Neural Control of Walking in People with Parkinsonism

Abstract

People with Parkinson's disease exhibit debilitating gait impairments, including gait slowness, increased step variability, and poor postural control. A widespread supraspinal locomotor network including the cortex, cerebellum, basal ganglia, and brain stem contributes to the control of human locomotion, and altered activity of these structures underlies gait dysfunction due to Parkinson's disease.

The ability to walk is severely impaired in people with Parkinson's disease (PD), and these impairments are associated with reduced quality of life, frequent falls, and complications from falls such as increased morbidity and mortality (6, 96). Recently, a widespread supraspinal locomotor network has been described, including premotor cortical, motor cortical, basal ganglia, cerebellar, and brain stem structures (66, 137). PD impairs structure and function in all of these locomotor regions, and these pathological and compensatory changes contribute to Parkinsonian gait, characterized by its slowness, variability, and poor postural control. In this review, we first discuss the basis for these three primary Parkinsonian gait impairments. Then, we discuss how altered structure and function of supraspinal locomotor regions contribute to gait impairments in people with PD.

Gait Impairments in People With PD

Gait impairments associated with PD can be broadly separated into continuous and transient disturbances (45). Continuous gait control problems have been recently characterized using principle component analysis as consisting of five independent domains: pace, rhythm, variability, asymmetry, and postural control (FIGURE 1B; Ref. 41). For the purposes of this review, we have consolidated these domains to three primary gait impairments: 1) gait slowness (pace, rhythm), 2) increased variability and asymmetry, and 3) poor postural control. These impairments have been shown to be relatively independent (22, 41), and we postulate each depends on partially distinct neural networks. Transient gait disturbances include occasional, context-specific festination (rapid, short steps) and freezing of gait (either akinesia or trembling of the legs) (126). In the current review, we will not discuss these transient disturbances but will refer to several recent reviews (34, 46, 102, 103).

FIGURE 1.

Continuous gait disturbances

A: continuous gait disturbances in people with PD. B: people with PD exhibit dysfunction in gait speed (pace/rhythm), variability and asymmetry, and postural control. This is depicted by a satellite plot showing deviations from control subjects (dotted line). SV, step velocity; SL, step length; Swi, swing time; ST, Step time; Sta, Stance time; Wid, Step width; sd, standard deviation (gait variability); as, asymmetry. *Differences between the control and PD group. Figure reproduced from Ref. 41 with permission from Movement Disorders.

Slowness of Gait

Gait slowness is common in people with PD (93) and can be observed throughout the course of the disease. The primary causes of gait slowness are 1) hypokinesia (reduced step size)/bradykinesia (increased step duration) and 2) rigidity/hypertonia.

Hypokinesia and bradykinesia of gait.

Parkinsonian gait is both hypokinetic (e.g., small movements such as reduced step amplitude and arm swing) and bradykinetic (e.g., slow movements such as slower step and arm swing velocity) (22, 41). However, reduced size of steps may be a more consistent contributor to slowed gait than slower steps (93). Similar to voluntary, upper extremity movement, most aspects of step kinematics and kinetics (e.g., joint angles, ground reaction forces, arm swing, etc.) are diminished in PD (92, 93). These findings suggest that, in the absence of transient disturbances such as freezing, the primary pathology of PD gait is inadequate scaling of motor output rather than the coordination of locomotor patterns (27, 28, 87). One notable exception is the increase in co-contraction of agonists and antagonists in the shank during walking (30). Hypokinesia and bradykinesia of gait is also apparent in the small and slow arm swing and lack of axial trunk rotation in people with PD, even after controlling for gait speed (156). Despite hypokinetic arm swing, coordination of arm and leg motion is largely intact in people with PD (29).

Rigidity/hypertonia of gait.

Rigidity is a specific type of hypertonicity observed in people with PD. Both axial and limb rigidity are observed in people with PD (84, 148), and both likely contribute to gait slowness. Axial tone can be measured using a device that slowly rotates different portions of the body, e.g., hips, trunk, and neck, during quiet stance (48). The resistance to these slow rotations provides a measure of passive resistance to external movements while an individual is actively maintaining stance posture. This technique shows hip, trunk, and neck tone to be elevated 30–50% in people with PD compared with age-matched control subjects (148). Axial rigidity contributes to gait slowness as hip rigidity interferes with hip hyperextension, a primary factor in step length and gait speed (3). Furthermore, excessive neck tone is related to functional mobility, such as ability to curve gait into a figure eight or roll over (40). Finally, changing gait direction with a turn is particularly slow in people with PD, even when gait speed is normal (156). Increased trunk stiffness and resistance to twisting may contribute to the more en block turning style of people with PD (55, 61) and the reduced turning speed.

People with PD also exhibit abnormal tone of the limbs, including the knees and ankles. This increased tone contributes to the flexed postural alignment of people with PD as well as gait slowness. Excessive flexor muscle actively pulls the hips, knees, and ankles into flexion, resulting in flexed spinal abnormalities, stooped posture, and reduced lower limb joint torques (58, 62). This abnormal posture also pushes the center of mass forward over the feet (125) and contributes to short, shuffling steps common in this population. In addition, excessive tonic muscle activity in ankle flexors and extensors increases co-contraction and joint stiffness (58). The elevated tone may be compensatory as it slows the speed and extent of body center of mass displacements and stabilizes joints. However, increased tone and phasic antagonist co-contraction reduces the net joint torques for control of gait and posture. Specifically, reduction in net ankle plantarflexion torque, a primary propulsive force during gait, results in smaller steps and slower gait velocity (71). Another factor that may play a role in the reduced ankle plantar flexion torque, and thus reduced gait velocity, in people with PD is a proximal distribution of joint torques from the ankle to the hip (130).

Increased Variability and Asymmetry of Gait

Gait variability includes “stride-to-stride fluctuations in walking” (53), and is consistently elevated in people with PD (41, 52). Variability in the medio-lateral and anterior posterior planes may have distinct sources. Collins and Kuo (20) demonstrated that step width variability, which occurs in the medio-lateral plane, is related to active step-to-step adjustment by the central nervous system to maintain balance during gait. This suggests that elevated lateral step variability observed in people with PD may partially reflect impaired control of lateral postural equilibrium. In contrast to medio-lateral variability, anterior-posterior variability of steps is closely related to fluctuations in self-selected walking speed. That is, anterior-posterior variability co-varies with gait speed and, in particular, the slow fluctuations in gait speed that are likely unrelated to step-to-step adjustments (20). Variability of steps is larger in people with PD in both the anterior-posterior and medio-lateral directions (52). Interestingly, increased variability of steps and reduced step length seem to be somewhat distinct phenomenon, since increased step variability appears before reduced step length, and step length is improved (increased) with levodopa, whereas variability is levodopa resistant (4, 5). These results suggest the possibility of different underlying causes for these changes in parkinsonian gait.

Excessive temporal and spatial left-right asymmetry of stepping parameters has also been consistently observed in people with PD (41, 109). For example, step length (119) and step time (41, 109) have been shown to be more asymmetric in people with PD than in healthy older adults. This likely relates to the asymmetric onset of bradykinesia and rigidity in lower limbs as well as in upper limbs. Temporal (156) and spatial (82) asymmetry of the arms during walking also occurs in people with PD. In fact, one of the earliest signs of abnormal gait in PD is asymmetric arm swing amplitude (156). Even postural control asymmetry, measured as the velocity of center of pressure movement under the left and right feet during quiet stance, has been shown to be elevated in people with PD (44). Furthermore, gait asymmetry, along with variability, has been suggested to underlie more serious gait impairments, such as freezing of gait and falls (110). Thus asymmetry represents an important and independent gait impairment in people with PD.

Poor Postural Control

Postural control involves maintaining, achieving, and restoring a state of balance during movements and posture (111). Indeed, all three components of postural control are impaired in people with PD and affect their gait.

Maintaining balance during standing or walking activities requires precise control of the head-arms-trunk (HAT) segments (145). During standing, people with PD have increased area, velocity, and jerkiness of postural sway and reduced limits of stability, defined as the maximum center of mass displacement possible without changing one's base of support. Limit of stability is especially impaired in people with PD in the backward direction (57) and is observed even in very early PD, before taking any antiparkinsonian medication (59, 86). In fact, the characteristic flexed, inflexible postural alignment of people with PD results in forward position of the body center of mass, possibly to protect against backward falling (125). Maintenance of balance during standing may worsen when on levodopa, possibly due to levodopa-induced dyskinesia (18, 22). Lateral control of balance is particularly affected in people with PD, as lateral trunk sway is especially elevated during quiet stance (1, 85) and while walking with and without obstacles (42, 134). While some amount of lateral motion is necessary for stability and optimized energetics (79), too much lateral motion is likely detrimental and is related to increased falls (120).

Achieving balance during voluntary movements such as gait initiation or fast, upper extremity movements is also impaired in people with PD. For example, anticipatory postural adjustments (APAs), defined as involuntary displacement of the center of pressure in preparation for voluntary movements, are bradykinetic in people with PD (86). A common example of an APA is the lateral and posterior movement of the center of pressure toward the swing foot before initiating a step from quiet stance. This lateral movement of the center of pressure is necessary to lift the stepping foot. The slow, small APAs observed in people with PD have been linked to delayed step initiation (76) and reduced step width, since smaller step widths permit smaller lateral weight shifts during locomotion. In fact, people with PD have difficulty scaling up the size of their lateral APAs before a step when their step width increases, which can result in inability to take a step (akinetic freezing) (116).

Small APAs may be due parkinsonian motor bradykineisa since slow movements require smaller APAs. Alternatively, APA dysfunction may be related to poor temporal coupling of posture (e.g., trunk lean) and gait (i.e., stepping). The coupling of posture and gait during walking or gait initiation is highly complex and requires precise control of the center of mass trajectory (83, 89, 146, 147). Unsurprisingly, people with PD have difficulty with posture-gait coupling (for review, see Ref. 88), and this has been suggested to contribute to gait challenges including delayed step-time and freezing of gait (65). Interestingly, APAs can be increased to normal levels when a step is initiated in response to an external cue or when on levodopa (15). Recently, it has been shown that rehabilitative interventions to assist lateral displacement of the APA may be beneficial for improving steps in people with PD (88).

Restoring balance after a slip, trip, or other external perturbation is also dysfunctional in people with PD. For example, people with PD exhibit smaller steps with larger subsequent displacements of the center of mass than healthy adults in response to large surface translations that trigger automatic compensatory stepping responses (23, 76, 131). Similarly, people with PD exhibit hypokinetic postural responses with reduced rate of rise of reactive torque in response to smaller perturbations (58, 60). The size of stepping responses to external perturbations, however, can be increased when people with PD can see their legs as they step toward a visual target, suggesting a deficit in proprioception (evaluation of limb position) and/or kinesthesia (evaluation of limb movement) (63).

The ability to quickly adapt postural responses based on task and environmental context is also altered in people with PD (16, 60). Chong and colleagues showed that, if perturbation direction was unexpectedly altered, healthy adults immediately adapted their postural responses by changing postural muscle activation patterns. People with PD, however, took several trials before exhibiting the appropriate postural response to the new perturbations. This represents a relative “inflexibility” to alter postural responses to match changes in task conditions or contexts (24). However, despite potentially slowed rates of adaptation and learning, people with PD can eventually adapt gait and stepping patterns with repetition (11, 56, 90, 118).

Summary of Gait Impairments

Slowed gait, increased variability, and poor postural control are the primary gait impairments in people with PD. While some gait impairments in people with PD are related to primary pathophysiology (such as bradykinesia), others may be compensatory in nature. For example, reduced step length and walking speed can be protective against falls as more time is spent with both feet on the ground at slower gait speeds. Similarly, while increased rigidity contributes to slowed gait (81, 99), it may also represent a compensatory mechanism to increase joint stiffness and joint stability. Age-related impairments in mobility (e.g., reduced muscle strength, reduced proprioception) also reduce the ability to control the trunk or increase walking speed. These effects are further compounded by physical inactivity and sarcopenia that are especially pronounced in people with PD (for review, see Ref. 33). Therefore, while we focus our review on the possible neural underpinnings of each gait impairment, it is important to note that compensatory mechanisms and secondary, age-related musculoskeletal impairments also contribute to gait impairments in people with PD, in addition to neurodegeneration.

Neural Underpinnings of Gait Dysfunction in People with PD

A definitive diagnosis of PD includes death of dopaminergic cells within the substantia nigra pars compacta that project to the striatum (i.e., caudate and putamen) of the basal ganglia. Often overlooked, however, is that numerous other brain regions also show altered structure and function in people with PD. For instance, areas that release acetylcholine (ACh), such as the pedunculopontine nucleus (PPN) and nucleus basalis of Meynert (nbM), and noradrenaline, such as the locus coeruleus, exhibit alpha-synuclein deposition very early in the disease (14). Cortical motor and nonmotor structures show considerable deposition later in the course of the disease. Cerebellar function is also altered in PD (149). Indeed, every critical node in the central locomotor network (66) likely plays a role in PD gait dysfunction. Table 1 summarizes studies investigating the neural activity during gait or gait-like tasks, showing reduced or increased activity in a number of brain regions (or nodes) within the locomotor/postural network in people with PD. In the following section, we will discuss how these changes in supraspinal activity may contribute to the gait dysfunction described above: gait slowness, variability/asymmetry, and impaired postural control.

Table 1.

Summary of recent imagining studies showing increased or decreased activity of supraspinal locomotor nodes during actual or imagined locomotion in people with Parkinson's disease

Reduction of Activity in PD	Reference(s)	Increase in Activity in PD	Reference(s)
Motor
Supplementary motor area (L)	5
Dorsal premotor (R)	142
Primary motor cortex	142
Cortical
Precuneus (L)	21, 51	Temporal cortex (R)	51
Precuneus (R)	21, 51, 142	Insula (R)	51
Parieto-occipital region	21, 142	Superior frontal gyrus (L)	51
Posterior hippocampus (L)	21	Cingulate cortex (L)	51
Superior parietal lobule (R)	21, 132
Anterior cingulate cortex (R)	21, 132
Anterior cingulate cortex (L)	21
Inferior parietal lobule (R)	142
Occipital	142
Lingual gyri	21
Basal ganglia
Globus pallidus (R)	105
Putamen (R,L)	104
Brain stem
PPN/MLR	21
Cerebellum
Cerebral hemisphere (L)	21, 51	Cerebellar locomotor region	51
Cerebellar locomotor region	21

L, left; R, right; PPN, pedunculopontine nucleus; MLR, mesencephalic locomotor region.

Neural Underpinnings of Slowness in PD Gait

Hypokinesia and bradykinesia.

Hypokinesia and bradykinesia contribute to slowed gait and may be explained partially by the long-standing “rate-model” of basal ganglia pathology in people with PD (2, 26, 43). Although this model has been adapted over time (25), it remains a commonly used model of motor dysfunction in PD and supports a critical role of the basal ganglia in scaling self-initiated movement.

FIGURE 2 depicts the “rate model” of basal ganglia pathophysiology in people with PD. This model suggests that neural degeneration of the substantia nigra pars compacta results in increased inhibition of the globus pallidus external segment and reduced inhibition of the globus pallidus internal segment. Together, this leads to overexcitation of the globus pallidus internal segment and thus more inhibition of the thalamus and PPN. Interestingly, recent research suggests that increased inhibitory output from the pons [an area affected early in the course of PD (14)] may further exacerbate these changes by suppressing substantia nigra output (32).

FIGURE 2.

Rate model of basal ganglia dysfunction

Rate model of basal ganglia dysfunction in normal (A) and parkinsonian (B) states. Over activity of the indirect and underactivity of the direct pathways result in more inhibitory output from the basal ganglia output structures (GPi) to the thalamus and the brain stem, and ultimately reduced amplitude of movements, including gait. Green arrows represent excitatory and red arrows inhibitory projections. Arrow thickness represents the relative firing rate of projections, and dashed arrows indicate the relative reduction of the SNpc D1 and D2 dopaminergic projections to the striatum. SNpc, substantia nigra pars reticulate; GPe, globus pallidus external segment; GPi, globas pallidus internal segment; STN, subthalamic nucleus.

These alterations in basal ganglia output result in over-inhibition of the ventral anterior/ventral lateral thalamus and reduced excitation of cortical motor structures such as the supplementary motor area (SMA) and primary motor cortex (72, 135) (FIGURE 3). Given the importance of these motor and premotor regions for movement scaling and planning (97), the reduced excitation of the SMA could lead to hypokinetic gait, APAs, and automatic postural responses (49, 136, 139). This hypothesis is supported by the fact that gait and APA hypokinesia is improved by levodopa in the early stages of the disease. Indeed, we recently showed that hypokinesia (e.g., reduced step length, gait velocity, and arm swing) was the gait impairment most improved by levodopa. These results are shown in FIGURE 4, which depicts the effects of levodopa on different gait domains (e.g., pace, arm and trunk movement, dynamic stability, etc.) in 104 individuals with idiopathic PD. Interestingly, other measures of gait dysfunction, including gait timing and postural sway, were unaffected or even worsened, respectively, by levodopa (22).

FIGURE 3.

Framework for supraspinal control of locomotion in people with PD

Alterations in activity of the basal ganglia (1) and brain stem (4) contribute to gait slowness and increased postural instability, respectively, and increased cerebellar activity may partially compensate for these alterations (2). Increased volitional control (i.e., cortico-spinal) and reduced automatic control (3) may contribute to increased gait variability and asymmetry. See text box above for more information. PPN, pedunculopontine nucleus; MLR, mesencelphalic locomotor region; PMRF, pontomedulary reticular formation; SMA, supplementary motor area.

FIGURE 4.

The dysfunction in speed, variability and asymmetry, and postural control in PD are differently affected by levodopa

Rightward (positive) values represent a positive impact of levodopa, whereas leftward (negative) values represent a negative impact of levodopa. Gait speed and size of limb movements are improved by levodopa (top). Postural sway can be worsened by levodopa (bottom). Although few variability or asymmetry variables were available in this particular analysis, asymmetry of arm swing and temporal coordination, such as double support time, were not consistently improved by levodopa. A value larger than 0.20 represents small, 0.50 moderate, and 0.80 large responsiveness. *Differences between the control and PD group. Figure reproduced from Ref. 22 with permission from Movement Disorders.

The notion that altered basal ganglia-thalamo-cortical function is related to gait slowness is supported by recent neuroimaging studies (Table 1). For example, Hanakawa and colleagues (51) showed that SMA activity was reduced in people with PD with respect to healthy adults during actual walking. More recent studies have shown that the amount of activation in the SMA during imagined walking was correlated to overground gait function in people with PD (21, 105). In addition, basal ganglia activity (105) and the availability of dopamine in the nigrostriatal system (104) are reduced in people with PD during locomotion with respect to healthy adults.

Dysfunction of locomotor network regions outside of the basal ganglia-thalamo-cortical system, including areas related to cholinergic activity, have also been implicated in gait slowness in people with PD. ACh in the brain comes from three primary sources: cholinergic interneurons in the striatum; the nbM, which is the primary source of ACh to the cortex and basal forebrain; and the PPN, which supplies cholinergic input to the thalamus and spinal cord (for review, see Ref. 154). Recent work suggests that the availability of cortical ACh, supplied by the nbM, may be specifically related to gait speed. For example, among a cohort of 125 people with PD, gait speed was not affected in people with PD with nigrostriatal denervation and normal cholinergic denervation. In contrast, gait speed was significantly reduced in individuals with both nigrostriatal and cholinergic denervation. This suggests that gait speed may be related to either cholinergic dysfunction (alone) or in conjunction with dopaminergic dysfunction (7). Furthermore, Rochester et al. (117) measured cortical cholinergic availability using transcranial magnetic stimulation, showing correlations between gait speed and step length with cortical cholinergic function. These findings suggest that impaired cortical ACh dysfunction (supplied primarily by the nbM) together with dopamine dysfunction may lead to reduced gait speed. Indeed, although the substantia nigra is commonly noted as the source of motor dysfunction in PD, the nbM is among the first structures to exhibit structural pathology in the course of PD (14). In addition to the nbM, the PPN may also play a role in hypokinesia, since PPN lesions in primates cause arm and leg hypokinesia (77, 95). Interestingly, The PPN may also play a role in akinesia, or the lack of movement. Jenkinson and colleagues (69) showed that levodopa in combination with PPN stimulation improved the mean number of movements per hour in a parkinsonian primate to normal levels, again suggesting that both dopaminergic and non-dopaminergic dysfunction are critical for movement production.

In contrast to reduced activity of cortical and brain stem areas, several investigations have shown increased activity in the cerebellum, possibly compensating for altered basal ganglia function. This compensatory hypothesis has been supported by several recent investigations (for review, see Ref. 149). A single proton emission computed tomography study during locomotion showed concomitant increased activity in midline cerebellum and decreased activity of the SMA in people with PD (51). In a follow-up study, Hanakawa and colleagues showed that, by using transverse lines on the floor (visual cues), gait in people with PD was improved and the cerebellum and lateral premotor cortex were especially active. The premotor cortex, which is partially regulated by the cerebellum, is related to externally cued movement (122). In contrast, the SMA, which receives considerable input from the basal ganglia, may preferentially contribute to internally driven movement (97). Therefore, in people with PD, the cerebellum may regulate gait using external cues, compensating for poorer internally generated movement via the basal ganglia and SMA (50). More recently, Festini and colleagues demonstrated that when off anti-Parkinson medication (levodopa), people with PD exhibit increased levels of cerebellar-whole brain and cerebellar-cerebellar connectivity. Furthermore, increased cerebellar activity was, in some cases, correlated to improved motor and cognitive performance in people with PD (35). The cerebellum plays a critical role in the generation (112) and timing (31) of movements, particularly during locomotion (67, 91), and is densely connected with the locomotor brain stem areas (such as PPN) (68), the basal ganglia (12, 13), and the cortex. Furthermore, this region has recently been shown to exhibit relatively high presynaptic cholinergic terminals, underscoring its importance within the cholinergic system (107). Thus, although the mechanism is not fully understood, increased activity of the cerebellum in PD may be an attempt to compensate for reduced activity of other structures, including the SMA. This may, however, result in overdependence on external cues (8, 51, 149) (FIGURE 3).

Hypertonicity/rigidity.

Despite their congruent role in gait slowness, hypokinesia/bradykinesia and rigidity may have somewhat distinct neural underpinnings (70). While hypo- and bradykinesia are related primarily to dysfunction of the cortico-thalamo-basal ganglia loop (see previous section), rigidity may be related to dysfunction of the interaction between the basal ganglia and deep brain structures, including the cholinergic region of the PPN (FIGURE 3). Converging animal and human experiments suggest that the PPN and MLR within the brain stem play critical roles for controlling axial postural tone. Specifically, the MLR, consisting primarily of the cuneiform nucleus, contributes to the excitatory muscle tone and rhythm-generating system (137). In animals, tonic activity of this region can elicit gait-like flexion and extension of hindlimbs. Conversely, the PPN inhibits extensor and flexor alpha motoneurons (138). Both the MLR and PPN receive considerable efferent, inhibitory signals from the globus pallidus internal segment (GPi) (137), and, as noted above, people with PD exhibit excessive inhibitory output from the GPi. Thus the elevated inhibition of the MLR in people with PD likely reduces the ability to initiate and maintain locomotion. In contrast, inhibition of the inhibitory system (i.e., the PPN) may result in hypertonus and rigidity (140). However, axial and limb rigidity are somewhat distinct due to the different pathways from dorsolateral (axial) and ventromedial (limbs) descending spinal systems (80). In fact, although levodopa is very effective in reducing limb rigidity (58), it does not reduce axial rigidity (40, 148). Thus primary dysfunction of the PPN, either prior to or secondary to nigrostriatal dysfunction, may contribute to increased axial tone.

Neural Underpinnings of Variability and Asymmetry in PD Gait

Gait variability and asymmetry in people with PD has been well characterized; however, the neural underpinnings of this impairment are not well understood. One hypothesis suggests that people with PD have difficulty controlling automatic movements and thus shift to more voluntary control of gait (8). This compensatory shift from automatic gait to voluntary stepping may result in more gait variability (19, 153) (FIGURE 3). Indeed, considerable research has shown that conscious control of typically overlearned tasks reduces performance and increases variability (153). Automatic tasks seem to rely more on subcortical structures, including the basal ganglia and brain stem, whereas voluntary tasks rely more on cortical activity and require more attention (19, 49, 140, 151, 152). People with PD exhibit larger than normal cortical activity during upper extremity motor tasks, both when the task is new and after overlearning has occurred (150, 152), suggesting more voluntary control of tasks in this population. Thus even highly overlearned tasks, such as walking, may rely more heavily on cortical structures in people with PD. This shift toward increased voluntary locomotor control may partially compensate for dysfunction of the basal ganglia and brain stem automatic pathways but may also increase variability.

Dual task paradigms provide further support of PD impairments in automatic control of gait leading to increased variability while walking (74). When completing gait with a secondary cognitive task, healthy adults typically demonstrate decrements in both the cognitive task (e.g., counting backward by 3s) and gait. These decrements in performance are noted as “dual task cost.” Because cognitive tasks are primarily supported by frontal structures and require attention, dual task cost suggests that the primary motor task (e.g., walking) also requires a level of attention and voluntary control. Interestingly, people with PD have more pronounced dual task cost during walking than age-matched adults (74). This increased dual task cost in people with PD suggests that gait requires more attentional, voluntary control than in healthy adults, resulting in increased variability (106, 108).

Gait asymmetry may be related directly to asymmetric neural dysfunction within the basal ganglia. Dopaminergic neuron loss is often asymmetric, and the side with greater dopamine loss corresponds to the opposite side with greater motor signs in humans (75, 78, 114, 141) and in animals (39). Similarly, asymmetry of rigidity and bradykinesia have been shown to relate to asymmetric limb movements during gait, asymmetric turning impairments, and asymmetric dopaminergic degeneration (144). Finally, a recent report suggests increased levodopa use may improve asymmetry of gait (41).

Neural Underpinnings of Poor Postural Control in PD Gait

Considerable research is pointing toward an important role of non-dopaminergic structures such as the PPN in postural dysfunction of people with PD (FIGURE 3). For example, trunk sway and variability in the mediolateral direction, common measures of postural control, are excessive in people with PD but do not correlate well with other PD signs, such as the UPDRS (85), and are not consistently improved by levodopa (113, 115). In fact, lateral sway during stance may be increased after taking levodopa; possibly due to reduced rigidity, with no concomitant improvement in postural control (116). Muller and colleagues recently demonstrated that thalamic cholinergic innervation, i.e., the amount of cholinergic neurons projecting from the PPN to the thalamus, was directly correlated to postural sway (94). Importantly, no correlations were observed between postural sway and cortical ACh innervation. In contrast, Rochester and colleagues showed that gait speed, but not step-width variability (a measure of postural control), was related to cortical ACh activity (117). Therefore, ACh in the thalamus, supplied primarily by the PPN, seems to be related to postural control (e.g., postural sway and sway variability) (94), whereas cortical cholinergic function, supplied by the nbM, may be related to gait speed and hypokinesia (7, 36).

Restoring balance after an external perturbation relies on cortical, basal ganglia, and brain stem structures (64). Like voluntary stepping, the hypokinetic steps in response to large, external perturbations may reflect altered basal ganglia-thalamo-cortical circuitry. However, levodopa does not improve compensatory stepping as it improves voluntary stepping, suggesting that brain regions outside of the basal ganglia, such as areas related to ACh, contribute to reactive postural control strategies (23, 76). Indeed, recent animal and human research suggests that automatic postural responses are stored in the brain stem (54, 100, 101, 133). Furthermore, ACh in the thalamus (provided primarily by the PPN) is correlated with falls in people with PD, whereas dopaminergic function is not (9). In addition, medication that improves ACh availability (donepezil and rivostigmine) may reduce falls in people with PD (17). Together, these results suggest that non-dopaminergic regions, including the brain stem, likely play a critical role for postural responses and fall prevention.

Several other converging lines of evidence support the role of PPN and ACh in postural control. Karachi et al. (73) recently showed that PPN cholinergic lesions, both with and without concomitant nigrostriatal dopaminergic losses resulted in postural instability in rhesus monkeys. Furthermore, patients with progressive supranuclear palsy show severe balance impairments and postural instability, specifically with spontaneous backward falls. Progressive supranuclear palsy is an atypical parkinsonian syndrome marked by PPN and thalamic cholinergic dysfunction and cell loss but relatively spared cortical cholinergic innervation (47, 129). Falls, a prominent outcome of postural instability, are correlated to cholinergic innervation from the PPN to the thalamus in people with PD. Bohnen and colleagues (9, 10) showed that fallers had reduced thalamic cholinergic function with respect to nonfallers, with no similar loss in nigrostriatal dopaminergic denervation. Furthermore, a recent imaging study showed that people with PD exhibited altered PPN activity during imagined walking with respect to healthy controls, providing a direct link between PPN activity and gait dysfunction in people with PD (21). Deep brain structures with dense connections to the PPN, including the pontomedulary reticular formation, have also been linked to generation of APAs and coupling of posture and gait (123). For example, recent electrophysiological studies in cats have shown that the pontomedulary reticular formation contributes to coupling of anticipatory postural adjustments and limb movements (123, 124). While the neural underpinnings of poor postural and locomotor coupling in PD is not well understood, people with PD exhibit early pathology in deep brain structures such as the pontomedulary reticular formation (14). Therefore, it is possible that this pathology contributes to the poor coupling of posture and locomotion.

Interestingly, deep brain stimulation in the subthalamic nucleus (STN) (98, 115) or GPi (115) has been shown to improve some measures of postural sway in PD, including mediolateral sway velocity. Given the intractability of postural sway to levodopa, these finding suggests that deep brain stimulation of the STN and GPi may influence non-dopaminergic pathways. Indeed, considerable bidirectional pathways between the STN and deep brain structures, such as the PPN, exist, and impaired signaling from the STN to the PPN has been suggested to be central to some postural control dysfunction in people with PD (128). In fact, a recent report by Weiss and colleagues demonstrated that STN deep brain stimulation may improve walking via direct activation of the PPN (143), further underscoring the importance of both the STN and PPN in postural control in people with PD.

Together, these reports underscore the role of structures both within and outside the basal ganglia, including the PPN, thalamus, and cerebellum, in gait dysfunction in PD. Specifically, the deep brain structures and their dense connections with the cortex, cerebellum, STN, and spinal cord contribute to gait dysfunction observed in people with PD (37, 38).

Conclusions and Future Directions

Gait impairments in people with PD include pronounced slowness, variability, and postural instability. Considerable work is unraveling the pathophysiology underlying these characteristics, demonstrating likely contributions from a number of supraspinal regions. Specifically, slowness of gait seems to be related to dysfunction of the basal ganglia-thalamo-cortical loop in conjunction with reduced cortical ACh, as well as rigidity. Gait variability may be related to a shift from automatic to voluntary control of locomotion in people with PD. Finally, postural instability seems to be related to alterations in ACh in brain stem structures such as the PPN and MLR. In addition to these changes, increased activation in the cerebellum and cortex may partially compensate for impaired activity in other brain regions but may result in overdependence on external cues.

Additional work is necessary to continue to elucidate mechanisms for gait dysfunction in people with PD. Continued investigation of neural activity during actual and imagined locomotion (for review, see Ref. 8) as well as structural and functional neural connectivity will provide further insight into how the brain is differently active in people with PD who have specific parkinsonian impairments such as freezing of gait, dyskinesia, and dystonia. Finally, new studies are focusing on the role of cognitive impairments in gait dysfunction (127). Impairments of gait and cognition may share neural underpinnings in the prefrontal cortex. For example, availability of cortical ACh may underlie attention and executive cognitive dysfunction in people with PD (154), so new trials are examining the effects of cholinesterase inhibitors at improving both cognitive and gait function in people with PD (121). Indeed, each of the gait characteristics discussed above (slowness, variability, and postural control) have been directly linked to cognitive declines, perhaps because of shared prefrontal circuitry (for review, see Ref. 155). Further work is necessary to understand how cognitive function contributes to control of locomotion and gait impairments in PD.