운동단위 motor unit.. 근육 활성 탐구 ... 2021

[문형철]

Distribution of motor unit properties across human muscles

Jacques Duchateau and 

Roger M. Enoka

21 Dec 2021https://doi.org/10.1152/japplphysiol.00290.2021

• This is the final version - click for previous version

• Sections

• PDF(870 KB)
• Tools
• Share

Abstract

Download figureDownload PowerPoint

The purpose of our review was to compare the distribution of motor unit properties across human muscles of different sizes and recruitment ranges. Although motor units can be distinguished based on several different attributes, we focused on four key parameters that have a significant influence on the force produced by muscle during voluntary contractions:

the number of motor units,

average innervation number,

the distributions of contractile characteristics, and

discharge rates within motor unit pools.

Despite relatively few publications on this topic, current data indicate that the most influential factor in the distribution of these motor unit properties between muscles is innervation number. Nonetheless, despite a fivefold difference in innervation number between a hand muscle (first dorsal interosseus) and a lower leg muscle (tibialis anterior), the general organization of their motor unit pools, and the range of discharge rates appear to be relatively similar. These observations provide foundational knowledge for studies on the control of movement and the changes that occur with aging and neurological disorders.

본 연구의 목적은 

다양한 크기와 활성화 범위를 가진 인간 근육의 

운동 단위 특성의 분포를 비교하는 것이었습니다. 

운동 단위는 

여러 다른 속성에 따라 구분될 수 있지만, 

우리는 자발적 수축 시 근육이 생성하는 힘에 큰 영향을 미치는 

네 가지 핵심 파라미터에 초점을 맞췄습니다: 

운동 단위 수, 

평균 신경 분포 수, 

수축 특성 분포, 및 

운동 단위 풀 내 방전 속도. 

the number of motor units,

average innervation number,

the distributions of contractile characteristics, and

discharge rates within motor unit pools.

이 주제에 대한 연구가 상대적으로 적음에도 불구하고, 

현재 데이터는 

이러한 운동 단위 특성 분포에 있어 근육 간 가장 영향력 있는 요소가 

신경 분포 수임을 나타내고 있습니다. 

innervation number

그러나 

손 근육(첫 번째 등간근)과 하퇴 근육(전경골근) 사이의 신경 분포 수에 

5배의 차이가 있음에도 불구하고, 

그들의 운동 단위 풀의 일반적인 조직 구조와 방출 속도 범위는 상대적으로 유사합니다. 

이러한 관찰 결과는 

운동 조절 및 노화 및 신경학적 장애와 관련된 변화에 대한 

연구의 기초 지식을 제공합니다.

INTRODUCTION

Although the anatomical and functional organization of the motor output from the spinal cord has been investigated during the century since the pioneering studies of Sherrington (1, 2), there remains a lack of information regarding many of the details. These gaps in knowledge on the final common pathway constrain our understanding of the motor system.

Liddell and Sherrington (2) laid the foundation for this field by introducing the concept of a “motor unit,” which is defined as the entire motor neuron, its dendrites and axon, and the muscle fibers innervated by the axon (3). The technology to record the activity of single motor units in human muscles during voluntary contractions was first developed by Adrian and Bronk (4). Their pioneering studies demonstrated that the force exerted by a muscle during voluntary contractions was the result of the concurrent recruitment of motor units and modulation of the rate at which they discharge action potentials. Subsequently, based on observations in an animal preparation, Henneman (5, 6) proposed that the recruitment order of motor units was determined primarily by motor neuron size, which became known as the size principle. Since these key findings, a great amount of work has been done on characterizing motor unit function (for review papers, see Refs. 3, 7–11), but many basic questions remain unanswered.

Among these questions, the distribution of motor unit properties across human muscles and the consequences for the control of motor output are largely unknown. We know that the force produced by a single muscle depends on the number of active motor units, the number and size of the muscle fibers innervated by motor neurons (i.e., innervation number), the contractile characteristics of the motor units, and the frequency at which the motor neurons discharge action potentials (9, 12), but little is known about the variation in these properties across muscles. In both animals and humans, there is a wide difference in absolute size and force capacity of muscles (13–15), but also among the motor units in a given muscle (16–18). Within-pool differences in the contractile characteristics of muscle fibers also explain the wide ranges of mechanical properties of the motor units in the muscles of animals, with up to a 50- to 100-fold range in maximal forces and a fivefold range in contraction speed (13). In addition, the recruitment range of motor units during slow contractions differs between human muscles: the upper limit of recruitment is <60% of maximal voluntary contraction (MVC) force in some hand muscles (19–22) but extends up to 90% MVC force in most limb muscles and intrinsic foot muscles (18, 21–24).

This review focuses on the distribution of motor unit properties across a few human muscles that differ in size and recruitment range. Although the attributes of motor units can be distinguished on the basis of anatomical, contractile, and biochemical characteristics, we limit our discussion to four key parameters that influence the force produced by muscle during voluntary contractions: the number of motor units, average innervation number, and the distributions of contractile characteristics and discharge rates within motor unit pools. This basic knowledge can inform our interpretation of the significance of the adaptations in motor unit properties that occur during conditions that compromise the integrity of the neuromuscular system.

소개

셔링턴(Sherrington)의 선구적인 연구(1, 2) 이후 1세기 동안 척수에서 나오는 운동 출력의 해부학적 및 기능적 조직에 대한 연구가 진행되어 왔지만, 많은 세부 사항에 대한 정보가 부족합니다. 이 최종 공통 경로에 대한 지식의 공백은 운동 시스템에 대한 이해를 제한하고 있습니다.

Liddell과 Sherrington(2)은

“운동 단위”라는 개념을 도입함으로써 이 분야의 기초를 마련했습니다.

운동 단위는

운동 신경세포, 그 수상돌기와 축삭,

그리고 축삭에 의해 신경이 분포된 근육 섬유로 구성됩니다(3).

인간 근육에서 자발적 수축 시 단일 운동 단위의 활동을 기록하는 기술은

Adrian과 Bronk(4)에 의해 처음 개발되었습니다.

그들의 선구적인 연구는

자발적 수축 시 근육이 발휘하는 힘이 운동 단위의 동시적 동원과

그들이 발화하는 활동 전위의 속도 조절에 의해 결정된다는 것을 보여주었습니다.

이후 동물 실험 관찰을 바탕으로 Henneman(5, 6)은

운동 단위의 동원 순서가

주로 운동 신경 세포의 크기에 의해 결정된다고 제안했으며,

이는 크기 원리(size principle)로 알려지게 되었습니다.

이 핵심 발견 이후,

운동 단위 기능 특성화에 대한 많은 연구가 진행되었습니다(리뷰 논문은 Ref. 3, 7–11 참조),

그러나

많은 기본적인 질문이 여전히 미해결 상태입니다.

이 질문 중 인간 근육 내 운동 단위 특성 분포와 운동 출력 조절에 미치는 영향은 대부분 알려지지 않았습니다.

우리는 단일 근육이 생성하는 힘이

활성 운동 단위 수,

운동 신경에 의해 신경 분포된 근육 섬유의 수와 크기(즉, 신경 분포 수),

운동 단위의 수축 특성,

운동 신경이 활동 전위를 방출하는 빈도에 따라 결정된다는 것을 알고 있습니다(9, 12),

We know that the force produced by a single muscle depends on

the number of active motor units,

the number and size of the muscle fibers innervated by motor neurons (i.e., innervation number),

the contractile characteristics of the motor units, and

the frequency at which the motor neurons discharge action potentials 

하지만 이러한 특성들의 근육 간 변이에 대한 지식은 매우 제한적입니다. 동물과 인간 모두에서 근육의 절대적 크기 및 힘 용량에 큰 차이가 존재합니다(13–15), 또한 특정 근육 내 운동 단위 간에도 차이가 있습니다(16–18). 근육 내 근육 섬유의 수축 특성 차이는 동물 근육의 운동 단위 기계적 특성 범위(최대 힘의 50~100배, 수축 속도의 5배 범위)를 설명합니다(13). 또한, 느린 수축 시 운동 단위의 모집 범위는 인간 근육 간에 차이가 있습니다: 일부 손 근육에서는 최대 자발적 수축력(MVC)의 60% 미만이 상한선이지만(19–22), 대부분의 사지 근육과 내재성 발 근육에서는 MVC의 90%까지 확장됩니다(18, 21–24).

이 리뷰는 크기 및 모집 범위에서 차이가 있는 몇 가지 인간 근육의 운동 단위 특성의 분포에 초점을 맞춥니다.

운동 단위의 특성은

해부학적, 수축적, 생화학적 특성에 따라 구분될 수 있지만,

우리는 자발적 수축 시 근육이 생성하는 힘에 영향을 미치는

네 가지 핵심 매개변수에 논의 범위를 제한합니다:

운동 단위 수,

평균 신경 분포 수,

운동 단위 풀 내 수축 특성 및 방전 속도의 분포입니다.

the number of motor units,

average innervation number, and

the distributions of contractile characteristics and

discharge rates within motor unit pools

이 기본 지식은 신경근육 시스템의 무결성이 손상된 조건에서

운동 단위 특성의 적응이 갖는 의미를 해석하는 데 기여할 수 있습니다.

MOTOR UNIT NUMBER

Morphological Studies

One of the first studies to assess the number of motor units in human muscles was the one by Feinstein et al. (25). Their approach involved cadaveric estimates of the number of α axons innervating a muscle and measuring the size of myelinated axons in cross-sectional nerve slices. Based on a bimodal distribution observed for most muscles, they distinguished between small-diameter and large-diameter axons. The percentage of afferent and efferent axons was then estimated from the number of large-diameter axons based on observations from cat studies that ∼43% of the large axons in tibialis anterior and ∼36% in medial gastrocnemius were afferent fibers (26). These data were further supported by findings from a single patient who suffered from poliomyelitis and experienced an almost complete de-efferentation (Fex and Wohlfart, 1954 cited in 25). In this patient, 35%–40% of the large-diameter axons in the nerve supplying the medial gastrocnemius were afferent fibers. It was therefore assumed that ∼60% of the large-diameter axons in peripheral nerves were efferent fibers (25, 27). Based on these proportions, Feinstein et al. (25) reported that the number of motor axons, and thus motor units, ranged from 95 in the first lumbrical to 1,096 in the platysma. For example, they estimated 119 motor units in the first dorsal interosseus (FDI), 445 in tibialis anterior, 579 in medial gastrocnemius, and 332 in brachioradialis.

The major limitations of these findings are the relatively few specimens examined in each study (25, 28, 29) and the inability to distinguish between afferent and efferent fibers in human nerves. A major advance for the field occurred when Houser et al. (30) reported in 1983 that the enzyme choline acethyltransferase (ChAT) was exclusively expressed in efferent neurons and, more recently, by the discovery of a polyclonal anti-ChAT antibody capable of selectively labeling motor axons in peripheral nerves (31, 32). After applying ChAT and neurofilament immunofluorescence to nerves supplying the upper extremities, Gesslbauer et al. (31) quantified the total number of both efferent and afferent axons (myelinated and unmyelinated fibers) in nine organ donors whose hearts were still beating at the time of harvest. Their approach involved acquiring transverse images of whole nerve sections from the human specimens and analyzing the labeled images at predefined harvesting sites along the brachial plexus and the terminal nerves innervating the arm and hand. The images indicated that the brachial plexus receives contributions from the ventral rami of the four caudal cervical (C5-C8) and the first thoracic (T1) spinal nerves. At this level of the spinal cord, the average number of ventral axons, representing all the axons innervating the human arm, was found to be 350,000 ± 43,000. A major observation was that the proportion of motor axons (efferent fibers) in the brachial plexus was less than 10% (22,000 ± 3,000 motor axons) of the total number of axons.

To further understand the distribution of the nerve fibers projecting to upper limb muscles, Gesslbauer et al. (31) investigated the terminal nerves of the brachial plexus at several defined sites in the shoulder, arm, and hand. Specimens were harvested from the axillary, musculocutaneous, radial, median, and ulnar nerves. These data indicated that the axillary nerve, which originates from the posterior cord of the brachial plexus and supplies the three heads of the deltoid, teres minor, and occasionally the long portion of the triceps brachii, contained 20,389 ± 2,724 afferent axons and 2,108 ± 317 efferent axons (9.4 ± 1.2%). The musculocutaneous nerve, which innervates the elbow flexors, comprised 24,303 ± 5,865 afferent axons and 1,601 ± 164 efferent axons (6.4 ± 1.3%). A total of 61,353 ± 10,408 afferent axons and 4,343 ± 365 efferent axons were counted in the radial nerve (6.7 ± 1.0%), which is responsible for the arm, wrist, and finger extension, as well as forearm supination. The median nerve, supplying most of the flexor muscles of the hand and fingers, and pronator muscles, contained 60,523 ± 6,280 axons, of which 3,374 ± 388 were efferent axons (5.6 ± 0.7%). A lesser number of axons was counted in the ulnar nerve, which innervates some forearm flexors (flexor carpi ulnaris and medial half of the flexor digitorum profundus) and intrinsic hand muscles (hypothenar muscles; 3rd and 4th lumbricals, dorsal and palmar interossei, adductor pollicis, and deep head of flexor pollicis brevis). The ulnar nerve contained 40,379 ± 5,638 axons at its origin, of which 2,670 ± 347 were efferent axons (6.7 ± 0.8%).

Although the total number of efferent axons at different levels (origin, arm, and wrist) was also counted, a direct comparison of these results with those of Feinstein et al. (25) is not possible. Despite Feinstein et al. (25) being able to identify large-diameter fibers, which excluded γ motor fibers (muscle spindles) and small-diameter sensory fibers (e.g., pain, thermal), they were not able to distinguish between efferent and afferent fibers. In contrast, Gesslbauer et al. (31) were able to separate efferent and afferent fibers (myelinated and unmyelinated), but the myelinated axons of γ neurons were included in the count of efferent axons. As the relative proportion of α and γ fibers innervating each muscle in humans is uncertain (33), it is not possible to estimate the number of motor units from the study of Gesslbauer et al. (31). Moreover, the fact that the relative proportion of efferent and afferent fibers (60% vs. 40%) used by Feinstein et al. (25) was likely too high (<10% efferent fibers) as found by Gesslbauer et al. (31), the estimated number of motor axons and thus motor units was likely overestimated by Feinstein et al. (25). Without being able to dissociate α fibers from γ fibers, it is therefore not currently possible to estimate accurately the number of motor units per muscle in humans on the basis of the available morphological data.

A major step forward for the field would be the combination of histological approaches (25, 28, 29) with immunofluorescence techniques, such as the one used by Gesslbauer et al. (31), to examine the terminal section of the nerve just before it reached each target muscle.

운동 단위 수

형태학적 연구

인간 근육의 운동 단위 수를 평가한 최초의 연구 중 하나는 Feinstein 등(25)의 연구입니다. 그들의 접근 방식은 시체 해부법을 통해 근육을 신경 분포하는 α 축삭의 수를 추정하고 신경 단면 슬라이스에서 마이엘린화된 축삭의 크기를 측정하는 것이었습니다. 대부분의 근육에서 관찰된 이분형 분포를 바탕으로, 그들은 직경이 작은 축삭과 직경이 큰 축삭을 구분했습니다. 그런 다음, 고양이 연구에서 전경골근의 큰 축삭 중 약 43%가 구심성 섬유이고, 내측 비복근의 약 36%가 구심성 섬유라는 관찰 결과를 바탕으로, 구심성 및 원심성 축삭의 비율을 큰 직경의 축삭 수에서 추정했습니다 (26). 이 데이터는 소아마비로 고통받고 거의 완전한 운동 신경의 소실(Fex and Wohlfart, 1954, 25에서 인용)을 경험한 한 환자의 연구 결과로 더욱 뒷받침되었습니다. 이 환자에서, 내측 비복근에 신경을 공급하는 큰 직경의 축삭 중 35%~40%는 구심성 섬유였습니다. 따라서 말초 신경의 대구경 축삭의 약 60%는 원심성 섬유라고 가정했습니다 (25, 27). 이러한 비율을 바탕으로, Feinstein 등 (25)은 운동 축삭의 수, 즉 운동 단위의 수가 제1 렘브르키알에서 95개, 평활근에서 1,096개에 달한다고 보고했습니다. 예를 들어, 그들은 첫 번째 등간근(FDI)에 119개의 운동 단위, tibialis anterior에 445개, medial gastrocnemius에 579개, brachioradialis에 332개를 추정했습니다.

이러한 연구 결과의 주요 한계는 각 연구에서 조사한 표본이 상대적으로 적고 (25, 28, 29), 인간 신경에서 구심성 섬유와 원심성 섬유를 구분할 수 없다는 점입니다. 이 분야의 주요 발전은 Houser 등(30)이 1983년에 효소 콜린 아세틸트랜스퍼라제(ChAT)가 원심성 뉴런에서만 발현된다는 것을 보고한 것과, 최근에 말초 신경의 운동 축삭을 선택적으로 표지할 수 있는 다클론 항 ChAT 항체가 발견된 것입니다(31, 32). ChAT 및 신경필라멘트 면역형광법을 상지 신경에 적용한 후, Gesslbauer 등(31)은 심장이 아직 뛰고 있는 상태에서 기증된 9명의 장기 기증자의 원심성 및 구심성 축삭(수초화 및 비수초화 섬유)의 총 수를 정량화했습니다. 이 접근법은 인간 표본의 전체 신경 단면에서 가로 단면 이미지를 획득하고, 상완 신경총과 팔 및 손에 신경 분포를 제공하는 말단 신경의 미리 정의된 채취 부위에서 표지된 이미지를 분석하는 것을 포함했습니다. 이미지는 상완 신경총이 4개의 후방 경추(C5-C8)와 첫 번째 흉추(T1) 척추 신경의 복부 가지로부터 기여를 받는다는 것을 보여주었습니다. 척수 이 수준에서 인간 팔을 신경 분포하는 모든 축삭을 대표하는 복측 축삭의 평균 수는 350,000 ± 43,000으로 확인되었습니다. 주요 관찰 결과 중 하나는 상완 신경총에서 운동 축삭(원심성 섬유)이 전체 축삭 수의 10% 미만(22,000 ± 3,000개의 운동 축삭)을 차지한다는 것이었습니다.

상지 근육으로 투사되는 신경 섬유의 분포를 더욱 이해하기 위해 Gesslbauer 등(31)은 어깨, 팔, 손의 여러 정의된 부위에서 상완 신경총의 말단 신경을 조사했습니다. 표본은 액와 신경, 근육피부 신경, 방사 신경, 중간 신경, 및 척골 신경에서 채취되었습니다. 이 데이터에 따르면, 상완 신경총의 후방 코드에서 시작하여 삼두근, 소원근, 그리고 때로는 삼두근의 긴 부분까지 공급하는 겨드랑 신경에는 20,389 ± 2,724개의 구심성 축삭과 2,108 ± 317개의 원심성 축삭(9.4 ± 1.2%)이 포함되어 있는 것으로 나타났습니다. 팔꿈치 굴근에 신경 분포를 하는 근피 신경은 24,303 ± 5,865개의 구심성 축삭과 1,601 ± 164개의 원심성 축삭 (6.4 ± 1.3%)으로 구성되어 있습니다. 팔, 손목, 손가락의 신장과 팔뚝의 외전 운동을 담당하는 요골 신경에서는 총 61,353 ± 10,408개의 구심성 축삭과 4,343 ± 365개의 원심성 축삭이 계산되었습니다 (6.7 ± 1.0%). 손과 손가락의 대부분의 굴근과 회내근에 신경을 공급하는 정중 신경에는 60,523 ± 6,280개의 축삭이 포함되어 있으며, 그 중 3,374 ± 388개가 원심성 축삭(5.6 ± 0.7%)이었습니다. 척골 신경에는 전완 굴곡근(척골 손목 굴곡근과 손가락 굴곡근의 내측 반쪽)과 손의 내재근(소지근; 3번째와 4번째 럼브리칼 근육, 등쪽과 손바닥쪽 간골근, 엄지 내전근, 엄지 굴곡근의 깊은 머리)을 신경 분포하는 축삭의 수가 적었습니다. 척골 신경은 기원에서 40,379 ± 5,638개의 축삭을 포함하고 있으며, 그 중 2,670 ± 347개는 원심성 축삭(6.7 ± 0.8%)이었습니다.

다른 수준(기원, 팔, 손목)에서 원심성 축삭의 총 개수도 계산되었지만, 이러한 결과를 Feinstein 등(25)의 결과와 직접 비교하는 것은 불가능합니다. Feinstein 등(25)은 γ 운동 섬유(근육 스핀들)와 작은 직경의 감각 섬유(예: 통증, 열)를 제외한 큰 직경의 섬유를 식별할 수 있었지만, 원심성 섬유와 구심성 섬유를 구분하지는 못했습니다. 반면, Gesslbauer 등(31)은 원심성 및 구심성 섬유(수초가 있는 섬유 및 수초가 없는 섬유)를 구분할 수 있었지만, γ 뉴런의 수초가 있는 축삭은 원심성 축삭의 수에 포함되었습니다. 인간의 각 근육을 자극하는 α 및 γ 섬유의 상대적 비율은 불확실하기 때문에 (33), Gesslbauer 등 (31)의 연구를 통해 운동 단위의 수를 추정하는 것은 불가능합니다. 또한, Feinstein 등 (25)이 사용한 원심성 및 구심성 섬유의 상대적 비율 (60% 대 40%)은 Gesslbauer 등 (31)이 발견한 것보다 너무 높았기 때문에 (원심성 섬유가 10% 미만), 운동 축삭의 추정 수와 운동 단위의 추정 수가 Feinstein 등 (25)에 의해 과대 평가되었을 가능성이 높습니다. (25)가 사용한 원심성 섬유와 구심성 섬유의 상대적 비율(60% 대 40%)이 Gesslbauer 등(31)이 발견한 것보다 너무 높았을 가능성이 높기 때문에(원심성 섬유가 10% 미만), Feinstein 등(25)이 추정 한 운동 축삭의 수와 운동 단위의 수는 과대 평가되었을 가능성이 높습니다. α 섬유와 γ 섬유를 분리할 수 없는 상황에서, 현재로서는 이용 가능한 형태학적 데이터를 바탕으로 인간의 근육당 운동 단위의 수를 정확하게 추정하는 것은 불가능합니다.

이 분야에서의 중요한 진전은 조직학적 접근법(25, 28, 29)과 Gesslbauer 등(31)이 사용한 면역형광 기술을 결합하여 신경이 각 목표 근육에 도달하기 직전의 말단 부위를 조사하는 것입니다.

Electrophysiological Studies

As an alternative approach to morphological estimates, several electrophysiological methods have been developed to determine the number of motor units in human muscles. Three main approaches have been developed (for reviews, see Refs. 34, 35). The principle of these methods is to compare the average value of a motor unit parameter (EMG amplitude or area; peak force) with the value for the entire muscle. The first method, introduced by McComas et al. (36), involved electrically elicited muscle contractions in which stimulus intensity applied to a motor nerve is increased gradually until a maximal response is reached. The assumption of the McComas method is that the small stepwise increments in response size to each stimulus represent the addition of a single motor unit action potential to the compound action potential (M wave). By comparing the ratio between the average surface EMG increments to that of the maximal M wave, an average number of motor units can then be estimated (36). The major limitation of this method is that each small increment of the compound muscle action potential likely does not indicate the activation of a single motor unit (37, 38).

The variability and overestimation in motor unit numbers with this method are likely due to the small sample size and the alternation of two or more motor axons with similar depolarization thresholds during graded stimulation of the whole muscle nerve (38–40). Despite technical refinement (35) and the use of multipoint stimulation along the course of the nerve (38) to reduce the problem of alternation, this method cannot be used in many muscles (e.g., biceps brachii, vastus lateralis, tibialis anterior) for which the proximal nerve is not accessible for stimulation at enough locations needed to collect an adequate sample of single motor unit action potentials.

The problem of alternation can be avoided by using a spike-triggered averaging (STA) method (37, 41). This method involves recording both the discharge times of single motor units with intramuscular electrodes and the concurrent EMG activity with surface electrodes. By triggering from the discharge times and averaging into the surface EMG, it is possible to estimate the average size of the surface-recorded motor unit potential (macro-EMG potential). The number of motor units can be estimated by dividing the maximal M wave by the average value of the surface-detected motor unit sample (28, 37). A potential problem with this method is a sampling bias toward low-threshold motor units when they are recorded during low-force contractions. To reduce this drawback and obtain a representative sample, it has been recommended to record motor unit discharges during moderate-intensity voluntary contractions. For example, the measurements are made during contractions at ∼25% of MVC force for tibialis anterior (42), but the optimal contraction intensity likely differs across muscles. Furthermore, to avoid the potential confounding influence of phase cancellation (43), the preferred approach is to analyze the amplitude or area of the negative peak of the surface-detected response (motor unit and M wave) instead of the whole peak-to-peak amplitude of the signal (34, 41).

Based on the same general methodological approach, alternative methods have been developed. These include extracting single motor unit discharges noninvasively from the F-wave response (44), using high-density surface EMG recordings to augment the number of identified motor units (45), mathematical models that simulate the M wave and surface EMG interference signals recorded at different forces to determine a motor unit number index (MUNIX) (46), and improving the McComas method by addressing more carefully motor unit alternation (47). These methods are not described in our review, but information can be obtained elsewhere (34, 47).

In addition to estimating the number of motor units based on the EMG recordings, Stein and Yang (40) proposed a similar approach based on the contractile properties of the motor units. Their method involves measuring the twitch force of single motor units, either by STA (48, 49) or by intramuscular stimulation of axonal branches (50), and whole muscle twitch force induced by maximal electrical stimulation of the motor nerve. As for the EMG methods, the estimated number of motor units is calculated by dividing muscle twitch force by the average force of a sample of single motor units. As it is technically challenging to obtain accurate measurements of the force produced by low-threshold motor units in big muscles, the method has been mainly used in hand, forearm, and one lower leg muscle (tibialis anterior).

Although these methods have contrasting sampling problems, Stein and Yang (40) observed good agreement between the estimated number of motor units for the STA and microstimulation methods in the thenar muscles: 135 and 122 based on EMG, and 130 and 116 based on the force for the two methods, respectively. The value obtained with the McComas method (36) was 170. There was no statistically significant difference in these estimates. These findings indicate that the estimated number of motor units is relatively similar regardless of the method (STA vs. microstimulation) and parameter (EMG vs. force) used, at least for the thenar muscles. These data also agree reasonably well with the average value (133 motor units) calculated from different electrophysiological studies published in the literature (Table 1).

Enlarge table

Table 1. Average motor unit number estimated from spike-triggered averaging of either EMG or force in different muscles

Enlarge table

Distribution across Muscles

In an attempt to compare the distribution of motor-unit properties in human muscles of different sizes, we examined muscles for which data from a minimum of three studies were available. Table 1 summarizes the number of motor units for five muscles. To increase the consistency of the results, we only included data recorded by the STA obtained with either EMG or force as the outcome measurement. When possible, we selected EMG values determined from the surface-detected amplitude of the negative phase (first phase) of the motor unit. As expected, these data show fewer motor units for hand muscles compared with limb muscles (Table 1). Among the limb muscles, biceps brachii displayed a greater number of motor units (312) compared with the tibialis anterior (200) and vastus lateralis (146). However, the coefficient of variation was greater for vastus lateralis and tibialis anterior compared with the other muscles. Relative to its size (Table 2), the estimated number of motor units in the vastus lateralis is low, likely due to limitations of the STA method in large muscles (56). No significant association was found between the average number of motor units and muscle anatomical cross-sectional area (CSA) (Fig. 1A) calculated for four muscles from data reported in Tables 1 and 2.

Figure 1.A: estimated numbers of motor units and muscle fibers for four human muscles. Each data point corresponds to the means ± SD for each muscle cited in both Table 1 and Table 2. The average number of fibers relative to the muscle anatomical cross-sectional area (CSA) has been corrected for each muscle by the specific ratio between physiological and anatomical CSA (see Comparison across Muscles for the procedure). The graph indicates the absence of a significant association between the two variables. B: relation between physiological cross-sectional area (PCSA) and innervation number in the same four muscles as in A. Each data point corresponds to the means ± SD for the studies cited in Table 2. The innervation number has been calculated for each muscle as the ratio between the average muscle fibers relative to the physiological CSA (PCSA) and the average motor unit number reported in Table 1. The graph shows a significant correlation (r = 0.988; P < 0.011) between the two variables. BB, biceps brachii; FDI, first dorsal interosseus; TA, tibialis anterior; VL, vastus lateralis.

Download figureDownload PowerPoint

Enlarge table

Table 2. Characteristics of selected muscles and associated muscle fibers

Enlarge table

In conclusion, except for hand muscles, there is considerable variability in the estimated numbers of motor units across studies for a given muscle and additional data are needed to determine with more certainty the number of motor units in human muscles. Although electrophysiological methods appear to provide a relatively consistent estimate for hand muscles, the results are more variable for larger muscles. As many muscles are not accessible to electrophysiological approaches, it seems that only the results from morphological methods can provide a more objective and complete view of the number of motor units within a motor unit pool. Nonetheless, it is necessary to improve electrophysiological methods as they offer the only viable approach to track motor neuron loss during aging or the progression of diseases, such as amyotrophic lateral sclerosis and spinal muscular atrophy (34, 35, 45).

AVERAGE INNERVATION NUMBER

A comparison of the distribution of motor unit properties across muscles requires information about both the average number of motor units within the pool and the number of muscle fibers innervated by each motor neuron. Findings from animal studies indicate that there is a strong correlation between the innervation number and the maximal tetanic force of a motor unit (76–78). The most direct approach used to determine innervation number in such studies is the glycogen-depletion technique (79). Briefly, the method involves depleting the glycogen in muscle fibers that belong to a single motor unit by prolonged electrical stimulation of its motor axon and then using classic histochemical techniques to count the number of muscle fibers in cross-sectional slices (76).

Such studies have shown substantial variation in the innervation number between motor units in a given muscle. In the cat and rat tibialis anterior muscles, for example, strong and fast contracting motor units (type FF) innervate an average of 2.7 (76) and 4.1 (77) times the number of muscle fibers compared with weak and slow-contracting motor units (type S). Due to the lesser variation in specific force (maximal tetanic force per unit of CSA) and muscle fiber diameter between motor units, these findings indicate that maximal motor unit force mainly depends on innervation number (77, 78).

Estimation from Cadaveric Measurements

The experimental approaches used in animal studies cannot be translated to humans and other methods are required. One approach is based on counting the total number of muscle fibers in a transverse section of a muscle from cadaver specimens. With this approach, Feinstein et al. (25) found that the average number of muscle fibers varied from ∼10,000 fibers in the first lumbrical to ∼1,100,000 fibers in medial gastrocnemius. By computing the ratio between the total number of muscle fibers and the estimated number of motor axons supplying the muscle, the average innervation number can be estimated. The average innervation number varies across muscles, ranging from ∼100 for the first lumbrical up to 2,000 for medial gastrocnemius (25).

In addition to the uncertainty about the exact number of motor axons as described in the preceding section, other limitations of this method are that not all muscle fibers extend along the length of a muscle, and differences in fiber CSA are not taken into account. Subsequent studies analyzed muscle slices after staining for mATPase to visualize type I and type II fibers (80). A total number of fibers in the muscle was counted as well as the fiber-type proportions for the muscle (81). Based on this method, the estimated total number of muscle fibers was ∼500,000 in vastus lateralis (67, 69) and ∼140,000 in tibialis anterior (72–74) (Table 2). These numbers varied between sexes with significantly fewer muscle fibers (∼20%) in the tibialis anterior of women (73) compared with men (72). As the number of motor axons was not quantified in these studies, it is not possible to estimate the average innervation number from these data.

Estimation from Muscle Biopsies

An alternative approach to cadaveric measurements is to determine the average CSA of individual muscle fibers obtained from muscle biopsy samples in humans and to measure whole muscle CSA with an imaging technique. After correcting for connective tissue content and assuming that most muscle fibers extend along the whole muscle length, the total number of muscle fibers was estimated by dividing the whole muscle anatomical CSA (i.e., area of the cross section perpendicular to muscle axis) by average muscle-fiber CSA (Table 2). With this approach, Sale et al. (65) obtained average values of ∼200,000 muscle fibers in biceps brachii of men and ∼140,000 for women. Similarly, Miller et al. (14) estimated a slightly greater number (not significant) of muscle fibers in biceps brachii of men (180,620) than in women (156,872), but not in vastus lateralis (451,468 in men vs. 465,007 in women). The data for vastus lateralis are quite similar to the values obtained from cadaveric specimens (∼500,000; 67, 69).

Unfortunately, no direct estimation of the number of motor axons was performed in these studies. However, Miller et al. (14) did use the McComas method (36) to estimate the number of motor units in the two muscles; they found 126 motor units in biceps brachii for men and 110 for women, representing average innervation numbers of ∼1,433 fibers per motor unit for men and 1,426 for women. The values for vastus lateralis were 282 for men and 229 for women, resulting in average innervation numbers of 1,600 and 2,030 fibers per motor unit, respectively. The average number of motor units obtained with the other electrophysiological methods yielded 312 units for biceps brachii and 146 for vastus lateralis (Table 1) suggesting that Miller et al. (14) may have underestimated the innervation number for biceps brachii but overestimated it for vastus lateralis.

Comparison across Muscles

As emphasized in the two preceding paragraphs, there are few data on innervation number in human muscles. To increase sample size and thereby approximate innervation number with more confidence, we first calculated the ratio between whole muscle anatomical CSA and average muscle-fiber CSA to estimate the number of muscles fibers (Table 2). However, the number of muscle fibers is likely underestimated in pennated and big muscles (tibialis anterior and vastus lateralis) with this approach because most muscle fibers do not extend along the length of the entire muscle. As these studies did not report muscle volume and the associated architectural parameters (pennation angle and fascicle length), it was not possible to express these values relative to the physiological CSA (i.e., area of the cross section perpendicular to the muscle fibers; 75) and to determine innervation numbers more accurately.

As an alternative approach, we multiplied the total number of muscle fibers in each muscle by the ratio between the physiological CSA and the maximal anatomical CSA reported in the literature: 1.1 for FDI (62) and 1.9 for tibialis anterior (75). This ratio was calculated for vastus lateralis from the average anatomical CSA value of Table 2 and the physiological CSA reported by Marzilger et al. (82). The resulting physiological CSA was ∼2 times greater than the anatomical CSA for vastus lateralis. Given that pennation angle for biceps brachii is close to zero (83) and fascicle length is difficult to measure precisely in vivo, we used a ratio of 1 for biceps brachii, which may have been underestimated slightly the true value. Despite some limitations with this approach (data from cadavers and in vivo measurements were mixed and the ratios were calculated slightly differently between studies), the estimated average innervation numbers were 346 for the FDI, 622 for biceps brachii, 1,845 for tibialis anterior, and 5,990 for vastus lateralis. These data indicate that innervation number was least for FDI and greatest for vastus lateralis. As illustrated in Fig. 1B, average innervation number was highly correlated (r = 0.988; P = 0.011) with average physiological CSA across the four muscles. Collectively, these data indicate that the force capacity of muscle (physiological CSA) is more related to the average innervation number than to the number of motor units in the pool.

DISTRIBUTIONS WITHIN A MOTOR UNIT POOL

Although the data in the preceding section are useful for an overall comparison between muscles, it is the range of innervation numbers within an individual motor unit pool that is more critical for understanding how the nervous system controls muscle force. In animal studies, the glycogen-depletion method has indicated that the innervation number varies exponentially within a motor unit pool, with low-threshold motor units containing lower values (76, 84). For example, the innervation number ranges from 18 to 830 in the cat medial gastrocnemius muscle, corresponding to a 50-fold range (85).

Contractile Characteristics

The range of innervation numbers in motor unit pools of humans can be approximated indirectly from measures of motor unit force based on the strong correlation between the innervation number and maximal motor unit force (76–78). Using such an approach, Enoka and Fuglevand (9) determined the range of tetanic forces and estimated the number of muscle fibers required to achieve these forces in the FDI. Based on the data provided by Feinstein et al. (25) for this muscle (120 motor units and 40,500 muscle fibers), they estimated that the innervation number ranged from 21 to 1,770, representing an 84-fold range. These data further indicated that the frequency distribution of the innervation numbers within a motor unit pool is typically characterized by an exponential function with many units generating small forces, whereas relatively few motor units exert large forces. Although the FDI contains a relatively similar percentage of type I and II fibers (86), the skewed distribution of innervation numbers resulted in ∼84% of the motor units comprising type I muscle fibers and 16% containing type II fibers (9). The key point of these estimates is the finding of a nonlinear relation between innervation number and motor unit size, as shown in animal studies (87).

These results raise the question of whether the distribution of motor unit properties is similar for muscles that differ in size, contractile characteristics, recruitment range, and function. Due to technical difficulties, there is relatively little information on the range of contractile characteristics of single motor units of humans. Three main methods have been used to assess these characteristics: intramuscular microstimulation (50), intraneural microstimulation (16, 88), and STA methods (48, 49). Each of these methods has its own advantages and limitations (89). The first two approaches involve electrical stimulation of a single motor axon with a microelectrode inserted into either a muscle or a nerve trunk and recording the contractile properties of isolated motor units in response to a single stimulus (twitch) and various trains of stimuli (tetanus). This method can be used to determine the force-frequency relation and the maximal tetanic force of motor units. In contrast, the STA method is performed during a sustained voluntary contraction and involves recording the intramuscular action potential of a single motor unit while discharging at a low steady rate and averaging into the force signal to estimate the mechanical contribution of the selected unit to the whole muscle force (49). The STA method can be used to determine recruitment threshold, twitch force, and time to peak twitch force for individual motor units.

Although a more thorough characterization of motor unit contractile properties can be obtained with electrical microstimulation, the amount of slack during the recording procedure may underestimate twitch forces. In contrast, summation of twitch responses and possible synchronized activity of concurrent active units may distort the motor unit twitches with the STA method, but the voluntary contraction attenuates the confounding influence of tissue compliance on the evoked responses. Among these three experimental approaches, only the STA method has been used in intrinsic hand muscles (FDI, adductor pollicis, thenar muscles; 20, 40, 48, 90), forearm muscles (wrist extensors; 91), and leg muscles (tibialis anterior; 18, 92). We, therefore, compare the contractile properties and discharge rates of motor units in two of these muscles, FDI and tibialis anterior, for which we had sufficient data acquired in similar experimental conditions from the same laboratory. Our analysis of the literature indicates that tibialis anterior has an average innervation number approximately 5-fold larger than that of FDI (Table 2; Fig. 1B).

The distributions of the recruitment threshold for these two muscles are illustrated in Fig. 2. The range of recruitment thresholds is much narrower for FDI than for tibialis anterior: it ranges from 0.4 to ∼35% MVC force (sample: 44 units) for FDI and from 1 to ∼90% MVC force (sample: 528 units) for tibialis anterior. Relative to the total sample, 50% of the units have recruitment thresholds of less than 10% MVC force in the FDI and 20% MVC force in tibialis anterior. Furthermore, the relative number of motor units in the 0%–5% MVC range for FDI is about twice that of the tibialis anterior, representing ∼37% and 18% of the total sample for these two muscles. The peak twitch force ranged from 11.8 to 170 mN with an average value of 62 ± 46 mN for the FDI (20), whereas peak twitch torque ranged from 1 to 171 mN·m with an average value of 70 ± 42 mN·m for the tibialis anterior (18). A relatively comparable distribution for the time to peak twitch force/torque was found for the two muscles: values ranged from 25 to 60 ms (mean: 39 ± 10 ms) for FDI and from 20 to 86 ms (mean: 46 ± 14 ms) for tibialis anterior.

Figure 2.Distribution of recruitment thresholds for 528 motor units (MUs) in the tibialis anterior (black bars) and 44 motor units in the FDI (white bars) recorded during slow ramp contractions. These experimental data are from Van Cutsem et al. (18) and Duchateau and Hainaut (20) with permission from John Wiley and Sons. FDI, first dorsal interosseus.

Download figureDownload PowerPoint

A more complete view of the motor unit pool for the two muscles can be obtained by assigning the measured mechanical properties (peak twitch force and time to peak twitch force) in a computational model (93). To that end, a pool of 100 and 200 motor units was determined for the FDI and tibialis anterior muscles, respectively, based on experimental findings (Table 1). As illustrated in Fig. 3A, peak twitch force for the motor units increases exponentially in both muscles with a rate constant of 0.045 for FDI and 0.029 for tibialis anterior, when expressed as the rank occupied (in %) by the motor units in each pool. However, low-threshold motor units develop proportionally more force in FDI than in tibialis anterior (Fig. 3A). For example, 50% of the motor unit pool produced a cumulative force of ∼18% in FDI, but only ∼10% in tibialis anterior (Fig. 3D).

Figure 3.Distribution of motor unit twitch forces in a model that comprised a pool of 200 units for tibialis anterior (TA) and 100 units for first dorsal interosseus (FDI). Motor unit forces, recorded by spike-triggered averaging (STA), are from Van Cutsem et al. (18) for TA and Duchateau and Hainaut (20) for FDI. A: motor unit (MU) force, expressed as a percentage of the greatest twitch force in the pool, is plotted as a function of the rank (in %) occupied by the unit in the pool. An exponential (e) growth function best fit the experimental data (r2 = 0.98 for TA and 0.97 for FDI). The equations are Y = 1.11·e(0.045·X) for TA and Y = 4.45·e(0.029·X) for FDI. B: motor unit (MU) force, expressed as a percent of the greatest twitch force in the pool, is plotted as a function of its time to peak force/torque (ms). A one-phase exponential decay best fit the data (r2 = 0.99 for TA and 0.97 for FDI). The equations are Y = 1,535·e(−0.117·X) + 3.5 for TA and Y = 1,548·e(−0.1135·X) + 3.4 for FDI. C and D: cumulative sum of MU forces, expressed as % of maximum, as a function of the rank occupied by the unit in the pool expressed either in absolute value (C) or in percentage (D). An exponential growth function fits these data (r2 ≥ 0.98): C equations were Y = 1.11·e(0.0225·X) for TA and Y = 3.1·e(0.035·X) for FDI, and the D equations were Y = 0.93·e(0.047·X) for TA and Y = 3.1·e(0.035·X) for FDI.

Download figureDownload PowerPoint

The shape of the association between peak force (expressed in % of the greatest motor unit peak force within the pool) and the time to peak force/torque (ms) was similar for the two muscles (rate constant: −0.114 for FDI and −0.117 for tibialis anterior; Fig. 3B). A key difference between the two muscles, however, was the longer time to peak twitch force/torque (≥ 60 ms) of many low-force units in tibialis anterior, presumably due to a greater amount of slack and compliance of the series elastic component of the dorsiflexor muscles (94) compared with FDI. Interestingly, when cumulative motor unit force is expressed as a function of the absolute number of motor units in the pool (Fig. 3C), the lower rate constant in tibialis anterior (0.022) compared with FDI (0.035) indicates a slower rate of force gradation (about half) due to a greater recruitment range. When cumulative motor unit force is expressed as a percentage of the total motor unit pool (Fig. 3D), the shape of the relation for these two muscles is quite similar, suggesting that differences in force control between the two muscles depend on their relative discharge rate characteristics.

Discharge Rate Characteristics

In addition to the number and size of the fibers innervated by a motor neuron, the rate at which action potentials are discharged (rate coding) has a substantial influence on the force that a motor unit can produce (95). A sigmoidal association between the increase in stimulation frequency and force is classically reported for individual motor units (16, 17, 96). When force increases during a voluntary contraction, three phases have been often described: a first phase during which rate coding increases rapidly, a second phase characterized by a more gradual linear increase in discharge rate, and finally, a sharp increase in rate when the contraction approaches maximal force (97–99).

This general force-frequency relation seems to vary among motor units. In some units, rate coding tends to plateau (saturation effect) after the first phase despite a continual increase in force (22, 100, 101). This effect has been attributed, at least partly, to intrinsic mechanisms that limit the capacity of some motor neurons to increase their discharge rate despite continual increases in the net excitatory synaptic input (11, 95, 101). However, there are currently not enough published data on the force-frequency relation for single motor units in either FDI or tibialis anterior to determine that relation with enough confidence.

In the absence of these data, rate coding is often characterized in terms of the minimal and peak discharge rates. The association between minimal discharge rate and recruitment threshold appears variable and there is no consensus on whether the last recruited (high threshold) motor units achieve the greater peak discharge rates during gradual increases in the applied force. Some studies reported that peak discharge rate recorded with intramuscular EMG reaches greater values for high- than low-threshold motor units (22, 97, 100, 102–104), whereas other studies reported the opposite (18–20, 105, 106) or similar values for the whole pool of motor units (99). Even the recent decomposition methods of multichannel recordings to identify the discharge times of many concurrently active motor units have not clarified this issue (24, 107, 108). Some of this uncertainty may be attributed to differences between muscles (19, 109) and the experimental tasks, such as a slow isometric ramp or trapezoidal contractions (18–20, 99), step contractions to different target forces (22, 110), sustained MVC (63, 98, 108, 111), and also differences in muscle-tendon length (94).

Given the variability in the shape of the force-frequency relation across motor units within a pool, we compared the average minimal and peak discharge rates for the FDI and tibialis anterior. Surprisingly, no difference was observed between these muscles for either the minimal or peak discharge, regardless of contraction mode (Table 3). Average minimal discharge rates at the recruitment threshold are 8.2 for FDI and 8.6 pps (pulses per second) for tibialis anterior. Although the variability across studies is greater for peak discharge rate, the average values are 32.4 for FDI and 29.3 pps for tibialis anterior. This indicates that the range of discharge rate modulation is relatively similar (∼21–24 pps;Table 3) between these two muscles, despite longer times to peak twitch force/torque for some low-force motor units in tibialis anterior (Fig. 3B).

Enlarge table

Table 3. Average minimal and peak discharge rates of motor units in FDI and tibialis anterior obtained during different recording modalities

Enlarge table

Although the range of motor unit recruitment and time to peak twitch force/torque differ between FDI and tibialis anterior, the general distribution of properties within the motor unit pool (peak force and time to peak twitch force/torque) and the activation capabilities (discharge-rate range) appear to be relatively similar. The main difference between the two muscles is that force gradation is distributed over a greater number of motor units in tibialis anterior, which may improve the capacity of the muscle to increase force smoothly during graded contractions despite greater innervation numbers.

CONCLUSIONS AND FUTURE DIRECTIONS

The objective of our review was to compare the distribution of motor unit properties across human muscles that differ in size and recruitment range. We summarized what is known about the number of motor units in various muscles, the average innervation number for each muscle, the distribution of contractile characteristics within selected pools of motor units, and the range of their discharge rates during voluntary contractions. The major conclusion to emerge from our analysis is that despite a large difference in innervation number between a hand muscle (FDI) and a lower leg muscle (tibialis anterior), the distribution of contractile characteristics and the range of discharge rates within each pool appears to be relatively similar.

Despite a limited amount of data in humans, the results indicate that the force capacity of muscle is more directly related to average innervation number than it is to motor unit number. Indeed, there was no association between the number of motor units and the total number of muscle fibers within the muscle (Fig. 1A). Notably, there is only a ∼2.5-fold difference in the number of motor units between hand muscles (FDI, thenar muscles) and an arm muscle (biceps brachii). In contrast, we found a ∼20-fold difference in the number of muscle fibers between FDI and vastus lateralis (Table 2), despite a similar number of motor units (127 vs. 146, respectively) in the two muscles. Although there are limitations with the electrophysiological methods used to estimate the number of motor units in large muscles (56), this factor could not explain the large difference in average innervation number. This difference in average innervation number between these two muscles likely reflects the difference in function: hand muscles are mainly involved in tasks requiring dexterity, whereas the vastus lateralis contributes to the forces involved in locomotion.

The main observation from the comparison of the FDI and tibialis anterior is that despite a large difference in muscle size and thus in maximal force, the general distribution of their motor unit properties within each pool appears relatively similar. Although tibialis anterior has a slightly greater number of low-force motor units with longer times to peak force/torque (Fig. 3B) and that 50% of the motor unit pool in the FDI produced a cumulative force almost twice that of the tibialis anterior (Fig. 3D), the shape of the relation between the relative force and the time to peak force/torque of motor units is comparable for the two muscles. Consistent with this observation, the range of discharge rates is similar for the two muscles. The absence of major differences between these parameters implies a similar control of motor output during voluntary actions. This observation contrasts with the alternative view (116) that motor unit number plays a critical role in determining the precision of muscle force, but this discrepancy could be examined more systematically with computational models based on the data summarized in our review.

Another key point of our review is that morphological studies do not provide an adequate estimate of the number of motor units in a muscle and should be used with caution. Although cadaveric studies can be used to estimate both motor unit and innervation numbers in most human muscles, future studies should combine different techniques to achieve this goal. In parallel, improvements are needed in the electrophysiological methods as they offer the only way to track motor unit changes during aging or the progression of neuromuscular diseases (34, 35, 45). Three specific issues need attention: the bias toward sampling low-threshold units, phase cancellation when recording EMG signals, and the distortion of mechanical force associated with motor unit synchronization in the STA technique. We also need to characterize the shape of the relation between force and discharge rate for individual motor units in more muscles. The recent development of high-density multielectrode arrays to obtain surface EMG recordings should enable reliable tracking of many concurrently active motor units over a wide range of force and thereby provide more information on the force-frequency relation. Critically, more work is required to determine the functional consequences of differences in the distribution of motor unit properties across muscles.

GRANTS

No funding was received for this review article.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.D. and R.M.E. conceived and designed the review; J.D. and R.M.E. analyzed data; J.D. and R.M.E. interpreted results of experiments; J.D. prepared figures; J.D. and R.M.E. drafted manuscript; J.D. and R.M.E. edited and revised manuscript; J.D. and R.M.E. approved final version of manuscript.

Distribution of motor unit properties across human muscles

Jacques Duchateau and 

Roger M. Enoka

21 Dec 2021https://doi.org/10.1152/japplphysiol.00290.2021

• This is the final version - click for previous version

• Sections

• PDF(870 KB)
• Tools
• Share

Abstract

Download figureDownload PowerPoint

The purpose of our review was to compare the distribution of motor unit properties across human muscles of different sizes and recruitment ranges. Although motor units can be distinguished based on several different attributes, we focused on four key parameters that have a significant influence on the force produced by muscle during voluntary contractions: the number of motor units, average innervation number, the distributions of contractile characteristics, and discharge rates within motor unit pools. Despite relatively few publications on this topic, current data indicate that the most influential factor in the distribution of these motor unit properties between muscles is innervation number. Nonetheless, despite a fivefold difference in innervation number between a hand muscle (first dorsal interosseus) and a lower leg muscle (tibialis anterior), the general organization of their motor unit pools, and the range of discharge rates appear to be relatively similar. These observations provide foundational knowledge for studies on the control of movement and the changes that occur with aging and neurological disorders.

INTRODUCTION

Although the anatomical and functional organization of the motor output from the spinal cord has been investigated during the century since the pioneering studies of Sherrington (1, 2), there remains a lack of information regarding many of the details. These gaps in knowledge on the final common pathway constrain our understanding of the motor system.

Liddell and Sherrington (2) laid the foundation for this field by introducing the concept of a “motor unit,” which is defined as the entire motor neuron, its dendrites and axon, and the muscle fibers innervated by the axon (3). The technology to record the activity of single motor units in human muscles during voluntary contractions was first developed by Adrian and Bronk (4). Their pioneering studies demonstrated that the force exerted by a muscle during voluntary contractions was the result of the concurrent recruitment of motor units and modulation of the rate at which they discharge action potentials. Subsequently, based on observations in an animal preparation, Henneman (5, 6) proposed that the recruitment order of motor units was determined primarily by motor neuron size, which became known as the size principle. Since these key findings, a great amount of work has been done on characterizing motor unit function (for review papers, see Refs. 3, 7–11), but many basic questions remain unanswered.

Among these questions, the distribution of motor unit properties across human muscles and the consequences for the control of motor output are largely unknown. We know that the force produced by a single muscle depends on the number of active motor units, the number and size of the muscle fibers innervated by motor neurons (i.e., innervation number), the contractile characteristics of the motor units, and the frequency at which the motor neurons discharge action potentials (9, 12), but little is known about the variation in these properties across muscles. In both animals and humans, there is a wide difference in absolute size and force capacity of muscles (13–15), but also among the motor units in a given muscle (16–18). Within-pool differences in the contractile characteristics of muscle fibers also explain the wide ranges of mechanical properties of the motor units in the muscles of animals, with up to a 50- to 100-fold range in maximal forces and a fivefold range in contraction speed (13). In addition, the recruitment range of motor units during slow contractions differs between human muscles: the upper limit of recruitment is <60% of maximal voluntary contraction (MVC) force in some hand muscles (19–22) but extends up to 90% MVC force in most limb muscles and intrinsic foot muscles (18, 21–24).

This review focuses on the distribution of motor unit properties across a few human muscles that differ in size and recruitment range. Although the attributes of motor units can be distinguished on the basis of anatomical, contractile, and biochemical characteristics, we limit our discussion to four key parameters that influence the force produced by muscle during voluntary contractions: the number of motor units, average innervation number, and the distributions of contractile characteristics and discharge rates within motor unit pools. This basic knowledge can inform our interpretation of the significance of the adaptations in motor unit properties that occur during conditions that compromise the integrity of the neuromuscular system.

MOTOR UNIT NUMBER

Morphological Studies

One of the first studies to assess the number of motor units in human muscles was the one by Feinstein et al. (25). Their approach involved cadaveric estimates of the number of α axons innervating a muscle and measuring the size of myelinated axons in cross-sectional nerve slices. Based on a bimodal distribution observed for most muscles, they distinguished between small-diameter and large-diameter axons. The percentage of afferent and efferent axons was then estimated from the number of large-diameter axons based on observations from cat studies that ∼43% of the large axons in tibialis anterior and ∼36% in medial gastrocnemius were afferent fibers (26). These data were further supported by findings from a single patient who suffered from poliomyelitis and experienced an almost complete de-efferentation (Fex and Wohlfart, 1954 cited in 25). In this patient, 35%–40% of the large-diameter axons in the nerve supplying the medial gastrocnemius were afferent fibers. It was therefore assumed that ∼60% of the large-diameter axons in peripheral nerves were efferent fibers (25, 27). Based on these proportions, Feinstein et al. (25) reported that the number of motor axons, and thus motor units, ranged from 95 in the first lumbrical to 1,096 in the platysma. For example, they estimated 119 motor units in the first dorsal interosseus (FDI), 445 in tibialis anterior, 579 in medial gastrocnemius, and 332 in brachioradialis.

The major limitations of these findings are the relatively few specimens examined in each study (25, 28, 29) and the inability to distinguish between afferent and efferent fibers in human nerves. A major advance for the field occurred when Houser et al. (30) reported in 1983 that the enzyme choline acethyltransferase (ChAT) was exclusively expressed in efferent neurons and, more recently, by the discovery of a polyclonal anti-ChAT antibody capable of selectively labeling motor axons in peripheral nerves (31, 32). After applying ChAT and neurofilament immunofluorescence to nerves supplying the upper extremities, Gesslbauer et al. (31) quantified the total number of both efferent and afferent axons (myelinated and unmyelinated fibers) in nine organ donors whose hearts were still beating at the time of harvest. Their approach involved acquiring transverse images of whole nerve sections from the human specimens and analyzing the labeled images at predefined harvesting sites along the brachial plexus and the terminal nerves innervating the arm and hand. The images indicated that the brachial plexus receives contributions from the ventral rami of the four caudal cervical (C5-C8) and the first thoracic (T1) spinal nerves. At this level of the spinal cord, the average number of ventral axons, representing all the axons innervating the human arm, was found to be 350,000 ± 43,000. A major observation was that the proportion of motor axons (efferent fibers) in the brachial plexus was less than 10% (22,000 ± 3,000 motor axons) of the total number of axons.

To further understand the distribution of the nerve fibers projecting to upper limb muscles, Gesslbauer et al. (31) investigated the terminal nerves of the brachial plexus at several defined sites in the shoulder, arm, and hand. Specimens were harvested from the axillary, musculocutaneous, radial, median, and ulnar nerves. These data indicated that the axillary nerve, which originates from the posterior cord of the brachial plexus and supplies the three heads of the deltoid, teres minor, and occasionally the long portion of the triceps brachii, contained 20,389 ± 2,724 afferent axons and 2,108 ± 317 efferent axons (9.4 ± 1.2%). The musculocutaneous nerve, which innervates the elbow flexors, comprised 24,303 ± 5,865 afferent axons and 1,601 ± 164 efferent axons (6.4 ± 1.3%). A total of 61,353 ± 10,408 afferent axons and 4,343 ± 365 efferent axons were counted in the radial nerve (6.7 ± 1.0%), which is responsible for the arm, wrist, and finger extension, as well as forearm supination. The median nerve, supplying most of the flexor muscles of the hand and fingers, and pronator muscles, contained 60,523 ± 6,280 axons, of which 3,374 ± 388 were efferent axons (5.6 ± 0.7%). A lesser number of axons was counted in the ulnar nerve, which innervates some forearm flexors (flexor carpi ulnaris and medial half of the flexor digitorum profundus) and intrinsic hand muscles (hypothenar muscles; 3rd and 4th lumbricals, dorsal and palmar interossei, adductor pollicis, and deep head of flexor pollicis brevis). The ulnar nerve contained 40,379 ± 5,638 axons at its origin, of which 2,670 ± 347 were efferent axons (6.7 ± 0.8%).

Although the total number of efferent axons at different levels (origin, arm, and wrist) was also counted, a direct comparison of these results with those of Feinstein et al. (25) is not possible. Despite Feinstein et al. (25) being able to identify large-diameter fibers, which excluded γ motor fibers (muscle spindles) and small-diameter sensory fibers (e.g., pain, thermal), they were not able to distinguish between efferent and afferent fibers. In contrast, Gesslbauer et al. (31) were able to separate efferent and afferent fibers (myelinated and unmyelinated), but the myelinated axons of γ neurons were included in the count of efferent axons. As the relative proportion of α and γ fibers innervating each muscle in humans is uncertain (33), it is not possible to estimate the number of motor units from the study of Gesslbauer et al. (31). Moreover, the fact that the relative proportion of efferent and afferent fibers (60% vs. 40%) used by Feinstein et al. (25) was likely too high (<10% efferent fibers) as found by Gesslbauer et al. (31), the estimated number of motor axons and thus motor units was likely overestimated by Feinstein et al. (25). Without being able to dissociate α fibers from γ fibers, it is therefore not currently possible to estimate accurately the number of motor units per muscle in humans on the basis of the available morphological data.

A major step forward for the field would be the combination of histological approaches (25, 28, 29) with immunofluorescence techniques, such as the one used by Gesslbauer et al. (31), to examine the terminal section of the nerve just before it reached each target muscle.

Electrophysiological Studies

As an alternative approach to morphological estimates, several electrophysiological methods have been developed to determine the number of motor units in human muscles. Three main approaches have been developed (for reviews, see Refs. 34, 35). The principle of these methods is to compare the average value of a motor unit parameter (EMG amplitude or area; peak force) with the value for the entire muscle. The first method, introduced by McComas et al. (36), involved electrically elicited muscle contractions in which stimulus intensity applied to a motor nerve is increased gradually until a maximal response is reached. The assumption of the McComas method is that the small stepwise increments in response size to each stimulus represent the addition of a single motor unit action potential to the compound action potential (M wave). By comparing the ratio between the average surface EMG increments to that of the maximal M wave, an average number of motor units can then be estimated (36). The major limitation of this method is that each small increment of the compound muscle action potential likely does not indicate the activation of a single motor unit (37, 38).

The variability and overestimation in motor unit numbers with this method are likely due to the small sample size and the alternation of two or more motor axons with similar depolarization thresholds during graded stimulation of the whole muscle nerve (38–40). Despite technical refinement (35) and the use of multipoint stimulation along the course of the nerve (38) to reduce the problem of alternation, this method cannot be used in many muscles (e.g., biceps brachii, vastus lateralis, tibialis anterior) for which the proximal nerve is not accessible for stimulation at enough locations needed to collect an adequate sample of single motor unit action potentials.

The problem of alternation can be avoided by using a spike-triggered averaging (STA) method (37, 41). This method involves recording both the discharge times of single motor units with intramuscular electrodes and the concurrent EMG activity with surface electrodes. By triggering from the discharge times and averaging into the surface EMG, it is possible to estimate the average size of the surface-recorded motor unit potential (macro-EMG potential). The number of motor units can be estimated by dividing the maximal M wave by the average value of the surface-detected motor unit sample (28, 37). A potential problem with this method is a sampling bias toward low-threshold motor units when they are recorded during low-force contractions. To reduce this drawback and obtain a representative sample, it has been recommended to record motor unit discharges during moderate-intensity voluntary contractions. For example, the measurements are made during contractions at ∼25% of MVC force for tibialis anterior (42), but the optimal contraction intensity likely differs across muscles. Furthermore, to avoid the potential confounding influence of phase cancellation (43), the preferred approach is to analyze the amplitude or area of the negative peak of the surface-detected response (motor unit and M wave) instead of the whole peak-to-peak amplitude of the signal (34, 41).

Based on the same general methodological approach, alternative methods have been developed. These include extracting single motor unit discharges noninvasively from the F-wave response (44), using high-density surface EMG recordings to augment the number of identified motor units (45), mathematical models that simulate the M wave and surface EMG interference signals recorded at different forces to determine a motor unit number index (MUNIX) (46), and improving the McComas method by addressing more carefully motor unit alternation (47). These methods are not described in our review, but information can be obtained elsewhere (34, 47).

In addition to estimating the number of motor units based on the EMG recordings, Stein and Yang (40) proposed a similar approach based on the contractile properties of the motor units. Their method involves measuring the twitch force of single motor units, either by STA (48, 49) or by intramuscular stimulation of axonal branches (50), and whole muscle twitch force induced by maximal electrical stimulation of the motor nerve. As for the EMG methods, the estimated number of motor units is calculated by dividing muscle twitch force by the average force of a sample of single motor units. As it is technically challenging to obtain accurate measurements of the force produced by low-threshold motor units in big muscles, the method has been mainly used in hand, forearm, and one lower leg muscle (tibialis anterior).

Although these methods have contrasting sampling problems, Stein and Yang (40) observed good agreement between the estimated number of motor units for the STA and microstimulation methods in the thenar muscles: 135 and 122 based on EMG, and 130 and 116 based on the force for the two methods, respectively. The value obtained with the McComas method (36) was 170. There was no statistically significant difference in these estimates. These findings indicate that the estimated number of motor units is relatively similar regardless of the method (STA vs. microstimulation) and parameter (EMG vs. force) used, at least for the thenar muscles. These data also agree reasonably well with the average value (133 motor units) calculated from different electrophysiological studies published in the literature (Table 1).

Enlarge table

Table 1. Average motor unit number estimated from spike-triggered averaging of either EMG or force in different muscles

Enlarge table

Distribution across Muscles

In an attempt to compare the distribution of motor-unit properties in human muscles of different sizes, we examined muscles for which data from a minimum of three studies were available. Table 1 summarizes the number of motor units for five muscles. To increase the consistency of the results, we only included data recorded by the STA obtained with either EMG or force as the outcome measurement. When possible, we selected EMG values determined from the surface-detected amplitude of the negative phase (first phase) of the motor unit. As expected, these data show fewer motor units for hand muscles compared with limb muscles (Table 1). Among the limb muscles, biceps brachii displayed a greater number of motor units (312) compared with the tibialis anterior (200) and vastus lateralis (146). However, the coefficient of variation was greater for vastus lateralis and tibialis anterior compared with the other muscles. Relative to its size (Table 2), the estimated number of motor units in the vastus lateralis is low, likely due to limitations of the STA method in large muscles (56). No significant association was found between the average number of motor units and muscle anatomical cross-sectional area (CSA) (Fig. 1A) calculated for four muscles from data reported in Tables 1 and 2.

Figure 1.A: estimated numbers of motor units and muscle fibers for four human muscles. Each data point corresponds to the means ± SD for each muscle cited in both Table 1 and Table 2. The average number of fibers relative to the muscle anatomical cross-sectional area (CSA) has been corrected for each muscle by the specific ratio between physiological and anatomical CSA (see Comparison across Muscles for the procedure). The graph indicates the absence of a significant association between the two variables. B: relation between physiological cross-sectional area (PCSA) and innervation number in the same four muscles as in A. Each data point corresponds to the means ± SD for the studies cited in Table 2. The innervation number has been calculated for each muscle as the ratio between the average muscle fibers relative to the physiological CSA (PCSA) and the average motor unit number reported in Table 1. The graph shows a significant correlation (r = 0.988; P < 0.011) between the two variables. BB, biceps brachii; FDI, first dorsal interosseus; TA, tibialis anterior; VL, vastus lateralis.

Download figureDownload PowerPoint

Enlarge table

Table 2. Characteristics of selected muscles and associated muscle fibers

Enlarge table

In conclusion, except for hand muscles, there is considerable variability in the estimated numbers of motor units across studies for a given muscle and additional data are needed to determine with more certainty the number of motor units in human muscles. Although electrophysiological methods appear to provide a relatively consistent estimate for hand muscles, the results are more variable for larger muscles. As many muscles are not accessible to electrophysiological approaches, it seems that only the results from morphological methods can provide a more objective and complete view of the number of motor units within a motor unit pool. Nonetheless, it is necessary to improve electrophysiological methods as they offer the only viable approach to track motor neuron loss during aging or the progression of diseases, such as amyotrophic lateral sclerosis and spinal muscular atrophy (34, 35, 45).

AVERAGE INNERVATION NUMBER

A comparison of the distribution of motor unit properties across muscles requires information about both the average number of motor units within the pool and the number of muscle fibers innervated by each motor neuron. Findings from animal studies indicate that there is a strong correlation between the innervation number and the maximal tetanic force of a motor unit (76–78). The most direct approach used to determine innervation number in such studies is the glycogen-depletion technique (79). Briefly, the method involves depleting the glycogen in muscle fibers that belong to a single motor unit by prolonged electrical stimulation of its motor axon and then using classic histochemical techniques to count the number of muscle fibers in cross-sectional slices (76).

Such studies have shown substantial variation in the innervation number between motor units in a given muscle. In the cat and rat tibialis anterior muscles, for example, strong and fast contracting motor units (type FF) innervate an average of 2.7 (76) and 4.1 (77) times the number of muscle fibers compared with weak and slow-contracting motor units (type S). Due to the lesser variation in specific force (maximal tetanic force per unit of CSA) and muscle fiber diameter between motor units, these findings indicate that maximal motor unit force mainly depends on innervation number (77, 78).

Estimation from Cadaveric Measurements

The experimental approaches used in animal studies cannot be translated to humans and other methods are required. One approach is based on counting the total number of muscle fibers in a transverse section of a muscle from cadaver specimens. With this approach, Feinstein et al. (25) found that the average number of muscle fibers varied from ∼10,000 fibers in the first lumbrical to ∼1,100,000 fibers in medial gastrocnemius. By computing the ratio between the total number of muscle fibers and the estimated number of motor axons supplying the muscle, the average innervation number can be estimated. The average innervation number varies across muscles, ranging from ∼100 for the first lumbrical up to 2,000 for medial gastrocnemius (25).

In addition to the uncertainty about the exact number of motor axons as described in the preceding section, other limitations of this method are that not all muscle fibers extend along the length of a muscle, and differences in fiber CSA are not taken into account. Subsequent studies analyzed muscle slices after staining for mATPase to visualize type I and type II fibers (80). A total number of fibers in the muscle was counted as well as the fiber-type proportions for the muscle (81). Based on this method, the estimated total number of muscle fibers was ∼500,000 in vastus lateralis (67, 69) and ∼140,000 in tibialis anterior (72–74) (Table 2). These numbers varied between sexes with significantly fewer muscle fibers (∼20%) in the tibialis anterior of women (73) compared with men (72). As the number of motor axons was not quantified in these studies, it is not possible to estimate the average innervation number from these data.

Estimation from Muscle Biopsies

An alternative approach to cadaveric measurements is to determine the average CSA of individual muscle fibers obtained from muscle biopsy samples in humans and to measure whole muscle CSA with an imaging technique. After correcting for connective tissue content and assuming that most muscle fibers extend along the whole muscle length, the total number of muscle fibers was estimated by dividing the whole muscle anatomical CSA (i.e., area of the cross section perpendicular to muscle axis) by average muscle-fiber CSA (Table 2). With this approach, Sale et al. (65) obtained average values of ∼200,000 muscle fibers in biceps brachii of men and ∼140,000 for women. Similarly, Miller et al. (14) estimated a slightly greater number (not significant) of muscle fibers in biceps brachii of men (180,620) than in women (156,872), but not in vastus lateralis (451,468 in men vs. 465,007 in women). The data for vastus lateralis are quite similar to the values obtained from cadaveric specimens (∼500,000; 67, 69).

Unfortunately, no direct estimation of the number of motor axons was performed in these studies. However, Miller et al. (14) did use the McComas method (36) to estimate the number of motor units in the two muscles; they found 126 motor units in biceps brachii for men and 110 for women, representing average innervation numbers of ∼1,433 fibers per motor unit for men and 1,426 for women. The values for vastus lateralis were 282 for men and 229 for women, resulting in average innervation numbers of 1,600 and 2,030 fibers per motor unit, respectively. The average number of motor units obtained with the other electrophysiological methods yielded 312 units for biceps brachii and 146 for vastus lateralis (Table 1) suggesting that Miller et al. (14) may have underestimated the innervation number for biceps brachii but overestimated it for vastus lateralis.

Comparison across Muscles

As emphasized in the two preceding paragraphs, there are few data on innervation number in human muscles. To increase sample size and thereby approximate innervation number with more confidence, we first calculated the ratio between whole muscle anatomical CSA and average muscle-fiber CSA to estimate the number of muscles fibers (Table 2). However, the number of muscle fibers is likely underestimated in pennated and big muscles (tibialis anterior and vastus lateralis) with this approach because most muscle fibers do not extend along the length of the entire muscle. As these studies did not report muscle volume and the associated architectural parameters (pennation angle and fascicle length), it was not possible to express these values relative to the physiological CSA (i.e., area of the cross section perpendicular to the muscle fibers; 75) and to determine innervation numbers more accurately.

As an alternative approach, we multiplied the total number of muscle fibers in each muscle by the ratio between the physiological CSA and the maximal anatomical CSA reported in the literature: 1.1 for FDI (62) and 1.9 for tibialis anterior (75). This ratio was calculated for vastus lateralis from the average anatomical CSA value of Table 2 and the physiological CSA reported by Marzilger et al. (82). The resulting physiological CSA was ∼2 times greater than the anatomical CSA for vastus lateralis. Given that pennation angle for biceps brachii is close to zero (83) and fascicle length is difficult to measure precisely in vivo, we used a ratio of 1 for biceps brachii, which may have been underestimated slightly the true value. Despite some limitations with this approach (data from cadavers and in vivo measurements were mixed and the ratios were calculated slightly differently between studies), the estimated average innervation numbers were 346 for the FDI, 622 for biceps brachii, 1,845 for tibialis anterior, and 5,990 for vastus lateralis. These data indicate that innervation number was least for FDI and greatest for vastus lateralis. As illustrated in Fig. 1B, average innervation number was highly correlated (r = 0.988; P = 0.011) with average physiological CSA across the four muscles. Collectively, these data indicate that the force capacity of muscle (physiological CSA) is more related to the average innervation number than to the number of motor units in the pool.

DISTRIBUTIONS WITHIN A MOTOR UNIT POOL

Although the data in the preceding section are useful for an overall comparison between muscles, it is the range of innervation numbers within an individual motor unit pool that is more critical for understanding how the nervous system controls muscle force. In animal studies, the glycogen-depletion method has indicated that the innervation number varies exponentially within a motor unit pool, with low-threshold motor units containing lower values (76, 84). For example, the innervation number ranges from 18 to 830 in the cat medial gastrocnemius muscle, corresponding to a 50-fold range (85).

Contractile Characteristics

The range of innervation numbers in motor unit pools of humans can be approximated indirectly from measures of motor unit force based on the strong correlation between the innervation number and maximal motor unit force (76–78). Using such an approach, Enoka and Fuglevand (9) determined the range of tetanic forces and estimated the number of muscle fibers required to achieve these forces in the FDI. Based on the data provided by Feinstein et al. (25) for this muscle (120 motor units and 40,500 muscle fibers), they estimated that the innervation number ranged from 21 to 1,770, representing an 84-fold range. These data further indicated that the frequency distribution of the innervation numbers within a motor unit pool is typically characterized by an exponential function with many units generating small forces, whereas relatively few motor units exert large forces. Although the FDI contains a relatively similar percentage of type I and II fibers (86), the skewed distribution of innervation numbers resulted in ∼84% of the motor units comprising type I muscle fibers and 16% containing type II fibers (9). The key point of these estimates is the finding of a nonlinear relation between innervation number and motor unit size, as shown in animal studies (87).

These results raise the question of whether the distribution of motor unit properties is similar for muscles that differ in size, contractile characteristics, recruitment range, and function. Due to technical difficulties, there is relatively little information on the range of contractile characteristics of single motor units of humans. Three main methods have been used to assess these characteristics: intramuscular microstimulation (50), intraneural microstimulation (16, 88), and STA methods (48, 49). Each of these methods has its own advantages and limitations (89). The first two approaches involve electrical stimulation of a single motor axon with a microelectrode inserted into either a muscle or a nerve trunk and recording the contractile properties of isolated motor units in response to a single stimulus (twitch) and various trains of stimuli (tetanus). This method can be used to determine the force-frequency relation and the maximal tetanic force of motor units. In contrast, the STA method is performed during a sustained voluntary contraction and involves recording the intramuscular action potential of a single motor unit while discharging at a low steady rate and averaging into the force signal to estimate the mechanical contribution of the selected unit to the whole muscle force (49). The STA method can be used to determine recruitment threshold, twitch force, and time to peak twitch force for individual motor units.

Although a more thorough characterization of motor unit contractile properties can be obtained with electrical microstimulation, the amount of slack during the recording procedure may underestimate twitch forces. In contrast, summation of twitch responses and possible synchronized activity of concurrent active units may distort the motor unit twitches with the STA method, but the voluntary contraction attenuates the confounding influence of tissue compliance on the evoked responses. Among these three experimental approaches, only the STA method has been used in intrinsic hand muscles (FDI, adductor pollicis, thenar muscles; 20, 40, 48, 90), forearm muscles (wrist extensors; 91), and leg muscles (tibialis anterior; 18, 92). We, therefore, compare the contractile properties and discharge rates of motor units in two of these muscles, FDI and tibialis anterior, for which we had sufficient data acquired in similar experimental conditions from the same laboratory. Our analysis of the literature indicates that tibialis anterior has an average innervation number approximately 5-fold larger than that of FDI (Table 2; Fig. 1B).

The distributions of the recruitment threshold for these two muscles are illustrated in Fig. 2. The range of recruitment thresholds is much narrower for FDI than for tibialis anterior: it ranges from 0.4 to ∼35% MVC force (sample: 44 units) for FDI and from 1 to ∼90% MVC force (sample: 528 units) for tibialis anterior. Relative to the total sample, 50% of the units have recruitment thresholds of less than 10% MVC force in the FDI and 20% MVC force in tibialis anterior. Furthermore, the relative number of motor units in the 0%–5% MVC range for FDI is about twice that of the tibialis anterior, representing ∼37% and 18% of the total sample for these two muscles. The peak twitch force ranged from 11.8 to 170 mN with an average value of 62 ± 46 mN for the FDI (20), whereas peak twitch torque ranged from 1 to 171 mN·m with an average value of 70 ± 42 mN·m for the tibialis anterior (18). A relatively comparable distribution for the time to peak twitch force/torque was found for the two muscles: values ranged from 25 to 60 ms (mean: 39 ± 10 ms) for FDI and from 20 to 86 ms (mean: 46 ± 14 ms) for tibialis anterior.

Figure 2.Distribution of recruitment thresholds for 528 motor units (MUs) in the tibialis anterior (black bars) and 44 motor units in the FDI (white bars) recorded during slow ramp contractions. These experimental data are from Van Cutsem et al. (18) and Duchateau and Hainaut (20) with permission from John Wiley and Sons. FDI, first dorsal interosseus.

Download figureDownload PowerPoint

A more complete view of the motor unit pool for the two muscles can be obtained by assigning the measured mechanical properties (peak twitch force and time to peak twitch force) in a computational model (93). To that end, a pool of 100 and 200 motor units was determined for the FDI and tibialis anterior muscles, respectively, based on experimental findings (Table 1). As illustrated in Fig. 3A, peak twitch force for the motor units increases exponentially in both muscles with a rate constant of 0.045 for FDI and 0.029 for tibialis anterior, when expressed as the rank occupied (in %) by the motor units in each pool. However, low-threshold motor units develop proportionally more force in FDI than in tibialis anterior (Fig. 3A). For example, 50% of the motor unit pool produced a cumulative force of ∼18% in FDI, but only ∼10% in tibialis anterior (Fig. 3D).

Figure 3.Distribution of motor unit twitch forces in a model that comprised a pool of 200 units for tibialis anterior (TA) and 100 units for first dorsal interosseus (FDI). Motor unit forces, recorded by spike-triggered averaging (STA), are from Van Cutsem et al. (18) for TA and Duchateau and Hainaut (20) for FDI. A: motor unit (MU) force, expressed as a percentage of the greatest twitch force in the pool, is plotted as a function of the rank (in %) occupied by the unit in the pool. An exponential (e) growth function best fit the experimental data (r2 = 0.98 for TA and 0.97 for FDI). The equations are Y = 1.11·e(0.045·X) for TA and Y = 4.45·e(0.029·X) for FDI. B: motor unit (MU) force, expressed as a percent of the greatest twitch force in the pool, is plotted as a function of its time to peak force/torque (ms). A one-phase exponential decay best fit the data (r2 = 0.99 for TA and 0.97 for FDI). The equations are Y = 1,535·e(−0.117·X) + 3.5 for TA and Y = 1,548·e(−0.1135·X) + 3.4 for FDI. C and D: cumulative sum of MU forces, expressed as % of maximum, as a function of the rank occupied by the unit in the pool expressed either in absolute value (C) or in percentage (D). An exponential growth function fits these data (r2 ≥ 0.98): C equations were Y = 1.11·e(0.0225·X) for TA and Y = 3.1·e(0.035·X) for FDI, and the D equations were Y = 0.93·e(0.047·X) for TA and Y = 3.1·e(0.035·X) for FDI.

Download figureDownload PowerPoint

The shape of the association between peak force (expressed in % of the greatest motor unit peak force within the pool) and the time to peak force/torque (ms) was similar for the two muscles (rate constant: −0.114 for FDI and −0.117 for tibialis anterior; Fig. 3B). A key difference between the two muscles, however, was the longer time to peak twitch force/torque (≥ 60 ms) of many low-force units in tibialis anterior, presumably due to a greater amount of slack and compliance of the series elastic component of the dorsiflexor muscles (94) compared with FDI. Interestingly, when cumulative motor unit force is expressed as a function of the absolute number of motor units in the pool (Fig. 3C), the lower rate constant in tibialis anterior (0.022) compared with FDI (0.035) indicates a slower rate of force gradation (about half) due to a greater recruitment range. When cumulative motor unit force is expressed as a percentage of the total motor unit pool (Fig. 3D), the shape of the relation for these two muscles is quite similar, suggesting that differences in force control between the two muscles depend on their relative discharge rate characteristics.

Discharge Rate Characteristics

In addition to the number and size of the fibers innervated by a motor neuron, the rate at which action potentials are discharged (rate coding) has a substantial influence on the force that a motor unit can produce (95). A sigmoidal association between the increase in stimulation frequency and force is classically reported for individual motor units (16, 17, 96). When force increases during a voluntary contraction, three phases have been often described: a first phase during which rate coding increases rapidly, a second phase characterized by a more gradual linear increase in discharge rate, and finally, a sharp increase in rate when the contraction approaches maximal force (97–99).

This general force-frequency relation seems to vary among motor units. In some units, rate coding tends to plateau (saturation effect) after the first phase despite a continual increase in force (22, 100, 101). This effect has been attributed, at least partly, to intrinsic mechanisms that limit the capacity of some motor neurons to increase their discharge rate despite continual increases in the net excitatory synaptic input (11, 95, 101). However, there are currently not enough published data on the force-frequency relation for single motor units in either FDI or tibialis anterior to determine that relation with enough confidence.

In the absence of these data, rate coding is often characterized in terms of the minimal and peak discharge rates. The association between minimal discharge rate and recruitment threshold appears variable and there is no consensus on whether the last recruited (high threshold) motor units achieve the greater peak discharge rates during gradual increases in the applied force. Some studies reported that peak discharge rate recorded with intramuscular EMG reaches greater values for high- than low-threshold motor units (22, 97, 100, 102–104), whereas other studies reported the opposite (18–20, 105, 106) or similar values for the whole pool of motor units (99). Even the recent decomposition methods of multichannel recordings to identify the discharge times of many concurrently active motor units have not clarified this issue (24, 107, 108). Some of this uncertainty may be attributed to differences between muscles (19, 109) and the experimental tasks, such as a slow isometric ramp or trapezoidal contractions (18–20, 99), step contractions to different target forces (22, 110), sustained MVC (63, 98, 108, 111), and also differences in muscle-tendon length (94).

Given the variability in the shape of the force-frequency relation across motor units within a pool, we compared the average minimal and peak discharge rates for the FDI and tibialis anterior. Surprisingly, no difference was observed between these muscles for either the minimal or peak discharge, regardless of contraction mode (Table 3). Average minimal discharge rates at the recruitment threshold are 8.2 for FDI and 8.6 pps (pulses per second) for tibialis anterior. Although the variability across studies is greater for peak discharge rate, the average values are 32.4 for FDI and 29.3 pps for tibialis anterior. This indicates that the range of discharge rate modulation is relatively similar (∼21–24 pps;Table 3) between these two muscles, despite longer times to peak twitch force/torque for some low-force motor units in tibialis anterior (Fig. 3B).

Enlarge table

Table 3. Average minimal and peak discharge rates of motor units in FDI and tibialis anterior obtained during different recording modalities

Enlarge table

Although the range of motor unit recruitment and time to peak twitch force/torque differ between FDI and tibialis anterior, the general distribution of properties within the motor unit pool (peak force and time to peak twitch force/torque) and the activation capabilities (discharge-rate range) appear to be relatively similar. The main difference between the two muscles is that force gradation is distributed over a greater number of motor units in tibialis anterior, which may improve the capacity of the muscle to increase force smoothly during graded contractions despite greater innervation numbers.

CONCLUSIONS AND FUTURE DIRECTIONS

The objective of our review was to compare the distribution of motor unit properties across human muscles that differ in size and recruitment range. We summarized what is known about the number of motor units in various muscles, the average innervation number for each muscle, the distribution of contractile characteristics within selected pools of motor units, and the range of their discharge rates during voluntary contractions. The major conclusion to emerge from our analysis is that despite a large difference in innervation number between a hand muscle (FDI) and a lower leg muscle (tibialis anterior), the distribution of contractile characteristics and the range of discharge rates within each pool appears to be relatively similar.

Despite a limited amount of data in humans, the results indicate that the force capacity of muscle is more directly related to average innervation number than it is to motor unit number. Indeed, there was no association between the number of motor units and the total number of muscle fibers within the muscle (Fig. 1A). Notably, there is only a ∼2.5-fold difference in the number of motor units between hand muscles (FDI, thenar muscles) and an arm muscle (biceps brachii). In contrast, we found a ∼20-fold difference in the number of muscle fibers between FDI and vastus lateralis (Table 2), despite a similar number of motor units (127 vs. 146, respectively) in the two muscles. Although there are limitations with the electrophysiological methods used to estimate the number of motor units in large muscles (56), this factor could not explain the large difference in average innervation number. This difference in average innervation number between these two muscles likely reflects the difference in function: hand muscles are mainly involved in tasks requiring dexterity, whereas the vastus lateralis contributes to the forces involved in locomotion.

The main observation from the comparison of the FDI and tibialis anterior is that despite a large difference in muscle size and thus in maximal force, the general distribution of their motor unit properties within each pool appears relatively similar. Although tibialis anterior has a slightly greater number of low-force motor units with longer times to peak force/torque (Fig. 3B) and that 50% of the motor unit pool in the FDI produced a cumulative force almost twice that of the tibialis anterior (Fig. 3D), the shape of the relation between the relative force and the time to peak force/torque of motor units is comparable for the two muscles. Consistent with this observation, the range of discharge rates is similar for the two muscles. The absence of major differences between these parameters implies a similar control of motor output during voluntary actions. This observation contrasts with the alternative view (116) that motor unit number plays a critical role in determining the precision of muscle force, but this discrepancy could be examined more systematically with computational models based on the data summarized in our review.

Another key point of our review is that morphological studies do not provide an adequate estimate of the number of motor units in a muscle and should be used with caution. Although cadaveric studies can be used to estimate both motor unit and innervation numbers in most human muscles, future studies should combine different techniques to achieve this goal. In parallel, improvements are needed in the electrophysiological methods as they offer the only way to track motor unit changes during aging or the progression of neuromuscular diseases (34, 35, 45). Three specific issues need attention: the bias toward sampling low-threshold units, phase cancellation when recording EMG signals, and the distortion of mechanical force associated with motor unit synchronization in the STA technique. We also need to characterize the shape of the relation between force and discharge rate for individual motor units in more muscles. The recent development of high-density multielectrode arrays to obtain surface EMG recordings should enable reliable tracking of many concurrently active motor units over a wide range of force and thereby provide more information on the force-frequency relation. Critically, more work is required to determine the functional consequences of differences in the distribution of motor unit properties across muscles.

GRANTS

No funding was received for this review article.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.D. and R.M.E. conceived and designed the review; J.D. and R.M.E. analyzed data; J.D. and R.M.E. interpreted results of experiments; J.D. prepared figures; J.D. and R.M.E. drafted manuscript; J.D. and R.M.E. edited and revised manuscript; J.D. and R.M.E. approved final version of manuscript.