High Fidelity of Mouse Models Mimicking Human Genetic Skeletal Disorders

The 2019 International Skeletal Dysplasia Society nosology update lists 441 genes for which mutations result in rare human skeletal disorders. These genes code for enzymes (33%), scaffolding proteins (18%), signal transduction proteins (16%), transcription factors (14%), cilia proteins (8%), extracellular matrix proteins (5%), and membrane transporters (4%). Skeletal disorders include aggrecanopathies, channelopathies, ciliopathies, cohesinopathies, laminopathies, linkeropathies, lysosomal storage diseases, protein-folding and RNA splicing defects, and ribosomopathies. With the goal of evaluating the ability of mouse models to mimic these human genetic skeletal disorders, a PubMed literature search identified 260 genes for which mutant mice were examined for skeletal phenotypes. These mouse models included spontaneous and ENU-induced mutants, global and conditional gene knockouts, and transgenic mice with gene over-expression or specific base-pair substitutions. The human X-linked gene ARSE and small nuclear RNA U4ATAC, a component of the minor spliceosome, do not have mouse homologs. Mouse skeletal phenotypes mimicking human skeletal disorders were observed in 249 of the 260 genes (96%) for which comparisons are possible. A supplemental table in spreadsheet format provides PubMed weblinks to representative publications of mutant mouse skeletal phenotypes. Mutations in 11 mouse genes (Ccn6, Cyp2r1, Flna, Galns, Gna13, Lemd3, Manba, Mnx1, Nsd1, Plod1, Smarcal1) do not result in similar skeletal phenotypes observed with mutations of the homologous human genes. These discrepancies can result from failure of mouse models to mimic the exact human gene mutations. There are no obvious commonalities among these 11 genes. Body BMD and/or radiologic dysmorphology phenotypes were successfully identified for 28 genes by the International Mouse Phenotyping Consortium (IMPC). Forward genetics using ENU mouse mutagenesis successfully identified 37 nosology gene phenotypes. Since many human genetic disorders involve hypomorphic, gain-of-function, dominant-negative and intronic mutations, future studies will undoubtedly utilize CRISPR/Cas9 technology to examine transgenic mice having genes modified to exactly mimic variant human sequences. Mutant mice will increasingly be employed for drug development studies designed to treat human genetic skeletal disorders. Significance Great progress is being made identifying mutant genes responsible for human rare genetic skeletal disorders and mouse models for genes affecting bone mass, architecture, mineralization and strength. This review organizes data for 441 human genetic bone disorders with regard to heredity, gene function, molecular pathways, and fidelity of relevant mouse models to mimic the human skeletal disorders. PubMed weblinks to citations of 249 successful mouse models are provided.


INTRODUCTION
Rare human genetic diseases cumulatively affect about 1 in 200 individuals and involve an estimated 7,000 genes. Major research efforts are underway to identify these mutant genes and characterize their disease phenotypes. Knowledge gained can guide therapies and provide hypotheses to develop future treatments. As recently summarized (1), "Genome sequencing has revolutionized the diagnosis of genetic diseases. Close collaborations between basic scientists and clinical genomicists are now needed to link genetic variants with disease causation. To facilitate such collaborations, we recommend prioritizing clinically relevant genes for functional studies, developing reference variant-phenotype databases, adopting phenotype description standards, and promoting data sharing." Rare human genetic skeletal dysplasias affect about 1 in 5,000 individuals (2) and account for 5% of all birth defects (3). The International Skeletal Dysplasia Society (ISDS, https://www.isds. ch), promotes scientific progress in the field of skeletal dysplasias and dysostoses, meets every second year, and published skeletal nosology summaries during 2001 (4), 2006 (5), 2010 (6), 2015 (7), and 2019 (8). There are presently 441 skeletal nosology genes, with an average of 20 new genes identified yearly (Figure 1). The classification aims to (i) identify metabolic pathways active in cartilage and bone, and their regulatory mechanisms; (ii) identify cellular signaling networks and gene expression sequences implicated in skeletal development; (iii) identify candidate genes for genetic disorders; (iv) facilitate integration of data coming from spontaneous and genetically engineered mouse mutants; (v) help in developing diagnostic strategies; (vi) stimulate the design and exploration of new therapeutic possibilities; and (vii) provide a knowledge framework accessible to physicians as well as to basic scientists and thus to facilitate communication between clinical genetics and pediatrics and the basic sciences (4).
The objectives of the present review include further characterizations of these 441 skeletal nosology genes and evaluating the reliability of mutant mouse models to mimic these human skeletal disorders.

HISTORICAL HIGHLIGHTS
Short stature and other visually obvious skeletal dysplasias were apparent throughout human history (9). The discovery of X-rays by Wilhelm Röntgen (10) was quickly followed by the description of osteopetrosis by Albers-Schönberg (11) and many skeletal dysplasias during the following decades (12). Dual-energy Xray absorptiometry (DXA) technology, developed during the 1980s (13), permitting quantitation of bone mineral density (BMD), and continued advances in computed tomography (CT), providing 3 dimensional images, lead to increasing sophisticated understanding of bone dysmorphology. The first nosology gene identified was CA2 (carbonic anhydrase 2, osteopetrosis), initially in 1983 using electrophoretic, enzymatic and immunologic techniques on red blood cell extracts (14), and subsequently by genetic mutation analysis in 1991 (15). The first genetic mutation for any human disease to be identified by WES was DHODH (dihydroorotate dehydrogenase), responsible for postaxial acrofacial dysostosis, in 2010 (16).

NOSOLOGY
Nosology is the classification of diseases, which in its simplest form involves symptoms and pathogenic mechanisms. No classification system is perfect and there are often multiple ways to classify a given disorder. At the extremes, "lumpers" and "splitters" prefer few and many categories, respectively (17). Heredity can be X-linked, autosomal dominant, or autosomal recessive. Skeletal dysplasias can affect the skeleton only, or be part of pleiotropic syndromes affecting multiple organs. Mutations of various genes within a molecular pathway can each produce similar phenotypes. Loss-of function (LoF) mutations completely disrupt the activities of their encoded proteins but hypomorphic mutations allowing reduced protein activities occur. Gain-of-function (GoF) mutations increase the activities of enzymes and receptors and produce different phenotypes than LoF mutations. Dominant-negative mutations adversely affect functions of wild-type proteins. Mutations can occur within the protein-coding region of the genome (exome), within introns, or between gene coding regions. Mutations include deletions, duplications, and inversions.
The 2019 edition of the ISDS Nosology and Classification of Skeletal Disorders database organizes mutant human skeletal phenotypes into 42 groups, based on clinical observations and known gene/phenotype relationships (8). A total of 461 disorders and 441 genes are provided, when all 10 genes listed within the Notes sections of the tables ( Table 1) are included. Updated HGNU gene symbols for 11 genes ( Table 2) are employed. Supplemental Table 1 provides an alphabetical list in spreadsheet format of all 441 genes, with information on heredity, gene function and mouse model status. Genetic disorders are not listed, as mutations in many genes result in multiple phenotypes. Inheritance patterns are 242 autosomal recessive, 135 autosomal dominant, 34 autosomal recessive or autosomal dominant depending upon the exact mutation in the gene, 21 X-linked and 11 non-inherited, somatic mutations. Three genes can have either germline or somatic mutations.
RMRP encodes an RNA regulating DNA transcription, RNU4ATAC encodes an RNA that is a component of an enzyme  There are no orthologous mouse genes for human ARSE (arylsulfatase E) and RNU4ATAC (RNA, U4atac small nuclear, U12-dependent splicing). Supplemental Table 1 summarizes published data on the availability and fidelity of mouse models for the 439 human rare bone disease genes. Mutant mice with bone phenotypic data exist for 260 of the 439 genes (59%) with similar bone phenotypes observed for 249 (96%) genes. Supplemental Table 2 contains PubMed hyperlinks to publications for all 249 genes provided in Supplement Table 1 having mutant mouse bone phenotypes. These two supplemental tables should provide a major resource for the bone research community. Mutant mouse bone data are inconsistent with human skeletal phenotypes for 11 genes (Ccn6, Cyp2r1, Flna, Galns, Gna13, Lemd3, Manba, Mnx1, Nsd1, Plod1, Smarcal1). There are no obvious explanations for or commonalities among these humanmouse phenotype inconsistencies. For 97 genes (22%) mutant mice have been generated and examined, but no skeletal data were reported. Mutant mice do not appear to have been examined for 82 genes (19%) and 36 (8%) of these genes belong to the understudied Ignorome/Dark Genome (18)(19)(20). Individual laboratories and/or consortia are encouraged to examine these genes, now known to contribute to poorly understood human rare bone diseases.

SKELETAL DISORDER VIGNETTES
This section briefly summarizes selected skeletal disorders resulting from various mutations, highlighting the wide range of transcription and translation events that can be disrupted.
• Mutations can be benign with healthy nutrition but produce disease when key nutrients are lacking. All humans have an inactivating mutation in GULO, encoding an enzyme involved in the synthesis of ascorbic acid, and develop scurvy without sufficient dietary intake of vitamin C. The ascorbate synthetic pathway, involving aldehyde and aldose reductases, was only fully characterized in 2010 (31). Ascorbic acid is a required cofactor for the hydroxylation of proline and lysine residues in collagen and disruption of the mouse gulonolactone oxidase gene results in spontaneous bone fractures (32). Similarly, human and mouse HAAO and KYNU genes are involved in the synthesis of the enzymatic cofactor NAD and inactivating mutations in these human and mouse genes can result in congenital malformations (33). • X-linked human mutations comprise 6% of the total skeletal disorders. X-inactivation of one of the two X chromosomes in women by long non-coding RNA specific transcript XIST occurs, but about 20% of X chromosome genes escape this inactivation (34). AMER1 and PORCN are X-linked genes that code for components of the WNT signaling pathway, with dominant mutations in women causing osteopathia striata with cranial sclerosis and focal dermal hypoplasia (including osteopathia striata), respectively. Due to developmental lethality male patients are extremely rare, but a few males having post-zygotic mosaic mutations have been identified (35,36). Amer1 mutations in mice disrupt bone architecture (37) and treating adult mice with inhibitors of the PORCN enzyme reduces bone mass (38).

HEREDITY OF BONE MASS WITHOUT SKELETAL DYSPLASIA
Osteoporosis is a common skeletal disease in which reduced amounts of otherwise normal bone lead to fragility and fractures. Adult bone mass, even within the normal range, has a strong heredity influence (60, 61) and identifying the genes involved in bone mass accumulation during growth and loss during aging has received great interest within the context of the etiology and treatment of osteoporosis. GWAS studies over the past decade described an increasing number of genes affecting BMD, with 518 loci identified in the 2019 UK Biobank analysis (62). Juvenile osteoporosis, although not a true dysplasia as bone architecture is normal, usually has genetic causes (63,64).
There are healthy subjects with unexplained high bone mass (65,66) and attempts are underway to identify the genes responsible. Recent discoveries of such genes include LRP6 (67) and SMAD9 (68).

MOUSE MODELS
All models are wrong, but some are less imperfect than others, and many are useful -George Box Mouse models make important contributions to understanding and treating human diseases (69)(70)(71)(72), including skeletal disorders (73,74). Mutant mice that model human phenotypes also model successful drugs (75), help identify genes responsible for human genetic disorders and can provide insights for osteoporosis drug development (76). Bone mass and architecture vary in healthy humans and among laboratory mouse strains, with the most commonly studied C57BL/6 mouse strain an outlier having limb bones with high diameters and low cortical thickness (77)(78)(79)(80)(81).
The majority of mouse data summarized in this review involve individual investigator-initiated studies examining possible skeletal phenotypes in transgenic mice with specific alterations in genes chosen by the investigator. This approach, known as reverse genetics, utilizes the expertise of the laboratories involved.
In contrast, human studies involve forward genetics, with genes responsible for known skeletal phenotypes identified. Forward genetics is also employed in mouse studies, as genes responsible for spontaneous and mutageninduced skeletal malformations are identified. The Jackson Laboratories (JAX), with a long history of studying mouse strains, recently employed WES to identify 14 genes having spontaneous mutations causing bone phenotypes (82,83). Several laboratories employed N-ethyl-N-nitrosourea (ENU) in chemical mutagenesis campaigns to produce mouse lines having a wide-range of phenotypes. This approach yielded 41 genes having mutations causing bone phenotypes similar to the corresponding human skeletal disorders. These 41 genes with relevant citations are provided in Supplemental Table 3.
Two high-throughput mouse reverse genetics gene knockout phenotyping campaigns have been undertaken (84). The International Mouse Phenotyping Consortium (IMPC, www.mousephenotype.org) aims to characterize knockout mouse phenotypes for all 20,000 genes (74,85). Lexicon Pharmaceuticals' Genome5000 TM effort examining the druggable genome confirmed known bone phenotypes for 23 genes and identified 11 genes, including Notum (86), for which bone phenotypes were not previously characterized (87). Importantly, skeletal phenotypes were described for Fam20c (non-lethal Raine syndrome), Lrrk1 (osteosclerotic metaphyseal dysplasia), Pappa2 (short stature), Sfrp4 (Pyle's disease), and Slc10a7 (skeletal dysplasia) prior to knowledge of the human skeletal dysplasias when mutated in humans (84). For the 439 mouse genes discussed in this review, 149 genes have been examined by the IMPC, yielding 63 viable adult homozygous mouse mutants. Skeletal phenotypes (either body BMD or radiological dysmorphology) were observed for 28 genes. Results from the IMPC phenotyping campaign are summarized in Table 3. Mouse models of human genetic disorders are employed to evaluate potentially beneficial skeletal actions of therapies approved for other disease indications. Teriparatide treatment increases bone mass in Lrp5 KO mice mimicking humans with osteoporosis pseudoglioma syndrome from loss of function LRP5 mutations (88,89). Similarly, anti-sclerostin antibody treatment increases bone mass in mutant mouse models with low bone mass from gene disruptions (90)

MOUSE STUDY PRECAUTIONS
Several experimental pitfalls should be avoided when performing mouse studies (100).
• Knockout of individual genes can disrupt the functions of neighboring genes (101). Examples include the presence of orofacial defects resulting from a hypomorphic Pax9 allele during knockout of the neighboring Slc25a21 gene (102) and glycosaminoglycan accumulation resulting from reduced expression of the Naglu gene during knockout of the neighboring Hsd17b1 gene (103). • Transgenic Cre mouse lines are invaluable for conditionally activating or inactivating genes of interest. Several reporter genes are available for visualizing bone cells at different stages of development (104). But not all Cre lines are as specific as originally believed (105)(106)(107). Understanding these limitations is critical for experimental design and interpretation. • Quantitative PCR methods are often not optimized and MIQE (Minimum Information for the publication of qPCR Experiments) guidelines have been established (108,109). Selection of the appropriate reference gene(s) is important (110)(111)(112). • Many antibodies suffer from a lack of specificity resulting from cross-reactivity to similar epitopes present on multiple proteins. Clifford Saper in 2005, as Editor-in-Chief of The Journal of Comparative Neurology, repeatedly received ". . . distressed communications from authors . . . to withdraw papers because an antibody against a novel marker was found to stain tissue in knockout animals . . . " (113). Excellent reviews (not cited here) provide guidelines for successful antibody validation and the purposeful joviality in their titles ("Antibody  (115,116). MC3T3-E1 cell subclones vary as models of osteoblast biology (117).

LARGE ANIMAL AND ZEBRAFISH MODELS
Large animals can have advantages over rodents for understanding human genetic disease and drug development. Hypophosphatasia occurs in sheep (118) and dogs (119) having mutations in ALPL. Canine genetic skeletal disorders include mutations in ADAMTSL2-geleophysic dysplasia (120), COL1A2-osteogenesis imperfecta (121), DVL2-Robinow syndrome (122), HES7-spondylocostal dysostosis (123), and SERPINH1-osteogenesis imperfecta (124). Spontaneous mutations in chicken KIAA0586 (125) and LMBR1 (126) genes result in the expected bone phenotypes. Zebrafish are increasing contributing to our knowledge of skeletal genomics (127,128) and advantages over mouse models include acquiring data more rapidly. Zebrafish mutants have been described for several of the 441 genes in this review. One complication of zebrafish studies is that zebrafish underwent a teleost-specific whole genome duplication and have more than 26,000 protein-coding genes (129). There is a one-toone relationship between 47% of human genes and a zebrafish ortholog. There are multiple zebrafish genes associated to a single human gene, and vice versa.

DRUG DEVELOPMENT
Exciting advances are being made in developing drug treatments for patients with genetic skeletal disorders (130,131) and mouse models invariably contribute to this progress. These advances are best reviewed by the laboratories involved, but three examples are illustrative. An antibody to NOTCH2 reverses osteopenia in a mouse model of Hajdu-Cheney syndrome (132). Cinacalcet corrects hypercalcemia in a mouse model of familial hypercalcemia type 2 (133). ENPP1 enzyme replacement therapy improves blood pressure and cardiovascular function in a mouse model of generalized arterial calcification of infancy (134).
Understanding genetic skeletal disorders provides key knowledge for developing osteoporosis therapies (76,135). Disruptions in genes coding for proteins in the RANK-RANKL-osteoprotegerin signaling pathway involved in osteoclast generation cause human skeletal disorders. The RANKL neutralizing antibody denosumab is a successful osteoporosis therapy. The recently approved anabolic osteoporosis treatment romosozumab, a sclerostin neutralizing antibody, was developed with knowledge gained from subjects with osteosclerosis resulting from SOST gene mutations. Subjects with pinocytosis have mutations in the cathepsin K coding gene CTSK. Treatment with odanacatib, an inhibitor of cathepsin K in osteoclasts, reduced bone fractures in postmenopausal women but cardiovascular side effects precluded regulatory approval.

FUTURE DIRECTIONS
Since many human disorders involve hypomorphic, gain-offunction, dominant-negative and intronic mutations, future studies will undoubtedly utilize CRISPR/Cas9 technology and other evolving techniques to examine transgenic mice having genes modified to exactly mimic variant human sequences (72,136). RNA sequencing will increasingly be employed for diagnosis and mechanistic understanding of genetic diseases (137)(138)(139)(140)(141).
The IFMRS (International Federation of Musculoskeletal Research Societies), in collaboration with the Broad Institute, is establishing a Musculoskeletal Genomics Knowledge Portal (MGKP) to integrate, interpret and present human data linked to musculoskeletal traits and diseases (http://www.kp4cd.org/ about/bone).

AUTHOR CONTRIBUTIONS
RB performed the literature search, data analyses, and prepared the manuscript. CO provided helpful suggestions and reviewed the manuscript.