Introduction

Skeletal muscles undergo large changes in size and structure during childhood development. Transverse growth of the whole muscle, i.e. growth in the plane perpendicular to muscle fibres, must be accompanied by generation of new fibres (hyperplasia), transverse growth of existing fibres (fibre hypertrophy), or both. In rats and mice, postnatal muscle growth appears to originate almost exclusively from fibre hypertrophy and not hyperplasia¹, but the mechanisms of muscle growth in humans remain uncertain.

In this study we used diffusion tensor imaging (DTI) to study macroscopic and microscopic features of skeletal muscle in children aged 5 to 17 years. We estimated fibre numbers from DTI-based measurements of physiological cross-sectional area² (PCSA; a macroscopic muscle property equal to the summed cross-sectional areas of all fibres in a muscle) and mean fibre cross-sectional area (FCSA; estimated from measurements of diffusion rates and diffusion simulations^3,4). We hypothesised that human muscles grow exclusively through hypertrophy, not hyperplasia, so that fibre number does not change with age.

Methods

We re-analysed previously published data on muscle PCSA, mean axial diffusivity (λ₁), radial diffusivity (λ₂ and λ₃) and fractional anisotropy (FA) obtained with mDixon and DTI scans of the lower legs of 39 children². Data were available for 59 medial gastrocnemius muscles: both sides of 20 typically developing children and the less-affected side of 19 children with unilateral spastic cerebral palsy.

DTI scans of the lower legs were obtained on a 3T Philips Achieva TX with the following settings: DT-EPI with spectral pre-saturation with inversion recovery (SPIR) fat suppression, TR/TE 8715/63 msec, field of view 180x180 mm, slice thickness 5 mm, 16 gradient directions on a hemisphere, NSA 4, b=500 s/mm2 (b₀ image with b=0 s/mm²), diffusion gradient time Δ/δ = 30.4/8.2 msec and scan time 568 sec.

Muscle volume was derived from manual segmentation of the medial gastrocnemius muscles on mDixon scans. Using anatomically constrained DTI tractography⁵, fascicle length was measured for, on average, 3,750 fascicles per muscle (details in ref 2). PCSA was calculated by dividing muscle volume by median fascicle length. Diffusion properties were calculated per fibre tract by interpolating maps of diffusion properties at the fibre tract points. The diffusion properties of a muscle were defined as the median value of all tracts in that muscle.

To determine the relationship between fractional anisotropy and FCSA, we used random walk simulations of the restricted diffusion process in skeletal muscle^3,4. A total of 20,000 walkers were placed on a two-dimensional square Voronoi diagram⁶ resembling a cross-section of muscle tissue with 127 cells (muscle fibres; Fig. 1a). The coefficient of variation of FCSA (Voronoi cell area) was 18%. Walkers took 2,000 steps in randomly chosen directions with a step size r=√(4D₀Δ/2000) , where D₀, the unrestricted diffusion coefficient, was set as the mean λ₁ in our dataset of 1.82 μm²/msec and Δ, the total diffusion time, was set at 30.4 msec, corresponding to the diffusion time of the DTI acquisition. Walkers attempting to cross cell membranes succeeded with a probability of 2rκ/D₀ where κ, the permeability of muscle cell membranes, was set at 0.018 μm/msec ⁷. λ₂ and λ₃ were derived by eigenvalue decomposition of the diffusion tensor (covariance of walker displacements divided by 2Δ). FA was calculated using λ₁=1.82 μm²/msec. Simulations were done with different scale factors for the Voronoi diagram so that the mean FCSA ranged from 79 to 7854 μm², equivalent to the areas of cylindrical fibres with diameters ranging from 10 to 100 μm.

The simulated FA-FCSA relationship was used to estimate FCSA from DTI-measured FA values, after which fibre number (= PCSA/FCSA) was calculated. Linear mixed models with participants as random factors were used to regress PCSA against diffusion properties, FCSA and fibre number. We also regressed age against FCSA and fibre number.

Results

Children’s muscles with larger PCSAs had higher radial diffusion rates (λ₂ and λ₃), higher mean diffusivity, and lower fractional anisotropy (Fig. 2), indicative of larger FCSAs. There was no effect of PCSA on axial diffusivity (λ₁). The simulated relationship between FCSA and FA (Fig. 1b) was used to estimate mean FCSA for all muscles (Fig. 3 and 4).

Muscles with larger PCSAs also had larger mean FCSAs (Fig. 3a) and more muscle fibres. Compared to a muscle with a PCSA of 20 cm², a muscle with a 3× larger PCSA (60 cm²) was estimated to have 1.5× larger FCSA (661 vs 1,016 μm²) and twice as many fibres (3.1·10⁶ vs 6.4·10⁶). Fibre number also increased with age (Fig. 4).

Discussion

We describe the use of DTI to study fundamental mechanisms of muscle growth during childhood development. Our findings suggest that growth in human muscles originates both from increases in fibre cross-sectional areas and addition of new fibres.

The FA-FCSA relationship derived here corresponds well to findings of another simulation study⁴ (Fig. 1b), but further validation is required. Cadaver measurements of FCSA in gastrocnemius muscles of seven children⁸ (Fig. 4A) agree reasonably well with our data in younger muscles (5-6 years), but older muscles had larger diameters than we estimated. Future studies could estimate fibre diameters more accurately using time-dependent DTI⁹ and simulations which include the extracellular matrix^4,10.

Acknowledgements

The authors acknowledge the facilities and scientific and technical assistance of NeuRA Imaging, a node of the National Imaging Facility, a National Collaborative Research Infrastructure Strategy (NCRIS) capability. The study was supported by the Australian National Health and Medical Research Council (NHMRC; Program Grant APP1055084).