Recent studies using insulin-like growth factor I (IGF-I) knockout mice demonstrate that IGF-binding protein (IGFBP)-5, an important bone formation regulator, itself is a growth factor with cellular effects not dependent on IGFs. Because IGFBP-5 contains a nuclear localization sequence that mediates transport of IGFBP-5 into the nucleus, we propose that IGFBP-5 interacts with nuclear proteins to affect transcription of genes involved in bone formation. We therefore undertook studies to identify proteins that bind to IGFBP-5 using IGFBP-5 as bait in a yeast two-hybrid screen of a U2 human osteosarcoma cDNA library. Five related clones that interacted strongly with the bait corresponded to the FHL2 gene, which contains four and a half LIM domains. Co-immunoprecipitation studies with lysates from U2 cells overexpressing FHL2 and IGFBP-5 confirmed that interaction between IGFBP-5 and FHL2 occurs in whole cells. In vitro interaction studies revealed that purified FHL2 interacted with IGFBP-5 but not with IGFBP-3, -4, or -6. Northern blot analysis showed that FHL2 was strongly expressed in human osteoblasts. Nuclear localization of both FHL2 and IGFBP-5 was evident from Western immunoblot analysis and immunofluorescence. The role of FHL2 as an intracellular mediator of the effects of IGFBP-5 and other osteoregulatory agents in osteoblasts will need to be verified in future studies.

IGF binding protein-5 (BP-5) is an important bone formation regulator. Therefore, elucidation of the identity of IGF binding protein-5 (BP-5) protease produced by osteoblasts is important for our understanding of the molecular pathways that control the action of BP-5. In this regard, BP-5 protease purified by various chromatographic steps from a conditioned medium of U2 human osteosarcoma cells migrated as a single major band, which comigrated with the protease activity in native PAGE and yielded multiple bands in SDS-PAGE under reducing conditions. N-Terminal sequencing of these bands revealed that three of the bands yielded amino acid sequences that were identical to that of R2 macroglobulin (R2M). Although R2M was produced by human osteoblasts (OBs), it was not found to be a BP-5 protease. Because R2M had been shown to complex with ADAM proteases and because ADAM-12 was found to cleave BP-3 and BP-5, we evaluated if one of the members of ADAM family was the BP-5 protease. On the basis of the findings that (1) purified preparations of BP-5 protease from U2 cell CM contained ADAM-9, (2) ADAM-9 is produced and secreted in high abundance by various human OB cell types, (3) purified ADAM-9 cleaved BP-5 effectively while it did not cleave other IGFBPs or did so with less potency, and (4) purified ADAM-9 bound to R2M, we conclude that ADAM-9 is a BP-5 protease produced by human OBs.