Skeletal Muscle Mitochondrial Function Is Associated with Longitudinal Growth Velocity in Children and Adolescents

Abstract

Context:

Periods of rapid growth require an increase in energy use and substrate formation. Mitochondrial function contributes to each of these and therefore may play a role in longitudinal growth.

Methods:

Twenty-nine children and adolescents of ages 8–15 yr were enrolled in a comprehensive longitudinal assessment of glucose homeostasis and mitochondrial function. Fasting laboratory studies and an estimate of mitochondrial function (as assessed by the time to recovery of phosphocreatine (PCr) concentration after submaximal quadriceps extension/flexion exercise using ³¹P magnetic resonance spectroscopy) were obtained at baseline and annually for 2 yr.

Results:

Data were complete for 23 subjects. Subjects were 11.3 ± 1.9 (sd) yr old at the beginning of the study; 61% were male. Average annualized growth velocity at 1 yr for boys was 7.1 ± 1.5 cm/yr and for girls 6.5 ± 1.7 cm/yr. More rapid recovery of PCr concentration, suggestive of greater skeletal muscle oxidative phosphorylation capacity at baseline, was associated with faster growth velocity in the subsequent year (r² = 0.29; P = 0.008). In multivariate modeling, baseline mitochondrial function remained significantly and independently associated with growth (R² for model = 0.51; P = 0.05 for effect of phosphocreatine recovery time constant), controlling for age, gender, Tanner stage, body mass index Z-score, and height Z-score.

Conclusions:

We report a novel association between time to recovery of PCr concentration after submaximal exercise and faster annual linear growth in healthy children. Future studies are needed to determine the physiological mechanisms and clinical consequences of this observation.

Periods of rapid growth require an increase in energy use and substrate formation. Mitochondrial function both depends on and contributes to each of these (1) and therefore may be related to longitudinal growth velocity. To our knowledge, there are no studies in healthy children and adolescents that relate mitochondrial function to linear growth or to GH action.

Mitochondrial disease is associated with poor growth (2). In a retrospective study of 24 individuals with clear evidence of mitochondrial dysfunction, average height was 1.97 sd score (SDS) below the mean, and short stature was stable from childhood to adulthood, in contrast to body mass index (BMI) SDS, which continued to decline over that time (3). A separate study reported similar findings with respect to short stature in a cohort of 53 patients with mitochondrial disease (4). Short stature was also present in a cohort of patients with Barth syndrome, an X-linked disorder caused by mutations in the taffazin gene that produce, among other deficits, defects in mitochondrial function (5).

In healthy children and adolescents, pubertal increases in GH/IGF-I system activity may affect mitochondrial function. GH has been demonstrated to affect mitochondrial function, both in vitro and, more recently, in vivo as well, although the extent and mechanisms of action are not yet well understood (6–8).

The objective of the present study was to characterize the relationship between linear growth velocity and an in vivo assessment of skeletal muscle mitochondrial function as assessed by the time to recovery of phosphocreatine (PCr) concentration (reflecting ATP regeneration) after submaximal exercise in a cohort of healthy prepubertal and early- to midpubertal children and adolescents.

Materials and Methods

Study design

The current study was approved by the Partners, Massachusetts General Hospital (MGH), and Massachusetts Institute of Technology (MIT) Institutional Review Boards. Written informed consent or parental consent and participant assent were obtained from all participants. The investigation was conducted according to the principles of the Declaration of Helsinki. Participants were recruited by local advertisements, from pediatricians in the community, and from the obesity programs at MGH and the Children's Hospital Boston from June 2007 through April 2009. All participants were healthy individuals between the ages of 8 and 15 yr old, inclusive, without contributory chronic medical illness, without a history of smoking, and without medical condition, syndrome, or medication known to be related to obesity status. Participants were excluded for use of medications known to alter glucose metabolism and a personal or family history of diabetes mellitus in a first-degree relative. Overweight status was determined by BMI percentile as 85th percentile or greater for age and sex based on Centers for Disease Control 2000 growth charts. Overweight subjects were overrepresented in this longitudinal study of glucose homeostasis and mitochondrial function. Waist circumferences percentiles were calculated by imputation using the National Health and Nutrition Examination Survey III data (9). A small number (n = 3 pairs) of siblings were permitted to enroll and included in the final analysis. Similar results were obtained when analyses were performed with data from only one participant per family (data not shown). Twenty-nine subjects were initially evaluated at baseline; three of these had initial magnetic resonance imaging (MRI) scans that were not usable, and an additional three were lost to follow-up at the 1-yr time point. As a result, 23 subjects were included at the baseline time point and had 1-yr follow-up visits; all 23 had usable initial MRI scans, and 21 of them also had a usable MRI scan at the 1-yr time point. Of these 21 subjects, four were lost to follow-up after the 12-month visit, seven had an 18-month visit only, and 10 had both 18- and 24-month visits with height measurements. If subjects had an 18-month visit but not a 24-month visit, annualized growth velocity from yr 1 to yr 2 was calculated based on the 18-month visit (seven subjects).

Baseline data, but not longitudinal data, from this cohort have been published previously (10, 11). Participants underwent a baseline evaluation, including a complete medical and family history; physical examination, including Tanner staging by a pediatric endocrinologist (A.F.); blood pressure by oscillometry; dietary and exercise questionnaires; and baseline laboratory testing. Fasting laboratory testing was performed during midmorning for all subjects. In addition, glucose and insulin values were obtained at baseline under fasting conditions and every 30 min for 120 min after a 1.75 g/kg (up to a maximum of 75 g) oral glucose load. ³¹P magnetic resonance spectroscopy (MRS) studies to evaluate mitochondrial function via PCr recovery rate were obtained as described in detail below. Exercise questionnaires were completed. Anthropometric measurements were performed and dietary records were reviewed by nutrition staff. Standing height was measured three times, and the median value was taken. Height, weight, and BMI Z-scores were calculated using Centers for Disease Control 2000 growth charts. All baseline measures were repeated annually, with fasting laboratory studies and anthropometric measurements performed at interim visits every 6 months, for up to 2 yr. Annualized growth velocity was calculated for each participant from baseline to 1 yr, and 1 yr to 2 yr.

Homeostasis model assessment for insulin resistance (HOMA-IR) was calculated for each participant (12). HOMA-IR measures basal insulin resistance and in children correlates well with the gold standard hyperinsulinemic-euglycemic clamp (13). The whole-body index of insulin sensitivity (WBISI) was also measured. This index is defined as 10,000/[square root of (fasting glucose × fasting insulin) × (mean glucose × mean insulin during OGTT)] and is closely associated with the rate of glucose disposal as assessed by euglycemic clamp (14).

³¹P MRS protocol

We used ³¹P MRS to measure the time to PCr recovery at baseline and annually for up to 2 yr, which provides a noninvasive measure of mitochondrial function. Prolonged time to PCr recovery implies relatively decreased mitochondrial oxidative function (15, 16). Mitochondrial function was determined using ³¹P MRS to assess PCr resynthesis after exercise as previously reported (10, 17). ³¹P MRS was performed at the MGH/MIT Athinoula A. Martinos Center for Biomedical Imaging using a 60-cm bore, Siemens 3.0T Tim Trio System with a ³¹P operating frequency of 49.879 MHz. Scans were performed in the late afternoon, and participants were instructed to refrain from caffeine intake and vigorous physical activity during the afternoon before the study. Maximum voluntary contraction (MVC) for the quadriceps was determined before commencement of the exercise protocol, based on the maximum weight that could be lifted. The determination of MVC was performed using a lifted lever inside the bore of the MRI scanner. The extension exercise was identical to that used for the 3 min of continuous exercise. After determination of MVC, resting ³¹P spectra were collected every 15 sec for 2 min to provide calibration data for correction for partial saturation. The exercise protocol consisted of 2 min rest followed by 3 min repetitive bilateral quadriceps contractions, lifting a predetermined, individualized load (30% MVC) followed by 5 min recovery, during which all participants returned to their previous baseline level of PCr. ³¹P spectra were obtained every 2 sec over this 10-min period. An 8-cm-diameter double-resonant radiofrequency surface coil, tuned for ³¹P, was fastened proximal to the superior aspect of the patella over the anteromedial aspect of the right thigh, keeping the coil's long dimension perpendicular to the z-axis of the magnet. This allowed for isolation of the quadriceps muscle. The height of the knee, position of the padded bar to be lifted with quadriceps extension, and position of the surface coil within the magnet were identical for each participant and were assured with a sliding patient tray that measures position to the millimeter. The range of motion at the knee was constrained by the inner bore dimensions, and full extension/relaxation and frequency of extension times (0.5 Hz, every 2 sec for 3 min) were assured by a study investigator. Mitochondrial function was determined from the ³¹P MRS spectra by fitting the PCr concentration vs. time during exercise recovery to a single exponential function to determine the recovery time constant (τ). PCr concentrations were obtained from integrated areas, corrected for partial saturation, and normalized with respect to the β-ATP peak area using a value of 8.2 mm for the ATP concentration, during the postexercise rest phase. Intracellular pH was calculated by comparing the chemical shift difference between the PCr and inorganic phosphate (Pi) peaks in parts per million, using the following equation: intracellular pH = 6.85 + log [(δ − 3.56)/(5.64 − δ)].

Questionnaires

Three-day food records were collected at the baseline evaluation and reviewed by nutrition staff with analysis by Nutrition Data Systems, Minneapolis, MN. Self-reported levels of physical activity were assessed with pediatric-specific modifications made to the Modifiable Activity Questionnaire (18).

Assays

The serum insulin was measured using one of two commercially available RIA (one from Diagnostic Products Corp., Los Angeles, CA, or Access Immunoassay System from Beckman Coulter, Chaska, MN). In same-sample comparisons, interassay correlation was excellent (r = 0.99), using identical linear scales, without systemic differences in the results of the assays by Bland-Altman analysis (19). Serum glucose, lipids, and chemistries were measured using standard methodologies in the medical laboratories of MIT and MGH. IGF-I was measured by enzyme immunoassay (ALPCO Diagnostics, Salem, NH). An extraction step is used in which IGF-I is separated from its binding protein. Serum estradiol (girls) and testosterone (boys) levels were measured using Access Immunoassay Systems (Beckman Coulter, Fullerton, CA), and the lower limit of detection was 20 pg/ml for estradiol and 10 ng/dl for testosterone.

Statistical analysis

Statistical analyses were performed using JMP SAS-based software. Student's paired t tests were used to evaluate within-individual differences between outcome variables obtained at baseline and after 1 yr. Parametric (Pearson's) and nonparametric (Spearman's) correlations were used as appropriate for the data to evaluate the relationship between the outcome variable (annualized growth velocity) and explanatory variables. Differences were further assessed using multivariate regression analyses controlling for demographic, anthropometric, and metabolic factors as well as other potential biomarkers of growth capacity such as IGF-I and sex steroids.

Results

Subject characteristics

Twenty-three subjects had complete data at the 1-yr time point. Subject characteristics are summarized in Table 1. Subjects were 11.3 ± 1.9 (sd) yr old at the initiation of the study; 61% of subjects were male. By design we selected children representing a range of prepubertal and early-to-mid-pubertal Tanner stages for enrollment into the study, and 26% (6) were Tanner stage I, 39% (9) were Tanner stage II, and 35% (8) were Tanner stage III. Initial average height Z-score was 0.62 SDS and increased significantly over the year to 0.78 SDS (P = 0.0007). At the baseline visit, serum estradiol or testosterone levels were detectable in 52% of subjects (12 of 23), and at the 12-month visit, these were detectable in 82% of subjects (18 of 22) for whom these studies were performed. As expected, estradiol and testosterone levels, respectively, were increased in girls and boys in later Tanner stages and associated with higher IGF-I levels (data not shown). Recruitment was designed to overrepresent overweight subjects to investigate the relationship between obesity-related insulin resistance and mitochondrial function during growth and pubertal maturation in this longitudinal cohort. Thirty-five percent of subjects (n = 8) were normal weight (BMI <85th percentile), and 65% (n = 15) were overweight (BMI ≥85th percentile). Average annualized growth velocity at 1 yr for boys was 7.1 ± 1.5 cm/yr and for girls, 6.5 ± 1.7 cm/yr. Growth velocity, presented according to age and Tanner stage, is shown in Table 2. There was no significant difference in growth velocity between subjects who were normal weight and overweight.

Table 1.

Subject characteristics

	Baseline	yr 1	Average within-individual change over 1 yr ± sem, P value
Age (yr)	11.3 ± 1.9	12.3 ± 1.9
Male/female (n)	14/9	14/9
Tanner stage [%, (n)]
I	26 (6)
II	39 (9)	17 (4)
III	35 (8)	52 (12)
IV		22 (5)
V		9 (2)
Height (cm)	150.4 ± 12.7	157.4 ± 12.5	7.0 ± 0.3, <0.0001
Height Z-score	0.62 ± 0.77	0.78 ± 0.77	0.15 ± 0.04, 0.0007
Weight (kg)	59.7 ± 20.7	67.5 ± 23.1	7.8 ± 0.9, <0.0001
Weight Z-score	2.42 ± 2.09	2.49 ± 2.16	0.08 ± 0.07, NS
BMI (kg/m2)	25.8 ± 7	26.8 ± 7.6	1 ± 0.3, 0.0093
BMI Z-score	2.80 ± 2.39	2.73 ± 2.45	−0.06 ± 0.1, NS
Systolic blood pressure (mm Hg)	102 ± 9	109 ± 14	8 ± 4, 0.056
Diastolic blood pressure (mm Hg)	62 ± 9	67 ± 7	4 ± 2, 0.060
HOMA-IR	1.73 ± 1.15	1.73 ± 0.91	<0.01 ± 0.14, NS
WBISI	6.60 ± 4.12	5.62 ± 3.23 (n = 22)	−1.11 ± 0.6, 0.090
IGF-I (ng/ml or μg/liter)	165 ± 52	201 ± 62 (n = 22)	36 ± 14, 0.019
Alkaline phosphatase (U/liter)	291 ± 93	284 ± 82	−7 ± 11, NS
PCrτ (sec), time to recovery of PCr concentration after submaximal exercise	31.5 ± 10	34.8 ± 10.8 (n = 21)	3.1 ± 2.5, NS

Unless indicated otherwise, results are shown as mean ± sd. Mean within-individual differences over 1 yr (by matched pairs t test) are shown, along with sem and P value. NS, Not significant.

Table 2.

Growth velocity during each of the two study years, by sex and Tanner stage

	Growth velocity (cm/yr)
	Girls	Boys
yr 1
Tanner I (n = 4 girls, 2 boys)	7.0 ± 0.5	6.7 ± 1.2
Tanner II (n = 2 girls, 7 boys)	7.9 ± 1.8	7.2 ± 1.5
Tanner III (n = 3 girls, 5 boys)	4.9 ± 1.8	7.4 ± 1.8
yr 2
Tanner II (n = 2 girls, 2 boys)	5.6 ± 0.8	6.5 ± 2.2
Tanner III (n = 6 girls, 5 boys)	6.1 ± 2.5	8.0 ± 1.9
Tanner IV (n = 4 boys)		7.2 ± 2.8

Results are shown as mean ± sd.

Relationship of growth velocity to mitochondrial function as assessed by PCrτ

More rapid skeletal muscle oxidative phosphorylation, as suggested by a shorter time to PCr recovery after submaximal exercise, was significantly associated with growth velocity in the subsequent year (r² = 0.29; P = 0.008). This result was significant in the first year. A similar relationship was seen between baseline mitochondrial function and growth in the second year of follow-up, although it did not reach statistical significance in the second year because the sample size was smaller (Fig. 1). This result persists after controlling for known experimental parameters that affect estimation of PCrτ, including the change in the PCr concentration that occurs with exercise (ΔPCr) as well as the end-of-exercise intracellular pH (20, 21). No effect of season was detected on either PCrτ or the relationship between PCrτ and growth (R² for model = 0.45; P = 0.02 for effect of PCrτ). The relationship of PCrτ with growth velocity was also persistent in multivariate regression analysis controlling for other baseline factors including age, gender, Tanner stage, BMI Z-score, and height Z-score (R² for model = 0.51; P = 0.05 for effect of PCrτ).

Fig. 1.

Mitochondrial and growth velocity in the subsequent year.

Mitochondrial function and energy intake and expenditure

This relationship between growth velocity and skeletal muscle oxidative phosphorylation remained significant after controlling for indices of self-reported energy intake (kilocalories per kilogram of body weight per day) and physical activity (number of episodes of vigorous activity per week) (R² for model = 0.39; P = 0.047 for effect of PCrτ). Also, controlling for insulin resistance as represented by either HOMA-IR (R² = 0.33; P = 0.014) or WBISI (R² = 0.3; P = 0.015) did not alter the association between shorter PCrτ and more rapid future growth velocity.

Relationship of growth velocity to IGF-I

IGF-I level, obtained at the baseline visit, did not demonstrate a significant relationship with growth velocity in the subsequent year. However, a rise in IGF-I over the year was associated with a faster growth velocity (r² = 0.25; P = 0.019). Every 83-ng/ml rise in IGF-I over the year translated to 1-cm/yr additional growth. Subjects with the highest initial IGF-I values tended to have a smaller increase or a frank decrease in IGF-I levels over the subsequent year (r² = 0.23; P = 0.022).

Additional multivariate regression analyses were used to assess the role of IGF-I as a potential modulator of the association of mitochondrial function and growth velocity. The relationship between mitochondrial function and growth velocity persisted after controlling for baseline IGF-I (R² = 0.35; P = 0.011). We found that shorter PCrτ at baseline (P = 0.04) and a rise in IGF-I from baseline to the 1-yr measurement (P = 0.07) each seemed independently associated with more rapid growth between the baseline to the 1-yr measurements. Furthermore, we found that shorter PCrτ at the 1-yr measurement (P = 0.07) and a rise in IGF-I from baseline to 1 yr (P = 0.03) also each seemed independently associated with more rapid growth between the 1- and 2-yr evaluations. These results illustrate that an IGF-I level measured in a pre- or early-to-mid-pubertal child does not completely capture the magnitude and effect of GH secretory patterns that produce growth. This may also explain why in the present study we fail to detect a potential association between GH and mitochondrial function. In contrast, growth velocity itself is the biological index of GH function and relates significantly to mitochondrial function.

Relationship of growth velocity to sex steroid levels

Detectable levels of sex steroids were not associated with more rapid growth velocity (P = 0.36 by Wilcoxon rank-sum test) or associated with mitochondrial function (P = 0.12 by Wilcoxon rank-sum test). Additional multivariate regression analyses were used to assess the role of sex steroids as potential modulators of the association of mitochondrial function and growth velocity. The relationship between mitochondrial function and growth velocity persisted after controlling for the presence or absence of detectable sex steroid levels (R² = 0.3; P = 0.011).

Discussion

We report a novel association between more rapid recovery of skeletal muscle PCr concentration after submaximal exercise and increased linear growth velocity in healthy normal-weight and overweight children and adolescents. This effect remained highly significant after controlling for other clinical characteristics likely to affect growth, including age, height, BMI, sex, and Tanner stage. This relationship also persisted after controlling for factors related to energy intake and expenditure and indices of insulin resistance and was independent of IGF-I level (in part representing GH axis activity) or the presence of detectable levels of sex steroids.

There are several mechanisms that could underlie the observed association between skeletal muscle oxidative phosphorylation and linear growth velocity. An increase in mitochondrial capacity might occur along with growth acceleration in response to ongoing demand for energy. Indeed, previously identified factors that affect mitochondrial function include exercise, cold, and starvation, as reviewed in Echave et al. (22). These are all circumstances associated with negative energy balance, where promoting mitochondrial biogenesis and/or stimulating mitochondrial function would be adaptive. It is possible to speculate that the mitochondria of children and adolescents undergoing puberty may be receiving similar, as of yet unknown, signals in response to the pubertal growth spurt.

The increased energy use during rapid growth may lead directly to a greater capacity for skeletal muscle oxidative phosphorylation. Wallace (1) describes how ongoing use of ATP at a high rate promotes mitochondrial efficiency by driving the production of a proton gradient and the activity of ATP synthase. In the present study, potential associations between more rapid skeletal muscle oxidative phosphorylation with self-reported caloric intake and self-reported frequency of vigorous exercise were seen, although these did not reach statistical significance. Even after controlling for these estimates of energy intake and expenditure, the positive relationship between annualized growth velocity and this index of mitochondrial function persisted. This suggests that current energy balance may not be the sole determinant of how skeletal muscle oxidative phosphorylation relates to linear growth velocity, although certainly a comprehensive estimate of energy use was not undertaken. Additional studies are required to better understand how energy demand affects mitochondrial function in growing children.

The GH/IGF-I system is another possible candidate system for growth-related regulation of mitochondrial function. The relationship between GH and mitochondrial function has been the subject of several recent investigations. Makimura et al. (8) demonstrated a positive association between mitochondrial function and GH action in adults. Short et al. (7) note increased ATP production (as assessed in vitro after skeletal muscle biopsy) in response to an iv infusion of GH. However, the relationship remains an area of investigation. GH deficiency has been observed in individuals with mitochondrial disease but seems not to be the primary etiology of poor growth in this population (23).

In the present study, we measured serum IGF-I. Although circulating levels of hepatic IGF-I are in large part determined by GH stimulation, there are other GH-independent determinants of IGF-I as well. In our subjects, those with the highest initial values of IGF-I were most likely to see levels decline or rise more slowly in the subsequent year, perhaps suggesting that GH secretion had already begun to slow. These dynamics limit our ability to make inferences about whether the observed relationship between growth velocity and mitochondrial function may be mediated by GH. In our subjects, multivariate regression analysis demonstrated that shorter PCrτ and a greater rise in IGF-I each seemed independently associated with more rapid growth, and no interaction appeared to exist between these terms. In addition, rise in IGF-I over the preceding year (another representation of GH secretion) was not associated with PCrτ at the 1-yr time point, and each of these independently predicted growth velocity in the subsequent year. Despite evidence suggesting that GH has the capacity to promote mitochondrial function, we were not able to demonstrate a direct relationship between GH activity and skeletal muscle oxidative phosphorylation in the present study. More in-depth investigations of GH secretion and mitochondrial function in adolescents could yield additional insights into this relationship. In addition, although we found no association between circulating sex steroid levels and either mitochondrial function or growth velocity, a larger sample size would permit consideration of potential interactions between GH and sex steroid levels affecting mitochondrial function and growth.

We examined skeletal muscle oxidative phosphorylation in the present study, representing one dimension of mitochondrial function. ³¹P MRS is a validated technique for noninvasive measurement of mitochondrial function (15, 16); our findings also hold after adjusting for known experimental confounders of this measurement and after considering possible seasonal effects (20, 21). Although muscle biopsy would have provided additional evidence and information about other changes in mitochondrial function, its invasiveness can be a barrier to its use, especially in healthy pediatric populations. ³¹P MRS provides a viable alternative for longitudinal assessments. In addition, two thirds of the study group were overweight children, which might limit the generalizability of these findings. However, the relationship between growth and skeletal muscle oxidative phosphorylation remained highly significant when controlling for BMI and also when the analysis was restricted to the subset of normal-weight children. In addition, growth velocity was not significantly related to BMI. Finally, in the second year, annualized growth velocity was extrapolated from the 18-month visit for seven of 17 subjects, which may introduce additional error in these rapidly growing children. Sample size may have also limited our ability to detect a statistically significant association between growth velocity and mitochondrial function in the second year.

In summary, in this study, we demonstrate for the first time that more rapid PCr recovery in skeletal muscle after sub-maximal exercise predicts subsequent longitudinal growth velocity. It is not possible to infer causality given the study's design, but the robust independent relationship we demonstrate in a prospective, longitudinal study between apparent muscle mitochondrial capacity and growth velocity suggests a possible role for mitochondrial function in linear growth during childhood and adolescence. Additional studies are warranted to further elucidate the etiology and implications of this novel finding.