CRISPR-mediated conditional mutagenesis of Smad1/5/8 reveals BMP/GDF signaling restricts postnatal bone overgrowth

The BMP/GDF branch of TGF-{beta} signaling regulates diverse aspects of skeletal biology, from skeletal development to maintenance and repair. However, the complexity, redundancy, and pleiotropy of BMP/GDF signaling have hamstrung a genetic dissection of its activities in different cell types over time. Here, we tested the feasibility of a three-transgene system using CRISPR/Cas9 to conditionally mutate six target sites, two each in the receptor-mediated Smad1, Smad5, and Smad8 transcriptional effectors of BMP/GDF signaling. Briefly, we used Prx1-cre to activate a conditional Cas9 transgene by recombination in early limb bud mesenchyme; this endonuclease then complexes with gRNAs expressed from a polycistronic tRNA-gRNA array for targeted mutagenesis. Slower than expected accumulation of gRNA-directed mutations in each Smad produced an unexpected postnatal skeletal phenotype. Beginning around one month after birth, all animals developed hyperostosis on the surface of all long limb bones, which progressively worsened with age. This woven bone expansion occurred through proliferation of RUNX2+ osteoprogenitor cells in the cambium layer of the periosteum, producing an abundance of periosteal osteoblasts. Endosteal osteoblasts did not increase in number but increased their mineralizing activity. As a result, the marrow cavities narrowed, and the patella and carpal elements, which have no periosteum, increased internal bone mass without altering shape and size. Thus, while BMP/GDF signaling is known to promote early postnatal bone growth, these data support an additional homeostatic role during late postnatal osteogenesis by regulating both periosteal and endosteal osteoblasts. Although this genetically simple approach requires further optimization to improve efficiency, combining three transgenes produced more than 160 conditionally mutagenized animals with a fully penetrant and reproducible phenotype. This is an advance over traditional cre/lox systems that scale in complexity with the number of target loci, and it highlights the potential to model a wide range of genetically complex traits and disorders.