To the editor—In the March issue of Nature Medicine, Horwitz and colleagues presented results of bone marrow transplantation (BMT) in three children with osteogenesis imperfecta1 (OI). Their results should be interpreted with caution.

Growth, bone density, histology and fracture number are important outcome parameters, but are not simple to measure accurately and reproducibly in children with severe OI. For example, using simple crown-to-heel measurements, it is difficult to distinguish increased length due to growth and increased length due to decreased contractures after physical therapy. Their patient 3 was followed independently at the NIH Clinical Center. Lumbar bone density (L1–L4) measured 4 months before and 9 months after BMT shows no significant change (0.133 gm/cm2 to 0.115 gm/cm2; –7.4 s.d. compared with age-matched controls).

The most troubling aspect of this article, however, is the contrast between the low level of osteoblast engraftment achieved and the dramatic changes reported in skeletal parameters. The iliac crest-derived osteoblasts for patients 1 and 2 were culture-expanded in my laboratory and tested at p1. Although I am confident of the 1.5% engraftment in patient 1 and of the mesenchymal type of the cells, I am doubtful that the presence of 1.5–2% normal osteoblasts could lead to a fourfold increase in osteoblast number, a 45–78% increase in bone mineral content and substantial growth.

Although osteoblast replacement is a valid approach for the treatment of OI, the level of engraftment required for clinical success is unknown. This assertion is based on data from mosaic carriers of OI. These individuals have a post-zygotic mutation, as evidenced by the occurrence of the collagen mutation in some but not all of their cells. As they are themselves clinically unaffected or have only very mild symptoms, they are usually recognized when they produce children with full expression of the mutation and a severe skeletal phenotype. The mosaic parents have a great deal to teach us about developmental patterns and the goals of gene therapy at the bone level. Although parental mosaicism is not rare in OI, molecular data on mosaic individuals is, having been reported in only 18 cases. Three of these have a high percent (approaching 100%) mutant cells in dermal fibroblasts, four have intermediate levels (50–75%) and five have undetectable levels2,3,4. Significantly, each mosaic individual has variation in the level of mutant cells in different tissues. Clearly, to determine the levels of normal cells required to affect skeletal phenotype, one must examine bone. However, neither bone tissue nor osteoblasts have been studied in any mosaicism cases. Thus, we lack the human context required before we can conclude what extent of osteoblast engraftment should be the goal of OI therapy. Bone data from mosaic parents can answer this important question.

Horwitz et al. suggest that normal collagen fibrils will be preferentially incorporated into and retained by bone matrix, allowing a small number of cells to dominate the composition of the bone matrix. The limited in vitro and in vivo data on bone matrix composition in OI reveals a complex structure with quantitative abnormalities of several noncollagenous proteins5, as well as structurally abnormal collagen. Normal collagen may not have a selective survival advantage compared with mutant collagen at the bone matrix level. In several cases, mutant collagen was more efficiently incorporated into bone and into matrix deposited by osteoblasts in culture than into dermal matrix6. We have unpublished data showing that normal and mutant collagen 'chase' from cultured osteoblast matrix at the same rate (A. Forlino and J.C.M., unpublished data), consistent with the random assortment of collagen helices to form fibrils in the extracellular matrix.