Paradoxical kinesia may no longer be a paradox waiting for 100 years to be unraveled

: Parkinson ’ s disease (PD) is a progressive neurodegenerative disorder mainly characterized by bradykinesia and akinesia. Interestingly, these motor dis-abilities can depend on the patient emotional state. Disabled PD patients remain able to produce normal motor responses in the context of urgent or externally driven situations or even when exposed to appetitive cues such as music. To describe this phenomenon Souques coined the term “ paradoxical kinesia ” a century ago. Since then, the mechanisms underlying paradoxical kinesia are still unknown due to a paucity of valid animal models that replicate this phenomenon. To overcome this limitation, we established two animal models of paradoxical kinesia. Using these models, we investigated the neural mechanisms of paradoxical kinesia, with the results pointing to the inferior colliculus (IC) as a key structure. Intracollicular electrical deep brain stimulation, glutamatergic and GABAergic mechanisms may be involved in the elab-oration of paradoxical kinesia. Since paradoxical kinesia might work by activation of some alternative pathway bypassing basal ganglia, we suggest the IC as a candidate to be part of this pathway.


Introduction
Parkinson's disease (PD) is the second most common neurodegenerative disorder after Alzheimer's (Dorsey et al. 2007). PD typically develops between the ages of 55 and 65 years and occurs in 1-2% of people over 60 years, rising to 3.5% at age 85-89 years (de Lau and Breteler 2006;Twelves et al. 2003). In general, PD is mainly characterized by bradykinesia (slowness of movement), akinesia (loss of movement), rigidity and postural instability, all of which contribute to gait variability and walking difficulty (Kemoun and Defebvre 2001). These motor impairments are attributed to a dysfunction of the basal ganglia-motor cortex circuits due to the loss of meso-striatal dopamine (DA). Abnormal firing of the basal ganglia motor output nuclei, i.e. substantia nigra pars reticulata (SNpr), and internal segment of globus pallidum, produces functional changes in subcortical and cortical structures and is thought to cause bradykinesia (Berardelli et al. 2001;Marsden and Obeso 1994;Wichmann and DeLong 1996). PD is characterized by two major pathologic processes: (i) premature selective loss of dopaminergic neurons, and (ii) the accumulation of Lewy bodies, composed of ∝-synuclein, which become misfolded and accumulate in multiple systems. Pathologic studies demonstrated that a step-wise degeneration of neurons occurs over many years (Braak et al. 2003). The motor symptoms become evident when there is 30-70% cell loss in the substantia nigra (SN; Dauer and Przedborski 2003). Although motor features define the disorder, various non-motor features are typically seen, including autonomic dysfunction, cognitive and psychiatric changes, sensory symptoms and sleep disturbances. The basis of pharmacotherapy aims to replace DA loss with dopaminergic medications.
Important historical references to PD-like motor symptoms can be found in early Egyptian papyrus from the 12th century B.C. to the notebooks of the renaissance genius Leonardo da Vinci (McMurrich 1930). Traditional Indian texts from approximately 1000 BC and ancient Chinese documents also provide descriptions of symptoms that would currently be considered as PD (Manyam 1990;Zhang et al. 2006). These observations did not receive the credit they deserved for contributing to the recognition of an important clinical entity probably because this disease tends to occur in persons whose age exceeds the prevailing life expected of that period (McMurrich 1930). It was only after James Parkinson published "An Essay on The Shaking Palsy" in 1817 (Parkinson 2002) and Charcot (1872) expanded the former's original description, that PD gained recognition as a distinctive clinical entity.
Exactly a century ago, French neurology flourished due to the end of the First World War and the need to deal with millions of soldiers with mutilations and neurological and psychological disorders. Working at the Pitié-Salpêtrière Hospital, among several notable names of that time, Alexandre-Achille Cyprien Souques (1860Souques ( -1944 distinguished himself and is remembered for his research on Parkinsonism. In 1921, he published an essay entitled "Rapport sur les syndromes parkinsoniens" (Souques 1921), in which he described for the first time an intriguing phenomenon, namely "a sudden and brief period of mobility typically seen in response to emotional or physical stress" (Souques 1921) in patients with advanced PD. Souques coined the term "paradoxical kinesia" to describe this phenomenon, which will be addressed in this review.

Paradoxical kinesia noticed in patients with Parkinson's disease
An interesting aspect of PD is the dynamic nature of the motor symptoms. For example, several impressive clinical reports show that bradykinesia can depend on the patient's emotional state (Jankovic 2008(Jankovic , 2005Naugle et al. 2012). For instance, Bonanni et al. (2010) reported that fourteen PD patients affected by severe akinesia were able to get up and flee during the strong L'Aquila earthquake in 2009, and many helped caregivers and members of their families to escape. Also, there is evidence from case reports: One akinetic PD patient was able to suddenly escape when someone screamed ''fire'' (Jankovic 2008), whereas another was suddenly able to run to save a child from an approaching car (Daroff 2008). Sometimes, this dramatic condition reaches such a degree that a previously incapacitated patient cannot be distinguished from a perfectly healthy person. Another interesting aspect of this phenomenon is that quickly after the event, the patient's akinesia returns (Daroff 2008). On the other hand, Sacks (2012) reported that the duration of paradoxical kinesia might vary from a few seconds to minutes and sometimes even hours. Paradoxical kinesia, however, is not restricted to stressful or even life-threatening events. Non-aversive visual and auditory cues have also been shown to be effective in improving gait (Azulay et al. 1996(Azulay et al. , 1999Georgiou et al. 1993; Morris et al. 1996;Suteerawattananon et al. 2004), movement velocity, movement accuracy (Majsak et al. 1998), reaction times (Kühn et al. 2004) and freezing (Kompoliti et al. 2000). For instance, patients can move effectively, like kicking a ball (Asmus et al. 2008) and riding a bicycle without any obvious cause of strong aversive emotion (Snijders and Bloem 2010;Snijders et al. 2011). Luria (1932) found that his PD patients could only walk across the floor when pieces of paper were placed on it "to stimulate each step". Regarding auditory cues, the effect of music on movement disorders is well known and used therapeutically (Harrison et al. 2018;Nombela et al. 2013;Sihvonen et al. 2017;Young et al. 2014). The benefit of rhythmic stimulation has already been reported in the 19th century when Charcot (1892) observed that after long carriage, train or horseback rides, patients with PD experienced marked symptom amelioration. Hence, he developed a device to provide rhythmic movement by an electrically powered "shaking chair" (fauteuil trépidant) and advocated vibratory therapy for the management of the disease (Charcot 1892).
Over the following years, paradoxical kinesia was undervalued as a phenomenon with an obscure significance to the clinical practice. More recently, Sacks described cases of paradoxical kinesia in patients with Parkinsonism caused by Encephalitis lethargica (Sacks 2012). In his book "Awakenings", Sacks reported that familiar music can induce paradoxical kinesia, i.e. "temporary awakenings", in patients with severe akinesia. Since then, many clinical studies have shown that music or cueing sounds can be therapeutically valuable for PD patients to treat freezing of gait and akinesia (Harrison et al. 2018;Nombela et al. 2013;Sihvonen et al. 2017;Young et al. 2014).
Despite that, the mechanisms and neural circuitry driving this remarkable phenomenon have remained enigmatic since its first description by Souques a century ago (1921). However, several proposals regarding the mechanism underlying such cue-related improvements in motor function were made: Some authors suggested that while motor function capacity is preserved, movement deficits occur as a result of impaired motor control (Rubinsten et al. 2002), i.e., PD patients have intact motor programs but have difficulty accessing them without an external sensory stimulation (Arias and Cudeiro 2008;Baker et al. 2007;Clark et al. 2009;Jankovic 2008;Matsui et al. 2006;Willems et al. 2007). On the other hand, activation of basal nuclei reserves, i.e., "surviving" dopaminergic neurons, or activation of alternative pathways, for example involving cerebellar circuits to improve motion somehow, could explain the phenomenon of paradoxical kinesia (Glickstein and Stein 1991). Here, two questions arise: (i) Which component of the triggers induces paradoxical kinesia and (ii) which is (are) the alternative neural pathway(s). The present review will address these questions.

Which components of the triggers induce paradoxical kinesia?
Several external triggers (cues) can cause paradoxical improvement of motor function. It is known that visual (transverse serial lines or laser cane; Hanakawa et al. 1999;McCandless et al. 2016) and auditory (music or metronome; McIntosh et al. 1997) cues may induce paradoxical movements in some PD patients. Regarding visual cues, floor markers were reported to be effective in improving the gait of PD patients (Azulay et al. 1999). In some cases, the patient is instructed to walk over each marker to achieve the desirable stride length for each step. Visual cues probably help because the patients focus on gait (Azulay et al. 1999;Morris et al. 1996). However, once the patient concentrates on walking, it is no longer an automatic task being processed through the defective basal ganglia. Indeed, when complex secondary tasks are interspersed between evaluations or when the patients do not know they are being evaluated and consequently are not focusing on their stride, their gait parameters degrade again (Morris et al. 1996;Scandalis et al. 2001). In any case, it seems desirable that the patient reestablishes his movements rhythmically and automatically without using attentional mechanisms, as in healthy individuals. The present review focuses on the auditory system because (i) unlike visual stimuli, beat perception is independent of attention or explicit musical training (Bouwer et al. 2014); (ii) the auditory system has a strong bias to detect temporal patterns of periodicity and structure, compared to other sensory systems (Thaut et al. 1999); (iii) auditory stimuli, for instance, music, have the remarkable ability to drive rhythmic, metrically organized motor behaviour (Patel et al. 2005;Repp and Penel 2004) and (iv) the auditory system has rich fiber connections with motor centers ranging from the spinal cord up to the brain stem, subcortical and cortical levels (Felix et al. 2011;Schmahmann et al. 2007).

Paradoxical kinesia as a result of auditory-to-motor transformation in humans
During the last years, the scientific literature has clearly shown the power of music in modifying neural activity. This effect is not limited to the auditory system but propagates to the motor system and beyond (Chen et al. 2008;Fujioka et al. 2012;Grahn 2012;Grahn and Brett 2007). Basic neurophysiology and biophysics of sensorimotor connectivity have shown an important interaction between the auditory and the motor system (Thaut 2005). Auditory rhythmic cues can synchronize with motor responses revealing a universal phenomenon named entrainment, defined by a temporal locking process in which one system's motion or signal frequency entrains the frequency of another system. Specifically, rhythmic entrainment refers to the ability to align motor actions with an auditory beat, a complex cognitive phenomenon that seems to depend on a dynamic interaction between auditory and motor systems in the brain (Merchant and Honing 2014;Patel 2014). For instance, finger and arm movements instantaneously entrain to the period of a rhythmic stimulus (music or a metronome beat) and stay locked even when subtle tempo changes are induced into the metronome that are consciously not perceived (Thaut et al. 1998). The finger and arm entrainment may suggest that the auditory system is specially constructed to detect temporal patterns in auditory signals with extreme precision and speed (Moore 2013) and, importantly, that the auditory system can influence motor output (Casseday and Covey 1996). Two early electrophysiological studies aimed at explaining the neural basis for such auditory-motor entrainment showed how sound signals and rhythmic music can prime and time muscle activation via reticulospinal pathways (Rossignol and Jones 1976). Beyond that, some motor networks, for instance, the basal ganglia, are active during beat perception even when no movement is involved (Merchant et al. 2015), suggesting that the mere perception of a beat involves interactions between auditory and motor areas in the brain (Zatorre et al. 2007). The interaction between these two areas indicates that the human mind has access to several distinct levels of periodicity, one of which can be selected at any given time as the beat (Drake et al. 2000;Phillips-Silver and Trainor 2007). More than that, an internal representation of rhythm must exist since rhythmic behaviours in humans also can emerge without a cueing sensory stimulus, as occurs when a musician plays alone (Mendoza and Merchant 2014) or when we keep the pace of a walk. Healthy individuals use internal oscillators (a non-specific cue to trigger movement) provided by the basal ganglia to release successive movement sequence stages (Grahn 2009;Harrington et al. 1998). Deficiencies of these internal oscillators due to basal ganglia degeneration (Grahn 2009;Georgiou et al. 1993;Harrington et al. 1998;Harrison et al. 2018) may lead to increased gait variability (Moon et al. 2016), as observed in PD. Indeed, two of the hallmark features of PD, freezing and festination, may both be manifestations of a problem with the maintenance of an internal gait rhythm (Giladi et al. 2001a(Giladi et al. , 2001b(Giladi et al. , 1992 due to basal ganglia dysfunction. Fortunately, despite this basal ganglia dysfunction, the affected brain can also access rhythmic entrainment mechanisms, probably activating alternative pathways. Early studies of gait training in PD confirmed the existence of rhythmic entrainment processes in patients (McIntosh et al. 1997;Thaut et al. 1996). Endorsing this evidence, Oliver Sacks, in his book named Musicophilia, demonstrated that music has the power to provide sequences and can even achieve this when other forms of organization fail (Sacks 2008). For some PD patients, music can be as powerful as any therapeutic drug in relieving motor symptoms (Sacks 2008). Here, it is worth pointing out that there is no single "music center" in the human brain, but that music processing involves around a dozen scattered networks throughout the brain (Sacks 2008) that have initially developed for other purposes (Pinker 1997).
Another point that remains to be clarified is whether the sensory-motor transformation observed during paradoxical kinesia may also encompass the motivational aspects of the trigger. In that regard, Mazzoni et al. (2007), introduced the idea that the motor system has its own motivation circuit, which operates analogously to but separately from explicit motivation. These authors proposed that DA from the nigrostriatal pathway triggers this "motor" motivation and thus, less DA may account to bradykinesia in PD patients (Mazzoni et al. 2007). So, it seems rational to think that this "motor" motivation may also be triggered by relevant auditory stimulus. A link to motivational aspects is particularly relevant because in searching to explain paradoxical kinesia, we should remember that both aversive (the scream "fire") and appetitive (music) auditory stimuli can induce this phenomenon. The emotional/motivational value of the stimulus may generate an internal process of moving oneself away from what one does not want (escape behaviour) or bringing one closer to what one wants (approach behaviour), thus defining the direction of the movement. Therefore, when searching for an alternative pathway that could be activated by auditory triggers inducing paradoxical kinesia, it would be plausible to look for brain areas that can process both auditory and emotional components of the stimuli and generate or support some form of motor output. In this sense, the inferior colliculus (IC) rises as a potential candidate.

Inducing paradoxical kinesia via activation of the inferior colliculus
The IC is a classical auditory nucleus in the brainstem known for being involved in processing auditory information and represents a major hub for integrating ascending and descending auditory pathways. As a source of projections to several areas, it is a site of convergence, receiving projections from the cochlear nucleus, superior olivary complex and the lateral lemniscal nuclei, and feedback projections from the auditory cortex (Oliver and Huerta 1992). The IC is comprised of a central nucleus (ICc), the main tonotopically organized subdivision (Malmierca et al. 2008;Romand and Ehret 1990) and the principal source of information ascending to the thalamus , which then proceeds toward the cortex. Surrounding the ICc are the lateral (ICl), dorsal (ICd), and caudal cortices of the IC (ICca), which are collectively referred to as the shell regions (ICs). These subdivisions are primarily non-tonotopically organized (Malmierca et al. 2008;Romand and Ehret 1990) and send ascending projections to the thalamus (Calford and Aitkin 1983;Kudo and Niimi 1980;Wenstrup 2005) as well as the superior colliculus (SC; Druga and Syka 1984;Harting and Van Lieshout 2000;Van Buskirk 1983), and descending information to the auditory brainstem (including the ICs; Huffman and Henson 1990). In the IC, the two main neurotransmitter classes most neurons belong to are glutamatergic and GABAergic (Ito and Oliver 2012). Converging anatomical and physiological evidence indicates that neurons within the IC are sensitive to visual, oculomotor, eye position and somatosensory information as well as signals relating to behavioural context and reward (Gruters and Groh 2012). Hence, the IC could be understood as a highly interconnected sensory, cognitive, and motor network that synthesizes higher-order auditory perception.
Although it is fundamentally an auditory area, the IC also serves motor functions that must be considered when contemplating its various roles (Casseday et al. 1976, Covey andCarr 2005). Ramon y Cajal's idea of the IC as an auditory reflex pathway was incomplete in the sense that it has functions that cannot be viewed as a "reflex" in the classic sense yet must undoubtedly be construed as a motor function; in that, it serves as a route to enable motor action via telencephalic auditory centers or deep superior colliculus or pontine gray/cerebellar circuits (Casseday et al. 2005). For instance, the ICs receive inputs from the ipsilateral caudal globus pallidus (GP; Shammah-Lagnado et al. 1996;Shinonaga et al. 1992;Yasui et al. 1990) as well as bilateral GABAergic inputs from the substantia nigra pars lateralis (SN1; Coleman and Clerici 1987;Moriizumi et al. 1992;and Yasui et al. 1991). The IC also receives bilateral projections from the ventral tegmental area (VTA) (Herbert et al. 1997), while both the ICc and ICs receive ipsilateral input from the basal nucleus of the amygdala (Hopkins and Holstege 1978;Marsh et al. 2002). All these regions have a substantial body of literature implicating them in various motor (Gruters and Groh 2012;Shammah-Lagnado et al. 1996) and motivational (Ono et al. 2000) behaviours, and thus auditory perception and behaviour are likely to be shaped by these non-auditory inputs to the IC.
In addition to transmitting patterns of changing air pressure detected by the cochlea, the IC is distinguished from other auditory nuclei in the brainstem by its output connections to motor pathways and movement coordination systems (Casseday and Covey 1996). These connections may explain why spinal reflexes can be elicited in decerebrated cats in response to acoustic stimuli, which disappeared after IC bilateral ablation (Adams 1983;Chu 1970;Gernandt and Ades 1964;Wright and Barnes 1972). The IC sends auditory information to motor centers that participate in the organization of behaviours, such as prey catching, predator avoidance, and orientation to a novel stimulus (Casseday and Covey 1996). Besides, the IC has long been implicated as part of the brain aversive system mediating aversive stimulus-induced emotional states. It is well known that electrical or chemical stimulation of the IC induces defensive behaviours such as arousal, freezing and escape responses (an explosive motor behaviour) that mimic reactions to fear elicited by environmental challenges (Brandão et al. 1988(Brandão et al. , 1993(Brandão et al. , 2005Moreira et al. 2003;Melo and Brandão 1995;Melo et al. 1992). More specifically, the ventral region of the IC is predominantly involved in filtering acoustic stimuli of aversive nature (Ferreira-Netto et al. 2007).
The IC sends rich projections to the cerebellum (via the dorsolateral pontine nuclei; Thaut et al. 2015), which is activated during sensorimotor synchronization tasks (Grahn et al. 2011;Stephan et al. 2002). Additionally, it was demonstrated that neural activity in the IC can synchronize to a rhythmic auditory stimulus (Tierney and Kraus 2013). This representation of timing information in the IC may have an important function in auditory-to-motor transformations during rhythmic entrainment (see the previous section), indicating an involvement of this auditory structure.
Corroborating the assumption that the IC is responsible for the integration of emotional/sensory/motor information, Casseday and Covey (1996) elaborated the theory that the neural mechanisms for perception and motor performance, having evolved together, are inseparably linked to one another. This theory states that a major function of the auditory midbrain is to select biologically important information and distribute it at a temporal pace appropriate for motor performance. Indeed, anatomically, the IC is uniquely positioned to integrate precise information from descending and ascending auditory information, emotion, and motor output. Thus, it seems rational that the IC be recruited during paradoxical kinesia, playing an important role in sensorimotor gating activated by emotional auditory stimuli.
Although the hypothesis to explain paradoxical kinesia looks convincing, a deeper investigation of the neural mechanisms underlying this phenomenon has been limited due to a scarcity of suitable animal models replicating this clinical phenomenon. In order to overcome this limitation, we developed two animal models of paradoxical kinesia induced by either aversive or appetitive triggers.
Before presenting our animal models of paradoxical kinesia, it is worth clarifying some important aspects of an animal model. Following a classical categorization strategy, animal models can be grouped into "fidelity models" that reproduce the maximum number of characteristics of the original and "discrimination models" (Russell and Burch 1959) that reproduce only one particular characteristic of the original. One might think that a fidelity model is always superior to a discrimination model; however, the latter can help by teasing out irrelevant factors from the important ones. The separation of relevant from irrelevant factors was the case in our studies. In order to reproduce paradoxical kinesia in rats, first akinesia had to be induced. For that, we choose a classical highly discriminative/poor fidelity model termed haloperidol-induced catalepsy, which models the akinesia and lack of spontaneous motor activity common in some PD patients. In rodents, systemic or intrastriatal administration of haloperidol induces a behavioural state in which the animals cannot correct externally imposed postures (Sanberg 1980). Such cataleptic rats exhibit inactivity, decreased responsiveness to stimuli, and a tendency to maintain an immobile posture. The cataleptic effects of haloperidol are mediated by the blockade of postsynaptic DA D2 receptors in the neostriatum (Sanberg 1980), mainly located on GABAergic projection neurons and cholinergic interneurons (Johnson et al. 2014;Kharkwal et al. 2016).

Intracollicular neural substrate involved in paradoxical kinesia
Given that catalepsy can be counteracted by systemic administration of glutamate antagonists (Elliott et al. 1990;Hauber and Schmidt 1990;Riederer et al. 1991), we investigated whether this immobility state would be affected by interfering with glutamatergic neurotransmission in the IC. The results showed that intracollicular administration of glutamate receptor antagonists (AP7 and MK-801) and the agonist NMDA influences systemic haloperidol-induced catalepsy in rats. The antagonists attenuate the catalepsy whereas the agonist potentiates it (Melo et al. 2010; see Figure 1). Due to the hypothesis that the neostriatum is the main brain structure responsible for catalepsy induced by neuroleptics (Costall et al. 1972;Honma and Fukushima 1978;Ossowska et al. 1990), we next investigated if glutamatergic neural circuitry in the IC can also influence the catalepsy induced by haloperidol microinjected directly into the neostriatum. We found that prior bilateral microinjection of NMDA into the IC induced a pro-cataleptic effect. MK-801 microinjected into the same treatment regimen, however, significantly reduced catalepsy induced by bilateral microinjection of haloperidol into the neostriatum (Medeiros et al. 2014). Intracollicular glutamate therefore has a proven role in catalepsy; likewise, the widely distributed GABAergic neurotransmitter is also involved. Systemic haloperidolinduced catalepsy is enhanced by IC microinjections of midazolam (Tostes et al. 2013) or diazepam (Tonelli et al. 2018b), benzodiazepine and GABA receptor agonists, respectively. The GABAergic antagonist bicuculline produced a biphasic effect (Tostes et al. 2013) which is particularly relevant since one role of GABAergic inhibition is to regulate IC output so motor centers do not receive excessive excitatory input (Casseday and Covey 1996).
Neurotransmitters other than glutamate and GABA may also be involved in the modulation exerted by the IC on motor centers. For instance, the endocannabinoid system modulates glutamatergic neurotransmission and dopaminergic activity in the basal nuclei (García et al. 2016). The activation of the endocannabinoid system causes hypopolarization of presynaptic neurons through Ca 2+ influx blockade. Consequently, the Ca 2+ blockade decreases the release of neurotransmitters such as GABA or glutamate in the synaptic cleft (Felder et al. 1993; Fride and Mechoulam Figure 1: Graphical representation summarizing haloperidol-induced catalepsy, and its modulation by glutamatergic drugs microinjected into the IC. (A) Rats received haloperidol (i.p.) and AP7 or NMDA in the IC a week after stereotactic surgery for microinjection cannulas implantation. Then, the bar test was performed and the step-down latency (catalepsy time) was assessed. (B) Catalepsy induced by haloperidol (1.0 mg/kg, i.p.) is attenuated by prior microinjection of the glutamatergic antagonist AP7 (10 or 20 nmol/0.5 μL) into the IC. (C) After intracollicular NMDA microinjection an increase of the catalepsy induced by haloperidol (0.5 mg/kg, i.p.) was observed. Bars represent step-down latency after AP7 (10 or 20 nmol/0.5 μL), NMDA (20 or 30 nmol/0.5 μL) and Sal (0.5 μL) microinjections in the IC of rats exhibiting catalepsy. Data are expressed as mean ± SEM. *p < 0.05 compared to Sal group. Dashed lines represent catalepsy time from additional control groups, i.e. rats receiving saline (i.p.) and AP7, NMDA or saline, into the IC. 1993; Mackie and Hille 1992;Mackie et al. 1995). Therefore, cannabinoid receptors in a given nucleus suggest an endogenous neuromodulatory system, playing a role in synaptic regulation there (Gerdeman and Lovinger 2001;Szabo et al. 2000). Proof for endocannabinoid involvement in the IC has been determined through immunohistochemical methods (Moldrich and Wenger 2000). Thus, it seemed plausible to investigate whether endocannabinoids play some neuromodulatory role in the IC concerning haloperidol-induced catalepsy. Indeed, rats receiving IC microinjection of the endogenous cannabinoid anandamide showed an increase in the duration of systemic haloperidolinduced catalepsy. In contrast, pretreatment of the IC with the CB1 cannabinoid receptor-selective antagonist AM251 significantly decreased the duration of catalepsy induced by peripheral haloperidol (Medeiros et al. 2016). Thus, intracollicular endocannabinoids may be considered a potential pharmacological target for treating movement disorders. This statement is supported by anatomical connections between the IC and the SN (Olazábal and Moore 1989) and other structures that are part of motor circuits. See the more detailed discussion about circuitries involving IC and motor structures in section "5" in this review.

Inducing paradoxical kinesia in rats
3.1 Animal model of paradoxical kinesia induced by appetitive stimulus: 50 kHz ultrasonic vocalizations Given that auditory (aversive or appetitive) stimulation can induce paradoxical kinesia in patients, we predicted that exposing rats to species-relevant auditory stimulation would reduce haloperidol-induced motor deficits. At this point, the first questions that came to mind were: Which auditory stimulation is relevant for rats? Which "song" do they like (or dislike) to hear? It has been shown that juvenile and adult rats emit 50 kHz ultrasonic vocalizations (USV) to communicate appetitive arousal and the presence of positive emotional states to conspecifics (e.g. Simola and Brudzynskic 2018). Briefly, 50 kHz calls serve important socio-affective communicative functions in regulating social interactions in rats and are thought to reflect a positive emotional state of the sender ("rat laughter"; Panksepp 2005). In line with a pro-social communicative function, 50 kHz USV are known to induce behavioural activation and approach behaviour towards the sound source in the receiver (Wöhr and Schwarting 2007). Both effects are closely related to DA function in brain reward centers like the nucleus accumbens (Burgdorf et al. 2001;Willuhn et al. 2014). In contrast, 22 kHz USV are known as "alarm cries" (Litvin et al. 2007), and are usually emitted in aversive situations, such as social defeat, predatory exposure, or fear conditioning (Brudzynski 2007;Wöhr and Schwarting 2010). For our studies, therefore, it seemed appropriate to use USV as auditory stimulation because USV emissions in rats are valence-specific (Brudzynski 2013), and are considered measures of many aspects of emotional and motivated behaviour.
To test our predictions, USV were presented to rats under haloperidol-induced catalepsy, and their response to the calls was assessed. In our study, the horizontal bar classically used to test catalepsy was combined with a novel setup developed by our group (see Figure 2A). Corroborating our predictions, when playback of 50 kHz USV was applied, cataleptic rats rapidly started to move and stepped down from the bar. Most of them even walked toward the sound source showing typical approach behaviour ( Figure 2B and C). Interestingly, soon after the 50 kHz USV presentation ended, the rats displayed catalepsy again (Tonelli et al. 2018a). This is reminiscent of PD patients, who revert to akinesia after an auditory or visual stimulus has induced paradoxical kinesia, strongly supporting the face validity of our animal model. Thus, for the first time, it was demonstrated that playback of 50 kHz USV is efficient to release rats from haloperidol-induced catalepsy. Our paradigm has led to a completely new method for evaluating paradoxical kinesia using appetitive auditory stimulation in rats.
Further studies were conducted to investigate the neural mechanisms underlying paradoxical kinesia induced by this new method. Since auditory stimulation was used as a trigger, the next logical step was to investigate whether pharmacological manipulation in the IC could interfere with the behavioural response to 50 kHz USV playback in rats under haloperidol-induced catalepsy. Indeed, the glutamatergic agonist NMDA, microinjected into the IC, increased catalepsy time (Medeiros et al. 2014;Melo et al. 2010) and blocked paradoxical kinesia induced by 50 kHz playback presentation (Tonelli et al. 2018b). Although potentiation of catalepsy can also be observed after intracollicular microinjection of the GABA/BZD agonist diazepam, it did not affect paradoxical kinesia induced by 50 kHz calls, suggesting that this phenomenon involves glutamatergic rather than GABAergic neurotransmission in the IC (Tonelli et al. 2018b). We assume that the motor improvement (stepping down from the bar and approach behaviour) induced by 50 kHz USV is the outcome of some form of sensorimotor transformation that encompasses motivational aspects of the trigger. This sensorimotor transformation occurs in the IC and involves glutamatergic mechanisms there. We can also speculate that 50 kHz playback may have induced the expectation of a social partner, and interestingly, previous studies in rats demonstrated that the activity of IC cells was found to increase as an expected rewarding (Nienhuis and Olds 1978) or aversive (Ruth et al. 1974) stimulus drew near.
Unlike appetitive 50 kHz calls, playback of 22 kHz USV did not significantly alter catalepsy time, since rats did not step down from the bar when exposed to the 22 kHz calls. This was not due to an auditory deficit, because most of the rats, while on the bar, showed pinna reflex and head movement towards the sound source, implying that they perceived the auditory stimuli. At first sight, this result might indicate that these aversive acoustic stimuli are not effective in reducing the cataleptic state, which is in contrast to classical clinical findings in PD patients (Bonanni et al. 2010;Schlesinger et al. 2007) and our previous experiments with aversive DBS of the IC (Melo-Thomas and Thomas 2015). However, one must consider that the typical response to playback of 22 kHz USV in most rat strains is reduced activity or transient immobility (Wöhr and Schwarting 2010). Therefore 22 kHz USV playback might not be compatible with the current model of testing haloperidol-induced catalepsy.
Since PD patients suffer from chronic motor impairments, typically slowness of movement, we considered it important to assess the long-term symptom relief in rats. For that, we addressed whether 50 kHz USV playback ameliorates psychomotor deficits induced by haloperidol in a subchronic dosing regimen (7 days), which is implemented by drug administration via osmotic mini-pumps in rats. This treatment significantly reduces locomotor activity, resembling a bradykinesia state in rats (Amato et al. 2011). Here is worth defining akinesia and bradykinesia. These are interrelated terms and characterize a negative symptom of Parkinsonism, namely the reduction of motor activity. "Akinesia" means immobility, however, except in the case of extremely severe states, Parkinsonian humans, as well as Hatched areas indicate proximal and a distal zone, i.e. one close to and the other opposite from the sound source, respectively. Each zone (24% of the arena) was divided into 6 quadrants. Briefly, an hour after haloperidol i.p. microinjection rats exhibiting catalepsy (bar test) were challenged with 50 kHz USV playback presentation (respective controls are not represented here). (B) Catalepsy times (i.e. durations until rats stepped down from the bar) under haloperidol-induced catalepsy and playback of different auditory stimuli. Bars represent means + SEM during 50 kHz USV, noise, background noise and silence playback presentation. *p < 0.05, as compared with noise, background and silence. Dashed line represents the criterion of 30 s to start playback presentation. (C) Exploratory behaviour displayed by rats that stepped down from the bar when exposed to playback of 50 kHz USV, as measured by counting quadrant crossings in the zone distal from and proximal to the active ultrasonic speaker, respectively. Data are expressed as means + SEM. *p < 0.05, as compared with the distal zone. Figure  animals, are not totally unable to move. Therefore, this term usually refers to the impaired ability to initiate movements. "Bradykinesia" indicates that movements in Parkinsonian individuals are slower than observed in healthy controls (Berardelli et al. 2001). In rodents, acute haloperidol administration induces an immobility state (limbs tend to remain in whatever position they are placed), and for this reason, haloperidol-induced catalepsy models the akinesia, which is common in some PD patients (Castagne et al. 2009;Sanberg et al. 1988;Wadenberg 1996). In turn, bradykinesia (movement time increases) is modeled by sub-chronic haloperidol administration  Thus, using a sub-chronic haloperidol treatment regimen, we demonstrated for the first time that 50 kHz USV playback could overcome bradykinesia in rats . It has been shown that the state of bradykinesia induced by sub-chronic haloperidol treatment can be assigned to decrease extracellular DA levels in the medial prefrontal cortex (mPFC; Amato et al. 2011). In contrast, electrical stimulation of the IC produces a longlasting increase in DA levels in the prefrontal cortex (Cuadra et al. 2000). It is also known that the output of the auditory cortex reaches broad areas of the prefrontal cortex (Jones and Powell 1970;Fuster 1989;Pandya and Kuypers 1969;Van Eden et al. 1992), and auditory-motor integration takes place at cortical levels as well. Based on these reports, we can speculate that during the processing for the auditory stimulation used in our study (50 kHz USV), the IC may increase DA levels in the prefrontal cortex. Such an increase may compensate for reduced levels of extracellular DA induced by sub-chronic haloperidol treatment. In addition, since paradoxical kinesia might work to improve motion by activation of some alternative pathways bypassing basal ganglia (Glickstein and Stein 1991), we can suggest that the IC is an important candidate to be part of this pathway.
3.2 Animal model of paradoxical kinesia induced by aversive stimulus: high frequency deep brain stimulation of the IC Deep brain stimulation (DBS) is an invasive and reversible treatment by which electrical pulses are applied to a neural structure within the brain through a chronically implanted electrode (Naskar et al. 2010). The goal of DBS is to influence neurons in their functioning and signaling and thereby relieve symptoms of patients with several brain disorders, such as PD. We hypothesized that DBS of the IC would activate the neural circuitry of fear and consequently induce explosive motor behaviour, which would temporarily counteract the catalepsy induced by dopaminergic impairment in the neostriatum; thus, providing an animal model to study paradoxical kinesia. The haloperidol-induced catalepsy model was again used to reproduce akinesia in rats. As expected, a significant reduction of catalepsy was seen in rats given haloperidol and receiving DBS at the IC. We showed that high frequency (830 Hz) intracollicular DBS at a specific current amplitude which elicits escape behaviour (escape threshold; Melo and Brandão 1995;Melo et al. 1992), also induces vigorous escape responses in rats treated with haloperidol, thereby releasing the cataleptic state (Melo-Thomas and Thomas 2015). This effect was observed during DBS, but it was temporary since the rats exhibited catalepsy again when stimulation was turned off. To the best of our knowledge, this was the first study suggesting that electrical stimulation of the IC can interfere with catalepsy induced by neuroleptics in rats. We proposed intracollicular DBS as a suitable animal model to study paradoxical kinesia, in this case, induced by aversive stimulation. Moreover, the fact that this effect was temporary, i.e., after showing explosive motor behaviouur rats exhibited catalepsy again once DBS was switched off, strongly supports the face validity of our animal model ( Figure 3A and B). Indeed, this temporary motor improvement seems to be reminiscent of PD patients, who revert to akinesia after an event has induced paradoxical kinesia.
Since intracollicular DBS improved motor deficits, one can suggest that the IC can be a non-conventional and alternative DBS target to treat PD. However, electrical stimulation induced aversive behaviour in rats, evidently limiting its use in clinical settings. In order to overcome this limitation, we performed further experiments on rats refining DBS parameters in the IC. Since typical DBS stimulation frequencies in humans range between 20 and 180 Hz (Anderson et al. 2005;Khoo et al. 2014;Stefani et al. 2007), we asked whether low-frequency DBS (30 Hz) in the IC could reduce haloperidol-induced catalepsy without having aversive side effects. We chose 30 Hz as the stimulation frequency because clinical evidence suggests that lower DBS frequencies might be more effective than higher ones (Khoo et al. 2014). In PD patients, increased effectiveness of DBS at lower frequencies is particularly evident for structures located outside the basal ganglia. For example, in the pedunculopontine nucleus, another brainstem structure involved in motor behaviour, the optimal stimulation frequencies range between 20 and 80 Hz (Stefani et al. 2007;Thevathasan et al. 2018).
We found that low-frequency 30 Hz-DBS (current amplitude 150-300 μA) targeted at the IC strongly ameliorates haloperidol-induced catalepsy (Engelhardt et al. 2018; Figure 3C). Notably, the motor improvements occurred without evidence of stimulation-induced escape behaviour, indicating that our protocol is not aversive in rats. Additionally, 30 Hz-DBS of the IC induced no place avoidance in the conditioned place preference test (Engelhardt et al. 2018), produced no anxiety-related behaviour on the elevated plusmaze (Engelhardt et al. 2018;Ihme et al. 2020), and did not trigger any other signs of aberrant motor reactions in the open field. Thus, we obtained no evidence of aversive or anxiogenic side effects across several independent measures. This finding corroborates our hypothesis that the IC has a modulatory role in motor function and can be targeted with DBS to alleviate akinesia-like deficits in an acute pharmacological model relevant to PD symptoms. Of note, we observed marked improvements despite deep catalepsy, although only unilateral stimulation was used. Clinical studies showed that bilateral DBS at basal ganglia structures is usually superior to unilateral DBS in treating PD (Kumar et al. 1999). Thus, our protocol might be even more effective when using a bilateral stimulation approach.
Aiming to fine tune DBS parameters to avoid aversive effects during IC stimulation, not only the stimulation frequency but also the current amplitude should be taken into account. We found that intracollicular DBS at 600 μA current amplitude combined with 30 Hz frequency, did not induce any evidence of aversive behaviour. Interestingly, these stimulation parameters induced an anxiolytic effect in rats submitted to the elevated plus-maze test and preserved the motor improvement assessed by the bar test (catalepsy; Ihme et al. 2020). As far as we know, our study is the first to show that the behavioural outcome changes from anxiogenic to anxiolytic by only changing the current parameters used to stimulate the same point at the same area (see Table 1). These are surprising results because electrical stimulation of the IC has so far only been associated with aversive behaviour (Brandão et al. 2003;Melo and Brandão 1995;Melo et al. 1992). Although additional studies are needed, our results not only corroborate the hypothesis that the IC can be a non-conventional DBS target in treating PD but also suggest that it could be beneficial to treat comorbidities like anxiety. Additionally, these outstanding results show that despite the clinical effectiveness of DBS, the underlying physiological mechanisms of the effects of DBS are far from being fully understood. By varying the stimulus parameters, the clinical outcome may lead us to a better understanding of the DBS mechanisms and allow clinicians to improve DBS system tuning.
Besides the abovementioned studies in awake, freelymoving rats, we tested anesthetized ones and demonstrated that high-frequency DBS in the IC induced immediate motor responses in specific body parts (hind-and fore-paws, vibrissae, tail and trunk; Melo-Thomas and Thomas 2015). This result is noteworthy because the explosive motor behaviour induced by electrical stimulation of the IC in awake rats has so far been ascribed to anxiety (Brandão et al. 2003;Melo and Brandão 1995;Melo et al. 1992), that is, the motor response was thought to be the consequence of aversive emotion. However, motor responses elicited by high-frequency IC DBS in anesthetized rats may not involve such emotional aspects (Melo-Thomas and Thomas 2015). Instead, the motor effect could be due to direct activation of some descending projection from the IC to motor areas, such as those located in the brainstem and cerebellum (see section "5"). These data could help us understand how paradoxical kinesia occurs clinically in response to a nonemotional auditory stimulus, such as the sound from a metronome.
Here it is worth mentioning that although DBS has emerged as a promising technique to treat motor disorders, an increasing number of patients have had unsatisfactory results from DBS surgery (Herrington et al. 2016;Le Goff et al. 2015;Okun et al. 2005). For instance, DBS of the globus pallidus significantly improved all Parkinsonian features but generated dyskinesias, even in the "off" drug condition (Bejjani et al. 1998). DBS of the subthalamic nucleus (considered the basal ganglia pacemaker), although significantly improving bradykinesia, can provoke a decline in verbal fluency performance in some patients (Costentin et al. 2019;Okun et al. 2005;Witt et al. 2013). Such worsening of verbal fluency may be related to microlesions due to the electrode trajectories. The subthalamic nucleus has the disadvantage of being situated deep in the brain. In order to reach it, the electrodes need to cross the inferior longitudinal fasciculus that plays a role in semantic processing (Costentin et al. 2019;Duncan et al. 2016). In contrast, the IC is located in the dorsal midbrain, a region of comparably easy access. Therefore, to reach the IC there is no need for the electrode to cross other major regions (apart from the overlying neocortex), avoiding microlesions due to the electrode trajectories. Again, by aiming to minimize DBS side effects in PD, the IC emerges as a promising non-conventional alternative target.
In summary, our studies showed that aversive high frequency (830 Hz) DBS in the IC of rats exhibiting haloperidol-induced catalepsy could represent an animal model of paradoxical kinesia. Our novel technique can be useful to investigate the neural mechanisms underlying this phenomenon in animals. In addition, we also proposed that IC could be an alternative nonconventional target for DBS in PD patients when using low-frequency stimulation (30 Hz) Down arrows indicate a reduction of catalepsy and + indicates that paradoxical kinesia was induced.
L. Melo-Thomas and R.K.W. Schwarting: Neural mechanisms underlying paradoxical kinesia which seems to be devoid of aversive effects. Currently, conventional DBS protocols targeted at basal ganglia sites can be ineffective for some PD patients (Olanow et al. 2009) or even bring side effects (Okun et al. 2005), warranting the search for alternative targets. Thus, according to our studies, the IC arises as a promising candidate since we showed that low-frequency (30 Hz) intracollicular DBS ameliorates motor impairments induced by haloperidol and also induces an anxiolytic effect in rats.

Other animal models of paradoxical kinesia
Previous studies by Yntema and Korf (1987) reported that stress (exposure to cold, handling and forced immobilization) significantly reduced catalepsy in rats injected with haloperidol. Colpaert (1987) showed a reduction of rigidity, tremor and hypokinesia induced by reserpine when he exposed the rats to stress by placing them in water. Rats exhibiting severe Parkinsonism induced by intraventricular 6-hydroxydopamine injections showed a rapid release of movement when placed among a colony of cats and escaped from a shallow floating ice bath (Keefe et al. 1989;Marshall et al. 1976). Furthermore, Keefe et al. (1989) demonstrated that the temporary motor improvement was not abolished by pretreating the animals with DA antagonist's haloperidol or SCH-23390, suggesting that paradoxical kinesia is not a consequence of DA release from residual dopaminergic neurons.
4 Evidence of IC modulating motor behaviour in animal models of Parkinsonism other than haloperidol-induced catalepsy

L-NOARG induced Parkinsonism
Another way to induce catalepsy in rats is by interfering with nitric oxide (NO) production in the neostriatum. There, NO modulates dopaminergic neurotransmission playing an important role in regulating cortico-striatal synaptic transmission (Del Bel et al. 2005;West and Tseng 2011). NO is a short-lived, highly diffusible free radical gas and serves as a messenger molecule in the nervous system. It is synthesized from l-arginine by a nitric oxide synthase (NOS; Garthwaite and Boulton 1995;Moncada et al. 1991). Rats and mice treated with various NOS inhibitors show impairments in fine motor control (Araki et al. 2001;Del Bel et al. 2002;Dzoljic et al. 1997;Star and Star 1995;Uzbay 2001). Systemic or intrastriatal administration of N G -nitro-l-arginine (l-NOARG), an inhibitor of NOS, induces catalepsy after intraperitoneal, intracerebroventricular or intrastriatal administration (Cavas and Navarro 2002;Del Bel et al. 2002Marras et al. 1995).
Based on this, we investigated whether IC glutamatergic mechanisms could modulate catalepsy induced by systemic l-NORARG. The results showed that glutamatergic antagonist (AP7) microinjected into the IC attenuated and the agonist (NMDA) potentiated the catalepsy induced by l-NOARG administered intraperitoneally (Iacopucci et al. 2012). These data show that glutamatergic neurotransmission in the IC influences not only systemic haloperidol-induced (Melo et al. 2010) but also L-NOARG-induced (Iacopucci et al. 2012) catalepsy.

MPTP Parkinsonism model
1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) is a commonly used toxin for inducing both rodent and primate PD models. Exposure of humans to MPTP causes a syndrome that mimics the core neurological symptoms and the dopaminergic neurodegeneration of PD (Langston et al. 1983). Studies in non-human primates identified that selective destruction of the nigro-striatal dopaminergic neurons is the pathological basis behind the motor deficits (Burns et al. 1983;Jenner et al. 1984;Langston et al. 1984). Regarding nonprimates, many species, including rats, are insensitive to the toxic effects of MPTP (Johannessen et al. 1985). However, specific strains of mice (notably black C57) are particularly sensitive to it (Sonsalla and Heikkila 1988) and have enabled the development of the MPTP mouse model of PD. That said, we addressed whether intracollicular DBS or intracollicular microinjection of the glutamatergic antagonist MK-801 in black C57 mice could improve motor deficits induced by this neurotoxic Parkinsonism animal model. MPTP was injected systemically, which induced a bilateral degeneration of the striatal DA systems. We demonstrated that unilateral microinjection of MK-801 or high-frequency DBS applied in the IC effectively improved MPTP-induced motor deficits in mice as tested on the rotarod and in the open field . Although it has not been tested, our protocol might be even more effective when using a bilateral stimulation approach. These results support previous reports from our group (Engelhardt et al. 2018;Ihme et al. 2020;Medeiros et al. 2014;Melo et al. 2010;Melo-Thomas and Thomas 2015) and reinforce the claim that the IC may provide a critical site able to activate motor circuits, even when striatal DA transmission is chronically impaired by MPTP treatment, which as a neurodegenerative model is more comparable to PD symptoms in humans.

6-OHDA model
The neurotoxin 6-hydroxydopamine (6-OHDA) is a widely used and valuable tool used to model PD in rats. Unlike MPTP (above), 6-OHDA does not efficiently cross the blood-brain barrier and requires direct injection into the brain. Usually, 6-OHDA is injected into the nigrostriatal tract at one of three locations: into the substantia nigra pars compacta (SNpc), the median forebrain bundle, or the terminal region of the neostriatum. Although the exact mechanism behind 6-OHDA toxicity is still not fully understood, the current understanding is that, once inside dopaminergic neurons, 6-OHDA initiates degeneration through a combination of oxidative stress and mitochondrial respiratory dysfunction. As a consequence, reduced levels of striatal DA and tyrosine hydroxylase (TH; rate-limiting step of DA biosynthesis) are observed (mimicking some biochemical features of PD), in addition to robust spontaneous and drug-induced behavioural phenotypes. The behavioural changes due to neurodegeneration following 6-OHDA lesions can be assessed by a standard rotation test from the first-week postsurgery. Briefly, after unilateral lesion, rats exhibit lateralized deficits in spontaneous behaviour, such as ipsiversive (i.e., towards the side of the lesion) rotations Huston 1996a, 1996b;Ungerstedt and Arbuthnott 1970). In addition, contraversive rotations (i.e., away from the lesion side) are produced by the systemic injection of the D1/D2 agonist apomorphine (APO). Such injections create an imbalance in striatal dopaminergic transmission, favouring excess stimulation of the ipsilateral (lesioned) side where denervated DA receptors have become supersensitive due to nigral cell destruction (Barneoud et al. 1995;Hefti et al. 1980;Hudson et al. 1993).
Using this animal model of Parkinsonism, we investigated whether such behavioural lateralizations (ipsi-and contraversive turning) can be attenuated by intracollicular DBS or glutamatergic manipulations in the IC. The results showed that when considering the specific site of electrode placement within the IC, 30 Hz DBS in the ventral (vCN) but not dorsal part of the IC central nucleus (dCN) attenuated the contraversive asymmetry induced by apomorphine. Interestingly, attenuation was dependent on the side of stimulation since the number of contralateral rotations was reduced only when the electrode was placed in the vCN of the IC ipsilateral to 6-OHDA lesion. This is particularly important since we demonstrated that intracollicular electrical stimulation in the vCN but not the dCN induced immediate motor responses in anesthetized rats (Melo-Thomas and Thomas 2015) and the ventral part of the IC is more sensitive to electrical stimulation eliciting wild running in awake rats (Bagri et al. 1991).
Regarding the influence of the glutamatergic neural network in the IC, we demonstrated that the ipsiversive asymmetry disappeared after microinjection of the glutamatergic antagonist, MK-801, into the IC, i.e., there was no difference when comparing the number of ipsiversive and contraversive rotations performed by 6-OHDA rats irrespective of the side of IC microinjection (ipsi-, contra-or bilaterally to the lesion). The absence of asymmetry was observed when the glutamatergic agonist NMDA was microinjected in the IC contralateral to the lesion. It is important to note, that NMDA microinjected in the IC ipsilateral to lesion produced a clear contraversive asymmetry, i.e., it reverted the lateralization ( Figure 4A-C). These findings speak for a modulation of the motor systems by the IC, strengthening the assumption that the motor and auditory system may act together to alleviate motor deficits in PD.

How can the inferior colliculus be involved in paradoxical kinesia? Anatomical considerations
In this section, we will discuss how the IC interacts with motor pathways at multiple levels including the brainstem, basal ganglia, thalamus and neocortex with implications to paradoxical kinesia.

The anatomical relationship of the inferior colliculus and mesencephalic locomotor region
The original studies describing locomotion following low amplitude stimulation of the posterior midbrain were carried out in precollicular decerebrate (Shik et al. 1969) and intact (Sterman and Fairchild 1966) cats. A functional region termed the mesencephalic locomotor region (MLR) was described and represents one of the supraspinal structures that command locomotor circuits localized in the spinal cord, which in turn control the coordination of muscle activity (Grillner 2003;Goulding 2009). However, the MLR does not reside in only one nucleus. Instead, it overlaps between the cuneiform nucleus (CnF) and the pedunculopontine nucleus (PPN) (Roseberry et al. 2016). Notably, the PPN is a key component of the reticular activating system. In humans, the MLR topography acquired through functional imaging techniques shows that imagining oneself walking is enough to activate the MLR and that its location closely fits data obtained in mice, centered in the CnF and PPN (Karachi et al. 2012). More recently, optogenetic manipulations have shown that stimulation of glutamatergic neurons in the PPN induces locomotion in mice (Roseberry et al. 2016). Furthermore, glutamatergic neurons there promote explorative locomotion, whereas those in the CnF promote escape (Caggiano et al. 2018), suggesting that these neurons may act in conjunction to control alternating-gait locomotion. Interestingly, the IC sends projections to glutamatergic PPN and CnF neurons (Caggiano et al. 2018), providing sensory (auditory) information to lower-level motor structures that generate behavioural output (i.e., the MLR; Figure 5A). Indeed, the PPN may be involved in initiating locomotion under certain conditions, such as when an animal is startled by an auditory stimulus (Garcia-Rill 1991).
PPN neurons also receive rich projections from basal ganglia nuclei (Roseberry et al. 2016;Dubuc 2013, 2017), suggesting that dysregulation of glutamatergic neurons in the PPN may have important roles in locomotor disability related to PD (Caggiano et al. 2018). The dysregulation may be at least partially compensated during auditory stimuli since they generate evoked potentials in the PPN, which correlate with enhancements in peak force and reductions in reaction time in PD patients (Anzak et al. 2016). This auditory input can reach the PPN via IC-PPN projections, through which the IC provides more than pure auditory information. Instead, as part of the brain aversive system, the IC may supply the PPN with auditory-emotional/motivational features leading to an additional movement gain important to producing paradoxical kinesia. The auditory-emotional/ motivational features are particularly important because the PPN is a key component of the reticular activating system (RAS). Through IC-PPN projections, proper auditory stimulation may exert an "energizing" influence on the motor system. Indeed, a loss of this "motor energy" in PD has been a theme in the literature for over a century (Flowers 1975;Hallett and Khosbin 1980;Schwab et al. 1959;Schwab and Zieper 1965;Trousseau 1921). Moreover, it has been proposed that paradoxical kinesia is produced by "energizing" relevant action systems in the brain which are otherwise insufficiently activated (Ballanger et al. 2006). Future manipulation of these IC-PPN projections exerting an "energizing" influence on motor circuitry may provide a novel approach for a Figure 4: Schematic illustration of the apomorphine-induced rotation modulated by glutamatergic neural substrate in the IC of rats. (A) Hemi-Parkinsonian rats were created by a unilateral injection of 6-OHDA in the right medial forebrain bundle (MFB) during a stereotaxic surgery. Immediately after the microinjection, the same rats were also implanted with microinjection guide-cannula in the IC. Apomorphine rotation test was performed a week later in rats receiving NMDA or MK-801 microinjection in the IC ipsi-, contra or bilaterally to lesion; (B) Lesion rats receiving physiological saline showed a clear ipsiversive asymmetry which disappeared after MK-801 intracollicular microinjection, regardless of the microinjection side (ipsi-, contra-or bilaterally) in the IC. There was neither effect of IC treatment nor any asymmetry in rats with sham lesions (data not showed); (C) after receiving Apo s.c., 6-OHDA rats exhibited a clear contraversive asymmetry. This asymmetry was significantly potentiated by MK-801 microinjected in the IC contralateral to the lesion side. The sham groups that received either MK-801 or physiological saline in the IC did not exhibit any asymmetry (data not showed). Bars represent means and vertical lines indicate S.E.M.; *p < 0.05; **p < 0.01, according to ANOVA, followed by Bonferroni's multiple comparisons post hoc test. non-dopaminergic/non-invasive (e.g. auditory stimulation) treatment in patients with motor disorders such as PD.

Pontocerebellar auditory pathway
The pontine nuclei (PN) are the largest of the pre-cerebellar nuclei and are the major provider of mossy fibers to the cerebellar hemispheres (Glickstein 1997;Lee and Mihailoff 1990;Schuller et al. 1991;Wiesendanger and Wiesendanger 1982a). The PN receives projections from all areas of the auditory cortex (Wiesendanger and Wiesendanger 1982b) and also glutamatergic projections from the IC (Saint Marie 1996) (pontocerebellar auditory pathway), suggesting an important auditory influence on the cerebellum.
In the cerebellum, sensory information is coordinated with motor feedback from the eyes, head and body to control movements (Schuller et al. 1991). Therefore, the IC is likely involved in modulating proprioceptive adjustments to sound by descending projections to the cerebellum via the PN. These projections permit auditory access to many brainstem and spinal pathways that affect muscle tone and movement initiation (Thompson 2005), enabling coordinated responses to sound. Accordingly, appropriate auditory stimuli or electrical stimulation of the IC activating this descending pathway could induce paradoxical kinesia circumventing basal ganglia deficits.

Nigrotectal and tecto-nigro-striatal pathways
In addition to receiving inputs from the auditory relay nuclei of the lower brainstem, the IC also receives descending projections from the SNl (Castellan-Baldan et al. 2006;Moriizumi and Hattori 1991), suggesting that the basal ganglia (A) Auditory information after being processed in the IC reach the MLR, an important mesencephalic supraspinal motor center that generate behavioural output (Caggiano et al. 2018;Garcia-Rill 1991;Goulding 2009;Grillner 2003). In turn, neurons from MLR projects to the RF, that are thought to mediate locomotion (Jordan et al. 2008), and from there to the spinal cord that execute locomotion (Kiehn 2016). This might be the way how auditory information reaches lower-level motor structures that generate behavioural output and produces paradoxical kinesia during auditory stimulation. Besides that, a descending dopaminergic pathway originating from the SNc innervates the MLR and was shown to increase locomotor output (Caggiano et al. 2018;Roseberry et al. 2016;Dubuc 2013, 2017). Degeneration of this dopaminergic pathway leads to a dysregulation of glutamatergic neurons in the MLR and may play a role in the locomotor deficits reported in PD. This dysregulation may be at least partially compensated during auditory stimulation. (B) When an emotional and motivational state is involved to produce paradoxical kinesia, as it happens during intracollicular DBS or 50 kHz playback presentation, an auditory-amygdalar feedback may be recruited. Indeed, projections from the IC to the MGB followed by its projections directly to the amygdala and auditory cortex have been already described (LeDoux et al. 1985). In turn, the IC receives projections from both amygdala and auditory cortex representing the auditory-amygdalar feedback (Marsh et al. 2002). In addition, the premotor cortex is another significant target receiving the major outputs from MGN (Cappe et al. 2009;Wenstrup 2005), suggesting that both auditory and motor representations of information may be transmitted by the MGN (Becker et al. 2002). The primary auditory and the motor cortex can also process multiple representations of an acoustic stimulus concurrently to encode different aspects of the sound (Cappe et al. 2009;Zotova et al. 2000), as for instance, generate the motor consequences of the acoustic signal, as it happens during paradoxical kinesia. Abreviations: inferior colliculus (IC), amygdala (Am), auditory cortex ( may modify collicular function. Indeed, behavioural evidence has shown that SN lesions can enhance the aversiveness of the IC stimulation, suggesting that defense motor patterning mechanisms recruit GABAergic fibers from SN that project to the IC (Bolam et al. 2000;Castellan-Baldan et al. 2006;Maisonnette et al. 1996). However, the IC is one of several brainstem sensorimotor structures, which are not only indirectly connected to the basal ganglia via sensorimotor loops, but also structures where basal ganglia outputs converge into a final common motor path to generate behavioural output (Moriizumi and Hattori 1991;Olazábal and Moore 1989;Redgrave et al. 2010). Moreover, the nigrotectal projections are structured to influence head movements. The SNpr is involved in visual orienting movements of the head and eyes via its projections to deep layers of the superior colliculus (SC). The IC plays a role in sound orientation (Olazábal and Moore 1989) through its projections to deep layers of the SC (Thompson 2005). This descending IC pathway is the route by which the IC influences the reflexive movement of the eyes, head, and pinna to a sound stimulus (Thompson 2005) . Although the SC is more related to visual orientation, it can also mediate these orienting movements to auditory stimuli (Syka and Radill-Weiss 1971) because a map of auditory space is thought to be encoded in this structure (Middlebrooks and Knudsen 1984). However, animals with large SC lesions can retain tonal conditioned eye blink and orientation to a sound source (Tunkl 1980), suggesting that the IC may influence auditory orienting behaviours independently of the SC.
We showed that NMDA but not diazepam microinjected into the IC prevented paradoxical kinesia during 50 kHz USV playback presentation (Tonelli et al. 2018b). Interestingly, the treatment did not interfere with the pinna reflex and head movement to the sound source in both cases, suggesting that the rats perceived the sound. These results also show that the route by which the IC influences the reflexive movement of the eyes, head, and pinna to a sound stimulus (Thompson 2005), i.e., the descending IC projections to deep layers of the SC, were not affected by NMDA or diazepam microinjected in the IC. Thus, the acousticomotor behaviour mediated by the IC may be more than only a mere reflexive orienting behaviour, reinforcing the assumption that the IC plays a more complex role in controlling orientation to sound modulating paradoxical kinesia.
In addition to the nigrotectal projections, a tecto-nigrostriatal projection has also been described (Castellan-Baldan et al. 2006). Auditory stimulation could activate this pathway resulting in increased striatal activity. This activation could explain paradoxical kinesia induced by acoustic stimulation since this phenomenon can be assigned to activating DA basal ganglia reserves in PD. However, this assumption needs more investigation since it is thought that paradoxical kinesia is not mediated by actions on striatal DA in rats (Keefe et al. 1989). We also demonstrated that intracollicular microinjections of NMDA and MK-801 did not interfere with the firing rate and striatal rhythmicity in anesthetized rats (Melo-Thomas et al. 2022). Accordingly, the facilitating effect of loud auditory tones is similar whether patients are off or on dopaminergic medication, suggesting a potentially nondopaminergic basis for the phenomenon (Naugle et al. 2012;Rogers et al. 2011). Although paradoxical kinesia has previously been attributed to behavioural energization through the release of "DA reserves" by intense stimuli (de la Fuente-Fernández and Stoessl 2002), any independence of the phenomenon from the dopaminergic state would heighten the relevance of the underlying neural systems as a potential novel target for therapeutic manipulation in PD.

The auditory thalamic relay
To explain how the IC modulates paradoxical kinesia, one should also consider the ascending projections from the IC to the medial geniculate nucleus (MGN), the main auditory thalamic relay. The IC projects to the thalamus and has abundant motor and premotor targets (Carr and Code 2000;Covey and Casseday 1995). Besides developing selectivity for biologically relevant sounds, the IC likely transforms a high rate of auditory input into a slower rate of output that matches the speed of motor performance (Casseday and Covey 1996).
These projections are excitatory and inhibitory (Peruzzi et al. 1997) and represent the primary afferent input to the MGN (LeDoux et al. 1987), which receives, modifies, and then transfers sensory information to specific regions of the neocortex. The auditory cortex (Wenstrup 2005) receives the major outputs from the MGN. However, the premotor cortex is another significant target (Cappe et al. 2009;Wenstrup 2005), suggesting that both auditory and motor representations of information may be transmitted by the MGN (Becker et al. 2002). In turn, the primary auditory cortex can generate multiple representations of an acoustic stimulus concurrently to encode different aspects of the sound (Zotova et al. 2000), for instance, the motor consequences of the acoustic signal when used as a CS for conditioning.
Contributing to the assumption that auditory-motor integration takes place at cortical levels, anatomical studies have shown that the output of the auditory cortex reaches broad areas of the prefrontal cortex (Jones and Powell 1973;Van Eden et al. 1992). The IC may contribute to this cortical auditory-motor integration since electrical stimulation of the IC produces a long-lasting increase in DA levels in the frontal cortex (Cuadra et al. 2000). Therefore, the IC-MGN projections may represent an alternative pathway that integrates precise information throughout the auditory system and influences motor output at thalamic and cortical levels (Tierney and Kraus 2013; Figure 5B).

The auditory-amygdalar feedback
The MGN, which receives major input from the IC, sends significant output to subcortical limbic forebrain structures like the amygdala (LeDoux et al. 1985), reinforcing the assumption that the role of the MGN may be more than auditory . The thalamo-amygdaloid pathway is unique to the auditory system (LeDoux et al. 1985) and is essential for associative learning based on auditory cues (LeDoux et al. 1986). In addition, a substantial direct projection from the basal nucleus of the amygdala was found, with terminals distributed widely throughout the IC, including most of the central nucleus, i.e., the major recipient of ascending auditory brainstem input (Marsh et al. 2002; see Figure 5B). These projections are excitatory (McDonald 1996) and may provide direct excitatory effects and inhibitory effects, mediated through collicular interneurons (Marsh et al. 2002). In turn, the IC receives projections from both the amygdala and auditory cortex, representing auditory-amygdalar feedback (Marsh et al. 2002), a dynamic network crucial for the emotional perception of sound. This network may be recruited when an emotional and motivational state is involved, for instance, the cataleptic state with an additional 50 kHz USV playback (see section "3.1"). Indeed, the presence of an auditoryamygdalar feedback circuit involving the IC may modify sound processing early in the ascending auditory pathway based on an animal's emotional or motivational state. In a recent study we showed that glutamatergic, but not GABAergic, intracollicular neurotransmission affected the motor response (catalepsy) to a highly emotional and motivational auditory stimulus (Tonelli et al. 2018b). We can speculate that NMDA, a glutamatergic agonist, microinjected into the IC impaired the auditory-amygdalar feedback by preventing the effect of 50 kHz USV playback to induce paradoxical kinesia. The auditory-amygdalar feedback may modulate the descending projection from the IC to the MLR to ensure an appropriate motor response. This modulation is particularly important since emotions play an important role in inducing paradoxical kinesia in humans (see section "2.2"). As our studies showed, the above anatomical connections may help explain how the intracollicular pharmacological manipulation induces paradoxical kinesia.

A new perspective on paradoxical kinesia
It is well known that as the brain evolved from primitive vertebrates to amphibians to mammals, new areas were added on top of the existing brain, but old ones were retained. New levels of complexity were created but basic abilities as old circuitries were preserved and remained functional. For instance, the neocortex must use circuits in both the brainstem and the spinal cord to perform independent movements of the arms and fingers, such as in tying a shoelace. Regarding locomotion, it involves periodic or "rhythmic" movements. Of note, rhythmicity characterizes some of the most basic, evolutionarily conserved types of movements (Moore et al. 2014). Following this line of reasoning it seems rational to think that locomotion as a very basic (old) ability may be preserved in old circuitries. Under given circumstances, when properly stimulated, these old locomotor circuits may be able to take over (producing paradoxical kinesia) regardless of basal ganglia defect as it happens in PD. Interestingly, these old circuitries can be activated by auditory stimulation regardless of newer circuits. Pioneering experiments performed in decerebrate cats by Sherrington and colleagues reported reactions such as retraction of the pinna, turning of the head, lashing of the tail, flexion, and extension movements of the limbs in response to acoustic stimulation (Forbes and Sherrington 1914). Spinal reflexes can be elicited in decerebrated cats in response to acoustic stimuli, which disappeared after IC bilateral ablation (Adams 1983;Chu 1970;Gernandt and Ades 1964;Wright and Barnes 1972). These classical experiments can help to explain paradoxical kinesia induced by auditory stimulation in PD patients and show that effective use can be made of these old (alternative) motor circuits when the "more recent" motor system fails. Thus, there is no real paradox in paradoxical kinesia if we consider that there is more than one motor system (Duysens and Nonnekes 2021). External stimuli may lead to paradoxical kinesia by "energizing" some of these old (alternative) motor systems. Moreover, it may also be that paradoxical kinesia is not a hallmark of Parkinson's disease but a general property of the motor system (Ballanger et al. 2006).
The ephemeral feature of paradoxical kinesia is a limitation for those who intend to use it as a rehabilitation tool and improve patients' quality of life in the long term. However, the underlying mechanisms of paradoxical kinesia are beginning to be revealed. Uncovering how these alternative motor pathways work and how to manipulate them in the long term may benefit patients with PD and other motor disorders. The two behavioural animal models of paradoxical kinesia (DBS in the IC and 50 kHz USV playback presentation) established by our group overcame a major limitation to understanding the neural mechanisms involved in this phenomenon, i.e., the lack of an animal model that replicates key aspects of paradoxical kinesia. New experiments in our lab using optogenetic techniques associated with both animal models will provide a deeper analysis of the mechanisms involved in the paradoxical kinesia phenomenon.
Author contribution: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission. Research funding: This work was supported by Deutsche Forschungsgemeinschaft Grants ME4197/2 and ME4197/3 (L. M.-T.). We would like to thank Theresa Kisko for comments on the English language styling. Conflict of interest statement: The authors have no financial interests to disclose.