We found that positive regulators of TGFβ superfamily signaling, including the activated R-Smads, pSmad1 and pSmad2, the co-Smad, Smad4, the cytoplasmic scaffold proteins, SARA and C184M, and the negative regulators of Smad signaling, Smad7, TGIF, and c-Ski, localized on early endosomes in vivo. Activated R-Smads were also present on late endosomes. When BMP signaling was reduced, the level of pSmad1 on endosomes and in the fiber cell nuclei decreased.
The localization on endosomes of activated R-Smads, co-Smad and negative regulators of Smad-dependent transcription, such as TGIF and c-Ski are novel findings. These, together with the co-localization of c-Ski and its binding protein, C184M on cytoplasmic vesicles, suggests that endosomes act as platforms for the assembly of both positive and negative components of the Smad signaling pathway. Understanding the functional significance of these observations will require methods that have not yet been widely applied to cells in vivo.
Previous studies localized TGFβ superfamily signaling in vivo by detecting the distribution of pSmad1 and pSmad2 in chicken and Xenopus embryos [30, 31]. These studies used histochemical detection in paraffin-embedded tissue sections. Although this is a sensitive method, it may obscure the endosomal localization of pSmads and other mediators of TGFβ superfamily signaling. Thus, while these authors showed abundant pSmad staining in the cytoplasm of all tissues examined, this staining was not obviously localized to vesicular structures. Using confocal microscopy to view thick, detergent-permeabilized tissue sections that had not been embedded in paraffin demonstrated that pSmads and other mediators of TGFβ superfamily signaling localized to punctate cytoplasmic structures, along with EEA1 and Rab5B or Rab7.
Differences between results of the present in vivo and previous in vitro studies
The results reported in this study contrast in many ways to those obtained from the study of cultured cells treated with ligands of the TGFβ superfamily. For example, activated Smads are rarely described in the cytoplasm of cultured cells, although we found abundant endosome-associated Smads in vivo. In fact, in lens epithelial cells, activated Smads were easily seen on endosomes, but did not accumulate to appreciable levels in nuclei. These differences in localization may be cell type specific, as studies of the endosomal localization of signaling components have not previously been performed on lens cells. They may also be related to the different manner in which cells are typically exposed to growth factors in vivo and in vitro.
Smad7 has previously been found associated with cytoplasmic membranes, but these vesicles were caveolin-1-positive compartments (caveolae), not endosomes [15, 24]. These authors concluded that the pathway for the Smad7-dependent degradation of TGFβ receptors (via caveolae) is distinct from the pathway by which these receptors activate downstream components (via endosomes). When we assessed the relative distribution of Smad7 on endosomes and caveolae, our results showed that Smad7 associates to an appreciable degree with endosomes in vivo, but not with caveolae. It is possible that Smad7 is regulated via a different pathway in lens cells. Another study showed strong co-localization of the polyoma virus VP1 protein with EEA1 and of EEA1 with Caveolin-1 in mouse fibroblasts. This observation suggests that Caveolin-positive vesicles carrying the virus fuse with EEA1-positive early endosomes . This observation raises the possibility that these pathways overlap in some cell types.
We are aware of no studies describing the localization of negative transcriptional regulators of TGFβ signaling, like c-Ski and TGIF, to endosomes. Similarly, the association of Smad4 with activated R-Smads has been assumed to occur in the cytoplasm, since Smad4 has not previously been detected on endosomes. Finally, although C184M was previously found exclusively in the cytoplasm, it was not shown to associate there with endosomes . Some of these differences may be due to the imaging methods used in the present study. For instance, cytoplasmic staining for pSmad1 and 2 has been shown previously in embryonic tissues, but it was not evident that this staining was associated with vesicles (see [28, 30] for examples).
The differences between our work and previous in vitro studies may be due to the methods that have been classically used to study growth factor signaling in cultured cells. In vitro, cells are often first 'starved' for growth factors and then exposed to levels sufficient to saturate all cell surface receptors. Following this acute exposure, signaling intermediates rapidly translocate to the nucleus. In vivo, especially during development and other non-traumatic events, cells are likely to be exposed to ligand levels that increase gradually over minutes or hours, as a growth factor is synthesized and diffuses to its target. After cells are exposed to a stimulus, they are likely to activate feedback mechanisms to modulate their response to stimulation. This may account for the presence of Smad7 on endosomes in vivo, but not in cultured cells. In vitro studies suggest that phosphorylated R-Smads move rapidly from the receptor to the nucleus and do not reside in the cytoplasm for an appreciable time. Our observations show that, in vivo, a substantial fraction of the total pSmad1 and pSmad2 is, at any time, associated with endosomes in the cytoplasm.
At the time they were removed from the eye, the lens cells studied in the present work had been chronically exposed to BMPs and other members of the TGFβ superfamily for days [16, 17, 19–21]. Therefore, the localization of signaling components and complexes is likely to reflect their steady-state distribution in the cells. Our results suggest that this steady state is characterized by the endosomal association of active R-Smads, I-Smad, co-Smad and Smad effector molecules. Most of these proteins are thought to have their primary function in the nucleus. Since at steady state, the distribution of molecules within different cell compartments reflects the amount of time they spend in these compartments, our observations suggest that components of the TGFβ signaling pathway spend a substantial proportion of their time on endosomes. Understanding the functions of these endosome-associated complexes may provide a more complete picture of Smad signaling and its regulation.
In vitro studies permit the analysis of signaling pathways using sophisticated analytical methods, most of which are not yet practical for in vivo studies. Conversely, in vivo studies reveal aspects of signaling that may not be appreciated using cultured cells. The current study identified several aspects of TGFβ superfamily signaling in vivo that are not typical of what has been seen for in vitro studies. These warrant further study to determine whether they are due to differences in cell type, differences between cells in vivo and in vitro, or some of both.
Differential localization of TGFβ signaling components to the cytoplasm and nucleus of lens epithelial and fiber cells
Lens epithelial cells showed high levels of all Smad signaling components, including pSmads, in their cytoplasm, but not their nuclei. This observation suggests that there are factors that regulate whether activated Smads primarily localize to the cytoplasm or the nucleus. Smad distribution could be regulated by altering the relative rates of nuclear import and export of activated R-Smads [33, 34]. Alternatively, recent studies showed that Sno-N (ski-related novel gene), a transcriptional repressor that is related to c-Ski, suppresses TGFβ signaling by sequestering Smads in the cytoplasm . Their cytoplasmic location in vivo raises the possibility that c-Ski and C184M may play a similar role in regulating the distribution of Smads between the cytoplasm and nucleus. Our studies detected the I-Smad, Smad6, in the epithelial cell cytoplasm, but not in the fiber cells. Smad6 may, therefore, prevent the nuclear localization of Smad signaling complexes in lens epithelial cells. Finally, lens epithelial and fiber cells abut different ocular compartments, with epithelial cells exposed to aqueous humor and fiber cells to vitreous humor. Different amounts of TGFβ family members or other growth factors in these compartments might account for the differences seen in the subcellular localization of activated Smads and other TGFβ signaling components in epithelial and fiber cells.
An example of the potential complexity of cytoplasmic signaling is provided by MAPK signaling in Drosophila eye development. MAPK is activated (phosphorylated) early in eye development, but is held in the cytoplasm. When the cells are later exposed to BMP and hedgehog ligands, the activated MAPK translocates to the nucleus, where it regulates development . Although the mechanisms that regulate the subcellular distribution of activated signaling molecules are not yet well understood, in Drosophila or vertebrates, we suggest that such mechanisms function downstream of TGFβ superfamily receptors in the lens in vivo.