezoid body), STR = striatum, THAL = thalamus (C=caudal, P=posterior). TS = torus semicircularis (P=principal, L=laminar, M=magnocellular).

The bat auditory cortex

The AC of echolocating bats is organized similarly to other mammals except that it contains functional areas that are specialized to process biosonar signals. Typically, echolocating bats are grouped into two categories: those that emit short-duration downward frequency-modulated (FM) sweeps (FM bats) and those that emit a combination of constant-frequency (CF) and FM pulses (CF-FM bats). The most well-studied of these bats is the mustached bat (Pteronotus parnellii), whose AC comprises multiple fields. The primary AC (AI) is tonotopically organized and subdivided into three regions according to the frequency range that is processed (Suga and Jen 1976). Low frequencies (< 61 kHz) are represented caudally in posterior AI (AIp) and high frequencies (> 61 kHz) rostrally in anterior AI (AIa). Between them is the foveal region with an overrepresentation of the CF component of the second harmonic. The second and third harmonics of FM components have smaller representations within AI and are located dorsal to the tonotopic region (Suga and Jen 1976; Suga et al. 1978; O’Neill and Suga 1982). Other areas such as the FM-FM and dorsal fringe areas respond to specific delays between pulses and echoes and are likely important for target ranging. In these areas, neurons are organized in a topographic fashion based on the delays that elicit maximum spike counts (O’Neill and Suga 1982, 1979; Riquimaroux et al. 1991; Suga et al. 1978; Schuller et al. 1991).

The auditory cortices in FM bats are varied in their organization schemes. Many FM bats contain neurons with simple and complex response properties residing within the same tonotopic regions, and many of these bats will have one or two additional fields, some of which contain mirror-image tonotopy (Shannon-Hartman et al. 1992; Dear et al. 1993; Macías et al. 2014; Macías et al. 2009). In contrast, the short-tailed fruitbat, Carollia perspicillata, contains six functional fields (Esser and Eiermann 1999). In another fruit-eating FM bat, Phyllostomus discolor, four physiologically and anatomically distinct fields have been described (Hoffmann et al. 2008).

The neocortex of bats shows the typical mammalian six-layered structure (Fitzpatrick and Henson 1994; del Campo et al. 2014; Macias et al. 2019). There are currently limited data on the laminar-specific connectivity of bat AC. The deep layers of the echolocating bat AC contain the requisite large pyramidal neurons (Fitzpatrick and Henson Jr 1994), are retrogradely labeled after tracer injection into IC (Ito et al. 2019; Marsh et al. 2002) and can be stimulated to modify the IC (Sun et al. 1996), and therefore seem to represent the source of descending projection to the thalamus and subcortical centers as they are in other mammals but details on their projection pattern remain to be established. However, despite the small amount of anatomical information that has been published, electrical stimulation or pharmacological suppression of the AC has been found to be crucial in modulating responses in the IC, suggesting that corticofugal projections in bats are substantial in number.

The bat corticocollicular system: cortical modulation of the inferior colliculus

The roles of descending corticofugal projections in auditory processing in echolocating bats have been investigated using mainly two approaches (see Suga (2012) for review): 1) effects on response properties of IC neurons after electrical stimulation of AC have been examined and 2) these effects have been examined after locally inactivating the AC. In most cases, electrophysiological recordings of IC acoustic response properties have been performed to determine the impact of cortical manipulations on IC response properties (Bajo and King 2013). As will be described below, corticofugal projections have been found to have an important effect on the neuronal sensitivity to frequency, intensity, duration and location of IC neurons (Yan and Suga 1998; Ma and Suga 2005, 2001b; Zhou and Jen 2005; Jen and Zhou 2003).

Cortical modulation of frequency tuning in the bat

Auditory neurons usually respond maximally in magnitude to a certain frequency, commonly known as the best frequency (BF). Electric stimulation of the DSCF area of the AC of the mustached bat has been shown to facilitate responses of similarly-tuned neurons in the IC. Further, stimulation of the DSCF area will shift the BF of unmatched IC neurons such that their BF will migrate towards that of the stimulated cortical area. These tuning shifts can last several hours after less than 10 minutes of cortical stimulation. This type of modulation is referred to as “egocentric selection” such that it increases the midbrain representations of the range of frequencies housed in the stimulated cortical area (Zhang and Suga 2000). Conversely, inactivating specific cortical regions increases acoustically-driven responses of BF-unmatched IC neurons, thus diminishing the relative representation of the silenced cortical area – an effect known as a “centripetal shift” in tuning (Zhang et al. 1997). Similar egocentric and centripetal effects have been seen in FM bats such as the big brown bat (Ma and Suga 2001b), though considerably less work has been done in these species. In addition, local IC stimulation also induced egocentric and centripetal frequency tuning shifts in other IC neurons. Interestingly, these shifts were blocked by inactivation of the AC (Zhang and Suga 2005). The authors interpreted this effect to be mediated by corticofugal projections, which facilitated plasticity of IC neurons, which is a theme that we will return to in sections 5 and 9 below.

Cortical modulation of delay and duration tuning in the bat

As previously described, pulse-echo delay tuned neurons are found in the bat AC and are topographically distributed. Although delay-tuned neurons were initially described in the midbrain (Hopkins and Holstege 1978) and have been characterized in the IC (Mittmann and Wenstrup 1995; Portfors and Wenstrup 1999, 2001), it is not yet clear how delay-tuning in the IC is constructed or modulated. Electrical stimulation of auditory cortical regions tuned to particular pulse-echo delays increases the delay-tuned responses of similarly delay-tuned neurons in the IC. Conversely, cortical stimulation diminishes the responses of neurons with unmatched best delays, mirroring the egocentric/centripetal shifts described in the previous section for frequency tuning (Yan and Suga 1996). Lidocaine application to inactivate cortical regions containing FM-FM neurons produced opposite effects to those produced by electrical stimulation. Specifically, lidocaine application to FM-FM regions reduced the auditory responses of matched collicular neurons and broadened their delay tuning curves while enhancing the acoustic responses of unmatched collicular neurons.

Duration tuning is also found in the IC and AC of echolocating bats (Pinheiro et al. 1991; Casseday et al. 1994; Casseday et al. 2000; Covey et al. 1996; Ehrlich et al. 1997; Fuzessery and Hall 1999; Faure et al. 2003; Fremouw et al. 2005; Macías et al. 2015; Mora and Kössl 2004; Galazyuk and Feng 1997). Ma and Suga (2001a) reported that electric stimulation of duration-tuned regions in the big brown bat AC modulated IC duration tuned neurons in their frequency and duration tuning, but only when the IC and auditory cortical region were matched in best duration and frequency. Tuning for duration became sharper, and best duration responses were facilitated when the two regions were matched in best duration. In unmatched neurons, there was a shift in the direction of the best duration of the stimulated cortical neuron that was linearly related to the difference in best duration between collicular and cortical neurons. Thus, similar egocentric/centripetal patterns were seen for frequency, pulse-echo delay and duration tuning across both CF-FM and FM bats.

An amygdalofugal projection in bats

A direct projection from the basolateral amygdala to the IC has been described in mustached bats, pallid bats and pipistrelles (Marsh et al. 2002; Ito et al. 2019), which are all echolocators. This pathway appears to target all regions of the IC, and in mustached and pallid bats the numbers of cells projecting from the amygdala to the IC appears to be on the same order as the total number of corticocollicular cells, which is generally considered to be a large source of modulatory input to the IC. The presence of such direct projections from the amygdala to the IC in species other than bats has not been consistently demonstrated, though a small number of amygdalo-collicular fibers have been described in the cat (Hopkins and Holstege 1978). Given the important functions of the amygdala in signaling the presence of highly aversive or other emotionally-laden acoustic stimuli, such as social stimuli (Wenstrup et al. 2020), it is likely that this projection permits rapid adjustments of acoustico-motor behavior in an animal species that relies on very fast patterns of emitted pulses and echoes to guide echolocation. Given the small projection described in cats, it may be that the mammalian brain bauplan contains an amygdalo-collicular projection that was elaborated upon to support echolocation while flying. It therefore would be interesting to examine the presence and size this pathway in non-echolocating species of bats.

Auditory corticocollicular modulation in other mammals

Though most of the early work on the corticocollicular system was done on echolocating bats, primarily by Suga and his colleagues, more recent work has been focused on other experimental animals with less specialized auditory systems. This change in species has provided some opportunities for comparisons between specialized and non-specialized species when the paradigms were similar, though most recent work in the corticocollicular system does not use the classic paradigm used by Suga.

To the extent that similar paradigms have been used, similar results have been obtained. For example, Yan and Ehret found in the mouse that stimulation of auditory corticocollicular projections shifted the best frequencies of IC neurons towards those of the stimulated site of the cortex (Yan and Ehret 2002; Yan et al. 2005; Yan and Ehret 2001), thus recapitulating the main findings seen in echolocating bats. There has also been a great deal of work examining other aspects of the corticocollicular system. For example, lesioning of the corticocollicular system impairs the normal plastic changes that occur in the IC after unilateral deafening, and corticocollicular projections appear to have a role to play in adjusting gain control of IC neurons after deafening (Asokan et al. 2018; Bajo et al. 2010). These effects may reflect a general role of corticocollicular projections in supporting plasticity at the level of the IC, as described above for echolocating bats (Zhang and Suga 2005) and potentially supported by the broad tuning of corticocollicular neurons and alterations in IC stimulus selectivity seen after corticocollicular stimulation (Blackwell et al. 2020; Williamson and Polley 2019).

Of interest from an evolutionary perspective are the recent findings that the corticocollicular projections may trigger escape behavior. Escape is a highly conserved behavior and midbrain circuits have been strongly implicated in escape responses (Branco and Redgrave 2020; Brandão et al. 1994). Zhang and colleagues found that activation of corticocollicular pathways triggers rapid escape behavior in mice that mimics their escape responses to loud sound (Xiong et al. 2015). This finding is thought to involve AC to IC projections, which then project to superior colliculus and periaqueductal gray to mediate the motor components of the response. It is interesting that the corticocollicular pathway is involved in this process, since it is usually argued that tectally-mediated escape responses exist because they are more rapidly engaged than behavioral responses requiring cortical processing. It is certainly possible (indeed likely) that the experimentally-induced cortically-triggered escape responses reflect a layer of control superimposed on a hardwired motor reflex to ensure that escape is not triggered under inappropriate circumstances. To that end, it will be interesting to determine what the impact of lesioning of the corticocollicular pathway would be on acoustically-driven escape responses.

Mammalian auditory thalamotectal projection:

A projection from the auditory thalamus, nearby posterior thalamic regions and brachium of the IC to the IC has been identified in mouse, rat, echolocating bat, cat and monkey (Adams 1980; Winer et al. 2002; Winer 2005; Kuwabara 2012; Kuwabara and Zook 2000; Senatorov and Hu 2002; Patel et al. 2017; Ito et al. 2019), herein referred to as the thalamotectal pathways. The neurons that comprise these pathways are found primarily in the more medial regions of the auditory thalamus containing the medial division of the MGB, posterior intralaminar nucleus and peripeduncular nucleus. There are also dense projections from the nearby nuclei of the brachium of the IC. These thalamotectal neurons have many features that distinguish them from canonical thalamocortical neurons. They do not have the typical bitufted morphology of thalamocortical neurons, they tend to not burst, they do not project to cortex and they do not express typical calcium binding proteins found in thalamocortical cells (parvalbumin, calbindin and calretinin), and a subset of these neurons are GABAergic (Patel et al. 2015; Patel et al. 2017; Winer et al. 2002). Like the evolutionarily more newly-derived corticocollicular pathway, these neurons target the non-lemniscal regions of the IC (dorsal and lateral cortices (Patel et al. 2017)). They also branch extensively to structures in the lower auditory brainstem (Kuwabara 2012; Kuwabara and Zook 2000). The degree to which corticocollicular projections branch to the lower auditory brainstem is not yet known, though early data using dual retrograde tracing techniques suggest that the degree of branching to the brainstem is small (Doucet et al. 2003). The finding that the mammalian thalamotectal system is derived from regions of the thalamus that have morphological, physiological and connectional similarities to thalamic neurons found in amphibians (see below), suggested to several authors that this system may represent an evolutionarily ancient system within mammals, that operates in conjunction with newer pathways involving the neocortex (Patel et al. 2015; Winer 2005). This idea will be explored below.

Anuran auditory thalamotectal projection:

Several studies have established the presence of an auditory thalamotectal projection in anurans. Here we review the literature noting that not all of these studies were done the same family of frogs, limiting our ability to draw exact parallels between the studies. Early work using lesion studies in the thalamus of Rana pipiens identified broad areas of the tectum as receiving thalamic inputs (Trachtenberg and Ingle 1974). Later work utilized focal horseradish peroxidase injections into the TS of Rana pipiens to identify its inputs and revealed substantial projections from the thalamus (Feng and Lin 1991). Feng and Lin found that this thalamic projection densely innervates the non-principal nuclei of the TS: the laminar and magnocellular nuclei. This thalamotectal projection is derived primarily from the central and posterior nuclei of the thalamus. The utility of having such a projection is not yet known, however several studies have shed some light on its potential function. Endepols and Walkowiak used a unique isolated brain preparation to record from TS neurons in Discoglossus pictus and Bombina orientalis intracellularly and then to determine the impact of thalamotectal stimulation on TS responses to auditory nerve stimulation. They reported that the main nucleus that receives functional thalamic input is the laminar subnucleus and that the primary influence of stimulation of these projections is suppression of “acoustically” driven responses ((Endepols and Walkowiak 2001; Endepols and Walkowiak 1999), See Fig. 2). Similar suppression was seen in a whole animal preparation using extracellular recording in Rana pipiens and Hyla cinerea, though inhibition was not uniformly seen and in some cases depended on the nature of the acoustic stimulus (Ponnath and Farris 2014). These data that are suggestive of a net inhibitory effect are consistent with the finding described above that a subset of thalamotectal neurons in mammals are GABAergic, and that lesioning of this region in amphibians may disinhibit tectally-mediated behaviors, at least for visual stimuli (Finkenstädt 1980; Ewert 1968).

Fig. 2:

The impact of thalamic stimulation on midbrain responses to fictive acoustic stimulation in the frog. A) Top: location of stimulating electrode in the thalamus, Bottom: location of the stimulating electrode in the eighth nerve. B) Top: Impact of electrical stimulation of eighth nerve on spiking in a midbrain neuron, recorded intracellularly. Arrowhead = stimulus artifact. Middle: Impact of stimulation of the thalamus, showing a hyperpolarizing response after an initial spike. Bottom: simultaneous stimulation of eighth nerve and thalamus, showing a diminished number of spikes compared to eighth nerve stimulation alone. C) Summary data in terms of action potentials (APs, top, n=5 cells) or post-synaptic potentials (PSPs, bottom, n=6 cells) in response to combined bottom-up and top-down stimulation. Figures from Endepols and Walkowiak 2001, with permission.

Amphibian auditory striatotectal projection:

Anatomical studies have revealed reciprocal connectivity between the striatum and TS of amphibians and the descending projection appears to target all three subdivisions of the TS (Marín et al. 1997; Vesselkin et al. 1980). Electrophysiological studies in the whole-brain ex-vivo anuran preparation described above revealed that a smaller proportion of TS neurons received striatal input than thalamic input. Most of these TS neurons received isolated inhibitory post-synaptic potentials after striatal stimulation, though about half of the responsive neurons demonstrated an increase in responsiveness to eighth nerve stimulation after striatal stimulation. This finding is distinct from the responses to thalamic stimulation described in section 7.0 which primarily suppressed TS responsiveness (Endepols and Walkowiak 1999; Endepols and Walkowiak 2001). In addition, unlike the thalamotectal system, which has clear parallels in the mammalian brain, to our knowledge a striatotectal system has not been described in mammals, though a pallido-tectal projection has been reported (Yasui et al. 1990; Shammah-Lagnado et al. 1996; Shinonaga et al. 1992).

Evolution of corticofugal systems and conclusions

To summarize the main findings with respect to descending projections to the IC: 1) Mammals have extensive sets of descending projections to the IC emanating from the neocortex and smaller projections emanating from the amygdala (which show high species-specificity), globus pallidus and posterior thalamus/pretectal regions, 2) Mammalian corticocollicular projections appear to be important in supporting plastic changes in the IC and in triggering evolutionarily-conserved escape behavior and 3) Amphibians do not have a corticocollicular projection (and in fact, may not have a homolog of mammalian cortex), but have a thalamotectal projection that may be important in modulating otherwise stereotyped tectally-mediated behavior. These findings leave us with several unanswered questions. First, what are the roles of the various descending projections with respect to normal hearing? Second, and related, at what point in evolution did corticocollicular projections arise - and what new behaviors or environmental pressures coincided with this arrival? Third, why do mammals need so many ways to modulate the IC? Finally, what are the relationships between the various descending pathways that modulate IC?

Anurans have to navigate acoustically-cluttered environments. For example, a female frog may have to find her mate in the dark using acoustic cues only, but has to do so in the midst of hundreds of other chorusing frogs, often calling with sound pressure levels that approach 100 dB SPL (Gerhardt 1975; Loftus-Hills and Littlejohn 1971; Penna and Solís 1998; Narins 1982). This is a very challenging signal processing problem that is functionally equivalent to the famous “cocktail party” problem faced by humans. This type of behavior is seen across multiple families of frogs (Gerhardt and Klump 1988; Mecham 1971) and also seen in toads (Arak 1983), and similar capacities have been found in male frogs (Gerhardt and Bee 2007), suggesting that the ability to navigate complex acoustic environments may be shared across anurans and in both sexes. Clearly anurans have mastered this difficult task (Blair 1958; Feng et al. 1976; Gerhardt 1982), and have done so without the benefit of a massive corticocollicular system to provide high-level feedback to shape tuning of IC neurons. These facts suggest that corticocollicular projections may not be necessary for this kind of task, or at least the frog has solved this problem in a different way than mammals have. At the very least, it suggests that corticofugal projections are not uniquely required to perform complex attention-requiring tasks (Krauzlis et al. 2018; Merker 2007). Given that top-down information may be very useful in solving this problem (i.e. using a “representation” of a mate’s calls as a Bayesian prior to help disambiguate incoming noisy sound streams), the only apparent sources for this descending information in anurans are the thalamotectal and striato-tectal (Marín et al. 1997) systems as well as descending systems emanating from the TS. In support of a potential role for thalamotectal projections, anuran thalamic neurons have acoustic response properties after exposure to behaviorally-relevant stimuli that are sufficiently complex to potentially carry such information (Brown and Marks 1977; Hall and Feng 1987; Mudry et al. 1977), but it is not yet known if thalamic neurons are necessary for solving this cocktail party problem in frog. If selective lesioning of the frog thalamotectal system were to interfere with mate-finding behavior, that would be very suggestive that thalamotectal neurons play a role in helping to disambiguate noisy signals. Relevant to this question, a study was performed involving lesions of the thalamus and/or TS of female gray treefrogs, and subsequent investigation of any resulting changes in phonotaxis behavior. It was found that thalamic lesions had relatively little impact on phonotaxis compared to TS lesions. However, the behavioral task did not require the frog to single out a particular call in a loud chorus (i.e., it did not require solution to the “cocktail party” phenomenon), and, as pointed out by the others, thalamic (and therefore thalamotectal) involvement may be required for more challenging tasks that require selective attention (Endepols et al. 2003). Future work should investigate the impact of selective and reversible silencing of thalamotectal projections in a frog faced with a challenging task requiring disambiguation of a complex sound.

When did corticocollicular projections first appear? This question has not been answered satisfactorily in the auditory system. However, there are visual corticofugal (both corticothalamic and corticotectal) projections that are first seen in the amniote lineage, specifically reptiles (Hall et al. 1977; Ulinski 1986; Elprana et al. 1980; Isabekova 1974), though these animals have a three-layered cortex so that the superficial appearance of these neurons bears little resemblance to what is seen in mammalian neocortex. To date, too little work has been done on these projections to know if they represent a homolog of mammalian corticofugal projections. To that end, it would be instructive to determine the degree to which these neurons share physiological and molecular characteristics of mammalian layer 5 and layer 6 corticocollicular neurons, which carry cortical information to the mammalian IC (Stebbings et al. 2014). If reptile corticofugal projections do represent the first arrival of the corticocollicular system, they arose in a group of species that had unique environmental challenges relative to amphibians. Amphibians, by definition, have adapted to land and aquatic habitats, but their land options are limited to damp environments near bodies of water. Given their predilection to remain on land, including arid regions, primitive reptiles had to survive in a variety of landscapes, possibly requiring the behavioral flexibility afforded by corticofugal projections, though this point remains speculative.

A final consideration is whether the more ancient IC-modulating pathways interact with the more recently-arrived pathways. Given that evolution has often been referred to as a “tinkerer” rather than developing new pathways de novo, one would assume that newer corticocollicular pathways would be built upon older thalamotectal pathways. If this is indeed the case, then one would predict that a subset of corticothalamic projections synapse on thalamotectal neurons and modulate the IC via this indirect pathway. Indeed, at least in the mouse, there are dense projections from the AC to the thalamus in the zones where thalamotectal neurons originate, and indeed many of these terminals are quite large (deemed “driver” synapses by Sherman and Guillery, (Llano and Sherman 2008)). If this indirect pathway exists that connects AC to IC, then it may be necessary to re-interpret previous studies that manipulated the corticocollicular pathway in light of the potential that manipulations of the AC indirectly impacted the IC based on projections through the thalamus. Given modern circuit-tracing techniques, it is possible answer the question of whether an integrated AC→MGB→IC circuit exists, or if corticocollicular and thalamotectal projections remain segregated (Fig. 3) using trans-synaptic retrograde or anterograde tracing approaches.

Fig. 3:

Diagrammatic representations of the potential relationships between corticocollicular projections and evolutionarily-ancient thalamotectal projections. Left: A fully integrated model whereby corticothalamic axons activate thalamotectal neurons via a branch of a corticocollicular axon. Both systems synapse on and influence the same neuron in the colliculus. Right: A fully segregated model whereby thalamotectal and corticocollicular projections do not converge on any common targets.

Thus, given that frogs have the capacity to solve complex acoustic tasks and recognize conspecific calls in the face of dense acoustic clutter, we conclude that the corticocollicular pathway is not uniquely necessary for this type of behavior. The corticocollicular pathway appears well-suited to contribute to behavioral flexibility and plastic changes in the IC, two attributes that are generally ascribed to the mammalian neocortex. We also suggest that it would be fruitful to conduct studies in both amphibians and mammals to 1) determine if the thalamotectal pathways are likely homologs (by looking for parallels in their molecular and physiological phenotypes), 2) investigate the behavioral impacts of manipulation of the thalamotectal projection in both sets of species, 3) determine the degree to which corticocollicular and thalamotectal pathways interact, which may help frame some of the heterogeneous effects that have been seen with previous studies of the corticocollicular system in mammals and 4) determine if these characteristics are shared in different families within a species or across both sexes, given some of the differences noted above.