OBJECTIVES: It remains difficult to characterize the source of pain in knee joints either using radiographs or magnetic resonance imaging (MRI). We sought to determine if advanced machine learning methods such as deep neural networks could distinguish knees with pain from those without it and identify the structural features that are associated with knee pain.
METHODS: We constructed a convolutional Siamese network to associate MRI scans obtained on subjects from the Osteoarthritis Initiative (OAI) with frequent unilateral knee pain comparing the knee with frequent pain to the contralateral knee without pain. The Siamese network architecture enabled pairwise learning of information from two-dimensional (2D) sagittal intermediate-weighted turbo spin echo slices obtained from similar locations on both knees. Class activation mapping (CAM) was utilized to create saliency maps, which highlighted the regions most associated with knee pain. The MRI scans and the CAMs of each subject were reviewed by an expert radiologist to identify the presence of abnormalities within the model-predicted regions of high association.
RESULTS: Using 10-fold cross-validation, our model achieved an area under curve (AUC) value of 0.808. When individuals whose knee WOMAC pain scores were not discordant were excluded, model performance increased to 0.853. The radiologist review revealed that about 86% of the cases that were predicted correctly had effusion-synovitis within the regions that were most associated with pain.
CONCLUSIONS: This study demonstrates a proof of principle that deep learning can be applied to assess knee pain from MRI scans.
KEY POINTS: • Our article is the first to leverage a deep learning framework to associate MR images of the knee with knee pain. • We developed a convolutional Siamese network that had the ability to fuse information from multiple two-dimensional (2D) MRI slices from the knee with pain and the contralateral knee of the same individual without pain to predict unilateral knee pain. • Our model achieved an area under curve (AUC) value of 0.808. When individuals who had WOMAC pain scores that were not discordant for knees (pain discordance < 3) were excluded, model performance increased to 0.853.
Dentine- and enamel-forming cells secrete matrix in consistent rhythmic phases, resulting in the formation of successive microscopic growth lines inside tooth crowns and roots. Experimental studies of various mammals have proven that these lines are laid down in subdaily, daily (circadian), and multidaily rhythms, but it is less clear how these rhythms are initiated and maintained. In 2001, researchers reported that lesioning the so-called master biological clock, the suprachiasmatic nucleus (SCN), halted daily line formation in rat dentine, whereas subdaily lines persisted. More recently, a key clock gene (Bmal1) expressed in the SCN in a circadian manner was also found to be active in dentine- and enamel- secretory cells. To probe these potential neurological and local mechanisms for the production of rhythmic lines in teeth, we reexamined the role of the SCN in growth line formation in Wistar rats and investigated the presence of daily lines in Bmal1 knockout mice (Bmal1 ). In contrast to the results of the 2001 study, we found that both daily and subdaily growth lines persisted in rat dentine after complete or partial SCN lesion in the majority of individuals. In mice, after transfer into constant darkness, daily rhythms continued to manifest as incremental lines in the dentine of each Bmal1 genotype (wild-type, Bmal , and Bmal1 ). These results affirm that the manifestation of biological rhythms in teeth is a robust phenomenon, imply a more autonomous role of local biological clocks in tooth growth than previously suggested, and underscore the need further to elucidate tissue-specific circadian biology and its role in incremental line formation. Investigations of this nature will strengthen an invaluable system for determining growth rates and calendar ages from mammalian hard tissues, as well as documenting the early lives of fossil hominins and other primates.
During human evolution, the knee adapted to the biomechanical demands of bipedalism by altering chondrocyte developmental programs. This adaptive process was likely not without deleterious consequences to health. Today, osteoarthritis occurs in 250 million people, with risk variants enriched in non-coding sequences near chondrocyte genes, loci that likely became optimized during knee evolution. We explore this relationship by epigenetically profiling joint chondrocytes, revealing ancient selection and recent constraint and drift on knee regulatory elements, which also overlap osteoarthritis variants that contribute to disease heritability by tending to modify constrained functional sequence. We propose a model whereby genetic violations to regulatory constraint, tolerated during knee development, lead to adult pathology. In support, we discover a causal enhancer variant (rs6060369) present in billions of people at a risk locus (GDF5-UQCC1), showing how it impacts mouse knee-shape and osteoarthritis. Overall, our methods link an evolutionarily novel aspect of human anatomy to its pathogenesis.
Growth and Differentiation Factor 5 (GDF5) is a key risk locus for osteoarthritis (OA). However, little is known regarding regulation of Gdf5 expression following joint tissue damage. Here, we employed Gdf5-LacZ reporter mouse lines to assess the spatiotemporal activity of Gdf5 regulatory sequences in experimental OA following destabilisation of the medial meniscus (DMM) and after acute cartilage injury and repair. Gdf5 expression was upregulated in articular cartilage post-DMM, and was increased in human OA cartilage as determined by immunohistochemistry and microarray analysis. Gdf5 expression was also upregulated during cartilage repair in mice and was switched on in injured synovium in prospective areas of cartilage formation, where it inversely correlated with expression of the transcriptional co-factor Yes-associated protein (Yap). Indeed, overexpression of Yap suppressed Gdf5 expression in chondroprogenitors in vitro. Gdf5 expression in both mouse injury models required regulatory sequence downstream of Gdf5 coding exons. Our findings suggest that Gdf5 upregulation in articular cartilage and synovium is a generic response to knee injury that is dependent on downstream regulatory sequence and in progenitors is associated with chondrogenic specification. We propose a role for Gdf5 in tissue remodelling and repair after injury, which may partly underpin its association with OA risk.
Variation in pelvic morphology has a complex genetic basis and its patterning and specification is governed by conserved developmental pathways. Whether the mechanisms underlying the differentiation and specification of the pelvis also produce the morphological covariation on which natural selection may act, is still an open question in evolutionary developmental biology. We use high-resolution quantitative trait locus (QTL) mapping in the F generation of an advanced intercross experiment (LG,SM-G ) to characterize the genetic architecture of the mouse pelvis. We test the prediction that genomic features linked to developmental patterning and differentiation of the hind limb and pelvis and the regulation of chondrogenesis are overrepresented in QTL. We find 31 single QTL trait associations at the genome- or chromosome-wise significance level coalescing to 27 pleiotropic loci. We recover further QTL at a more relaxed significance threshold replicating locations found in a previous experiment in an earlier generation of the same population. QTL were more likely than chance to harbor Pitx1 and Sox9 Class II chromatin immunoprecipitation-seq features active during development of skeletal features. There was weak or no support for the enrichment of seven more categories of developmental features drawn from the literature. Our results suggest that genotypic variation is channeled through a subset of developmental processes involved in the generation of phenotypic variation in the pelvis. This finding indicates that the evolvability of complex traits may be subject to biases not evident from patterns of covariance among morphological features or developmental patterning when either is considered in isolation.
PURPOSE OF REVIEW: The goal of the review is to provide a comprehensive overview of the current understanding of the mechanisms underlying variation in human stature.
RECENT FINDINGS: Human height is an anthropometric trait that varies considerably within human populations as well as across the globe. Historically, much research focus was placed on understanding the biology of growth plate chondrocytes and how modifications to core chondrocyte proliferation and differentiation pathways potentially shaped height attainment in normal as well as pathological contexts. Recently, much progress has been made to improve our understanding regarding the mechanisms underlying the normal and pathological range of height variation within as well as between human populations, and today, it is understood to reflect complex interactions among a myriad of genetic, environmental, and evolutionary factors. Indeed, recent improvements in genetics (e.g., GWAS) and breakthroughs in functional genomics (e.g., whole exome sequencing, DNA methylation analysis, ATAC-sequencing, and CRISPR) have shed light on previously unknown pathways/mechanisms governing pathological and common height variation. Additionally, the use of an evolutionary perspective has also revealed important mechanisms that have shaped height variation across the planet. This review provides an overview of the current knowledge of the biological mechanisms underlying height variation by highlighting new research findings on skeletal growth control with an emphasis on previously unknown pathways/mechanisms influencing pathological and common height variation. In this context, this review also discusses how evolutionary forces likely shaped the genomic architecture of height across the globe.
OBJECTIVES: Little is known about how ilium cortical bone responds to loading. Using a mouse model, this study presents data testing the hypothesis that iliac cross-sectional properties are altered in response to increased activity. MATERIALS AND METHODS: The sample derives from lines of High Runner (HR) mice bred for increased wheel-running activity. Four treatment groups of female mice were tested: non-selected control lines housed without (N = 19) and with wheels (N = 20), and HR mice housed without (N = 17) and with wheels (N = 18) for 13 weeks beginning at weaning. Each pelvis was μCT-scanned, cross-sectional properties (cortical area-Ct.Ar, total area-Tt.Ar, polar moment of area, and polar section modulus) were determined from the ilium midshaft, and robusticity indices (ratio of the square root of Ct.Ar or Tt.Ar to caudal ilium length) were calculated. Mixed models were implemented with linetype, wheel access, and presence of the mini-muscle phenotype as fixed effects, replicate line nested within linetype as a random effect, and body mass as a covariate. RESULTS: Results demonstrate that the mouse ilium morphologically resembles a long bone in cross section. Body mass and the mini-muscle phenotype were significant predictors of iliac cross-sectional properties. Wheel access only had a statistically significant effect on Ct.Ar and its robusticity index, with greater values in mice with wheel access. DISCUSSION: These results suggest that voluntary exercise increases cortical area, but does not otherwise strengthen the ilium in these mice, corroborating previous studies on the effect of increased wheel-running activity on femoral and humeral cross-sectional properties in these mice.
In tetrapods, the scapular and pelvic girdles perform the important function of anchoring the limbs to the trunk of the body and facilitating the movement of each appendage. This shared function, however, is one of relatively few similarities between the scapula and pelvis, which have significantly different morphologies, evolutionary histories, embryonic origins, and underlying genetic pathways. The scapula evolved in jawless fish prior to the pelvis, and its embryonic development is unique among bones in that it is derived from multiple progenitor cell populations, including the dermomyotome, somatopleure, and neural crest. Conversely, the pelvis evolved several million years later in jawed fish, and it develops from an embryonic somatopleuric cell population. The genetic networks controlling the formation of the pelvis and scapula also share similarities and differences, with a number of genes shaping only one or the other, while other gene products such as PBX transcription factors act as hierarchical developmental regulators of both girdle structures. Here, we provide a detailed review of the cellular processes and genetic networks underlying pelvis and scapula formation in tetrapods, while also highlighting unanswered questions about girdle evolution and development.
In bone, sclerostin is mainly osteocyte-derived and plays an important local role in adaptive responses to mechanical loading. Whether circulating levels of sclerostin also play a functional role is currently unclear, which we aimed to examine by two sample Mendelian Randomisation (MR). A genetic instrument for circulating sclerostin, derived from a genome wide association study (GWAS) meta-analysis of serum sclerostin in 10,584 European-descent individuals, was examined in relation to femoral neck bone mineral density (BMD; n= 32,744) in GEFOS, and estimated BMD by heel ultrasound (eBMD; n=426,824), and fracture risk (n=426,795), in UK Biobank. Our GWAS identified two novel serum sclerostin loci, B4GALNT3 (standard deviation (SD)) change in sclerostin per A allele (β=0.20, P=4.6x10 ), and GALNT1 (β=0.11 per G allele, P=4.4x10 ). B4GALNT3 is an N-acetyl-galactosaminyltransferase, adding a terminal LacdiNAc disaccharide to target glycocoproteins, found to be predominantly expressed in kidney, whereas GALNT1 is an enzyme causing mucin-type O-linked glycosylation. Using these two SNPs as genetic instruments, MR revealed an inverse causal relationship between serum sclerostin and femoral neck BMD (β= -0.12, 95%CI= -0.20 to -0.05) and eBMD (β= -0.12, 95%CI= -0.14 to -0.10), and a positive relationship with fracture risk (β= 0.11,95%CI= 0.01 to 0.21). Colocalization analysis demonstrated common genetic signals within the B4GALNT3 locus for higher sclerostin, lower eBMD, and greater B4GALNT3 expression in arterial tissue (Probability>99%). Our findings suggest that higher sclerostin levels are causally related to lower BMD and greater fracture risk. Hence, strategies for reducing circulating sclerostin, for example by targeting glycosylation enzymes as suggested by our GWAS results, may prove valuable in treating osteoporosis. This article is protected by copyright. All rights reserved.
OBJECTIVES: Given the role of growth and differentiation factor 5 () in knee development and osteoarthritis risk, we sought to characterise knee defects resulting from loss of function and how its regulatory regions control knee formation and morphology. METHODS: The () mouse line, which harbours an inactivating mutation in , was used to survey how loss of function impacts knee morphology, while two transgenic reporter bacterial artificial chromosome mouse lines were used to assess the spatiotemporal activity and function of regulatory sequences in the context of clinically relevant knee anatomical features. RESULTS: Knees from homozygous mice () exhibit underdeveloped femoral condyles and tibial plateaus, no cruciate ligaments, and poorly developed menisci. Secondary ossification is also delayed in the distal femur and proximal tibia. mice have significantly narrower femoral condyles, femoral notches and tibial plateaus, and curvier medial femoral condyles, shallower trochlea, steeper lateral tibial slopes and smaller tibial spines. Regulatory sequences upstream from were weakly active in the prenatal knee, while downstream regulatory sequences were active throughout life. Importantly, downstream but not upstream regulatory sequences fully restored all the key morphological features disrupted in the mice. CONCLUSIONS: Knee morphology is profoundly affected by absence, and downstream regulatory sequences mediate its effects by controlling expression in knee tissues. This downstream region contains numerous enhancers harbouring human variants that span the osteoarthritis association interval. We posit that subtle alterations to morphology driven by changes in downstream regulatory sequence underlie this locus' role in osteoarthritis risk.
Vertebrate pelvic reduction is a classic example of repeated evolution. Recurrent loss of pelvic appendages in sticklebacks has previously been linked to natural mutations in a pelvic enhancer that maps upstream of . The sequence of this upstream enhancer is not conserved to mammals, so we have surveyed a large region surrounding the mouse gene for other possible hind limb control sequences. Here we identify a new pelvic enhancer, , that maps downstream rather than upstream of drives expression in the posterior portion of the developing hind limb, and deleting the sequence from mice alters the size of several hind limb structures. sequences are broadly conserved from fish to mammals. A wild stickleback population lacking the pelvis has an insertion/deletion mutation that disrupts the structure and function of , suggesting that changes in this ancient enhancer contribute to evolutionary modification of pelvic appendages in nature.
Background: Mechanistic understanding of tendon molecular and cellular biology is crucial toward furthering our abilities to design new therapies for tendon and ligament injuries and disease. Recent transcriptomic and epigenomic studies in the field have harnessed the power of mouse genetics to reveal new insights into tendon biology. However, many mouse studies pool tendon tissues or use amplification methods to perform RNA analysis, which can significantly increase the experimental costs and limit the ability to detect changes in expression of low copy transcripts. Methods: Single Achilles tendons were harvested from uninjured, contralateral injured, and wild type mice between three and five months of age, and RNA was extracted. RNA Integrity Number (RIN) and concentration were determined, and RT-qPCR gene expression analysis was performed. Results: After testing several RNA extraction approaches on single adult mouse Achilles tendons, we developed a protocol that was successful at obtaining high RIN and sufficient concentrations suitable for RNA analysis. We found that the RNA quality was sensitive to the time between tendon harvest and homogenization, and the RNA quality and concentration was dependent on the duration of homogenization. Using this method, we demonstrate that analysis of gene expression in single mouse tendons reduces the biological variation caused by pooling tendons from multiple mice. We also show successful use of this approach to analyze and gene expression changes in injured compared with uninjured control tendons. Discussion: Our work presents a robust, cost-effective, and straightforward method to extract high quality RNA from a single adult mouse Achilles tendon at sufficient amounts for RT-qPCR as well as RNA-seq. We show this can reduce variation and decrease the overall costs associated with experiments. This approach can also be applied to other skeletal tissues, as well as precious human samples.
Given GDF5 involvement in hip development, and osteoarthritis (OA) and developmental hip dysplasia (DDH) risk, here we sought to assess the role(s) of GDF5 and its regulatory sequence on the development of hip morphology linked to injury risk. The brachypodism (bp) mouse, which harbors a Gdf5 inactivating mutation, was used to survey how Gdf5 loss of function impacts the development of hip morphology. Two transgenic Gdf5 reporter BAC lines were used to assess the spatiotemporal expression of Gdf5 regulatory sequences. Each BAC line was also used to assess the functional roles of upstream and downstream sequence on hip morphology. bp/bp mice had shorter femora with smaller femoral heads and necks as well as larger alpha angles, smaller anterior offsets, and smaller acetabula, compared to bp/+ mice (p<0.04). Regulatory sequences downstream of Gdf5 drove strong prenatal (E17) expression and low postnatal (6 months) expression across regions of femoral head and acetabulum. Conversely, upstream regulatory sequences drove very low expression at E17 and no detectable expression at 6 months. Importantly, downstream, but not upstream Gdf5 regulatory sequences fully restored all the key morphologic features disrupted in bp/bp mice. Hip morphology is profoundly affected by Gdf5 absence, and downstream regulatory sequences mediate its effects by controlling Gdf5 expression during development. This downstream region contains numerous enhancers harboring risk variants related to hip OA, DDH, and dislocation. We posit that subtle alterations to morphology driven by changes in downstream regulatory sequence underlie this locus' role in hip injury risk.