Ankle muscle tenotomy does not alter ankle flexor muscle recruitment bias during locomotor-related repetitive limb movement in late-stage chick embryos

Abstract

In ovo, late-stage chick embryos repetitively step spontaneously, a locomotor-related behavior also identified as repetitive limb movement (RLM). During RLMs, there is a flexor bias in recruitment and drive of leg muscle activity. The flexor biased activity occurs as embryos assume an extremely flexed posture in a spatially restrictive environment 2–3 days before hatching. We hypothesized that muscle afferent feedback under normal mechanical constraint is a significant input to the flexor bias observed during RLMs on embryonic day (E) 20. To test this hypothesis, muscle afference was altered either by performing a tenotomy of ankle muscles or removing the shell wall restricting leg movement at E20. Results indicated that neither ankle muscle tenotomy nor unilateral release of limb constraint by shell removal altered parameters indicative of flexor bias. We conclude that ankle muscle afference is not essential to ankle flexor bias characteristic of RLMs under normal postural conditions at E20.

1 |. INTRODUCTION

Many animals assume flexed postures during embryogenesis, as evidenced by reconstructions of 3D and 4D ultrasound recordings of humans, cats, dogs, dolphins, elephants, and penguins (Abbas, 2006; Macdonald, 2005; Townend & Kern, 2009). Human fetuses in utero assume flexion in the arms by 13 weeks, and in multiple limb joints by 20–28 weeks (Fong, Buis, Savelsbergh, & de Vries, 2005; Jirásek, 2001; Larsen, Sherman, Potter, & Scott, 2001). Flexion posture predominates by 36–37 weeks, with the hands contacting the face, and legs contacting the trunk (Casaer, 1979;Fong et al., 2005;Vles, Kingma, Caberg, Daniels, & Casaer, 1989). Semiflexed postures, presumed residual effects of prenatal constraints, may persist 1–10 days after full term birth (Beintema, 1968; Casaer, 1979; Dubowitz, Dubowitz, & Mercuri, 1999). Rats assume a semiflexed posture in the elbows by embryonic day (E) 16, and in the legs between E16-E20 (Moessinger, 1983; Narayanan, Fox, & Hamburger, 1971). Flexed postures also have been documented in fetal kittens at 32–36 days (Knospe, 2002), fetal sheep by 43 days (Barcroft & Barron, 1939) and crocodiles at E53 (Vieira et al., 2011). Chick embryos assume an extremely flexed posture by E18 (Hamburger & Oppenheim, 1967). The body is folded over onto itself and legs are deeply flexed, with the toes contacting the head (Figure 1c1).

FIGURE 1.

Embryonic postures, kinematic preparations and EMG burst detection. (a) For tenotomy experiments, embryos were placed prone and egg shell was removed to access both legs. White kinematic markers approximated the location of left and right ankles from a posterior view. (b) Muscle bursts were identified in the rectified EMG using previously established criteria: baseline signal amplitude, amplitude threshold, burst duration, and interburst interval detected apriori by a computer program (see text for details). In this example, tibialis anterior (TA) EMG activity during an RLM is plotted, and four bursts detected by the program are identified by four upward deflections from the threshold plot line. Burst duration was equal to the time between burst onset and offset (b-a₁). Interburst interval was equal to the time between consecutive bursts (a₂-b). Cycle duration was equal to the time between consecutive onsets (a₂-a₁). (c) For experiments examining unilateral release of limb constraint in ovo, E20 embryos were positioned sidelying. (c1) Egg shell overlying the right leg was removed to prepare and record leg activity during typical in ovo posture. In this posture, the legs were tightly flexed, the head was rotated to the right, as indicated by the upward orientation of the beak, and the toes of the right foot contacted the back of the head. (c2) When the shell was removed anterior to the foot (foot-free), the right leg extended outside the egg during RLMs

Despite the prevalence of flexed prenatal postures, it remains unclear to what extent they are attributed to neural, mechanical and/or genetic mechanisms. Early work determined that pharmacologic blockade of muscle activity produced fused joints, suggesting that movement contributed to joint morphological development (Drachman & Coulombre, 1962). Studies of spontaneous motility concluded that somatosensation and experience during embryogenesis had little impact on motor development (Haverkamp & Oppenheim, 1986; Haverkamp, 1986; Landmesser & O’Donovan, 1984; Oppenheim, 1972; Ryu & Bradley, 2009). In chicks, neither deafferentation nor spinal transection during embryogenesis appeared to alter motility, though embryos needed assistance to hatch (Hamburger & Balaban, 1963; Hamburger & Oppenheim, 1967; Hamburger, Wenger, & Oppenheim, 1966; Landmesser & O’Donovan, 1984; Oppenheim, 1975). Thus, it appears the neural underpinnings of postures acquired during embryogenesis have not been addressed. It also appears no studies have sought to determine if constraints imposed by the prenatal environment, or muscle and joint mechanics significantly contribute to the flexed postures of fetuses under normal conditions. However, clinical studies have suggested that uterine restriction during breech positions may induce joint flexion (Fong et al., 2005), and oligohydramnios may increase spinal flexion (Albuquerque et al., 2002) or reduce movement amplitude (Sival, Visser, & Prechtl, 1990) that could limit the range of possible postures. Genetic mechanisms may also bias early posture, given their contribution to spinal curvature in familial idiopathic scoliosis (Miller, 2011). Knowledge of mechanisms that induce flexed postures is important because it can inform our understanding of the mechanisms that impact motor control development when posture and movement are altered by atypical events, such as premature birth.

Deeply flexed and spatially restrained during the last 3 days before hatching, chick embryos produce spontaneous repetitive limb movements (RLMs) in ovo accompanied by locomotor-related muscle activity. During RLMs, flexor, and extensor antagonist muscles are alternately active, as during the swing and stance phases of gait. Flexors in one leg are alternately active with flexors in the other leg, as during walking. Further, leg muscles are repetitively recruited at frequencies observed in posthatching locomotion (Bradley, Ryu, & Lin, 2008; Ryu & Bradley, 2009; Sindhurakar & Bradley, 2012). One feature that distinguishes RLM EMG activity from posthatching locomotion is a bias in recruitment and drive of flexor muscles (Sun & Bradley, 2017). Flexor muscles are more consistently recruited than the antagonist extensors during repetitive bursting. At least three sources of neural drive may underlie the flexor bias: descending pathways, spinal networks and/or sensory inputs. For example, reticulospinal pathways initiate locomotor activity in chick embryos (Valenzuela, Hasan, & Steeves, 1990). Isolated spinal locomotor networks exhibit flexor-biased organization in the neonatal mouse (Dougherty et al., 2013; Endo & Kiehn, 2008; Machado, Pnevmatikakis, Paninski, Jessell, & Miri, 2015). Further, imposed changes in muscle length can alter RLM EMG in chicks 1 to 3 days prior to hatching (Bradley, Ryu, & Yeseta, 2014). Though isolated spinal circuits produce a flexor bias during fictive locomotion, it appears that the mechanisms for flexor bias have not been examined in a neurologically intact embryo or fetus under normal late-stage prenatal conditions that precede postnatal locomotor behavior.

In neurologically intact chick embryos, sensory input can alter RLM muscle activity by E20. For example, restraint-induced stretch of ankle flexors increased RLM flexor burst amplitude (Bradley et al., 2014). During restraint, flexor RLM bursts were enhanced, but EMG was quiescent between bursts and consecutive RLMs, suggesting descending fusimotor drive may have contributed to the phasic response. During locomotion in adult cats, fusimotor drive produces cyclic activity in dorsal root recordings of muscle spindle afferents, coding muscle length changes and/or task demands (Loeb & Duysens, 1979; Taylor, Durbaba, Ellaway, & Rawlinson, 2000). Interestingly, repetitive hatching behavior also emerges in the final 3 days in ovo (Hamburger & Oppenheim, 1967). It is triggered by proprioceptive input associated with neck flexion, and is characterized by an EMG pattern that differs from RLMs, one of prolonged bursts in both leg extensor and flexor muscles (Bekoff & Kauer, 1984; Bekoff, Nusbaum, Sabichi, & Clifford, 1987). Thus, constraints on posture and movement in ovo might impose a different proprioceptive trigger that enhances flexor bias during RLMs (Sun & Bradley, 2017).

Muscle spindles play a dominant role in reflex modulation of movement (Mileusnic, Brown, Lan, & Loeb, 2006). Thus, in this study, we asked if muscle afferent feedback under normal mechanical constraints is a significant input underlying the flexor bias observed during RLMs at E20. Afferents from extensor muscles may be important because extensors appear to achieve maximum physiological length due to extreme flexion of the hip, knee, and ankle. Muscle afferent physiology has not been studied in chicks, however in cats, sustained stretch of ankle extensors activate group II afferents (Matthews, 1972; Nelson & Hutton, 1985). Extensor II afferents participate in the flexor reflex afferent circuit, inhibiting homonymous extensor motor pools, and exciting antagonist flexor motor pools (Jankowska, 1992; Lundberg, Malmgren, & Schomburg, 1987; Matthews, 1972). Extensor 1a afferents may also be active during homonymous muscle activity, however, extensor recruitment is relatively infrequent during most RLMs (Bradley et al., 2008; Sun & Bradley, 2017), and extensor 1a imposes only weak reciprocal inhibition on antagonist flexor motor neurons (Nichols, 1989). Alternatively, flexor 1a afferents may enhance flexor bias owing to the extremely small joint excursions during RLMs (Bradley et al., 2014), in that limited movement during vigorous flexor activity can impose isometric contractions that reinforce flexor 1a afferent activity (Edin & Vallbo, 1990; Hagbarth, Wallin, & Lofstedt, 1975).Also, high burst frequencies of many RLMs (4–10 Hz)may produce inertial lags in joint excursions (Hagbarth et al., 1975), activating flexor 1a afferents and exciting homonymous motor neurons while strongly inhibiting extensors (Edin & Vallbo, 1990; Hagbarth et al., 1975; Nichols, 1989). Thus, we reasoned posture- and movement-related feedback in ovo may trigger a net excitatory input to flexor motor pools and inhibitory input to extensor motor pools. The other major source of proprioceptive input, 1b afferents from Golgi tendon organs, monitors tension produced by muscle contraction. In the limb loading phase of locomotion, extensor 1b afferents excite motor neurons of the homonymous muscle (Pearson, Misiaszek, & Fouad, 1998). However, 1b actions vary markedly across muscles and tasks (Mileusnic & Loeb, 2006), often inhibiting motor neurons of the homonymous muscle (Pearson & Collins, 1993). Although Golgi tendon organs also appear to code contractile tension in birds, 1b circuits have not been studied (Maier, 1998; Maier & Mayne, 1990).

In this study, we hypothesized that spindle afference under normal mechanical constraint due to spatial restriction is a significant input producing RLM flexor bias at E20. We tested our hypothesis by performing ankle muscle tenotomies to reduce 1a and II afferent contributions to the recruitment of ankle muscles (Hyngstrom, Johnson, Miller, & Heckman, 2007; Vrbová, 1963; Yellin & Eldred, 1970). We assumed that tenotomy would have minimal impact on other somatosensory modalities during RLMs owing to the constancy of spatial restriction during movement in ovo pre to post tenotomy. Also, passive stretch pre-tenotomy or loss of stretch post-tenotomy would not likely activate 1b afferents (Jami, 1992), and brief phasic muscle recruitment during RLMs would be insufficient to significantly drive III and IV muscle afferents (Windhorst, 2007). We further tested our hypothesis by conducting a new analysis of published data in which shell restricting leg posture and movement was removed (Bradley et al., 2008; Ryu & Bradley, 2009). Our findings provide evidence that a bias in flexor recruitment and drive persists after ankle muscle tenotomy and after shell removal. We discuss how these findings further our understanding regarding the role of sensory experience due to environmental constraint during locomotor-related behavior just prior to hatching. Preliminary findings were previously published in abstract form (Sun & Bradley, 2016).

2 |. METHODS

2.1 |. Subjects

Fertile Leghorn chicken (Gallus gallus) eggs were incubated under standard conditions (37.5°C, 62% humidity) and exposed to light (426–2160 lux) 12 hr daily. At E20, eggs were maintained in heated and humidified chambers for both surgical preparation and behavioral recording. Embryos were euthanized (Euthasol^®, 240 mg/kg) at the end of the recording. All procedures were approved by the University Animal Care and Use Committee.

2.2 |. Preparations for behavioral recording

Embryos were prepared for electromyography (EMG) and synchronized video recording of spontaneous leg movements in ovo. Procedures are briefly summarized, as detailed descriptions have been published (Sindhurakar & Bradley, 2012; Sun & Bradley, 2017). An analgesic (Buprenex^®, 0.01 mg/kg) was given prior to EMG implantation, and had no apparent impact on behavior following implantation, for embryos were typically active and often initiated RLMs upon placement in the recording chamber. Bipolar fine wire electrodes were inserted in the left and right ankle flexor (tibialis anterior, TA) and extensor antagonist (lateral gastrocnemius, LG). Also, to track 2-dimensional (2D) leg movement, a marker was fabricated from a minutien pin (#26002–10, Fine Science Tools, Foster City, CA) and inserted into the posterior surface of each ankle (Figure 1a). Accuracy of all EMG implantations was determined after euthanizing the embryo, by applying a cauterizing current through the electrodes and confirming both tips were seated within the muscle.

2.3 |. Tenotomy

During the course of the experiment, recording was temporarily suspended and the left LG or TA tendon was cut (e.g., tenotomy) to reduce muscle length and tension, thereby reducing proprioceptive afference generated during spontaneous leg movements (Hyngstrom et al., 2007; Vrbová, 1963; Yellin & Eldred, 1970). An anesthetic (Marcaine^®, 0.25% BW) was given before performing the tenotomy. A small surgical hook was then inserted deep to the LG or TA tendon to lift and stabilize the tendon as it was cut near the ankle using a scalpel. Bleeding was minimal. Total time to complete the tenotomy and resume recording was typically 10–15 min. Completeness of the tenotomy was verified by dissection after euthanizing the embryo. Neither Marcaine^® nor tenotomy had an apparent impact on neural functions associated with the quality or frequency of repetitive limb behavior following tenotomy (Figures 2b and 3b), and embryos remained active for the 4–6 hr of recording without any apparent change in behavior.

FIGURE 2.

RLM recruitment trends pre and post LG tenotomy. (a) RLM EMG activity is plotted for the left tibialis anterior (TA) and its antagonist, lateral gastrocnemius (LG), as well as the contralateral TA (TA-co) and LG (LG-co), prior to tenotomy of the LG. These RLM sequences exemplify the three common features of RLM muscle activity (see Results for details): (1) the TA was the first muscle recruited during RLM sequences; (2) a TA burst was detected in every cycle of the RLM; and (3) TA bursts outnumbered LG bursts. Vertical excursions in the kinematic traces for both the left ankle and contralateral ankle (Ankle, Ankle-co) indicated leg movements were generated bilaterally during the repetitive bursting. In this RLM, the average (±SD) integrated burst amplitudes were: TA 3.8 ± 1.1 mV · s, LG 2.6 ± 1.3 mV · s, TA-co 12.8 ± 6.3 mV · s, and LG-co 2.1 ± 1.1 mV · s. (b) EMG is plotted for an RLM following tenotomy of the LG (LGpost-t) during the same experiment as in A. Note that the three RLM features were still apparent in EMG traces after LG tenotomy. Further, vertical excursions in kinematic traces post LG tenotomy appeared similar to excursions pre tenotomy. Average integrated burst amplitudes also remained similar post tenotomy: TA 5.6 ± 2.2 mV · s, LG 2.1 ± 0.5 mV · s, TA-co 14.5 ± 5.6 mV · s and LG-co 3.2 ± 1.2 mV · s. C. RLM burst counts, normalized by recording duration and averaged across embryos (±SD) are plotted for each of the four muscles pre and post LG tenotomy. (c1). Burst counts did not differ from pre to post LG tenotomy for TA (filled circles pre and post tenotomy connected by a line) or LG (unfilled circles, same format). (c2). Also, burst counts for concurrent RLMs in the contralateral ankle did not differ from pre to post tenotomy for either TA-co (filled circles) or LG-co (unfilled circles). See text for details. (d) Burst counts for TA (filled circles) and LG (unfilled circles) are plotted for two experiments that included a subsequent tenotomy of the contralateral LG (Post LG-co tenotomy). TA burst count exceeded LG count in each phase of both experiments. Solid lines between symbols identify the LG and TA pair for one experiment; dashed lines identify the muscle pair for the other experiment

FIGURE 3.

RLM recruitment trends pre and post TA tenotomy. (a) RLM EMG activity is plotted for the left TA and LG, and contralateral ankle muscles (TA-co, LG-co) prior to TA tenotomy. As in Figure 2, the bias in TA vs LG recruitment is apparent, for example, a TA burst was detected every cycle producing a greater burst count compared to LG. In this RLM sequence, average (±SD) integrated amplitudes were: TA 2.6 ± 1.5 mV · s, LG 1.8 ± 0.7 mV · s, TA-co 1.8 ± 1.6 mV · s, and LG-co 7.6 ± 5.3 mV · s. (b) EMG activity is plotted for an RLM following TA tenotomy (TApost-t) during the same experiment as in A. The bias favoring TA recruitment was still apparent after tenotomy. In general, integrated burst amplitudes post tenotomy were similar to pre tenotomy averages, except for LG-co: TA 2.6 ± 0.6 mV · s, LG 1.9 ± 1.1 mV · s, TA-co 1.6 ± 0.7 mV · s, and LG-co 2.9 ± 2.7 mV · s. C. RLM burst counts, normalized by recording duration and averaged across embryos are plotted for each muscle pre and post TA tenotomy. (c1) Burst counts did not differ from pre to post TA tenotomy for TA (filled circles) or LG (unfilled circles). (c2) Burst counts for concurrent RLMs in the contralateral ankle did not differ from pre to post TA tenotomy for either TA-co (filled circles) or LG-co (unfilled circles). See text for details. (d) Burst counts for TA (filled circles) and LG (unfilled circles) are plotted for two experiments that included a subsequent tenotomy of the contralateral TA (Post TA-co tenotomy). TA burst count exceeded LG count in each phase of both experiments. Solid lines identify the LG and TA for one experiment; dashed lines identify the muscle pair for the other experiment

2.4 |. Recording procedures

EMG and video captured 4–6 hr of spontaneous leg movements in ovo while embryos were positioned prone within the recording chamber. At least 2 hr of spontaneous activity was recorded before the left TA or LG was tenotomized. An additional 2 hr or more of activity was recorded following the tenotomy. Given potential afferent contributions from ankle muscles contralaterally, the homologous LG or TA of the contralateral (right) ankle was then tenotomized (e.g., LG-co or TA-co tenotomy) in four embryos, and an additional 2 hr of activity was recorded. During data collection, EMG signals were analog filtered (band-pass filter 100–1,000 Hz) and amplified by 2,000 (Astro-Med Grass model P511 K, Natus Medical Inc., Pleasanton, CA). EMG signals were sampled at 4,000 Hz and stored to computer (Datapac 2K2, Run Technologies). The posterior view of both ankles was recorded at 99 fps (StreamPix^®, NorPix, Montreal, Quebec, Canada) from a video camera (Scout-f, Pylon, Basler, Ahrensburg, Germany) placed above the recording chamber. An event pulse was manually generated at 5 min intervals, activating an LED in the visual field and a voltage signal in the EMG recording to synchronize EMG and video data during data processing.

2.5 |. Identification of RLM sequences

EMG burst detection and verification of leg movements in the concurrent video were used to identify RLMs based on protocols established in previous RLM studies (Bradley et al., 2008; Ryu & Bradley, 2009; Sun & Bradley, 2017). In brief, EMG activity was determined to be an RLM if four or more bursts were rhythmically generated within a frequency range of 1–10 Hz in at least one of the ankle muscles, and the bursts were accompanied by displacements of the ankle marker (Bradley et al., 2008; Ryu & Bradley, 2009; Sindhurakar & Bradley, 2012). Burst sequences were identified from rectified EMGs after estimating baseline signal amplitude for each channel from an interval of quiescence at the beginning of the recording. A computer-automated program (Datapac 2K2, Run Technologies) identified burst sequences based on apriori criteria for three burst parameters: burst amplitude threshold, burst duration, and interburst interval (Figure 1b). Burst amplitude threshold was determined for each channel and the value applied to all analyses within channel across experimental conditions. Across experiments, thresholds ranged from 2× to 3× baseline channel noise. Burst duration captured envelopes of EMG activity remaining above threshold for 20–1,000 ms; and interburst interval identified EMG activity remaining below threshold for ≥ 20 ms. Together, the three parameters determined onset and offset of each burst (i.e., a₁ and b, Figure 1b).

Ankle markers in the concurrent video for a burst sequence were digitized to generate 2D kinematic coordinates and confirm leg movement was produced. Video sequences were de-interlaced and digitized at 198 pictures/s (Datapac 2K2, Run Technologies). The 2D coordinates were filtered (5th order Butterworth, cut-off 10 Hz) and vertical coordinates were plotted as a time series (Datapac 2K2, Run Technologies). Vertical oscillations (i.e., Ankle, Figure 2) confirmed the EMG sequence represented RLM behavior (Sindhurakar & Bradley, 2012).

2.6 |. EMG analyses

Four burst parameters quantified muscle recruitment and drive pre and post tenotomy: burst count, burst duration, cycle duration, and integrated burst amplitude. Burst count was normalized by time (bursts/hr) to control for variations in recording duration between conditions and across experiments. Burst duration and cycle duration are defined in Figure 1B. Integrated burst amplitude described the area of the rectified EMG signal between burst onset and offset, and was equal to the product of burst duration and amplitude. Integrated amplitude for each burst was normalized as a percentage of the maximum RLM integrated burst amplitude within muscle to control for differences in electrode sampling between muscles and across experiments.

We also examined the effect of tenotomy on flexor and extensor recruitment as a function of the presence or absence of LG bursts during RLMs. Under normal conditions in ovo, burst parameters for leg flexor and extensor antagonists vary with the presence or absence of extensor bursts (Sun & Bradley, 2017). Thus, based on established methods, RLMs were sorted into 1 of 3 LG recruitment patterns: no LG recruitment, inconsistent, and consistent LG recruitment (Sun & Bradley, 2017). No LG recruitment indicated that no LG bursts were detected. Inconsistent LG recruitment indicated a burst was detected during some, but not all cycles (e.g., LG, Figure 2a). Consistent LG recruitment indicated a burst was detected in every cycle.

The contribution of proprioception to the bias favoring flexor muscle activity during RLMs was tested by comparing the effects of tenotomy on TA and LG burst parameters using a within-subject design. RLMs typically vary from three to four cycles, but some exceed seven to nine cycles (Bradley et al., 2008); thus to avoid bias in selecting which RLMs were included in analyses and to ensure an adequate sample for all burst parameter means pre and post tenotomy, inclusion criteria were applied. Data for an experiment were retained if TA in the tenotomized (left) ankle produced ≥ 150 bursts during each phase (≥2 hr) of the experiment (pre and post TA or LG tenotomy). Burst parameters were averaged within embryo and averages within embryo were compared pre and post tenotomy. Shapiro-Wilk tests for normality indicated that across embryos, burst parameter averages were normally distributed, p > 0.05 (Ghasemi & Zahediasl, 2012). Two-way repeated measures ANOVAs were used to test for within-embryo differences in RLM burst parameters pre and post TA or LG tenotomy, and between TA and LG. Paired t-tests were used to test for differences in the proportion of TA-initiated RLMs within embryo before and after tenotomy. Significance level was set at 0.05, and a Bonferroni correction (p < 0.017) was applied to account for the interdependence of burst duration, cycle duration, and integrated burst amplitude calculations. Grand means and standard deviations (±SD) are reported.

2.7 |. Removal of shell wall constraint

In light of tenotomy results, we asked if afference during large amplitude movement alters flexor bias characteristic of spatially restricted RLMs in ovo. We applied analysis herein to RLMs recorded at E20 in a study that examined EMG and kinematics before and after removing shell anterior to the foot (Bradley et al., 2008; Ryu & Bradley, 2009). Data acquisition and processing were similar to methods for tenotomy experiments. However, embryos were positioned sidelying (Figure 1c) and only EMG and kinematics for the right leg were recorded. Parameters indicative of flexor bias during RLMs spatially constrained in ovo and unconstrained by shell removal were compared using the Wilcoxon signed rank test and Friedman test. The significance level (p < 0.05) and Bonferroni correction (p < 0.017) were applied.

3 |. RESULTS

We report results of RLM EMG analyses that tested if muscle afferent feedback under normal mechanical constraint due to spatial restriction is a significant input to circuits producing flexor bias in recruitment and drive of ankle muscle activity during RLMs in ovo at E20. Results summarize analyses for 10 LG and 10 TA tenotomy experiments. Tables 1 and 2 list RLM sample sizes for each experiment. Collectively, LG tenotomy analyses were based on 496 RLMs pre tenotomy and 637 post tenotomy. TA tenotomy analyses were based on 650 RLMs pre tenotomy and 765 post tenotomy. Analyses for contralateral ankle muscle activity are based on comparable samples. We first summarize effects of tenotomy on muscle recruitment and burst parameters. Tenotomy effects on neural drive modulating burst amplitudes are next examined. These analyses are supplemented by effects of a secondary tenotomy in the contralateral ankle. Effects of releasing the leg from spatial restriction imposed by the shell are also examined.

TABLE 1.

Total number of RLMs included in analyses pre and post LG tenotomy

Embryo ID	Pre	Post
Left ankle (Ankle)
1	41	25
2	33	55
3	50	44
4	50	88
5	32	112
6	39	100
7	44	69
8	93	45
9	40	33
10	74	66
Right ankle (Ankle-co)
1	35	28
2	38	53
3	68	61
4	37	115
5	31	82
6	38	96
7	36	63
8	79	66
9	49	52
10	77	80

TABLE 2.

Total number of RLMs included in analyses pre and post TA tenotomy

Embryo ID	Pre	Post
Left ankle (Ankle)
1	118	111
2	42	152
3	90	45
4	87	29
5	74	216
6	68	63
7	70	41
8	24	28
9	30	42
10	47	38
Right ankle (Ankle-co)
1	81	95
2	31	136
3	88	39
4	29	57
5	85	227
6	77	53
7	47	26
8	31	37
9	30	34
10	81	66

3.1 |. Effects of LG tenotomy on ankle muscle recruitment during RLMs

We first asked if muscle afferent feedback from LG under normal mechanical constraints in ovo is a significant input to the bias in recruitment and drive of flexors during RLMs by examining EMG activity pre and post LG tenotomy. Exemplary RLM sequences during one experiment are plotted in Figure 2. TA typically generated more extensive burst sequences than LG, both pre (Figure 2a) and post LG tenotomy (Figure 2b). TA was usually the first muscle recruited, and it produced a burst every cycle in most RLMs. LG was less consistently recruited, both pre and post LG tenotomy, dropping out for one or more cycles, or commencing to burst after one or more cycles of TA bursting. These trends were also observed during concurrent RLMs in the contralateral ankle (e.g., TA-co, LG-co) pre and post LG tenotomy (Figure 2).

Within-subject comparisons indicated that LG tenotomy did not alter the more prevalent recruitment of TA. Burst counts, normalized by time (bursts per hour) within embryo and averaged across embryos, are plotted pre and post LG tenotomy in Figure 2c. The two-way repeated measures ANOVA for LG tenotomy experiments indicated that TA burst counts (filled circles) exceeded LG counts (unfilled circles), F(1, 9) = 37.11, p < 0.001 (Figure 2c1). However, there was no difference in burst counts pre and post tenotomy, F(1, 9) = 0.71, p > 0.4, and no interaction, F(1, 9) = 0.81, p > 0.3. TA-co burst counts (filled circles, Figure 2c2) were also greater than LG-co counts (unfilled circles) for concurrent RLMs contralaterally and the difference was significant, F(1, 9) = 27.07, p < 0.002; yet burst counts did not differ pre to post LG tenotomy, F(1, 9) = 0.96, p > 0.3, and there was no interaction, F(1, 9) = 2.39, p > 0.1. Further, TA was the first muscle recruited in the majority of RLMs pre (85 ± 11% of RLMs) and post LG tenotomy (88 ± 7% of RLMs), and the trend did not differ pre to post tenotomy, t(9) = −1.42, p > 0.09, one-tailed. TA-co was the first recruited during concurrent RLMs contralaterally, pre (84 ± 8%) and post LG tenotomy (84 ± 9%), t(9) = −0.31, p > 0.3, one-tailed.

LG tenotomy did not alter the consistent recruitment of TA in most RLM sequences. Within embryo, the TA was recruited every cycle in 89 ± 8% of RLMs pre tenotomy and 86 ± 9% post. Whereas, LG was consistently recruited in only 37 ± 19% and 36 ± 16% of RLMs, pre and post tenotomy, respectively. A two-way repeated measures ANOVA for ankle muscles indicated that the difference in consistent recruitment within embryo was significant, F(1, 9) = 54.92, p < 0.001. However, LG tenotomy had no effect, F(1, 9) = 0.97, p > 0.3, and there was no interaction, F(1, 9) = 0.48, p > 0.5. TA-co and LG-co exhibited similar recruitment trends during concurrent contralateral RLMs. TA-co was recruited every cycle in 89 ± 7% of RLMs pre and post LG tenotomy, whereas LG-co was consistently recruited in only 40 ± 12% of RLMs pre and 42 ± 20% post tenotomy. The two-way repeated measures ANOVA indicated that consistent TA-co bursting exceeded LG-co bursting within embryo, F(1, 9) = 75.09, p < 0.001, that tenotomy had no effect, F(1, 9) = 0.35, p > 0.5, and there was no interaction, F(1, 9) = 0.07, p > 0.7.

The above LG tenotomy findings were contrary to our predictions. Thus, given locomotor-like alternating steps of left and right legs are expressed during RLMs by E20 (Ryu & Bradley, 2009; Sindhurakar & Bradley, 2012), we asked if muscle afference from the contralateral ankle compensated for the loss of ipsilateral proprioceptive afference. To explore possible contralateral compensation, LG-co was tenotomized in two embryos after completing data collection post LG tenotomy. Collectively, the embryos produced 47 RLMs after LG-co tenotomy. Results suggested that LG-co tenotomy did not alter the differences in TA and LG recruitment. TA burst count continued to exceed LG count (Figure 2d); TA bursts initiated the majority of RLMs in both embryos (76% and 96%), as did TA-co contralaterally (67%, 82%); and TA was still consistently recruited (96% of RLMs in both embryos).

3.2 |. Effects of TA tenotomy on ankle muscle recruitment during RLMs

Give the LG tenotomy results and previous evidence that TA is very responsive to stretch at E20 (Bradley et al., 2014), we asked if TA muscle afference, under normal mechanical constraints in ovo, is a significant input biasing flexor recruitment and drive during RLMs. Exemplary EMG sequences during one TA tenotomy experiment are shown in Figure 3. Comparison of sequences pre (Figure 3a) and post TA tenotomy (Figure 3b) suggested that flexor recruitment bias persisted; for example, TA produced relatively more bursts than LG, TA initiated most RLMs and typically produced bursts every cycle. Burst counts were normalized within embryo and averages across embryos are plotted pre and post TA tenotomy in Figure 3c. A two-way repeated measures ANOVA indicated the difference in TA and LG counts was significant, F(1, 9) = 22.55, p < 0.002. However, burst counts did not differ pre and post TA tenotomy, F(1, 9) = 0.50, p > 0.4, and there was no interaction, F(1, 9) = 0.03, p > 0.8. Similar results were obtained for TA-co and LG-co during concurrent contralateral RLMs (Figure 3c2). TA-co burst counts exceeded LG-co counts, F(1, 9) = 18.20, p < 0.003, TA tenotomy had no effect on burst counts, F (1, 9) = 1.70, p > 0.2, and there was no interaction, F(1, 9) = 4.61, p > 0.05. In two embryos, the TA-co was subsequently tenotomized. Analyses for 139 RLMs suggested the additional tenotomy did not alter the flexor bias in burst count (Figure 3d, post TA-co tenotomy).

TA tenotomy did not alter TA initiation of RLM burst sequences, or consistent recruitment of TA across RLM cycles. Normalized by RLM sample size within embryo, comparisons indicated that TA was the first muscle recruited pre (82 ± 9% of RLMs) and post TA tenotomy (78 ± 15%), and there was no difference between conditions, t(9) = 0.65, p > 0.2, one-tailed. During concurrent RLMs in the contralateral ankle, TA-co was the first muscle recruited pre (69 ± 16%) and post TA tenotomy (72 ± 11%), t (9) = −0.63, p > 0.2. Consistent recruitment of TA was again more prominent compared to LG. Within embryo, TA was recruited every cycle in most RLMs pre (92 ± 7% of RLMs) and post TA tenotomy (90 ± 7%), whereas consistent LG recruitment was far less common pre (28 ± 15%) and post tenotomy (31 ± 20%). Thus, two-way repeated measures ANOVA indicated that consistent TA recruitment was significantly greater compared to LG, F(1, 9) = 120.21, p < 0.001, the difference was not altered by TA tenotomy, F(1, 9) = 0.06, p > 0.8, and there was no interaction, F(1, 9) = 0.32, p > 0.5. These recruitment differences were mirrored in TA-co and LG-co EMG. TA-co was consistently recruited in most RLMs pre (86 ± 11% of RLMs) and post TA tenotomy (85 ± 14%), whereas consistent LG-co was less common pre (49 ± 17%)and post tenotomy (51 ± 18%).Further, these differences were significant, F(1, 9) = 38.98, p < 0.001, were not altered by TA tenotomy, F (1, 9) = 0.18, p > 0.6, and there was no interaction, F(1, 9) = 0.06, p > 0.8. TA-co tenotomy (two embryos) did not alter these trends; TA recruitment initiated most RLM sequences in both embryos (93% and 87%), and TA was still consistently recruited (96% and 90% of RLMs).

3.3 |. Effects of tenotomy on burst parameters

Tenotomy did not alter RLM burst durations, normalized integrated burst amplitudes or cycle durations. Two-way repeated measures ANOVA comparisons for burst parameters averaged within embryo pre and post tenotomy are detailed in Tables 3–5 (Bonferroni p < 0.017). Grand averages for burst duration ranged from 63 to 82 ms (Table 3). Normalized integrated amplitudes ranged from 11% to 17% of maximum integrated amplitude (Table 4). Average cycle durations ranged from 212 to 259 ms (Table 5). TA and LG burst parameters also appeared similar following LG-co or TA-co tenotomy (Supplementary material A and B).

TABLE 3.

Average within embryo burst durations (ms) for RLMs pre and post tenotomy (grand mean ± SD)

	LG tenotomy		TA tenotomy
	Pre	Post	Pre	Post
TA	78 ± 15	82 ± 12	78 ± 12	77 ± 14
LG	70 ± 10	78 ± 18	63 ± 15	67 ± 18

LG tenotomy: Pre, Post: F(1, 9) = 6.10, p > 0.03, TA, LG: F(1, 9) = 3.11, p > 0.1, Interaction: F(1, 9) = 0.30, p > 0.5.

TA tenotomy: Pre, Post: F(1, 9) = 0.93, p > 0.3 TA, LG: F(1, 9) = 5.29, p > 0.04 Interaction: F(1, 9) = 0.34, p > 0.5.

TABLE 5.

Average within embryo cycle durations (ms) for RLMs pre and post tenotomy (grand mean ± SD)

	LG tenotomy		TA tenotomy
	Pre	Post	Pre	Post
TA	227 ± 38	245 ± 101	256 ± 58	259 ± 66
LG	212 ± 26	252 ± 74	230 ± 59	230 ± 57

LG tenotomy: Pre, Post: F(1, 9) = 1.83, p > 0.2, TA, LG: F(1, 9) = 0.08, p > 0.7, Interaction: F(1, 9) = 4.63, p > 0.05.

TA tenotomy: Pre, Post: F(1, 9) = 0.03, p > 0.8 TA, LG: F(1, 9) = 6.79, p > 0.02 Interaction: F(1, 9) = 0.01, p > 0.9.

TABLE 4.

Average within embryo normalized integrated amplitudes (% of maximum) for RLMs pre and post tenotomy (grand mean ± SD)

	LG tenotomy		TA tenotomy
	Pre	Post	Pre	Post
TA	17 ± 5	14 ± 4	11 ± 3	11 ± 5
LG	14 ± 4	13 ± 6	14 ± 7	14 ± 5

LG tenotomy: Pre, Post: F(1, 9) = 3.98, p > 0.07, TA, LG: F(1, 9) = 0.83, p > 0.3, Interaction: F(1, 9) = 0.71, p > 0.4.

TA tenotomy: Pre, Post: F(1, 9) = 0.04, p > 0.8 TA, LG: F(1, 9) = 2.88, p > 0.1 Interaction: F(1, 9) = 0.29, p > 0.6.

3.4 |. Effects of tenotomy on burst amplitudes varying with LG recruitment

Both TA and LG integrated burst amplitudes co-varied with the consistency of LG recruitment and neither was altered by tenotomy (Figure 4). Two-way repeated measures ANOVA for LG tenotomy experiments indicated that within embryo, TA integrated amplitude was least when LG was not recruited, and greatest when LG was recruited in all cycles of an RLM, F(2, 8) = 6.45, p < 0.03 (Figure 4a); however, amplitude was not altered by tenotomy, F(1, 9) = 2.84, p > 0.1, and there was no interaction, F(2, 8) = 0.22, p > 0.8. LG burst amplitude was also least when LG was inconsistently recruited, and greatest when consistently recruited, F(1, 9) = 18.64, p < 0.003 (Figure 4b), yet was not altered by tenotomy, F(1, 9) = 0.66, p > 0.4, and there was no interaction, F(1, 9) = 0.05, p > 0.8. These trends also persisted after LG-co tenotomy (Supplementary material C).

FIGURE 4.

Scaling of TA and LG integrated burst amplitudes during RLMs pre and post LG tenotomy. Burst amplitudes varied as a function of LG recruitment (no LG recruitment, inconsistent recruitment, or consistent recruitment), but did not vary between LG tenotomy conditions (pre, post). Significant results for two-way repeated measures ANOVAs are indicated, see text for details. (a) TA amplitude was least when LG was not recruited and greatest when LG was consistently recruited, both pre and post LG tenotomy. TA integrated burst amplitude was similar pre and post tenotomy. (b) LG amplitude was greatest when it was consistently recruited, both pre and post LG tenotomy

Burst parameters for TA tenotomy experiments mirrored those above for LG tenotomy. A two-way repeated measures ANOVA indicated that TA integrated amplitude also varied across the three LG recruitment patterns (Table 6), F(2, 8) = 18.31, p < 0.002; was not altered by TA tenotomy, F(1, 9) = 0.22, p > 0.6; and there was no interaction, F(2, 8) = 1.18, p > 0.3. TA amplitude was least in the absence of LG recruitment and greatest when it was consistently recruited, both pre and post TA tenotomy. LG burst amplitude was also greatest when LG was consistently recruited (Table 6), F(1, 9) = 33.84, p < 0.001, but did not vary with tenotomy, F(1, 9) = 1.05, p > 0.3, and there was no interaction, F(1, 9) = 0.27, p > 0.6. Trends persisted following TA-co tenotomy (Supplementary material D).

TABLE 6.

Average within embryo normalized integrated amplitudes (%) as a function of LG recruitment for RLMs pre and post TA tenotomy (grand mean ± SD)

LG recruitment pattern	Pre	Post
TA
No LG recruitment	8 ± 3	8 ± 5
Inconsistent LG recruitment	10 ± 4	12 ± 6
Consistent LG recruitment	14 ± 5	14 ± 7
LG
Inconsistent LG recruitment	15 ± 9	11 ± 5
Consistent LG recruitment	22 ± 12	20 ± 10

3.5 |. Effects of unconstrained leg movements on muscle recruitment

Given the persistence of flexor bias following tenotomy, we speculated that any afferent input associated with mechanical constraint of leg movement in ovo might support the bias. For example, it is possible that other somatosensory inputs from body contact with the shell contributed to the bias in flexor activity. Thus, we asked if reducing mechanical constraint altered the bias in flexor recruitment and drive during RLMs. Relatively unrestrained RLMs following shell removal (compare Figure 1c1 and 1c2) likely augmented leg proprioceptive feedback due to greater linear and angular displacements at the ankle (Figures 5a1 and 5a2). Further, greater leg movements likely altered somatosensory feedback from body on body and skin to shell contact.

FIGURE 5.

Muscle recruitment during RLMs under control and foot-free conditions. (a) Two RLM sequences from a single experiment exemplify TA and LG activity during each condition. Upward deflections of the kinematic trace (Ankle) indicate ankle extension (e.g., plantar flexion). The postures typical of the two recording conditions are shown in Figure 1. (a1) This RLM was produced in ovo, for example, control conditions (Figure 1c1). (a2) This RLM was produced after shell restraint was removed, for example, foot free conditions (Figure 1c2). The greater ankle displacements after shell removal corresponded with significantly larger leg excursions that often extended the leg outside the egg (Ryu & Bradley, 2009). Nonetheless, patterns of consistent TA recruitment and inconsistent LG recruitment were similar between conditions. (b) Normalized burst counts for TA (b1) and LG (b2) during control and foot-free conditions are plotted. Each pair of symbols connected by a line represents data for one embryo (N = 5). TA and LG burst counts were similar between conditions. TA bursts outnumbered LG bursts in all embryos. Wilcoxon signed rank test results were non-significant (n.s.)

Data for five embryos from a study that examined coordination during restrained and unrestrained leg movement at E20 (Ryu & Bradley, 2009) met inclusion criteria for analyses. A total of 179 RLMs during normal restraint in ovo (control, Figure 1c1) and 175 RLMs after shell removal (foot-free, Figure 1c2) were analyzed. In this limited sample, we found no differences in either TA or LG recruitment within embryo between conditions despite significant differences in joint posture and excursion range (Ryu & Bradley, 2009). TA burst counts were similar between control and foot-free conditions in four of five embryos (Figure 5b1), Z = −0.41, p > 0.3 and LG burst counts were similar between conditions in all embryos (Figure 5b2), Z = −0.67, p > 0.2. Burst counts were greater for TA than LG in all experiments (compare Figures 5b1 and 5b2). TA was the first recruited in most control (87 ± 6%) and foot-free RLMs (84 ± 9%), Z = −0.74, p > 0.2. TA was consistently recruited in nearly all control (98 ± 3%) and foot-free RLMs (98 ± 2%), Z = 0.0, p > 0.4; whereas, LG was consistently recruited in only a small fraction of control (14 ± 11%) and foot-free RLMs (20 ± 9%), Z = −1.22, p > 0.1. Burst durations, normalized integrated amplitudes and cycle durations were similar between control and foot-free conditions within embryo (Tables 7, 8, and 9).

TABLE 7.

Average within embryo burst durations (ms) for RLMs during control and foot-free conditions (grand mean ± SD)

	Control	Foot-free	Z	p>
TA	68 ± 10	70 ± 8	0.54	0.3
LG	58 ± 7	62 ± 10	0.54	0.3

Z, p: one-tailed, Wilcoxon signed rank test.

TABLE 8.

Average within embryo normalized integrated amplitudes (% of maximum) for RLMs during control and foot-free conditions (grand mean ± SD)

	Control	Foot-free	Z	p>
TA	19 ± 3	19 ± 3	0.0	0.4
LG	19 ± 4	16 ± 3	1.62	0.06

Z, p: one-tailed, Wilcoxon signed rank test.

TABLE 9.

Average within embryo cycle durations (ms) for RLMs during control and foot-free conditions (grand mean ± SD)

	Control	Foot-free	Z	p>
TA	205 ± 54	174 ± 9	1.35	0.09
LG	173 ± 39	178 ± 39	0.0	0.4

Z, p: one-tailed, Wilcoxon signed rank test.

TA and LG integrated burst amplitudes scaled with LG recruitment within embryo, and were greatest when LG was recruited every cycle of an RLM, as summarized in Table 10. Scaling of TA normalized integrated burst amplitude fell short of significance after Bonferroni correction (p < 0.01) during both control (Friedman test, χ² = 7.60, d.f. = 2, p > 0.02) and foot-free RLMs (χ² = 8.32, d.f. = 2, p > 0.01), and Wilcoxon signed rank tests indicated the burst amplitudes did not differ between control and foot-free conditions within embryo (Table 10). LG burst amplitudes also scaled with LG recruitment during control RLMs, and Wilcoxon signed rank tests indicated scaling was significant for control RLMs (Z = −1.75, p < 0.01) but fell short for foot-free RLMs (Z = −2.02, p > 0.02), after Bonferroni correction (p < 0.0125). LG burst amplitudes did not differ between control and foot-free conditions (Table 10).

TABLE 10.

Scaling of normalized TA and LG integrated amplitudes (% of maximium) for RLMs during control and foot-free conditions (grand mean ± SD)

LG recruitment pattern	Control	Foot-free	Z	p>
TA
No LG recruitment	17 ± 3	15 ± 2	−1.21	0.1
Inconsistent LG recruitment	19 ± 4	17 ± 5	−0.67	0.2
Consistent LG recruitment	25 ± 6	26 ± 9	−0.41	0.3
LG
Inconsistent LG recruitment	11 ± 2	12 ± 2	−0.37	0.3
Consistent LG recruitment	20 ± 5	19 ± 4	−0.37	0.3

Z, p: one-tailed, Wilcoxon signed rank test comparing control and foot-free integrated amplitudes.

4 |. DISCUSSION

We hypothesized that muscle afferent feedback under normal mechanical constraint in ovo is a significant input to the flexor bias observed during RLMs at E20 in chick embryos. We tested the hypothesis by manipulating proprioceptive feedback in two ways, by tenotomy to selectively reduce ankle flexor or extensor afference, and by new analysis of published data for RLMs generated during spatially restricted and relatively unrestricted leg movement. Results indicated neither tenotomy nor removal of shell altered the flexor bias in RLM EMG.We discuss how the findings challenge our hypothesis and their implications for the control of flexor bias. We also discuss their relevance to control of flexed postures common to the vertebrate embryo and fetus. Finally, we consider how our findings advance knowledge regarding proprioceptive control of locomotor-related leg movement antecedent to posthatching locomotion.

4.1 |. Flexor bias persists after tenotomy and shell removal

Our key finding was that neither LG nor TA tenotomy altered three key features of flexor recruitment bias during RLMs. TA bursts outnumbered LG bursts; TA bursts initiated most RLMs; and TA bursts were consistently generated across all cycles in most RLMs. All three features were also present in concurrent RLMs of the contralateral ankle. Further, the flexor bias persisted after a secondary, contralateral tenotomy. In light of these surprising findings, we performed a retrospective analysis of RLMs pre and post shell removal to rule out other unidentified proprioceptive inputs coding mechanical constraint. We reasoned that significantly larger excursions of unrestrained movement likely enhanced dynamic muscle afference during RLMs. Again to our surprise, measures of recruitment bias were unaffected. Collectively, results indicate that ankle proprioceptive afference associated with spatial restraint in ovo does not significantly contribute to flexor recruitment bias during RLMs.

Another key finding was that neither tenotomy nor removal of spatial restraint altered the scaling of integrated burst amplitude during RLMs (Sun & Bradley, 2017). Normalized TA integrated amplitude during RLMs under all conditions was least in the absence of LG activity and greatest when LG was consistently recruited. Given that burst amplitude is an indicator of excitatory neural drive (Farina, Merletti, & Enoka, 2004; Staudenmann, van Dieen, Stegeman, & Enoka, 2014), we recently proposed that amplitude scaling and consistent flexor bursting across all levels of drive are indicative of an excitatory drive favoring flexor motor pools during RLMs (Sun & Bradley, 2017). Thus, current results lead us to conclude that muscle afference does not significantly contribute to the flexor bias in neural drive during RLMs in E20 chicks. Rather, we propose that the flexor bias is generated by descending pathways and/or intrinsic spinal networks. As our study focused only on parameters indicative of bias in recruitment and drive of ankle muscles, it remains to be determined if muscle afference modulates other aspects of RLM activity comparable to the modulation of reflexes and EMG during overground walking.

4.2 |. Neural mechanisms underlying flexor bias

Flexor bias during RLMs may indicate that descending pathways driving flexor motor pools mature earlier than those driving extensor pools and contribute to an important asymmetry in circuits and behavior during embryonic locomotor development. In adult mesencephalic cats, activation of rubrospinal and reticulospinal pathways increases flexor muscle activity during the swing phase of locomotion, whereas activation of the vestibulospinal pathway increases extensor activity during the stance phase (Orlovsky, 1972; Orlovsky, Deliagina, & Grillner, 1999). However, descending pathways are somewhat immature in the perinatal period. In neonatal rats, for example, drive to leg extensor muscles is insufficient to produce weight-supported stance, whereas L-dopa administration can produce weight-supported stepping (Navarrete, Slawinska, & Vrbova, 2002). In chicks, both reticulospinal and vestibulospinal pathways reach the lumbar spinal cord by E5–7 (Okado & Oppenheim, 1985), but their functional maturation is yet to be fully elucidated. At E20, descending pathways contribute to synchronous bilateral extensor activity during hatching (Bekoff & Kauer, 1982), for chicks fail to initiate hatching if decerebrated (Oppenheim, 1972). Thus, vestibulospinal and dopaminergic pathways likely activate extensor motor pools by E20, but activation may not be sustained, and/or descending inputs to extensor may be gated by inputs favoring flexor pools during RLMs.

Lack of specific sensory inputs, such as limb loading, may also account for the frequent extensor recruitment failure during many RLMs. Though 1b affferent circuits are yet to be determined in birds, it is well established in cats and humans that extensor 1b afferents are activated by limb loading and enhance extensor muscle activity during the stance phase of gait (Conway, Hultborn, & Kiehn, 1987; Duysens & Pearson, 1980; Guertin, Angel, Perreault, & McCrea, 1995; Ivanenko, Grasso, Macellari, & Lacquaniti, 2002; Musselman & Yang, 2007; Pang & Yang, 2000; Pearson, Ramirez, & Jiang, 1992; Whelan, Hiebert, & Pearson, 1995). In chick embryos, leg loading occurs during hatching (Bekoff, 1976; Bekoff & Kauer, 1984; Hamburger & Oppenheim, 1967), but during RLMs, the foot does not appear to contact the eggshell to load the leg (Bradley et al., 2014). Thus, it is possible that flexor bias during RLMs was not altered by tenotomy or shell removal because limb loading was not sufficiently different from control conditions. It is also possible that leg extensor afferents are less mature than flexor afferents at E20, but this remains to be explored.

Recruitment and drive bias in RLMs could be a product of the locomotor central pattern generator (CPG), for locomotor activity in isolated mouse spinal cord preparations suggests the CPG has a flexor bias. For example, during fictive locomotion, it was determined that excitatory conductances were greater in flexor motor neurons compared to extensor neurons, indicating a bias in CPG output favoring flexor motor neurons (Endo & Kiehn, 2008). In a subsequent study evidence indicated that rhythm generating interneurons expressing Shox2 formed more synaptic connections with flexor than extensor motor neurons (Dougherty et al., 2013). Further, more Shox2-expressing interneurons were rhythmically synchronized with flexor than extensor motor neurons. More recently, during locomotor-related activity in the isolated spinal cord of the neonatal mouse, calcium imaging indicated that the position (e.g., column) identity of motor neuron pools determined if neurons generated a flexor- or extensor-like firing pattern. However, failure to establish position identify by inactivation of FoxP1 expression led to all motor neurons firing in a flexor-like pattern (Machado et al., 2015). Our findings extend these work, demonstrating that flexor bias is also expressed both before and after reducing sensorimotor afferent input in an otherwise intact nervous system during nearly normal behavioral conditions.

4.3 |. Flexed posture—neural or mechanical?

Persistence of flexor bias following tenotomy and shell removal also raises the question if central neural drive or mechanical constraint imposed by the shell is a key determinant of flexed posture during embryonic development. It is possible, for example, that the bias in flexor recruitment and drive during RLMs is an adaptive mechanism for accommodating late-stage prenatal growth within a spatially limited environment, such as that imposed by the egg shell or uterus. Also, relatively less extensor drive may be advantageous during late embryogenesis, protecting immature extensor muscles and tendons from being overloaded, as they are maintained in a lengthened position by extreme flexion. Reduced extensor drive may also preserve the integrity of the embryos’ environment, including the eggshell, membranes, blood vessels, yolk stalk and yolk, or in mammals, the umbilical cord and placenta.

Nonetheless, spatial restraint likely contributes to flexed prenatal posture. In utero, the arms and legs of human fetuses are deeply flexed against the head and body (Casaer, 1979; de Vries, Visser, & Prechtl, 1982). Whereas, the limbs of the preterm neonate, exposed to gravity, extend away from the head and body (Brown, Omar, & O’Regan, 1997). Chick embryos in ovo also assume a posture of deeply flexed limbs and spine (Hamburger & Oppenheim, 1967). However, if the shell anterior to the foot is removed, the leg extends outside the egg (Bradley et al., 2008; Ryu & Bradley, 2009). Yet, despite the more extended leg posture and joint excursion range, we found that flexor bias persisted during unrestrained RLMs. Thus, we propose that both mechanical constraint and central neural drive synergistically contribute to the flexed posture to serve motor development prior to birth. That is to say, central neural drive preferentially activates flexor motor pools and induces the flexed posture in anticipation of emerging mechanical constraints that in turn further contribute to the flexed posture. We cannot rule out a reverse order of events, and the origins of prenatal posture remain to be pursued.

4.4 |. Role of proprioceptive inputs during locomotor-related movements prior to hatching

We hypothesized that muscle afferent feedback from TA and LG under normal mechanical constraints, is a significant input producing flexor bias during RLMs at E20 because previous studies indicated proprioceptive circuits are functional prior to hatching. Most notably, sustained stretch of ankle flexors increased TA burst and cycle durations and TA burst amplitude during RLMs (Bradley et al., 2014). Additionally, loss of spatial restraint after shell removal seemly destabilized muscle activity (Bradley et al., 2008). Also, pyridoxine-induced elimination of proprioceptive afferents decreased leg excursion amplitudes in ovo (Sharp & Bekoff, 2015). We reasoned that TA afferent input could be activated by at least two different mechanisms. One, limited TA shortening due to constrained joint motion could produce isometric contractions and activate 1a afferents (Edin & Vallbo, 1990; Hagbarth et al., 1975). Two, the high burst frequency of many RLMs (4–10 Hz) might produce inertial lags in joint excursions sufficient to stretch spindles and activate 1a afferents (Hagbarth et al., 1975). We also speculated extreme dorsiflexion in ovo placed the LG on sustained stretch, activating group II afferents to excite TA motor neurons and inhibit the LG motor neurons (Nelson & Hutton, 1985). We tested our hypotheses using the tenotomy, because it has been shown to reduce both 1a and II afferent activity and stretch-induced reciprocal inhibition of antagonist motor neurons in decerebrate cats (Hyngstrom et al., 2007; Vrbová, 1963; Yellin & Eldred, 1970). We predicted that the tenotomy would eliminate or reduce burst parameters indicative of flexor bias.

Contrary to our prediction, tenotomy did not alter flexor or extensor burst parameters indicative of flexor bias during RLMs. At least three mechanisms may account for the failure of tenotomy to alter burst parameters. One, proprioceptors may not code small changes in muscle length during RLMs in ovo. Two, the LG may not be sufficiently stretched in ovo to activate extensor group II afferents.LG in chicks is a biarticular muscle that extends across the knee and ankle; thus, knee flexion in ovo may reduce the stretch produced by dorsiflexion. Three, the muscle-tendon unit may be too compliant to maintain stretch-induced tension required for intrafusal fiber length transduction. Although few studies have investigated the development of muscle tissue mechanics in chicks (Nowlan, Murphy, & Prendergast, 2007), a study of tetanic forces in latissimus dorsi, an axial muscle in chick, indicated that contractile properties of muscle fibers are not yet mature at E18–20 (Reiser & Stokes, 1982). We discounted the possibility of compensatory proprioceptive input from other ipsilateral leg joints because flexor bias also persisted during unrestrained RLMs after shell removal, and evidence suggests that 1a afferent input is a local segmental mechanism. For example, in decerebrate cats it was found that 1a reciprocal inhibition in the ankle muscles was not impacted by changes in hip joint angle (Hyngstrom et al., 2007). Evidence also indicates that reciprocal inhibitory connections of 1a afferents are functionally specific in neonatal mouse (Wang, Li, Goulding, & Frank, 2008), suggesting the circuitry may be selective prior to birth. We also discounted other somatosensory afferents as providing significant compensation post tenotomy because RLM movement dynamics do not appear to be different pre and post tenotomy (i.e., kinematic traces, Figures 2b and 3b). Nonetheless, if other inputs were sufficiently compensatory, it remains that muscle afferents were not likely a significant determinant of flexor bias during RLMs.

In sum, this study provides the first experimental evidence that in neurologically intact E20 chick embryos, the flexed leg posture assumed in a spatially restrictive environment does not significantly contribute to the flexor bias observed in EMG recordings during spontaneously generated RLMs. Our findings suggest that the bias in flexor recruitment and drive during locomotor-related movements is largely produced by central neural mechanisms at E20. RLMs are a valuable model for study of motor development because unlike fictive models, RLMs are spontaneously generated in an intact neurological system, unencumbered by effects of drugs or electrical stimulation. Whether or not the centrally generated flexor bias in motor pool recruitment contributes to the flexed postures observed in most animals during development remains to be determined.