In the present study, we examined multi-muscle and multi-joint NWR modulation during the transition from quiet stance to walking.
The main results can be summarised as follows: i) the NWR is phase-dependent. The excitability of the reflex (in terms of probability rate and size) is increased in the hip and knee flexor muscles of the starting leg during the first subphase (Rs), i.e. just prior to the occurrence of any movement, and in the ipsilateral knee flexor muscles as soon as the leg is unloaded (Us); ii) the NWR is hip joint kinematics-dependent in a crossed manner, i.e. the excitability of the reflex is enhanced in extensor muscles of the standing leg during hip flexion of the contralateral leg, whereas it is enhanced in the hip flexors of the standing leg during hip extension of the contralateral leg.
NWR modulation related to subphases
The NWR probability rate and size showed significant subphase-related modulation in the BF and RF muscles of the starting leg and in the RF muscle of the standing leg. Furthermore, significant differences in the reflex probability rate were found between the homonymous muscles of the two legs in almost all the subphases.
During the first gait initiation subphase (RSs), bodyweight was transiently shifted towards the starting leg as shown by the early COP displacement (Figure 1). The movement of the COP towards the starting leg has been suggested to be an anticipatory mechanism serving to stabilise the body before the postural perturbation . Interestingly, the reflex probability rate and size in the RF muscle were higher in this subphase than in the others, while no significant differences were found in the same muscle of the standing leg (Figures 4 and 5). Furthermore, the reflex probability rate was significantly higher in the BF muscle of the starting leg than in that of the standing leg in the RSs. These results suggest that the descending pathways conveying the commands for the motor programming of gait initiation activate the spinal circuitries mediating the NWR by enhancing the withdrawal reflexes in the hip flexors (taking into account the bi-articular function of the RF muscle), as well in the knee flexors of the starting leg (Figure 6). This activation, which may serve to prepare and assist the leg in the first step, takes place just before any movement actually occurs, even though the COP shifts towards the starting foot, suggesting that the intentional descending commands interact with the spinal circuitry in accordance with the role that the CNS assigns to each leg (starting or standing) at the very beginning of gait.
A main finding of our study was that the NWR was clearly subphase-dependent in the BF muscle of the starting leg (Figures 4
5 and 6), the reflex values being higher in the unloading subphase (Us) compared to the other subphases. In the Us, the COP moved towards the standing leg (Figure 1), thus quickly unloading the starting leg during this double support configuration. This clearly indicates that the sudden unloading of the leg is the main event bringing about the increase of the withdrawal reflex in the knee flexor muscles. It is known that in lower mammals the response to loading is related to locomotor circuits in the spinal cord [41, 42]. There is evidence that polysynaptic cutaneous reflexes (both painful and non-painful) are load-dependent. Indeed, the excitability of cutaneous reflexes is modified in symmetrical stance, compared to seated or prone postures [43, 44], as well as in asymmetrical stance in the loaded compared to the unloaded leg . A rich vein of research has revealed a phase-dependence (swing versus stance) of noxious and non-noxious cutaneous reflexes aimed at achieving leg withdrawal or maintenance of balance in the context of the gait cycle (for review see 15 and 45). This modulation of the reflex may reflect, at least in part, the effects of leg loading-unloading on spinal cord neuron excitability during the gait cycle.
The reversal of behaviour, in terms of NWR probability rate, seen in the knee flexor (BF) compared to the knee extensor (VM) muscles of the two legs during the DSs (Figures 4 and 6), when the leg is reloaded, further strengthens the idea that the shift of bodyweight from one leg to the other is the main factor in bringing about the increase/decrease in reflex activity at knee joint level. This finding is also in line with previous reports, in humans, of a facilitation of the cutaneous reflexes in knee extensors (i.e. the VL) [45, 46] during the swing-to-stance phase, a finding which suggests that this is a safety mechanism serving to increase knee stiffness and reduce the possibility of leg collapse .
Surprisingly, no significant modulation of the reflex between subphases was found in the ankle muscles. Furthermore, the reflex probability rate in the TA muscle of the starting leg was low throughout the gait initiation task (Figure 3). Ankle muscles are known to be highly responsive to loading-unloading conditions [47, 48]. One possible explanation for this could be that these muscles are under wide supraspinal control during early gait initiation. Indeed, there is growing evidence that the corticospinal control over distal muscles might be stronger than in proximal muscles during walking [49, 50]. This descending control might be even more pronounced during gait initiation. In particular, the EMG activity is typically suppressed in the SOL muscle and enhanced in the TA muscle during gait initiation . This suggests that voluntary recruitment of motoneurons prevails over the reflex response in these muscles, most likely in order to avoid mechanical perturbations and instability. Indeed, McIlroy et al. (1999) and Bent et al. (2001) [52, 53] showed that, during the anticipatory postural adjustments (equal to the early part of gait initiation), the CNS is able to anticipate the shift of the COP in the mediolateral direction and to delay the occurrence of withdrawal reflexes in order to preserve balance.
Modulation related to joint kinematics
The present study provides evidence that, irrespective of subphase-related modulation, dynamic joint motion regulates reflex variation among muscles in both legs.
Hip joint flexion of the starting leg led to an increase in the reflex probability rate and/or size in the Gmax, RF and SOL muscles of the standing leg (hip, knee and ankle extensors, respectively). This finding is interpretable as a reflex-mediated multi-joint extensor synergy serving to preserve the balance of the standing leg. Conversely, hip joint extension of the starting leg led to an increase in the reflex probability rate and/or size in the RF muscle of the standing leg (hip flexor) and in the RF and VM muscles (knee extensors) of the starting leg. The first of these findings may be interpreted as an enhanced flexion response serving to prepare for and support the first step of the contralateral leg (at the end of the stance phase), by helping to initiate the swing phase; while the latter finding may constitute an enhanced extensor response helping the ipsilateral leg (at the end of the swing) to accept the bodyweight at early stance.
Hip joint flexion of the standing leg induced increased reflex activity only in the RF muscle (knee extensor) of the starting leg, probably serving to support the knee extensors when the starting leg makes contact with the ground.
From these results, the RF muscle certainly seems to be extensively modulated by both bodyweight shift and joint motion in accordance with its bi-articular function (hip flexor or knee extensor). This bi-articular function of the RF muscle is typically observed during steady-state walking .
The limited reflex modulation induced by hip joint motion of the standing leg may reflect the limited movement performed by the standing leg during the gait initiation task (Figure 2). These results suggest that hip proprioceptive inputs are important signals for controlling locomotor activities once the legs start moving, as revealed during fictive locomotion in spinal animals [54, 55].
Different afferent fibres, called flexor reflex afferents, may evoke a flexion reflex in both humans and animals [56–58]. Included in this group of afferents are cutaneous low-threshold mechanoreceptors, cutaneous nociceptive afferents, group II, III and IV muscle afferents, and joint afferents. In the present study, modulation of the NWR in the hip and knee muscles was clearly related to bodyweight shift and hip joint motion. Although a convergence from many afferent sources may be hypothesised (including inputs from labyrinth and neck receptors), group II ankle muscle afferents, cutaneous afferents conveying inputs from the plantar foot mechanoreceptors [59, 60], and hip joint afferents  may all play key roles in modulating the NWR during gait initiation. In particular, group II afferents from ankle muscles have been demonstrated to play an important role in the control of bipedal stance and gait . It has been shown that stimulation of the gastrocnemius medialis and tibialis anterior nerves evokes a group II-mediated EMG facilitation in thigh muscles, with enhanced responses while leaning forwards or backwards . During gait initiation, when the body leans forwards, the ankle muscles contract while stretched, and thus it is likely that there is strong discharge from group II fibres which could facilitate the NWR in the proximal muscles. Furthermore, sensory information coming from plantar cutaneous afferents appears to play an important role in regulating stepping during human gait, facilitating control of compensatory stepping reactions. In particular, mechanoreceptors responding to pressure on the sole of the foot may be involved in sensing and controlling heel contact and subsequent weight transfer during termination of forward steps, and in maintaining stability during the prolonged swing phase of lateral crossover steps .
In our study, the afferents from the hip joints seem to be determinant in modulating the NWR during gait initiation. Our findings are in line with previous studies documenting that hip position entrains the activity of flexors and extensors during fictive locomotion in spinal animals [54, 55]. In spinal cord injured humans, hip flexion and extension movements either suppress or enhance the excitability of the flexor reflex pathways [64–67], suggesting that the inputs from the hip region are fully integrated by interneuronal circuits associated with motor control.
The flexion withdrawal reflex in mammals is believed to incorporate interneuronal circuits that contain elements of the stepping generator (i.e. the CPG) [68, 69]. However, in humans, despite the finding of stepping in anencephalic infants  and of alternating flexor-extensor bursts in patients with spinal cord lesions [71, 72], the existence of a spinal CPG has been hypothesised on the basis of extrapolations from simpler, animal models [11, 73].
Our findings suggest that in gait initiation NWR modulation and, possibly, recruitment of the CPG in the starting leg follows a well-ordered sequence: descending inputs, leg unloading, hip joint motion. Thus, after selective and asymmetrical excitation of the spinal substrate mediating the NWR (hip and knee flexor muscles of the starting leg) by descending motor commands, leg unloading and hip joint motion work in concert to produce an alternating (right and left) and crossed (flexors and extensors) activation mainly of the hip and knee joint muscles, predisposing the legs to the cyclical pattern of steady-state walking.
Herein, we speculate that CPG activation does not occur statically, in advance of the joint motion and torque changes, but rather that it emerges dynamically through the movement and unloading themselves. Conceptually, the CPG could be viewed as a “dynamic centre” that “arises” from load perturbation and movement. From this perspective, it can be suggested that one role of the descending commands is to initiate an asymmetrical (left-right) and unbalanced (flexor-extensor) activation of the spinal cord system.
Such a relationship between loading and hip joint motion possibly occurs early in the development of gait. Indeed, in human babies  and lower mammals [75, 76], the duration of the stance phase and its associated extensor muscle activity has been found to depend on both the position of the hip joint and the load borne by the standing leg.