Myotubularin-related Protein 4 (MTMR4) Attenuates BMP/Dpp Signaling by Dephosphorylation of Smad Proteins*

Background: The intensity and duration of phosphorylation levels of R-Smads are required for precise control of BMP signaling. Results: MTMR4 associated with and dephosphorylated the activated R-Smads in cytoplasm. Conclusion: MTMR4 attenuates BMP signaling via its DUSP activity. Significance: This study describes a novel role of MTMR4 as a negative modulator essentially involved in homeostatic BMP signaling. Bone morphogenetic proteins (BMPs) signaling essentially regulates a wide range of biological responses. Although multiple regulators at different layers of the receptor-effectors axis have been identified, the mechanisms of homeostatic BMP signaling remain vague. Herein we demonstrated that myotubularin-related protein 4 (MTMR4), a FYVE domain-containing dual-specificity protein phosphatase (DUSP), preferentially associated with and dephosphorylated the activated R-Smads in cytoplasm, which is a critical checkpoint in BMP signal transduction. Therefore, transcriptional activation by BMPs was tightly controlled by the expression level and the intrinsic phosphatase activity of MTMR4. More profoundly, ectopic expression of MTMR4 or its Drosophila homolog CG3632 genetically interacted with BMP/Dpp signaling axis in regulation of the vein development of Drosophila wings. By doing so, MTMR4 could interact with and dephosphorylate Mothers against Decapentaplegic (Mad), the sole R-Smad in Drosophila BMP pathway, and hence affected the target genes expression of Mad. In conclusion, this study has suggested that MTMR4 is a necessary negative modulator for the homeostasis of BMP/Dpp signaling.


Bone morphogenetic proteins (BMPs) signaling essentially regulates a wide range of biological responses. Although multiple regulators at different layers of the receptor-effectors axis have been identified, the mechanisms of homeostatic BMP signaling remain vague. Herein we demonstrated that myotubularin-related protein 4 (MTMR4), a FYVE domain-containing
dual-specificity protein phosphatase (DUSP), preferentially associated with and dephosphorylated the activated R-Smads in cytoplasm, which is a critical checkpoint in BMP signal transduction. Therefore, transcriptional activation by BMPs was tightly controlled by the expression level and the intrinsic phosphatase activity of MTMR4. More profoundly, ectopic expression of MTMR4 or its Drosophila homolog CG3632 genetically interacted with BMP/Dpp signaling axis in regulation of the vein development of Drosophila wings. By doing so, MTMR4 could interact with and dephosphorylate Mothers against Decapentaplegic (Mad), the sole R-Smad in Drosophila BMP pathway, and hence affected the target genes expression of Mad. In conclusion, this study has suggested that MTMR4 is a necessary negative modulator for the homeostasis of BMP/Dpp signaling.
Bone morphogenetic proteins (BMPs) 4 belong to the TGF␤ super family, and have a diverse array of functions in the devel-opment of mutilcellular organisms (1). Recent evidence has also shown that BMP signaling regulates the self-renewal and proliferation of stem cells (2,3), and prolongs the lifespan of stem cells (4). The hierarchy and function of BMP signaling pathway are highly conserved in the metazoan organisms (5)(6)(7). After binding to BMPs, the serine-threonine kinase transmembrane receptors activate a signal cascade through intracellular regulatory Smad proteins (R-Smads). Phosphorylated R-Smads form a ternary complex with the common partner Smad (Co-Smad) and translocate to the nucleus to activate transcription of a spectrum of effector genes (8). There are two members of BMPs in Drosophila. Decapentaplegic (Dpp) is the homolog of BMP2/4 (9) and glass bottom boat (Gbb) is the homolog of BMP5/6/7 (10). These two Drosophila BMP members play a major role in cell proliferation and/or cell survival (11), and are required for the vein formation (12) during Drosophila wings development (13). Thickveins (tkv) serves as the type I BMP receptor that mediates Dpp signaling during wing morphogenesis and other stages of fly development (11,14,15), and Mathers against Decapentaplegic (Mad) turns out as the sole R-Smad in BMP signaling of Drosophila (16,17).
The intensity and duration of phosphorylation levels of BMP receptors and R-Smads are required for precise control of BMP signaling (18,19). A few nuclear phosphatases for R-Smads have thus far been identified, such as pyruvate dehydrogenase phosphatase (PDP), PPM1A (protein phosphatase magnesiumdependent 1A), and small C-terminal domain phosphatase (SCP2/Os4) (19 -21). However, the existence and roles of cytoplasmic phosphatases have remained elusive; it has been recently revealed by us that a new endosomal phosphatase, myotubularin-related protein 4 (MTMR4), can specifically temper TGF␤ signaling (22). MTMR4 belongs to myotubularin family and contains tyrosine/dual-specificity phospha-tase (DUSP) activity (23) and functions in early endosomes (22,24). MTMR4 specifically interacts with and dephosphorylates the activated R-Smads to keep TGF␤ signaling in homeostasis (22). It therefore would be highly interesting to determine whether MTMR4 is also critically involved in BMP signaling.
In this study, we managed to demonstrate that MTMR4 is also an essential negative regulator of BMP signaling pathway. MTMR4 directly bound to and dephosphorylated the activated Smad1 via the DUSP enzymatic domain. Thus, overexpression of MTMR4 inhibited BMP-induced gene expression by accelerating Smad1 dephosphorylation, whereas knockdown of MTMR4 by siRNA enhanced BMP signaling with sustained Smad1 phosphorylation. We further demonstrated in vivo that MTMR4 and its Drosophila homologous gene CG3632 critically modulated the vein formation of Drosophila wings, by specifically targeting Mad activation. Therefore, MTMR4 may possess a conserved attenuator activity on R-Smads activation that is required for homeostatic BMP/Dpp signaling.
Cell Culture, Transfection, and Reporter Assays-HEK293T, HeLa, and HepG2 cells were grown in DMEM (Invitrogen), and S2 cells were cultured in Schneider's insect medium (Sigma). All media were supplemented with 10% heat-inactivated FBS (Invitrogen), 1% penicillin, and streptomycin (Invitrogen). The lipofectamine method (Invitrogen) was used to transfect HeLa and Drosophila S2 cells and calcium-phosphate method for 293T cells. Transient co-transfection and Luciferase reporter assays were performed as previously described (28). Data represent the average of triplicate experiments (mean Ϯ S.D.). Recombinant BMP2 and Dpp ligand proteins were purchased from R&D (Minneapolis, MN).
Immunoprecipitation and Western Blotting-Immunoprecipitation was carried out exactly as previously described (22). Antibodies against pSmad1/5/8 (phosphor-SXS) and Smad1 were from Cell Signaling Technology (Beverly, MA). GFP antibody was from Zymed Laboratories Inc. (San Diego, CA) and MTMR4 antibody from Abgent (San Diego, CA). Flag and ␤-actin antibodies were from Sigma.
RNA Interference-The 21-nucleotide siRNA duplexes targeting MTMR4 and scrambled siRNA were synthesized and purified as previously described (22). The efficiency was measured by Western blot using MTMR4 antibody.
Fluorescence Microscopy-S2 cells in 35-mm plates were transfected with various RFP/GFP fusion plasmids indicated (1 g each). After 36 h, the transfectants were seeded at 80% confluence on glass bottom tissue culture dishes and serumstarved overnight. Cells were then exposed to Dpp (40 ng/ml) for 2 h before fixed in 4% formaldehyde for 15 min at room temperature. Images were taken using a Zeiss fluorescent microscope (Observer Z1, Germany) and analyzed by Imag J software. DAPI (1:1000 dilution) was used to counter stain the nuclei.
Drosophila Stocks and Experimental Genotypes-All stocks were cultured at room temperature in standard cornmeal/molasses/agar media. The Drosophila stocks used in this study were as follows: UAS-gbb and UAS-dpp (described in Flybase), UAS-tkvRNAi (II), UAS-tkvRNAi (III), and vg-gal4 (kind gifts from Matthew Gibson, Stowers Institute for Medical Research). The Gateway cloning system (Invitrogen) was used to generate UAS-CG3632, UAS-MTMR4, and UAS-MTMR4C407S constructs according to manufacturer's manual, after cg3632, mtmr4, and mtmr4-C407S gene fragments were PCR amplified. Different plasmid constructs were injected into w1118 embryos to generate transgenic lines. To determine whether a moderate reduction of BMP signaling and overexpression of MTMR4 has any effect on Drosophila wing patterning, the vg-gal4 virgins were crossed with UAS-tkvRNAi (II),

MTMR4
Attenuates BMP Signaling-Because BMP and TGF␤ both belong to TGF␤ superfamily, that MTMR4 attenuates R-Smads activation in TGF␤ pathway (22) prompted us to test whether the conserved DUSP activity of MTMR4 would also play a role in BMP signaling. To test this hypothesis, first we transiently transfected two BMP-responsive luciferase reporters in HeLa cells, namely BRE-lux and GCCG-lux which contained multiple BMP-specific R-Smads binding elements GCCG (25,29). The R-Smad responsive reporter activities increased by ϳ2.5-fold post-BMP2 treatment for 16 h, which was effectively suppressed by co-expressed MTMR4 (Fig. 1 Fig. S1B for the knockdown efficiency), as expected, cells became more responsive to BMP2, as demonstrated by much higher BRE/GCCG reporter activities than those treated with scrambled RNAi (Fig. 1, C and D). Therefore, MTMR4 would be required to attenuate BMP signaling. To substantiate this notion, we went on to further test whether MTMR4 affects the expression of BMP target genes. For this purpose, we examine the transcriptional levels of Id-1, a regulator of cell proliferation and differentiation (26,30), and p21, a cell cycle inhibitor (31), both of which are well-known target genes of BMP/R-Smads signaling. Quantitative PCR analysis showed that the expression levels of Id1 and p21 was effectively suppressed by overexpressed MTMR4 (Fig. 1, E and F) and enhanced by knockdown of endogenous MTMR4 (Fig. 1, G and H) after BMP2 stimulation. Taken together, these results suggested that MTMR4 could attenuate BMP signaling, possibly via directly interrogating R-Smad activation.
MTMR4 Physically Interacts with Smad1/5-To validate the aforementioned hypothesis, we set out to examine whether MTMR4 interacts with those R-Smads of BMP pathway. Coimmunoprecipitation experiments showed that YFP-MTMR4 transiently expressed in 293T cells readily associated with Flag-Smad1 ( Fig. 2A) or HA-Smad5 (Fig. 2B), two of R-Smads in BMP signaling pathway. Such intermolecular interactions also occurred under physiological conditions since the endogenous MTMR4 and Smad1 associated with each other, and interestingly enough, only after BMP2 stimulation (Fig. 2C). Moreover, the amount of MTMR4 associated with Smad1 reduced proportionally to the phosphorylation levels of Smad1, suggesting that MTMR4 might preferentially interact with the activated Smad1 (Fig. 2C).
To map the regions of contact between MTMR4 and Smad1, we further performed co-immunoprecipitation experiments using an array of truncational mutants of MTMR4 and Smad1 that transiently co-expressed in 293T cells (supplemental Fig.  S2, A and B for the scheme of different mutants). The results showed that MTMR4 utilized its N-terminal region (MTMR4⌬FYVE, amino acids 1-591), which contains the conserved DUSP phosphatase domain, to contact Smad1 (Fig. 2D). Reciprocally, the MH2 domain, but not MH1 domain or linker region, of Smad1 was involved in contact with MTMR4 (Fig.  2E). These results suggested that the phosphorylated MH2 domain of Smad1 might serve as a physiological substrate of MTMR4.
MTMR4 Accelerates Smad1 Dephosphorylation-The aforementioned physical interaction between MTMR4 and the acti- vated Smad1 suggested that BMP signaling could be modulated by MTMR4 via targeting Smad1 phosphorylation. To test this hypothesis, HeLa cells were co-transfected with MTMR4 and Smad1, and the phosphorylation status of Smad1 was measured by a phosphor-SXS antibody specific to phosphorylated S463/ S465 residues in the MH2 domain of Smad1, which are essential for Smad1 activation (6). In the absence of MTMR4, Smad1 was phosphorylated rapidly, and it reached a peak around 30 min post-BMP2 treatment. Then the phosphorylation levels of Smad1 gradually decayed up to 4 h. Overexpression of MTMR4 did not alter the onset of Smad1 phosphorylation, however, it accelerated Smad1 dephosphorylation just after the peak time (30 min) of BMP2 treatment (Fig. 3A). This enhanced dephosphorylation of Smad1 was not due to its proteasomal degradation, because the levels of Smad1 protein remained the same in the presence or absence of overexpressed MTMR4 (Fig. 3A). Moreover, siRNA knockdown of the endogenous Mtmr4 gene in HeLa cells significantly attenuated Smad1 dephosphorylation when compared with the scrambled siRNA control (Fig.  3B). Consistent with the action of MTMR4 after phosphorylation of Smad1, overexpressed MTMR4 affected little BMP-target gene expression in the early time points and became more inhibitive in the later phase (supplemental Fig. S3). Therefore, these results suggested that MTMR4 control the velocity of Smad1 dephosphorylation, which could serve as a necessary mechanism to keep BMP activation signaling in homeostasis.
To further detect the function of MTMR4 in dephosphorylation of Smad1, we then transiently expressed the "phosphatase dead" mutant (MTMR4-C407S) or the mutant defective in early endosome localization (MTMR4⌬FYVE) of MTMR4 (supplemental Fig. S2, A and C for the scheme of different mutants). Measurement of phosphorylation levels of Smad1 upon BMP2 stimulation indicated that neither the onset nor the decay of Smad1 phosphorylation was affected by MTMR4-C407S or MTMR4⌬FYVE (Fig. 3C). Indeed, both BRE and GCCG reporter assays showed that overexpressed MTMR4-C407S or MTMR4⌬FYVE (Fig. 3, D and E) could no longer suppress BMP activation as compared with wild type MTMR4 (Fig. 1, A and B). These results therefore suggested that Smad1 dephosphorylation would require an endosome-localized phosphatase activity of MTMR4.

MTMR4 Impacts Vein Formation of Drosophila Wings-
The regulatory specificity of BMP signaling can be conveniently scored by visual phenotypes of vein formation in Drosophila wings (13). Thus, to emphasize the biological significance of MTMR4 in the control of BMP signaling in vivo, we then assessed whether ectopically expressed MTMR4 might affect BMP signaling in the development of Drosophila wings. First we tissue-specifically knocked down BMP type I receptor tkv expression in wings by UAS-tkvRNAi driven by vg-gal4 (32). Roughly 63% (n ϭ 189) of vg-gal4/ϩ;UAS-tkvRNAi(III)/ϩ wings (Fig. 4B) showed branches at posterior crossvein (PCV) compared with normal wings of vg-gal4/ϩ (Fig. 4A) or UAS-tkvRNAi(III)/ϩ (supplemental Fig. S4A) controls (Fig. 4J for statistical analysis). The same branching phenotype was obtained in another independent strain, UAS-tkvRNAi(II)/vg-gal4 (81% (n ϭ 200), supplemental Fig. S4B, F for representative phenotype). These results would represent the first line of evidence that tkv was required for vein development in wings. We then generated three lines of transgenic flies, UAS-MTMR4, UAS-MTMR4-C407S, and UAS-CG3632, respectively. CG3632 is a Drosophila homologue of MTMR4 that shares all the conserved functional domains including the phosphatase domain (supplemental Fig. S2C) typified for the catalytic subgroup of MTM super family (23). Interestingly, wings containing overexpressed mtmr4 (UAS-MTMR4/vg-gal4) or cg3632 (UAS-CG3632/vg-gal4) also developed the similar branched PCV as to those in tkv knockdown lines, albeit at slightly lower rates (20.6% (n ϭ 126) for UAS-CG3632/vg-gal4 and 24.2% (n ϭ 120) for UAS-MTMR4/vg-gal4, respectively) ( Fig. 4, C, D, and J). In contrast, the wings developed normally in the UAS-MTMR4-C407S/vg-gal4 line (Fig. 4E, J, n ϭ 150). The phenotypic similarity in PCV development would suggest that MTMR4 might play a role in tkv signaling, and more importantly, by tempering Dpp activation signaling when MTMR4 was overexpressed.
MTMR4 Is a Conserved BMP Attenuator-The genetic interaction between MTMR4 and Gbb/Tkv signaling pathway pro- moted us to test whether MTMR4 could interact with Mad, the only R-Smad in Drosophila BMP pathway. We transiently cotransfected HEK293T cells with YFP-MTMR4 and HA-Mad.
Co-immunoprecipitation showed that there was indeed a intermolecular interaction between Mad and MTMR4 (Fig. 6A). We then analyzed whether the phosphorylation levels of the endogenous Mad was affected by MTMR4 in Drosophila S2 cells.
Immunoblotting with a phosphor-Smad1 antibody showed that Mad phosphorylation peaked at 30 min after Dpp treatment of S2 cells, and gradually decreased till 2 h (Fig. 6B). Overexpression of MTMR4 carried by a Drosophila expression vector (pAGW-MTMR4) did not apparently affect the onset of Mad phosphorylation. However, it enhanced Mad dephosphorylation in response to Dpp stimulation (Fig. 6B). Cotransfection of the DUSP null mutant of MTMR4 (pAGW-C407S), as expected, failed to alter the activation kinetics of Mad (Fig. 6B). In agreement with the effect of MTMR4 on Mad dephosphorylation, fluorescent microscopy showed that nuclear translocation of RFP-Mad in response to Dpp stimulation (Fig. 6C, top two

DISCUSSION
Homeostasis of BMP signaling is critical for numerous biological processes. Among various modulators, phosphorylation-dephosphorylation of R-Smads provides undoubtedly efficient fine tuning. The observations in this work suggest that MTMR4/CG3632 might serve as a critical endosomal phosphatase that controls activation status of R-Smads/Mad in the BMP signaling pathway. MTMR4/CG3632 probably binds to the phosphorylated SXS-motif of R-Smads, and such physical proximity allows DUSP domain to dephosphorylate the activated R-Smads/Mad and consequently, attenuate transcriptional activation of target genes of BMP/Dpp. Intriguingly, just because MTMR4 is early endosome localized, the physical constrain would prohibit MTMR4 from affecting the phosphorylation of R-Smads/Mad in cytosol, or the subsequent translocation and accumulation of phophorylated R-Smads/ Mad in endosomes. The supportive evidence was shown by observation that Smad1 or Mad phosphorylation occurred normally (30 min after BMP2/Dpp stimulation) even in the presence of overexpressed MTMR4. Thus, overexpressing MTMR4 does not alter the onset of R-Smads phosphorylation. This indicates that MTMR4 inhibits BMP signaling might be not due to the redistribution of Endofin or related proteins (37) although some MTMs could result in PI3P diminution (33). Such an endosomal attenuator would be advantageous for cells to avoid R-Smads from being overactivated. BMP activity gradients are critical for Drosophila wing patterning (38,39). The finding that MTMR4/CG3632 modulated BMP signaling in vein formation strongly suggested that an endosomal DUSP activity would be critical in fine tuning of R-Smad activation status for proper BMP signaling. The precise spectrum of MTMR4/CG3632 target genes remains to be investigated. It is worthy of note that a great proportion of "severe" vein deformity appears in UAS-MTMR4/vg-gal4;UAS-tkvRNAi(III)/ϩ wings. This is likely that the increased amount of dephosphorylated Mad could activate the canonical Wingless signaling, because Wingless and BMP signal pathways are by large partitioned by the phosphorylation status of Mad (40). Therefore, besides modulating the homeostasis of BMP/Dpp signaling, MTMR4 activity per se has to be tightly controlled to avoid unnecessary crosstalk with other signaling pathways. This speculation has yet to be experimentally confirmed.
The involvement of multiple protein phosphatases at multiple layers of R-Smads activation axis has presented a complex network of TGF␤/BMP signaling. It has become more complicated that some phosphatases are dual functional in both TGF␤ and BMP signalings while some are unique to certain pathway. For example, PDP is a Mad-specific phosphatase in Drosophila BMP signaling (21), but with no effect on Smad2 dephosphorylation in TGF␤ signaling pathway (21,41). In contrast, PPM1A can dephosphorylate all R-Smads members in the nucleus (19,41). MTMR4 functions on both Smad2/3 and Mad, suggesting a PPM1A-like dual player on both TGF␤ and BMP pathways. Of course, MTMR4 differs from PPM1A for its subcellular localization, thus provides different checkpoint of BMP activation. Moreover, the new role of MTMR4 on vein development of Drosophila revealed by this study would provide insight into better understanding of hypoxic response that MTMR4 might be involved, such as the malignancy of papillary thyroid cancer (42) and hypertensive fibrogenesis (43).