https://orcid.org/0000-0002-3685-0950
Abstract
Mutations in type I collagen or collagen-related proteins cause osteogenesis imperfecta (OI). Energy expenditure and body composition in OI could reflect reduced mobility or intrinsic defects in osteoblast differentiation increasing adipocyte development.
This study compares adiposity and resting energy expenditure (REE) in OI and healthy controls (HC), for OI genotype- and Type-associated differences.
We studied 90 participants, 30 with OI (11 COL1A1 Gly, 8 COL1A2 Gly, 4 COL1A1 non-Gly, 1 COL1A2 non-Gly, 6 non-COL; 8 Type III, 16 Type IV, 4 Type VI, 1 Type VII, 1 Type XIV) and 60 HC with sociodemographic characteristics/BMI/BMIz similar to the OI group. Participants underwent dual-energy x-ray absorptiometry to determine lean mass and fat mass percentage (FM%) and REE. FM% and REE were compared, adjusting for covariates, to examine the relationship of OI genotypes and phenotypic Types.
FM% did not differ significantly in all patients with OI vs HC (OI: 36.6% ± 1.9%; HC: 32.7% ± 1.2%; P = 0.088). FM% was, however, greater than HC for those with non-COL variants (P = 0.016). FM% did not differ from HC among OI Types (P values > 0.05).
Overall, covariate-adjusted REE did not differ significantly between OI and HC (OI: 1376.5 ± 44.7 kcal/d; HC: 1377.0 ± 96 kcal/d; P = 0.345). However, those with non-COL variants (P = 0.016) and Type VI OI (P = 0.04) had significantly lower REE than HC.
Overall, patients with OI did not significantly differ in either extra-marrow adiposity or REE from BMI-similar HC. However, reduced REE among those with non-COL variants may contribute to greater adiposity.
Osteogenesis imperfecta (OI) or “brittle bone disease” is a heterogenous, heritable dysplasia of bone matrix composition and homeostasis. The disorder is rare, with 1 in 15 000 to 20 000 births presenting with OI (1, 2). Primary symptomatic manifestations of OI are low bone mass and reduced bone material strength leading to increased bone fragility and brittleness that results in bone fractures and deformity (3, 4). OI was originally classified into types according to radiographic and clinical features by Sillence, as Types I through IV (5). The Sillence Types of OI are caused by structural or quantitative defects in type I collagen. Elucidation of the genetic alterations causing OI has both expanded the number of Types of OI and made clear that classification by skeletal severity does not always correlate to the gene that is affected (6).
OI is most frequently caused by function-altering collagen sequence variants (7), with the majority of OI cases due to dominantly inherited defects in type 1 collagen itself, specifically, in COL1A1 or COL1A2, which code for the alpha chains of type 1 collagen (8, 9). Glycine substitutions comprise more than 80% of the causative abnormalities in collagen (10) because of glycine’s essential role in triple helical folding. Thus, most OI cases display defects in type I collagen’s primary structure or quantity, leading to secondary abnormalities in collagen posttranslational modifications, folding, intracellular transport, and matrix incorporation that ultimately alter bone matrix (4, 7). Those with defects in the structure of type I collagen are classified as having Type II, III, or IV OI, which are the perinatal lethal, severe progressive deforming, and moderately severe phenotypes, respectively. However, within the last decade, genetic analysis has uncovered abnormalities in non-collagen genes that are causative for recessive forms of OI. These gene alterations most commonly result in deficiency of proteins that interact with type I collagen (4). Recessive OI genes involved in collagen biosynthesis, posttranslational modification, and processing include CRTAP, P3H1, PPIB, SERPINH1, FKBP10, PLOD2, and BMP1 (9). Mutations in IFITM5/BRIL and SERPINF1/PEDF affect osteoblast differentiation and bone matrix mineralization (9, 11). Mutations in SP7, TMEM38B, WNT1, CREB3L1, SPARC, and MBTPS2 affect osteoblast differentiation and function, while impairing collagen secretion, incorporation into matrix, and cross-linking.
OI can also affect other tissues, due either to the direct effects of collagen abnormalities or from reductions in physical mobility (7). Increased bone fragility per se often limits mobility and weight-bearing may be further suspended during recovery from fractures or surgery. The decreased activity and weight-bearing tend to decrease both bone and lean mass and may contribute to weight gain. In addition, there may be an intrinsic component to increased adipose tissue in OI. Osteoblasts and adipocytes share a common bone mesenchymal precursor and there is plasticity during differentiation between these cell types (12). Impaired differentiation of osteoblasts, as occurs in OI, may shunt precursors into other pathways, including adipocytes. In a murine model for a dominant mutation caused by a Gly349Cys substitution in Col1a1, the Brtl mouse, in vitro osteogenesis from mesenchymal stem cells is impaired while adipocytes from mesenchymal stem cells are increased in size and number (10). For recessive Type VI OI caused by absence of PEDF, the protein product of the SERPINF1 gene, the Serpinf1 null mouse has >50% increased total body adiposity and reduced bone mineral content vs wild-type littermates (13). Therefore, body composition and, as a result, energy expenditure, may potentially be intrinsically affected in both dominant and recessive OI.
Several prior studies have shown that patients with OI often have significant alterations in both body weight and stature; however, few studies have focused on adiposity or resting energy expenditure (REE). OI-specific growth curves for 100 children with Types III and IV OI demonstrate both short stature and excessive weight gain (14). Their body mass index (BMI) curves showed marked upward shifting as children with OI aged, with the 95th percentile OI curve shifted upwards from 2 years of age to adulthood and the 50th percentile OI curve greater than the 75th percentile for a healthy pediatric curve. The abnormalities appear more marked in Type III than IV, and in females than males. However, calculated BMIs may be problematic in cases with severe short stature predominantly due to shortened lower extremities. Transverse forearm computed tomography scans in 266 children with defects in COL1A1/2 compared with 255 healthy controls and adjusted for size demonstrated decreased forearm muscle mass and normal subcutaneous fat (7). A Brazilian study of pediatric OI Types I, IV, and III caused by collagen defects found increased BMI, especially in Type III OI, and a good correlation between body fat calculated by dual-energy x-ray absorptiometry (DXA) and skin fold thickness and classified most Type III OI patients as having obesity (8, 15).
Very few studies have examined energy metabolism in patients with OI and none have explored the effects of specific genotype on resting expenditure. A study from nearly 50 years ago, prior to either the Sillence classification or the identification of the genetic basis of OI, presented evidence of hypermetabolism in OI patients aged 4 to 18 years, based on low body weight, elevated serum thyroxine, oxygen consumption, and basal temperature (16). Their data on weight and thyroid function have not been corroborated in subsequent studies, and the metabolism corroboration has not been undertaken with contemporary methodology. Thus, further examination of body composition, adiposity, and REE in individuals with OI is of direct clinical importance. Our study compared adiposity and REE in individuals with OI and age/BMI-similar healthy controls and examined the correlation of OI genotype and clinical phenotype with adiposity and REE among several genetic subtypes of OI.
Methods
Study Population
The study population comprised 2 convenience samples: (a) individuals with a diagnosis of OI who are participating in a National Institute of Child Health and Human Development OI Natural History protocol that recruited patients with OI Types III-XVIII, but not type I OI, based on genetic testing or clinical diagnosis and gene testing; and (b) sex- and race-matched, age- and BMI-similar healthy volunteer controls (2 for each person with OI). All participants were evaluated at the National Institutes of Health Clinical Center in Bethesda, Maryland. Each participant underwent resting energy expenditure assessment and a DXA scan for whole body composition. The studies (NCT03575221, NCT00459992, and NCT00001522) were approved by the Institutional Review Board of the National Institutes of Health. Written informed consent, and assent when appropriate, were obtained prior to the assessments.
Height and weight were obtained using calibrated scales and stadiometers, and body mass index (BMI, kg/m2) was calculated and converted to age- and sex-specific z-scores (BMIz) based on reference data published by the Centers for Disease Control and Prevention (17). For those with age >18 years, BMIz for age 18 years was used.
Total body DXA lean and fat mass measurements (kg) were obtained using a Hologic QDR Discovery (Hologic Inc, Waltham, Massachusetts) instrument for most participants and using a Hologic QDR 2000 instrument in the pencil beam mode for 5 healthy controls. The device hardware remained unchanged during the study period and all scans were performed in the National Institutes of Health Clinical Center Nuclear Medicine Department. Total body % fat mass was calculated as 100 × fat mass/total body mass. Duplicate scans from 50 patients found excellent precision (% coefficient of variance < 2.0% for all indices) with similar results from repeated scans of phantoms. To account for possible issues resulting from surgical hardware present in some patients that might increase apparent bone and therefore total mass, we also ran analyses without bone mineral content (BMC) included in the total mass to calculate fat mass percentage (FM%) as fat mass/(fat mass + lean mass). Results all remained with the same direction and significance as found for analyses using total body % fat (data not shown). The QDR4500 DXA instrument data were reported using the calibration proposed by Schoeller et al to account for systematic underestimation of body adiposity (18, 19) as implemented by the National Health and Nutrition Examination Survey Body Composition Manual (20).
REE assessment was performed using a ventilated hood system, the ParvoMedics True 2400 indirect calorimetry cart (ParvoMedics, Salt Lake City, Utah, USA). The machine was calibrated before each test with a reference gas mixture and to ambient barometric pressure and temperature. The device hardware remained unchanged during the study period. Indirect calorimetry measures oxygen consumption (VO2) and carbon dioxide production (VCO2) to calculate resting energy expenditure (21) using the equations of Weir (22). The breath-by-breath measurements of VO2 and VCO2 were automatically averaged and recorded at 1-minute intervals. Average REE was calculated from measurements collected over a 25- to 30-minute period, deleting intervals with subject motion or difficulty staying awake. The first few minutes of data collection, as participants became accustomed to having the clear plastic hood over their head and shoulders, were also not used for calculation of REE. For all participants, assessments were done at approximately 7:30 am under resting and low-light conditions with all patients fasting for 8 to 10 hours prior to the test and resting supine for at least 30 minutes before the start of REE measurement.
Genotypic Mutation Analysis
Identification of OI-causative variants utilized methodology that changed over time with new sequencing developments, encompassing RNA-hybrid analysis, cDNA Sanger sequencing, and Next-Gen gDNA sequencing. Methods have been described in previous reports (7, 9, 23). Sillence Types III and IV OI were determined by radiographic and clinical findings in conjunction with molecular defects in COL1A1 or COL1A2; recessive OI Types were determined by molecular testing for defects in specific genes and by clinical phenotype. The defects were divided into COL1A1 and COL1A2 for analysis because the collagen triple helix has 2 copies of the A1 chain and 1 copy of the A2 chain. This means that 75% of helices have at least 1 mutant chain in cases of A1 defects, while a maximum of 50% of helices have a mutant chain for A2 defects. Some features of OI, such as perinatal lethality, have been found to have different proportions of occurrence depending on the chain in which the mutation is located.
Statistical Analyses
The major objectives of the study included (1) to compare adiposity in individuals with OI and healthy controls of the same BMI, as determined by percentage fat mass for total body DXA (total body % fat mass); and (2) to compare REE in individuals with OI and healthy controls. In addition, the OI cohort was grouped into subsets according to genotype to address (3) the specific question of whether there are OI-genotype-associated REE and adiposity differences. Categories included COL1A1 or COL1A2 glycine substitutions, non-glycine COL1A1 or COL1A2 structural variants, and non-COL variants. This was further broken down (4) to analyze differences in REE and adiposity between overall collagen glycine substitutions versus non-glycine structural variants. Finally, the OI cohort was grouped into subsets according to a genetic classification of OI Types (1, 5) to examine OI-Type-associated REE and adiposity differences. For this analysis, only Types III, IV, and VI were studied, because Types VII and XIV had only 1 participant each.
Differences among the groups that did not require controlling for covariates were tested for significance using unpaired t tests. Analysis of covariance (ANCOVA) was used to examine most differences and Bonferroni adjustment was used for post hoc analyses. For the comparisons of body composition and REE variables between groups (controls vs COL1A1 glycine, COL1A1 non-glycine, COL1A2 glycine, COL1A2 non-glycine, or non-COL, and controls vs Type III, Type IV, or Type VI), covariates included sex, age, and race. For REE, covariates included sex, age, race, total body lean mass, and percentage fat mass to account for group differences in these potential confounders. Additional analyses examined age × Type interactions. Pearson correlations were used to explore the relationship between variables. Nonnormally distributed data were log10-transformed prior to analysis. All tests were 2-tailed, and throughout the study P values of < 0.05 were considered significant. Calculations were performed using SPSS software (v 25.0, IBM Corp, Armonk, NY).
Results
There were 30 individuals with OI (17 females, 13 males; age range, 8-49 years) and 60 healthy controls (34 females, 26 males; age range, 6-55 years) (Table 1). Three members of the Type VI OI group were siblings, and 2 pairs of siblings were in the Type IV OI group. By design, there were no significant differences between those with OI and the healthy controls in sex, race, and mean age. Out of the 30 participants with OI, 43% were independently ambulatory and 57% utilized either a wheelchair or gait aide. These percentages reflect the patient’s status for at least the past several years, describing their primary form of mobility, and a regularly utilized second form, if applicable. The mean BMI and BMIz score, by design, also did not significantly differ between the 2 groups, although height was significantly lower in those with OI (P < 0.0001). The mean BMIz for the OI cohort was 0.94, which is almost +1 standard deviation above the mean (Table 1) such that 40% of the OI cohort had a BMIz of over 1.64, consistent with obesity and confirming the value of a further investigation of adiposity and REE.
Demographic and anthropometric data in people with osteogenesis imperfecta and healthy controls
| Osteogenesis imperfecta | Healthy controls | P value | |
|---|---|---|---|
| N | 30 | 60 | |
| Sex, % female | 56.7% | 58.3% | NS |
| Race, % non-Hispanic White | 73.3% | 76.7% | NS |
| Age, y | 23.4 ± 3.2 | 21.9 ± 3.3 | NS (P = 0.51) |
| Age range, y | 8-42 | 6-55 | |
| Height, cm | 126.3 ± 5.3 | 164.4 ± 4.0 | P < 0.001 |
| Body weight, kg | 47.6 ± 5.3 | 75.9 ± 5.4 | P < 0.001 |
| Fat mass, kg | 18.22 ± 3.9 | 26.31 ± 3.89 | P < 0.02 |
| Fat mass as a percentage of total mass, % | 36.6 ± 1.9 | 32.7 ± 1.2 | NS |
| Lean mass (kg) | 28.05 ± 3.76 | 47.0 ± 3.91 | P < 0.001 |
| BMI (kg/m2) | 29.5 ± 3.6 | 27.1 ± 2.9 | NS |
| BMI, z-score | 0.94 ± 1.21 | 1.08 ± 0.97 | NS |
| Obesity, % | 40% | 31% | NS |
| Overweight or obesity, % | 50% | 50% | NS |
| Respiratory quotient | 0.83 ± 0.19 | 0.84 ± 0.22 | NS |
| Osteogenesis imperfecta | Healthy controls | P value | |
|---|---|---|---|
| N | 30 | 60 | |
| Sex, % female | 56.7% | 58.3% | NS |
| Race, % non-Hispanic White | 73.3% | 76.7% | NS |
| Age, y | 23.4 ± 3.2 | 21.9 ± 3.3 | NS (P = 0.51) |
| Age range, y | 8-42 | 6-55 | |
| Height, cm | 126.3 ± 5.3 | 164.4 ± 4.0 | P < 0.001 |
| Body weight, kg | 47.6 ± 5.3 | 75.9 ± 5.4 | P < 0.001 |
| Fat mass, kg | 18.22 ± 3.9 | 26.31 ± 3.89 | P < 0.02 |
| Fat mass as a percentage of total mass, % | 36.6 ± 1.9 | 32.7 ± 1.2 | NS |
| Lean mass (kg) | 28.05 ± 3.76 | 47.0 ± 3.91 | P < 0.001 |
| BMI (kg/m2) | 29.5 ± 3.6 | 27.1 ± 2.9 | NS |
| BMI, z-score | 0.94 ± 1.21 | 1.08 ± 0.97 | NS |
| Obesity, % | 40% | 31% | NS |
| Overweight or obesity, % | 50% | 50% | NS |
| Respiratory quotient | 0.83 ± 0.19 | 0.84 ± 0.22 | NS |
Abbreviations: BMI, body mass index; NS, not significant.
Demographic and anthropometric data in people with osteogenesis imperfecta and healthy controls
| Osteogenesis imperfecta | Healthy controls | P value | |
|---|---|---|---|
| N | 30 | 60 | |
| Sex, % female | 56.7% | 58.3% | NS |
| Race, % non-Hispanic White | 73.3% | 76.7% | NS |
| Age, y | 23.4 ± 3.2 | 21.9 ± 3.3 | NS (P = 0.51) |
| Age range, y | 8-42 | 6-55 | |
| Height, cm | 126.3 ± 5.3 | 164.4 ± 4.0 | P < 0.001 |
| Body weight, kg | 47.6 ± 5.3 | 75.9 ± 5.4 | P < 0.001 |
| Fat mass, kg | 18.22 ± 3.9 | 26.31 ± 3.89 | P < 0.02 |
| Fat mass as a percentage of total mass, % | 36.6 ± 1.9 | 32.7 ± 1.2 | NS |
| Lean mass (kg) | 28.05 ± 3.76 | 47.0 ± 3.91 | P < 0.001 |
| BMI (kg/m2) | 29.5 ± 3.6 | 27.1 ± 2.9 | NS |
| BMI, z-score | 0.94 ± 1.21 | 1.08 ± 0.97 | NS |
| Obesity, % | 40% | 31% | NS |
| Overweight or obesity, % | 50% | 50% | NS |
| Respiratory quotient | 0.83 ± 0.19 | 0.84 ± 0.22 | NS |
| Osteogenesis imperfecta | Healthy controls | P value | |
|---|---|---|---|
| N | 30 | 60 | |
| Sex, % female | 56.7% | 58.3% | NS |
| Race, % non-Hispanic White | 73.3% | 76.7% | NS |
| Age, y | 23.4 ± 3.2 | 21.9 ± 3.3 | NS (P = 0.51) |
| Age range, y | 8-42 | 6-55 | |
| Height, cm | 126.3 ± 5.3 | 164.4 ± 4.0 | P < 0.001 |
| Body weight, kg | 47.6 ± 5.3 | 75.9 ± 5.4 | P < 0.001 |
| Fat mass, kg | 18.22 ± 3.9 | 26.31 ± 3.89 | P < 0.02 |
| Fat mass as a percentage of total mass, % | 36.6 ± 1.9 | 32.7 ± 1.2 | NS |
| Lean mass (kg) | 28.05 ± 3.76 | 47.0 ± 3.91 | P < 0.001 |
| BMI (kg/m2) | 29.5 ± 3.6 | 27.1 ± 2.9 | NS |
| BMI, z-score | 0.94 ± 1.21 | 1.08 ± 0.97 | NS |
| Obesity, % | 40% | 31% | NS |
| Overweight or obesity, % | 50% | 50% | NS |
| Respiratory quotient | 0.83 ± 0.19 | 0.84 ± 0.22 | NS |
Abbreviations: BMI, body mass index; NS, not significant.
Individuals with OI were divided into subgroups by their genotype and Type (Table 2). In total, there were 4 different genotypic subgroups, with the non-COL subgroup including all individuals with defects (mostly recessive null mutations) in collagen-related proteins (Table 2). Half the patients (50%) had a genetic alteration in COL1A1 and about a third of the patients (30%) had a genetic alteration in COL1A2, a representative distribution of mutation frequencies among people with OI. Close to 37% of the patients with OI had a genetic alteration resulting in an α1(I) glycine substitution, 13% caused a pro α1(I) non-glycine substitution, 27% had an α2(I) glycine substitution, 3% a pro α2(I) non-glycine substitution, and 20% an abnormality in a non-collagen (non-COL) gene (SERPINF1, TMEM38B, CRTAP, or BRIL).The non-glycine COL1A1 defects include 2 in the C-propeptide and 2 helical Y-position changes, while the non-glycine COL1A2 defect is located in the C-propeptide cleavage site. The characteristics of participants according to Sillence Type and genotypic subgroups are shown in Supplemental Tables 1 and 2 (24).
Participants by genotype and Sillence Type
| Protein affected | Description of defect | Subgroup | Sillence Type (% of total) | N |
|---|---|---|---|---|
| COL1A1 Glycine | Type I Collagen | COL1A1 Glycine | Type III (63.6%) Type IV (36.4%) | 11 |
| COL1A1 Non-Glycine | Type I Collagen | COL Non-Glycine | Type IV (100%) | 4 |
| COL1A2 Glycine | Type I Collagen | COL1A2 Glycine | Type III (12.5%) Type IV (87.5%) | 8 |
| COL1A2 Non-Glycine | Type I Collagen | COL Non-Glycine | Type IV (100%) | 1 |
| SERPINF1 | Collagen-Related Protein -PEDF deficiency | Non-COL | Type VI (100%) | 3 |
| TMEM38B | Collagen-Related Protein – TMEM38B null | Non-COL | Type XIV (100%) | 1 |
| CRTAP | Collagen-Related Protein - CRTAP null | Non-COL | Type VII (100%) | 1 |
| BRILa | Collagen-Related Protein - BRIL missense Het | Non-COL | Type VI (100%) | 1 |
| Protein affected | Description of defect | Subgroup | Sillence Type (% of total) | N |
|---|---|---|---|---|
| COL1A1 Glycine | Type I Collagen | COL1A1 Glycine | Type III (63.6%) Type IV (36.4%) | 11 |
| COL1A1 Non-Glycine | Type I Collagen | COL Non-Glycine | Type IV (100%) | 4 |
| COL1A2 Glycine | Type I Collagen | COL1A2 Glycine | Type III (12.5%) Type IV (87.5%) | 8 |
| COL1A2 Non-Glycine | Type I Collagen | COL Non-Glycine | Type IV (100%) | 1 |
| SERPINF1 | Collagen-Related Protein -PEDF deficiency | Non-COL | Type VI (100%) | 3 |
| TMEM38B | Collagen-Related Protein – TMEM38B null | Non-COL | Type XIV (100%) | 1 |
| CRTAP | Collagen-Related Protein - CRTAP null | Non-COL | Type VII (100%) | 1 |
| BRILa | Collagen-Related Protein - BRIL missense Het | Non-COL | Type VI (100%) | 1 |
aThis missense variant in IFITM5/BRIL (p.S40L) (30), presents as Type VI, not type V OI. All cases of type V OI are caused by a recurrent variant in BRIL 5’-UTR.
Participants by genotype and Sillence Type
| Protein affected | Description of defect | Subgroup | Sillence Type (% of total) | N |
|---|---|---|---|---|
| COL1A1 Glycine | Type I Collagen | COL1A1 Glycine | Type III (63.6%) Type IV (36.4%) | 11 |
| COL1A1 Non-Glycine | Type I Collagen | COL Non-Glycine | Type IV (100%) | 4 |
| COL1A2 Glycine | Type I Collagen | COL1A2 Glycine | Type III (12.5%) Type IV (87.5%) | 8 |
| COL1A2 Non-Glycine | Type I Collagen | COL Non-Glycine | Type IV (100%) | 1 |
| SERPINF1 | Collagen-Related Protein -PEDF deficiency | Non-COL | Type VI (100%) | 3 |
| TMEM38B | Collagen-Related Protein – TMEM38B null | Non-COL | Type XIV (100%) | 1 |
| CRTAP | Collagen-Related Protein - CRTAP null | Non-COL | Type VII (100%) | 1 |
| BRILa | Collagen-Related Protein - BRIL missense Het | Non-COL | Type VI (100%) | 1 |
| Protein affected | Description of defect | Subgroup | Sillence Type (% of total) | N |
|---|---|---|---|---|
| COL1A1 Glycine | Type I Collagen | COL1A1 Glycine | Type III (63.6%) Type IV (36.4%) | 11 |
| COL1A1 Non-Glycine | Type I Collagen | COL Non-Glycine | Type IV (100%) | 4 |
| COL1A2 Glycine | Type I Collagen | COL1A2 Glycine | Type III (12.5%) Type IV (87.5%) | 8 |
| COL1A2 Non-Glycine | Type I Collagen | COL Non-Glycine | Type IV (100%) | 1 |
| SERPINF1 | Collagen-Related Protein -PEDF deficiency | Non-COL | Type VI (100%) | 3 |
| TMEM38B | Collagen-Related Protein – TMEM38B null | Non-COL | Type XIV (100%) | 1 |
| CRTAP | Collagen-Related Protein - CRTAP null | Non-COL | Type VII (100%) | 1 |
| BRILa | Collagen-Related Protein - BRIL missense Het | Non-COL | Type VI (100%) | 1 |
aThis missense variant in IFITM5/BRIL (p.S40L) (30), presents as Type VI, not type V OI. All cases of type V OI are caused by a recurrent variant in BRIL 5’-UTR.
We first tested whether there were differences in FM% and REE between all participants with OI vs the healthy controls. The FM% (Fig. 1A) showed a trend toward higher adiposity in those with OI (mean ± standard error of the mean [SEM]: 36.6% ± 1.9%) that did not attain significance when compared with BMI- and BMIz-matched healthy controls (32.7% ± 1.2% P = 0.088). For the comparison of REE in all participants with OI vs matched healthy controls, REE did not differ significantly between the groups (OI: 1377.0 ± 28.5; healthy controls: 1376.5 ± 44.7 kcal/d; P = 0.349; Fig. 1B) when adjusted for covariates.
Mean ± SEM percentage total body fat mass (A) and resting energy expenditure (REE) adjusted for covariates including total lean body mass (B) in the entire cohort of individuals with OI (n = 30) and healthy controls (n = 60).
To evaluate the effect of OI genotype on body composition, we analyzed the OI cohort according to the type of underlying genetic abnormality: COL1A1/COL1A2 glycine or non-glycine substitutions and non-COL mutations. We compared FM% among OI genotypes and healthy controls after adjustment for sex, age, and race (Fig. 2A). Fat mass percentage was significantly greater among those with non-COL mutations compared with the healthy controls. We then analyzed the FM% in the OI cohort by type of protein defect, combining glycine substitutions in both chains, non-glycine substitutions in both chains for the collagen-related defects vs healthy controls, adjusting for sex, age, and race (Fig. 2B). Fat mass percentage did not differ significantly for those with amino acid substitutions in type I procollagen proalpha1 or proalpha2 chains but was increased in OI participants with non-collagen defects compared with healthy controls. Lastly, we analyzed the OI cohort according to the specific OI type. We compared FM% between each of OI Types III, IV, and VI, vs healthy controls after adjustment for sex, age, and race (Fig. 2C). Fat mass percentage did not differ between these 3 Types as compared with the healthy controls; however, Type VI OI trended toward having higher adiposity, although it was not significant (P = 0.097).
Mean ± SEM percentage total body fat mass in individuals with OI and healthy controls. (A) Grouped according to genotype. Gly: genetic variants altering the coding for glycine in COL1A1 (n = 11) or COL1A2 (n = 8); Non-Gly: genetic variants altering the coding for other amino acids in COL1A1 (n = 4) or COL1A2 (n = 1); Non-Col (n = 6): function-altering variants in genes that cause OI and do not code for collagen proteins; HC (n = 60): healthy controls. (B) Grouped by type of protein defect. COL Gly (n = 19): genetic variants altering the coding for glycine in COL1A1 or COL1A2; COL Non-Gly (n = 5): genetic variants altering the coding for other amino acids in COL1A1 or COL1A2 and HC (n = 60). (C): Grouped by Sillence Type: Type III (n = 8), Type IV (n = 16), Type VI (n = 4) and HC (n = 60). *P < 0.05, **P < 0.01, ***P = 0.001.
We next compared REE among OI genotypes and healthy controls. After adjustments for sex, age, race, percentage fat mass, and lean mass, we found that REE was significantly lower than in healthy controls only for those with non-COL mutations (Fig. 3A). When we compared REE by type of protein defect vs healthy controls (Fig. 3B), after adjustments for sex, age, race, fat mass percentage, and lean mass, we again found that REE was significantly lower than in healthy controls only for those with non-collagen-related defects. Lastly, we compared REE according to OI Type vs healthy controls (Fig. 3C). After adjustments for sex, age, race, FM%, and lean mass, we found that REE was significantly lower for those with Type VI OI than in healthy controls; however, REE for those with Types III (P = 0.78) and IV (P = 0.34) OI did not differ from healthy controls. Because lean mass is the major determinant of REE, the relationship between REE and lean body mass for each participant is shown in (Fig. 4A). Participants with non-COL mutations in CRTAP, SERPINF1, and BRIL were clustered at the low end of both lean mass and REE, indicating the importance of adjusting REE for body composition differences. In addition, we examined if there were age × Type interactions for REE. We found no evidence of differences in the relationship between age and REE according to Type (Fig. 4B, interaction P = 0.51), protein defect (interaction P = 0.95), or genotype (interaction P = 0.68). There were 3 pairs of individuals who shared the same collagen gene defect (1 COL1A1 glycine substitution; 2 COL1A2 glycine substitutions), including 1 pair not from the same family. Their adjusted REEs were all within 2.1% of their mean adjusted REE value. There were 3 siblings with the same SERPINF1 in-frame 9-bp duplication. Their adjusted REEs were all within 8.6% of their mean adjusted REE value.
Mean ± SEM resting energy expenditure (REE) adjusted for covariates including total lean body mass in individuals with OI grouped according to genotype and in healthy controls. (A) Mean ± SEM REE adjusted for covariates including total lean body mass in individuals with OI grouped by glycine-specific defect. Gly: genetic variants altering the coding for glycine in COL1A1 (n = 11) or COL1A2 (n = 8); Non-Gly: genetic variants altering the coding for other amino acids in COL1A1 (n = 4) or COL1A2 (n = 1); Non-Col (n = 6): function-altering variants in genes that cause OI and do not code for collagen proteins; HC (n = 60): healthy controls. (B) Mean ± SEM REE adjusted for covariates including total lean body mass in individuals with OI grouped by type of protein defect. COL Gly (n = 19): genetic variants altering the coding for glycine in COL1A1 or COL1A2; COL Non-Gly (n = 5): genetic variants altering the coding for other amino acids in COL1A1 or COL1A2 and HC (n = 60). (C) Mean ± SEM REE adjusted for covariates including total lean body mass in individuals with OI grouped by Sillence Type: Type III (n = 8), Type IV (n = 16), Type VI (n = 4) and HC (n = 60). *P < 0.05, **P < 0.01, ***P = 0.001.
Individual resting energy expenditure (REE) measurements. (A) REE versus total lean body mass in healthy controls and in individuals with OI according to genotype. Gly: genetic variants altering the coding for glycine in COL1A1 (n = 11) or COL1A2 (n = 8); Non-Gly: genetic variants altering the coding for other amino acids in COL1A1 (n = 4) or COL1A2 (n = 1); Non-Col (n = 6): function-altering variants in genes that cause OI and do not code for collagen proteins; HC (n = 60): healthy controls. The graphed regression line is for the relationship between total lean body mass and REE in healthy control participants. (B) REE (adjusted for covariates) versus age among Controls (n = 60), Type III OI (n = 8), and Type IV OI (n = 4). REE was adjusted for sex, race, lean mass, and fat mass percentage. Lines show linear fits for each group. The slopes for each line had a negative coefficient for all 3 groups and was significantly different from 0 only for controls (-8.35, P = 0.02), but not for Type III (-16.74, P = 0.09) or Type IV (-12.54, P = 0.067) OI. Comparisons among the groups found no significant differences for slopes (P = 0.65) or intercepts (P = 0.99).
Discussion
In this study, when all participants with OI were evaluated as one group, we found that fat mass percentage did not differ in those with OI from healthy controls matched for very similar BMI and BMIz. In addition, REE, adjusted for covariates, including lean mass, did not differ significantly between the entire cohort of individuals with OI and the healthy controls. This result would suggest there are no intrinsic energy expenditure abnormalities caused by OI contributing to their high BMI/adiposity. However, when separated into subgroups based on their respective genotypic causes, the combination of significantly higher adiposity and significantly lower REE than healthy controls was found only in those whose OI was caused by non-COL variants. When separated into subgroups based on their respective OI Type, REE was significantly lower in those with Type VI OI. Fat mass percentage vs BMI-matched controls did not attain significance in any specific OI type, but a trend to adiposity was found in the small group with Type VI, in agreement with studies of the Serpinf1 null mouse (13).
These results differ from previous published evidence from 2 smaller clinical studies of children with OI that reported increased total body FM%, specifically in those with moderate to severe forms of OI (8, 15). Specifically, a higher total body FM% would be expected in those with Type III, the most severe form of OI among those who survive infancy (1). Previous studies did not use BMI-matched controls or explore potential explanations for the increased adiposity. Some previous research suggested that those with OI were in a state of “hypermetabolism” with a higher metabolic rate and faster oxygen consumption (16). However, these early investigators did not have measurements of lean mass (or accurate measurements of body surface area in the participants with OI) and instead adjusted REE for total body weight, thus conflating metabolically active lean tissues with much less energy-requiring adipose tissue (25). The oim/oim mouse, often used to model severe OI, has a Col1a2 deletion that causes a frameshift in the carboxyterminal end that affects the chain alignment region and results in a nonfunctional proα2(I) collagen chain (26). Consequently, oim/oim assembles and secretes into the extracellular matrix a collagen homotrimer (α1(I)3) which is not normally present in bone and skin. Oim/oim has reduced gastrocnemius, plantaris, and tibialis anterior muscle mass compared with controls even when adjusted for total body weight (27). These mice have evidence for muscle mitochondrial dysfunction (reduced gastrocnemius oxygen consumption and fatty acid oxidation), but no mitochondrial dysfunction in hepatic or cardiac cells (28). Interestingly, oim/oim mice showed no significant differences from controls in total fat-free or fat mass (measured by the EchoMRI approach) when expressed as a percentage of total mass, but rather, showed trends toward reduced gonadal fat mass per total body weight (28). By indirect calorimetry adjusted for total body weight, these mice also showed increased daytime and nighttime energy expenditures (28). It remains unclear how energy expenditure, adjusted in ANCOVA models for lean and fat mass, would vary in such mice compared with wild-type mice, and indeed, how relevant the results from such mice are for OI patients who synthesize the normally present α1(I)2 α1(I)1 collagen heterotrimer, albeit with a mutation in one of the chains.
The group of OI participants with non-COL protein defects in SERPINF1, TMEM38B, CRTAP, and BRIL had low REE, even after adjustment for their reduced lean body mass. This could potentially help explain the increased adiposity seen in those with collagen-related protein defects. In addition, function-altering gene variants in SERPINF1 have been linked to increased adipocyte progenitors and to increased total body adiposity in knock-out mice (13). This coincides with the trending toward higher adiposity in our Type VI OI patients with SERPINF1 inactivating mutations. Mutations in both BRIL and TMEM38B can affect osteoblast differentiation; however, their effects on adiposity are still unclear, meriting further investigation (9, 29). Study participants with non-COL protein defects showed higher adiposity along with lower REE. Not surprisingly, those with Type VI, who comprise 4/5 participants with non-COL defects, trended toward higher adiposity, along with a lower REE, suggesting that lowered resting energy expenditure may influence BMI in such patients.
The present study has limitations that stem from the relatively small sample size of individuals with OI which limit the representativeness of the study. Moreover, the subgroups based on specific disease-causing genetic variants were small, such that individuals with the non-collagen gene abnormalities could not be studied separately. In addition, the age distribution of the sample is a limitation, as there were very few pre-pubertal individuals included and some prior data suggest younger patients might have greater differences in energy expenditure relative to controls (15). We also were not able to assess marrow fat in this study and did not collect data for fat distribution, such as waist and hip circumferences, or measures of dietary intake, in the patients with OI. In addition to the need for a larger sample size and better methods to assess fatness to answer some research questions, other limitations include the recruitment only of patients with OI who travel to Bethesda, MD to take part in a longitudinal study, and the fact that in some instances more than one sibling in a single family was enrolled in the OI study. Moreover, there were no available cross-calibration data between the 2 different DXA machines utilized in the study. Also, no power calculations were performed before this secondary analysis so there is always a chance of a Type I or Type II error. Strengths of the study include the careful determination of metabolic rate using indirect calorimetry and measurement of body composition by DXA, so that REE could be adjusted for lean mass and the use of controls matched for age, sex, race, and BMI/BMIz. However, because those with OI generally have shorter stature than age-matched controls, the matching was perforce incomplete.
In conclusion, the present study found that adiposity and resting energy expenditure do not differ in individuals with OI when examined as a single group against a BMI- and BMIz- similar group. However, there are genotype-specific and type differences affecting adiposity and energy expenditure in people with OI, notably for those with Type VI OI. Prospective studies with larger sample sizes including pre-pubertal children, and extension of studies to examine marrow adiposity are needed to further elucidate the factors that influence adiposity and resting energy expenditure in individuals with OI.
Abbreviations
- ANCOVA
analysis of covariance
- BMI
body mass index
- DXA
dual-energy x-ray absorptiometry
- FM%
fat mass percentage
- OI
osteogenesis imperfecta
- REE
resting energy expenditure
- SEM
standard error of the mean
Acknowledgments
We thank the participants and their families for their long-term involvement with this research project.
Financial Support: This research was supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, ZIAHD000408 and ZIAHD000641. The funding source was not involved in the study design, collection, analysis, or interpretation of data, in the writing of the report, or in the decision to submit the article for publication. The opinions and assertions expressed herein are those of the authors and are not to be construed as reflecting the views of the National Institutes of Health.
Author Contributions: The study was designed by J.C.M. and J.A.Y. Data were collected by K.L.B., N.T., S.K.T., M.M.K., and A.N.D.D. Statistical analyses were conducted by K.L.B. and J.A.Y., who, together with J.C.M., wrote the first draft of the manuscript. All authors provided critical review of the manuscript and had final approval of the submitted and published versions.
Additional Information
Disclosures: K.L.B., N.T., S.K.T., M.M.K., A.N.D.D., and J.C.M. have nothing to declare. J.A.Y. receives grant support for unrelated studies sponsored by Rhythm Pharmaceuticals, Inc. and Soleno Therapeutics, Inc for treatment of rare syndromes causing obesity.
Data Availability
Some or all datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request. Some or all data generated or analyzed during this study are included in this published article or in the data repositories listed in References. Data registered at https://clinicaltrials.gov as NCT03575221, NCT00459992, and NCT00001522.
