Factors associated with bone thickness: Comparison of the cranium and humerus

Cortical bone thickness is important for the mechanical function of bone. Ontogeny, aging, sex, body size, hormone levels, diet, behavior, and genetics potentially cause variations in postcranial cortical robusticity. However, the factors associated with cranial cortical robusticity remain poorly understood. Few studies have examined cortical robusticity in both cranial and postcranial bones jointly. In the present study, we used computed tomography (CT) images to measure cortical bone thicknesses in the cranial vault and humeral diaphysis. This study clearly showed that females have a greater cranial vault thickness and greater age-related increase in cranial vault thickness than males. We found an age-related increase in the full thickness of the temporal cranial vault and the width of the humeral diaphysis, as well as an age-related decrease in the cortical thickness of the frontal cranial vault and the cortical thickness of the humeral diaphysis, suggesting that the mechanisms of bone modeling in cranial and long bones are similar. A positive correlation between cortical indices in the cranial vault and humeral diaphysis also suggested that common factors affect cortical robusticity. We also examined the association of polymorphisms in the WNT16 and TNFSF11 genes with bone thickness. However, no significant associations were observed. The present study provides fundamental knowledge about similarities and differences in the mechanisms of bone modeling between cranial and postcranial bones.


Introduction
Bones provide structural support for the body and serve other biological functions with regard to blood cell production and metabolism. At the macrostructural level, bone is classified into two different types: cortical (or compact) and cancellous (or trabecular) [1]. Cortical bone thickness is important for the mechanical function of bone. Cortical robusticity in cranial and and 187 females; 20-76 years of age, 57.7 years of age on average) using a positron emission tomography (PET)/CT scanner (Biograth mCT 64Slice, Siemens Healthcare, Tokyo, Japan) at the Department of Radiology, University of the Ryukyus Hospital. To confirm the effect of the resolution of CT images on the measurements, high-resolution CT images for the head (slice resolution 0.5 mm for x-and y-axes; slice thickness 0.5 mm for z-axis) were also obtained from another 25 adults (7 males and 18 females; 23-59 years of age; 36.2 years of age on average) at Naha City Hospital (Aquilion TSX-101A, Toshiba Corp., Tochigi, Japan) or Doujin Hospital (Activion16 TSX-031A/1B, Toshiba Corp., Tochigi, Japan). These images were acquired for clinical purposes. All subjects were free of congenital and systemic diseases such as cleft lip or palate and jaw deformities. From the PET/CT subjects, we also obtained saliva specimens for DNA preparation and information regarding sex, age, height, weight, and the birthplaces of their four grandparents. All subjects provided written informed consent to participate in this study. The study was approved by the Ethics Committee of the University of the Ryukyus.

Measurement of bone thickness
Bone thickness was measured from the CT images using Stradwin 5.4 [40]. Using both low-and high-resolution CT images, we measured the cranial vault thickness (CVT) at temporal and frontal cranial regions (TCVT and FCVT). Since the inner surface of temporal region is uneven, the thinnest points of the left and right temporal regions were identified using the coronal section passing the mandibular fossa ( Fig 1A and 1B), and then TCVTs were measured at those points using the transverse section (Fig 1C). At these points, only cortical bone was contained in all the subjects. For FCVT, we measured cortical and full thicknesses (FCVT cortical and FCVT full ) at the lateral ends of slope of the frontal crest using the transverse section immediately above the frontal sinuses (Fig 1D and 1E). FCVT cortical denotes the total thickness of internal and external tables. Using low-resolution CT images, humeral cortical thickness (HCT) and humeral bone width (HBW) at the shaft immediately under the deltoid tuberosity were measured on the left and right bones (Fig 1F). Using a transverse section slice, measurements were performed along the short axis of the bone section, which roughly corresponds to the anatomically mediolateral aspect of the bone. It should be noted that due to the variation in the shape of the bone section, the anatomical direction of measurement can vary. HCT denotes the total thickness of two cortical walls of the section. Although the transverse section slice was not exactly perpendicular to the humerus, the errors were negligible: even Θ = 5 degrees of tilt theoretically yield only a 0.38% error (1/cosΘ). Values for the left and right sides were averaged before the analysis. All the measurements were conducted by the first author (S. G.). Intraobserver errors were evaluated using 20 subjects of low-resolution image: intraclass correlation coefficients were 0.965 for TCVT, 0.905 for FCVT cortical , 0.991 for FCVT full , 0.981 for HCT, and 0.957 for HBW.

Single nucleotide polymorphism (SNP) genotyping
Saliva specimens were collected and stored using an Oragene DNA self-collection kit (DNA Genotek, Ottawa, Ontario, Canada). Genomic DNA was extracted using a Gentra Puregene DNA Purification kit (Qiagen Japan, Tokyo, Japan). DNA specimens were available for 431 subjects. KASP genotyping assays (LGC genetics TW11 OLY, UK) were used to genotype two SNPs reportedly associated with tibial bone thickness in a European genome-wide association study: rs2707466 in WNT16 and rs9525638 near TNFSF11 [32].

Statistical analysis
Statistical analyses were performed using jmp 14 (SAS Institute Inc., Cary, NC, USA) and IBM SPSS (IBM Japan, Tokyo, Japan). In addition to the measurements, the cranial cortical index (CCI), defined as FCVT cortical /FCVT full , and humeral cortical index (HCI), defined as HCT/ HBW, were used as variables. Averages between two groups were compared using a t-test. Multiple regression analysis was performed to identify factors associated with the measurements or indices, including sex, age, height, weight, ancestry, and interaction between sex and age as explanatory variables. Here, sex was represented as male = 0 and female = 1. All the subjects were Japanese, and the ancestry variable denoted the number of grandparents who originated from the Okinawa prefecture (taking values 0, 1, 2, 3, and 4), which indicates the ancestry difference between the Ryukyuans and mainland Japanese. To examine the associations involving the SNPs, genotype (AA = 0, AD = 1, and DD = 2, where A and D are the ancestral and derived alleles, respectively) was also included as an explanatory variable. Correlation coefficients and partial correlation coefficients controlled by sex, age, height, weight, and ancestry were calculated between pairs of measurements/indices.

Results
Bone thickness varied markedly among individuals.  (Table 1). To confirm the differences in measured values depending on modality and resolution settings, we also measured TCVT, FCVT cortical , and FCVT full using high-resolution CT images. TCVT values in the low-resolution images (2.0 mm slice thickness) tended to be greater than those in the high-resolution images (0.5 mm slice thickness).
Because TCVT was measured in the thinnest part of the temporal bone, where the inner surface is uneven, it may be difficult to capture the thinnest part using low-resolution images. We concluded that although those TCVT values may not be accurate in terms of "the thinnest part", they could be used as an indicator of bone thickness. Males exhibited significantly higher HCT and HBW values than females. In contrast, females exhibited significantly greater CVT values than males. Fig 2A shows scatter plots for age and measurements, and Fig 2B shows graphs comparing values by sex and age. TCVT was positively correlated with age in females but not in males. However, FCVT full was negatively correlated with age in males but not in females, whereas FCVT cortical was negatively correlated with age in both sexes. HCT and HBW were negatively and positively correlated, respectively, with age in both sexes.
In the multiple regression analysis, the model including sex, age, ancestry, height, and weight as explanatory variables poorly explained the variation in the measurements/index for CVT (R 2 = 0.058-0.14). CVT was significantly associated with sex, age (except for FCVT full ), and ancestry (except for FCVT cortical ) but not with height or weight ( Table 2). In contrast, HCT and HBW were significantly associated with sex, age, height, and weight but not with ancestry ( Table 3). The HCI (HCT/HBW) was associated only with sex and age. As described above, TCVT and HBW exhibited a positive association with age, whereas FCVT cortical , CCI, HCT, and HCI exhibited a negative association with age. The interaction term between sex and age was positively associated with TCVT and FCVT full but negatively associated with CCI, HCT, and HCI (Tables 2 and 3). We also examined the partial correlations among measurements/indices controlling for sex, age, height, weight, and ancestry (Table 4). TCVT was correlated with both FCVT full and FCVT cortical (Table 4; Fig 3A). FCVT cortical was correlated with FCVT full , and HCT did with HBW. There was no correlation between FCVT cortical and HCT or between FCVT full and HBW (Table 4; Fig 3B). However, a significant positive correlation was observed between CCI and HCI (Table 4; Fig 3C). When subjects were classified based on sex and age, the correlation coefficient between CCI and HCI in females was higher than that in males, and that in the younger age group (age < 50 years) was higher than that in the older age group (age � 50 years) ( Table 5; Fig 3C).  Table 6 shows the results of genotyping for rs2707466 in WNT16 and rs9525638 in TNFSF11 genes that are reportedly associated with cortical thickness in the tibia [32]. Our study did not detect any significant association of these SNPs with the measurements (Table 7).

Sexual dimorphism in bone thickness
Because of sexual dimorphism in body size, males usually have a larger postcranial bone volume than females [28]. Some previous studies have shown that even after controlling for body  Values are partial correlation coefficient (P value) controlled by sex, age, height, weight, and ancestry *P < 0.05. size, males have greater bone volume and greater cross-sectional area in postcranial bones than females [41,42]. The present study clearly demonstrated that males have greater cortical thickness and greater width in the diaphysis of the humerus than females, even after controlling for other covariates, including height and weight (Fig 2 and Tables 1 and 3). It has been suggested that skeletal sexual dimorphism is due not only to differences in sex steroid secretion between males and females but also to complex interactions between many factors, such as sex hormones, the growth hormone and insulin-like growth factor-1 pathways, and mechanical loading [43][44][45][46]. Sexual dimorphism in CVT is a controversial subject. Some studies have observed no sexual difference in CVT [16,33,[47][48][49]. Other studies, however, have reported sex-related differences in CVT in particular cranial regions [14,15,[50][51][52][53][54][55][56][57][58][59][60]. Generally, it seems that males have greater CVT in the posterior region, whereas females have greater CVT in the anterior region [34]. The present study, which found that females have greater CVT in the frontal and temporal bones (Fig 2 and Table 1), supports the previous general findings. In addition, our regression analysis ( Table 2) showed that CVT values were not associated with body size (height and weight), in contrast to the humeral measurements. These data thus suggest that sex itself is a factor determining CVT and that it oppositely affects cranial and postcranial bones. It can be hypothesized that estrogen signaling plays an important role in the increased CVT of females. To elucidate the mechanism underlying sexual dimorphism in CVT, however, further studies will be needed.

Age-related changes in bone thickness
Various studies have suggested that growth in the width of long bones through periosteal apposition is retained throughout the human lifespan, with age-related loss of cortical  thickness via endosteal resorption occurring to a greater extent in females than males, primarily due to estrogen deficiency after menopause [24, 28, 29, 61]. Consistent with these observations, our study showed that HBW increases with age and that there is no difference in the age-related increase in HBW between males and females (Fig 2 and Table 3). The age-related decline in HCT was greater in females than males. In addition, a negative correlation between HBW and HCI indicated that the proportion of the medullary cavity increases as bone width increases (Table 4).
Only a few studies have examined age-related changes in CVT. These previous studies did not detect any significant age-related change in full CVT values in adults [33,34]. Lillie et al.
[35] reported a slight, but not significant, increase in full CVT with age and a significant decrease with age in the thickness of the inner and outer cortical tables. In the present study, we found a significant positive association between TCVT and age (Fig 2 and Table 2). In addition, multiple regression analysis demonstrated that the age-related increase in TCVT was greater in females than males and that the age-related effect on FCVT full also differed between males and females. We also detected an age-related decrease in FCVT cortical . Age-related decreases in FCVT cortical and HCT and age-related increases in TCVT and HBW suggest that the cranial and long bones share common mechanisms of bone resorption along the endosteal surface and bone formation along the periosteal surface. However, it is notable that the agerelated effect on FCVT cortical did not differ with sex, in contrast to long bones. Therefore, post-

Effects of Okinawan ancestry on CVT
In the present study, we examined only Japanese people living in Okinawa Prefecture. The participants included individuals of Okinawan (Ryukyuan) ancestry and of mainland Japanese ancestry. Previous anthropological and genetic studies have demonstrated that the Okinawan people are genetically and phenotypically differentiated from the mainland Japanese people [62][63][64]. Therefore, we examined the effects of ancestry on bone thickness. As results, individuals of Okinawan ancestry showed significantly higher values in TCVT and FCVT full than those of mainland Japanese ancestry. It has been shown that ancient Jomon skulls have larger CVT than modern Japanese skulls [16]. Since it has been also suggested that genetic contribution of Jomon to the Okinawans is larger than those to the mainland Japanese [65], the difference in CVT depending on ancestry may be attributed to Jomon-derived genetic variations.

Correlation between cortical thicknesses of the cranial vault and humeral diaphysis
Few studies have examined cortical thickness of both the cranial and postcranial skeleton. A previous study focusing on the link in cortical robusticity between the cranial and limb (humerus and femur) bones reported correlations in the proportional cortical thickness (R = 0.4) [36]. However, as that study used samples derived from a variety of populations, the correlations might have resulted from population stratification in the samples. In the present study, we calculated partial correlation coefficients, controlling for ancestry as well as sex, age, height, and weight as covariates. As a result, we found a low positive correlation between CCI and HCI (R = 0.20) (Table 4), which indicates the existence of common factors that affect cortical thickness in both cranial and postcranial bones. In addition, we observed that females and the younger age group exhibited a higher correlation coefficient between CCI and HCI than males and the older age group, respectively (Table 5; Fig 3C). This suggests that systemic factors involved in the variation in cortical robusticity play a more significant role in females than males and before reaching advanced age than after reaching advanced age.
Genetic factors are most likely associated with the systemic mechanisms that affect cortical robusticity in both cranial and postcranial bones. Molecules involved in bone formation and resorption tend to be common throughout the body. In particular, intramembranous ossification in cranial bones and appositional growth in limb bones share the same bone modeling mechanism. Therefore, genetic variations that alter the functions of related molecules are expected to have systemic effects. Meanwhile, cranial and limb bones differ in the responses to mechanical loading [37][38][39]. It has been hypothesized that the activity of limb bone cells depends on the strength of mechanical loading, whereas the activity of cranial cells is retained despite very low levels of mechanical loading. In the present study, we observed that CCI and HCI are highly correlated in young females ( Table 5). One of the reasons for this may be that, in the older age group, there is an only small correlation, if any, between the age-related effects on HCI and on CCI. Furthermore, meles may have a greater individual difference than females in the mechanical loading on limbs, depending on physical activity.
Alternatively, responses to exogenetic stimuli might also be involved in the systemic mechanisms affecting cortical robusticity. In a study using pigs and armadillos, Lieberman [10] demonstrated that regularly exercising animals exhibited significantly higher cortical robusticity in both the cranial and postcranial bones than non-exercising controls, suggesting that cortical robusticity in cranial bones is acquired via hormones such as growth hormone and insulin-like growth factors, but not directly through mechanical loading. Although Lieberman's observation was not replicated in a study on mice [66], levels of circulating factors such as hormones, growth factors, cytokines, and metabolites can nonetheless serve as non-genetic systemic factors [43,45,46].

Association of WNT16 and TNFSF11 polymorphisms with bone thickness
The molecular basis of bone development, remodeling, and aging has been well studied [30,67,68]. Genome-wide association studies have identified hundreds of genetic loci associated with osteoporosis and related traits [31]. A previous study reported that polymorphisms in the WNT16 (rs2707466) and TNFSF11 (RANKL) (rs9525638) genes were strongly associated with the cortical thickness of the tibial diaphysis [32]. WNT16 is a positive regulator of both cortical and trabecular bone mass and structure [69][70][71][72]. TNFSF11 is a key regulator of bone remodeling and essential for osteoclast differentiation, activation, survival, and enhancement of bone resorption [73][74][75][76][77]. In the present study, we examined polymorphisms in WNT16 (rs2707466) and TNFSF11 (RANKL) (rs9525638) as candidate genetic factors exhibiting systemic effects. We did not observe any significant association of these polymorphisms with the measurements of bone thickness, but there were a few instances where statistical tendencies were found in our analysis (Table 7). Especially in the analysis for HCI (P = 0.056 for rs2707466 and P = 0.14 for rs9525638), we confirmed that the direction of an allelic effect was the same as the previous study analyzing tibial cortical thickness. Bone thickness is a polygenic quantitative trait, and the effect size of each genetic variant on this trait is very small. Therefore, the sample size in this study may have been insufficient to detect an effect on bone thickness. Further studies with a larger sample size are thus needed to identify systemic genetic factors affecting cortical robusticity.

Concluding remarks
This study clearly showed that females have greater bone thickness than males in the cranial vault, in contrast to the humeral diaphysis. We also identified similarities and differences in age-related effects on cortical thickness between the cranial vault and humeral diaphysis. A positive correlation between CCI and HCI (R = 0.20) was observed after controlling for confounding factors, suggesting the existence of systemic factors that affect cortical robusticity. Our genetic analysis examining polymorphisms in the WNT16 and TNFSF11 genes did not detect any significant association between these polymorphisms and bone thickness. The present study, which adds insight into the differences in cortical robusticity between cranial and postcranial bones, enhances current understanding of the mechanisms of bone modeling.