Originally discovered as cell surface proteins with key developmental roles in the nervous system, recent data have confirmed a role for the RGM family in BMP signaling. Due to the many and varied functions of TGFβ superfamily ligands, multiple levels of regulation are required to tightly control cytokine function. In the BMP signaling pathway, a primary regulatory mechanism is modulation of ligand availability for its receptors. Preventing the BMP/receptor interaction is accomplished by a host of means, including extracellular BMP antagonists, receptor expression, and the presence of BMP pseudoreceptors. In contrast to these inhibitory mechanisms, the recent identification of the RGMs as BMP co-receptors offers a mechanism for enhancement of BMP signaling. Co-receptors, or accessory receptors, have been similarly involved in signaling for several other TGFβ superfamily members: Cripto has been shown to be essential for BMP16/Nodal signaling [18
], TGFβ type III receptor (also called betaglycan) is involved in TGFβ2 signaling [35
], and endoglin has been implicated as an accessory protein in multiple TGFβ superfamily receptor complexes [36
To better define the roles of the RGMs as co-receptors in BMP signaling, cellular mRNA for each RGM was knocked down via RNAi in cell culture and BMP responsiveness was examined. Additionally, TGFβ responsiveness was examined with the same approach. Interestingly, both BMP activity assays (BRE-luc and ALP enzyme activity) show that loss of either RGMa or RGMc dramatically decreases BMP2 and BMP12 responsiveness. The loss of RGMb has a more modest effect, though still significant. The same results were observed in RGM knockdown experiments that were carried out over a range of BMP2 concentrations (Figure ). Although it has been shown that BMP12 can signal through the BMP2 pathway [29
], the high dose needed to stimulate a BMP12 response in our experimental system suggests that we may be forcing BMP12 to use this pathway. It is possible that lower doses of BMP12 may be able to signal through an as-yet uncharacterized signaling pathway, and the role of the RGMs in this hypothetical pathway are unknown.
To be sure the observed effects were a result of specific RNA degradation, and not off-target effects mediated by the siRNA duplexes, individual experiments were carried out using four unique siRNA nucleotide sequence targets for each RGM co-receptor. Knockdown of RGMa, RGMb, and RGMc was successfully accomplished by each of the four siRNA duplexes targeting the proteins. Additionally, the observed assay results were virtually identical regardless of which of the four sequences was used to suppress mRNA levels (data not shown). As a negative control, ON-TARGET plus si CONTROL non-targeting pool siRNA duplexes (i.e., "scramble") were transfected in parallel with the RGM-targeting siRNA reagents. Taken together, the data indicate that all 3 coreceptors play roles in potentiating BMP signaling, as knockdown of each significantly reduces BMP response in cell culture.
Previous studies have shown that transient over-expression of RGMa, RGMb, or RGMc significantly increases the sensitivity of BMP-responsive cells in culture [20
]. Additionally, we observe that transient transfection of each RGM-Fc fusion protein in cell culture shows a repression of BMP activity (data not shown), demonstrating that the soluble forms of these co-receptors are capable of binding to and antagonizing BMP when not bound to the cell membrane.
Given their critical role in BMP signaling, experiments were carried out with the aim of quantifying the interactions between the RGM co-receptors and BMP2 or BMP12. The RGMs have been observed to bind qualitatively to BMP proteins previously, with a published equilibrium dissociation rate constant (kD
) only for the RGMb:BMP2 interaction [19
]. We therefore sought to quantitatively analyze the binding of BMP2 and BMP12 to each RGM protein using a Biacore platform. The binding results show that all three RGM proteins bind with very high affinity (equilibrium dissociation constants, kD
, in the low nanomolar range) to both BMP2 and BMP12. These tight affinities further confirm the observations that each RGM protein is capable of binding directly to the BMP ligands. This suggests that in their native GPI-bound form on the cell membrane, the RGM co-receptors are able to directly bind the BMP ligands; it is possible that this direct binding may lead to enhanced cytokine recruitment to the type I/II receptor complexes.
Perhaps the most interesting data from our studies is evidence that the RGM co-receptors are not able to compensate functionally for one another. All 3 RGMs are expressed in our cell system, C2C12; if the RGMs have redundant activities, knocking down levels of one co-receptor would not lead to a significant phenotypic result due to the presence of the other two. In such a case, knockdown of all three receptors would be needed before the signaling pathways are affected. However, our data show that reduced expression of a single RGM protein suppresses BMP activity in cells. The collective data suggest a mechanism in which the signaling BMP receptor complex requires two distinct RGM co-receptors for efficient signaling, though it is unclear whether they exist as a heterodimer or two distinct proteins. In this model, the possibilities are RGMa/RGMb, RGMa/RGMc, or RGMb/RGMc. In our cell system, RGMa/RGMc is the most effective combination, as suppression of either one significantly blocks BMP activity. Since knockdown of RGMb has the least significant effect, this suggests that complexes involving RGMb form less effective signaling complexes. Previous studies have shown that the RGM proteins have the ability to participate in BMP-receptor complexes. Thus, it appears that the presence of two RGM proteins enhances BMP ligand recruitment to the membrane-bound receptors, and helps generate an active signaling complex. Our data indicate that RGMa and RGMc form the most potent co-receptor pair in the context of our in vitro cell system. In accord with previous results, we observe no involvement of the RGM proteins in TGFβ signaling.
According to our proposed mechanism, in which the RGMa/RGMc combination of co-receptors is most potent at facilitating BMP signaling, cells that overexpress this combination should show improved BMP signaling versus cells that overexpress RGMa or RGMc individually. Experiments in which cells are transiently transfected with the same total amount of empty vector, RGMa, RGMc, or a RGMa/RGMc combination show a statistically significant (p value < 0.05, n = 4) increase in BMP activity (BRE-luc readout) when the RGMa/RGMc combination of co-receptors is expressed versus RGMa or RGMc individually. Overexpression of RGMa or RGMc individually also increase BMP signaling versus empty vector controls, consistent with previously published data [19
In addition to our cell-based assay data, several lines of evidence support our proposed mechanism. Analysis of endogenous RNA expression levels in a variety of BMP responsive cell lines (C2C12, C3H10 T 1/2, 3T3L1, and MLB13MYC Clone 14) shows that RGMa and RGMb RNA is abundant in each cell line, while RGMc RNA is appreciably expressed in C2C12 and MLB13MYC Clone 14 (data not shown). This expression data supports the hypothesis that two of the RGM co-receptors are needed for effective BMP signaling, since all tested BMP-responsive cell lines express at least two of the RGM proteins. Also, there is evidence that the RGM proteins do not functionally compensate for one another. It has been shown that mutations to the gene encoding RGMc are a leading cause of juvenile hemochromatosis, presumably due to loss of functional RGMc protein. The fact that RGMa and RGMb do not rescue this disease phenotype (when caused by mutated RGMc) suggests that they are not functionally redundant with RGMc protein. While all these results support our proposed mechanism, further experiments are required to more fully elucidate it. Alternatively, other explanations are possible. For example, the RGMs may act as monomeric coreceptors that can operate by different signaling routes; these pathways may be functionally distinct in certain BMP-induced activities (such as those measured in our experiments) but not in others. Additional studies are required to fully understand the mechanisms through which the RGMs regulate BMP signaling.