Endocytic deficiency induced by ITSN-1s knockdown alters the Smad2/3-Erk1/2 signaling balance downstream of Alk5

Recently, we demonstrated in cultured endothelial cells and in vivo that deficiency of an isoform of intersectin-1, ITSN-1s, impairs caveolae and clathrin-mediated endocytosis and functionally upregulates compensatory pathways and their morphological carriers (i.e. enlarged endocytic structures, membranous rings or tubules) that are normally underrepresented. We now show that these endocytic structures internalize the broadly expressed transforming growth factor β receptor I (TGFβ-RI or TGFBR1), also known as Alk5, leading to its ubiquitylation and degradation. Moreover, the apoptotic or activated vascular cells of the ITSN-1s-knockdown mice release Alk5-bearing microparticles to the systemic circulation. These interact with and transfer Alk5 to endocytosis-deficient endothelial cells, resulting in lung endothelial cell survival and phenotypic alteration towards proliferation through activation of Erk1 and Erk2 (also known as MAPK3 and MAPK1, respectively). We also show that non-productive assembly of the Alk5–Smad–SARA (Smad anchor for receptor activation, also known as ZFYVE9) signaling complex and preferential formation of the Alk5–mSos–Grb2 complex account for Erk1/2 activation downstream of Alk5 and proliferation of pulmonary endothelial cells. Taken together, our studies demonstrate a functional relationship between the intercellular transfer of Alk5 by microparticles and endothelial cell survival and proliferation, and define a novel molecular mechanism for TGFβ and Alk5-dependent Erk1/2MAPK signaling that is significant for proliferative signaling and abnormal growth.


INTRODUCTION
Acute lung injury (ALI) or mild acute respiratory distress syndrome (ARDS), according to the Berlin definition (Ranieri et al., 2012), are associated with excessive apoptosis of endothelial and epithelial cells (Henson and Tuder, 2008;Le et al., 2008;Predescu et al., 2013). Although apoptosis might induce pulmonary endothelial and epithelial barrier dysfunction leading to pulmonary edema, evidence suggests that apoptosis plays a beneficial role during ALI resolution owing to the proregenerative role of clearance of apoptotic cells Schmidt and Tuder, 2010). This effect is mediated through the production of growth factors including TGFb by macrophages engulfing apoptotic cells, or perhaps by other vascular cells Henson and Tuder, 2008). TGFb, owing to its anti-inflammatory properties confines the extent of septal injury and speeds recovery in ALI (D'Alessio et al., 2009).
We have recently shown that in vivo deficiency of ITSN-1s, an isoform of ITSN-1 that is highly prevalent in lung endothelium and deficiency of which is relevant to the pathology of ALI/ARDS Predescu et al., 2013), induces extensive lung endothelial cell apoptosis and injury; after only 7 days of ITSN knockdown (KD-ITSN), the remaining endothelial cells exhibited phenotypic changes including hyperproliferation and apoptosis resistance against ITSN-1s deficiency, leading to increased microvessel density, repair and remodeling of the injured lung. Under pathological conditions, dysfunctional endothelial cells also show altered intracellular trafficking and signaling of cell surface receptors, such as TGFb-RI, which is implicated in the pathogenesis of ALI/ARDS (Kranenburg et al., 2002;Morrell et al., 2001;Sehgal and Mukhopadhyay, 2007;Voelkel and Cool, 2003). Endocytic dysfunction and non-productive assembly of the endocytic machinery might alter canonical signaling pathways with detrimental consequences for endothelial cell function (Mukherjee et al., 2006;Sorkin and von Zastrow, 2009). Although endothelial cells alone are insufficient to cause ALI (Wiener-Kronish et al., 1991), their injury or dysfunction and activation, as well as their interaction with the alveolar epithelium are crucial not only for the onset of ALI/ARDS, but also for repair and remodeling of the injured lung.
Emerging in vivo and in vitro evidence has revealed a crucial role of circulatory microparticles as transcellular delivery systems and in the communication between different cell types; microparticles are present in healthy and pathological settings; they store important bio-effectors and induce endothelial modifications, angiogenesis or differentiation (Mause and Weber, 2010). Although the presence of microparticles in ALI/ ARDS has been reported (McVey et al., 2012), their in vivo relevance in the modulation of signaling pathways leading to improved endothelial and vascular functions in the setting of lung injury has not been explored. Given that ITSN-1s deficiency in cultured endothelial cells triggers mitochondrial apoptosis (Predescu et al., 2007a), whereas, in vivo, it leads to the emergence of proliferative and apoptosis-resistant endothelial cells , we hypothesized that the in vivo microparticles released by apoptotic or activated vascular cells in the systemic circulation of KD-ITSN mice might account for endothelial cell survival and alterations in their phenotype. We now demonstrate a functional relationship between the intercellular transfer of Alk5 by microparticles and endothelial cell survival and proliferation, and define a novel molecular mechanism for TGFb-Alk5-dependent Erk1 and Erk2 (also known as MAPK3 and MAPK1, respectively; hereafter referred to as Erk1/2 MAPK ) signaling, significant for the abnormal proliferation of pulmonary endothelial cells.

Endocytic deficiency caused by KD-ITSN modifies Alk5 endocytic trafficking and enhances its degradation
Recently, we investigated the in vivo effects of long-term ITSN-1s deficiency on pulmonary vasculature and lung homeostasis, using a KD-ITSN mouse model generated by repeated delivery of a specific small interfering (si)RNA targeting ITSN-1 (siRNA ITSN ; Bardita et al., 2013;Predescu et al., 2012). We have shown that acute ITSN-1s deficiency in the murine lungs results in a significant decrease in Erk1/2 MAPK pro-survival signaling, increased endothelial cell apoptosis and lung injury; at 24 days post siRNA ITSN initiation, the surviving endothelial cells showed reactivation of Erk1/2 MAPK and phenotypic changes towards proliferation. The threefold increase in mature TGFb expression at 10 days post siRNA ITSN treatment compared with that of control mice suggested that TGFb signaling might account for Erk1/2 MAPK activation in KD-ITSN mice. Because TGFb elicits its signaling by binding to its cell surface Ser/Thr kinase receptors, leading to the formation of heterocomplexes between Alk5 (also known as TGFBR1) and transforming growth factor b type II receptor (TGFb-RII or TGFBR2) (Lebrin et al., 2005), and because Alk5 expression might play an important regulatory role in TGFb signaling, we performed a timecourse analysis of Alk5 protein expression in the lung lysates of the KD-ITSN mice. At 72 h post siRNA ITSN delivery, Alk5 expression was 80% lower, compared with that of all controls [wild type (wt), empty-liposome-treated and siCONTROL non-targeting siRNA (siRNA Ctrl )-treated mice; Fig. 1A]. Later on, at 10 days, 15 days and 24 days post siRNA ITSN delivery, Alk5 expression showed a gradual increase, reaching values relatively close to those of controls. The expression of ITSN-1s protein was monitored at several time points after siRNA ITSN delivery by using western blotting of mouse lung lysates (Fig. 1B); at day 3, it was ,75% lower relative to expression in all control mice, and the knockdown was maintained for the next 21 days. Delivery of empty liposomes (Fig. 1B,  in the lung, was detected in the brain, although knockdown in the heart, kidneys and liver was less efficient . Because ITSN-1s deficiency functionally upregulates alternative transport pathways and their carriers to compensate for impaired endocytosis mediated by caveolae and clathrincoated vesicles (Predescu et al., 2012) involved in Alk5 intracellular trafficking (Derynck and Zhang, 2003), we also investigated Alk5 expression and internalization in cultured endothelial cells deficient for ITSN-1s (EC KD-ITSN ). The ITSN-1 gene was specifically and efficiently knocked down using an siRNA approach that has been described previously (Predescu et al., 2007a). EC KD-ITSN were used at 38-40 h post siRNA ITSN transfection, a time point when the protein expression is 50% lower compared with that of controls (Fig. 1C) and endothelial cells are not yet apoptotic (not shown), as determined by TUNEL as described previously (Predescu et al., 2007a). Actin served as the loading control (Fig. 1C). The transfection of endothelial cells with siCONTROL non-targeting siRNA did not affect the expression of ITSN-1s at 40 h post transfection. The expression of Alk5 protein in EC KD-ITSN was 40% of the levels observed in controls, as indicated by western blotting with an anti-Alk5 antibody followed by densitometry; actin was used as the loading control (Fig. 1D).
In addition, pre-embedding immuno-EM for Alk5 (Fig. 2D), indicated that in the lung endothelial cells of KD-ITSN mice, 8nm gold-conjugated Alk5 antibody labels the cell surface (d3) and is apparently internalized and associated with large endocytic tubulovesicular structures (d2). In wild-type mouse lung endothelial cells, Alk5 antibody labels caveolae and CCVs (d1), endothelial plasma membrane and occasionally endosomal structures (not shown). Taken together, the observations suggest that perturbation of caveolae-mediated endocytosis due to ITSN-1s deficiency upregulates cav1-dependent alternative endocytic pathways and their morphological structures or carriers, which are underrepresented under normal conditions (Doherty and McMahon, 2009;Predescu et al., 2012), and that these structures might be involved in Alk5 endocytic trafficking.
Because caveolae internalization sends Alk5 to the ubiquitylation machinery (Itoh and ten Dijke, 2007), we next investigated whether the decreased expression of Alk5 might be caused by increased ubiquitylation. Alk5 immunoprecipitation followed by immunoblotting with an antibody against ubiquitin applied to EC Ctrl and EC KD-ITSN (Fig. 3A), as well as on lung lysates of wild-type and KD-ITSN mice, 10 days post siRNA ITSN initiation (Fig. 3A), confirmed significant Alk5 ubiquitylation in ITSN-deficient specimens compared with that of controls. Moreover, double immunofluorescence for Alk5 and ubiquitin in EC KD-ITSN revealed a punctate pattern of Alk5 in the cytosol (Fig. 3B), no significant plasmalemma staining and prominent colocalization with ubiquitin immunofluorescence (Fig. 3B, merged image). EC Ctrl showed increased Alk5 immunoreactivity in the cytosol and at the plasma membrane compared with that of EC KD-ITSN (Fig. 3C,c.1), some colocalization of Alk5 and ubiquitin and a significant pool of Alk5 not colocalizing with ubiquitin ( Fig. 3C,c.2). The panels b.1 and c.2 show for comparison the magnified boxed areas in the merged images in Fig. 3B and Fig. 3C, respectively; although under control conditions, colocalization between Alk5 and ubiquitin is limited, in EC KD-ITSN , Alk5 and ubiquitin were significantly colocalized, consistent with Alk5 ubiquitylation and degradation; moreover, the decreased Alk5 immunoreactivity in EC KD-ITSN compared with that of EC Ctrl is consistent with decreased Alk5 expression in these ECs. In the ubiquitin-proteasome pathway, the HECT-type E3 ubiquitin ligases (Smurf1 and Smurf2) interact with the nuclear Smad7 (a negative regulator of TGFb signaling) and induce its nuclear export, followed by assembly of the Smad7-Smurfs-Alk5 complex and enhanced turnover of Alk5 by ubiquitylation (Murakami et al., 2010). Alk5 immunoprecipitation followed by immunoblotting with antibodies against Smad7 and Smurf1 applied on endothelial cell lysates indicated an increased association of both Smad7 and Smurf1 with Alk5 in EC KD-ITSN by comparison to EC Ctrl (Fig. 3D). However, because Alk5 amounts are ,40% lower in KD-ITSN samples, as estimated by densitometry, the ratios of Smad7:Alk5 (Fig. 3E) and Smurf1:Alk5 ( Fig. 3F) are significantly higher in EC KD-ITSN compared with EC Ctrl , consistent with increased Alk5 degradation. We also detected translocation of Smad7 from the nucleus to the cytosol in EC KD-ITSN , whereas the EC Ctrl showed significant Smad7 nuclear immunoreactivity (Fig. 3G). Taken together, these observations demonstrate that ITSN-1s deficiency alters the endocytic trafficking of Alk5, causing its enhanced degradation.
The apoptotic or activated circulating and vascular cells of KD-ITSN mice release elevated levels of microparticles comprising Alk5 into the bloodstream EM analyses of KD-ITSN mouse lungs revealed frequently in the lumen of the blood vessels the presence of microparticles with 0.5-1.0 mm diameter, many of them membrane-bound to endothelial cells (Fig. 4A,a1). Because microparticles might be a means to replenish endothelial cells with Alk5, we isolated the microparticles from the blood of KD-ITSN mice (MP KD-ITSN ) at 10 days post siRNA ITSN initiation and subjected them to negative-staining EM. MP KD-ITSN are abundant, display double-membrane morphology and notably undergo membrane fusion and communicate with each other (Fig. 4B,b1). In vivo MP KD-ITSN release, evaluated by quantification of the amount of the total protein in the isolated microparticles, indicated the highest amount, a ,44% increase compared with controls ( Fig. 4C), at day 10 post siRNA ITSN , when endothelial cell apoptosis was at its peak . Next, equal volumes of MP Ctrl and MP KD-ITSN (normalized to equivalent ml of blood) were analyzed for their Alk5 content; Alk5 expression was significantly higher in MP KD-ITSN (Fig. 4D), consistent with the idea that in the systemic circulation of the KD-ITSN mice there are more Alk5-positive microparticles compared with wild-type mice. MP KD-ITSN were also immunoreactive to the vascular endothelial growth factor receptor-2 and bone morphogenetic protein receptor-2 but with no detectable differences between MP Ctrl and MP KD-ITSN (not shown) and to TGFb-RII, but in this case, the amounts were 30% less in the MP KD-ITSN compared with the MP Ctrl (Fig. 4E). To get more accurate data regarding the abundance of microparticles and their Alk5 content, we labeled the microparticles with an APCconjugated antibody against Alk5 and analyzed them by flow cytometry (Fig. 5). Spherotech nano fluorescent size standard beads (0.45 mm, 0.88 mm and 1.35 mm) were used to confirm the MP KD-ITSN transfer Alk5 to EC KD-ITSN to restore Erk1/2 MAPK prosurvival signaling Next, we addressed whether MP KD-ITSN can interact with and transfer Alk5 to endothelial cells, using a microparticle transfer assay and fluorescent imaging. MP KD-ITSN , 10 days post siRNA treatment (used throughout the study) were either biotinylated followed by incubation with neutrAvidin conjugated to Alexa Fluor 594 or double labeled with neutrAvidin and Alk5 antibody, followed by streptavidin conjugated to Alexa Fluor 594 and a secondary IgG conjugated to Alexa Fluor 488, as described in Materials and Methods. Biotin-neutrAvidin gives a continuous, donut-shape labeling of microparticles ( Fig. 6A,a1). Double biotin and Alk5 antibody labeling revealed Alk5 immunoreactive puncta associated with the donut-shaped particles (Fig. 6B,b.1-b.6); on average, one to four clusters of Alk5 molecules were associated with the donut-shaped, biotin-labeled microparticles. Note also the high propensity of microparticles to fuse to each other (b.4-b.6). The arrow in Fig. 6B points to a large biotin and Alk5-labeled particle (4-5 mm diameter), most likely generated by fusion of two or three individual microparticles. Morphometric analyses indicated that ,19% of the MP KD-ITSN population is immunoreactive to Alk5 antibody ( Fig. 6C), in close agreement with flow cytometry data.
Next, we investigated the ability of MP KD-ITSN to interact (bind and incorporate) and transfer Alk5 to EC KD-ITSN . Briefly, MP KD-ITSN were labeled with anti-Alk5 and an Alexa-Fluor-594conjugated secondary antibody (referred to hereafter as anti-Alk5-Alexa-Fluor-594 pre-labeled MP KD-ITSN ), using a similar approach to that described above. Biotin-neutrAvidin labeling was omitted to shorten the experimental manipulation of the microparticles to preserve their properties and ability for interaction. Then, EC KD-ITSN at 48 h post siRNA transfection were grown on coverslips and exposed to anti-Alk5-Alexa-Fluor-594 pre-labeled MP KD-ITSN for 1 h on ice to allow binding and, subsequently, transferred to 37˚C for 10 min, 20 min and 60 min to allow internalization. Because TGFb signals through the heteromeric TGFb-RI-TGFb-RII receptor complex and because western blotting indicated that the microparticles contained both receptors, the cells were counterstained with a TGFb-RII antibody followed by an Alexa-Fluor-488-conjugated secondary antibody, to evaluate whether or not the Alk5-TGFb-RII complex can be detected morphologically on the microparticles interacting with endothelial cells. To this end, EC KD-ITSN exposed to anti-Alk5-Alexa-Fluor-594 pre-labeled MP KD-ITSN for 1 h on ice were subjected to three 10-min washing steps in phosphate buffered saline (PBS), to rule out the possibility of visualizing just the simple physical association of microparticles with the endothelial plasma membrane; then, cells were permeabilized and fixed with methanol at 220˚C for 7 min. The permeabilization and fixation step renders endothelial cells unable to internalize the microparticle-derived, pre-labeled Alk5-Alexa-Fluor-594. Fixed and permeabilized endothelial cells were quenched in 1% BSA in PBS and then incubated with TGFb-RII antibody followed by the appropriate Alexa-Fluor-488-conjugated secondary antibody as described in Materials and Methods. Given this experimental approach, anti-Alk5-Alexa-Fluor-594 labeling indicates only the Alk5 present on the microparticles, whereas the anti-TGFb-RII-Alexa-Fluor-488 detects both the microparticlederived and the endogenous receptor. The immunoreactivity for the two receptors is detected frequently colocalizing on the plasma membrane, very suggestive of their heterodimerization and residence on the same microparticle (Fig. 6D, yellow circles; Fig. 6E,e.1-e.6). A lower magnification of the field used to select the image in Fig. 6D is provided in supplementary material Fig.  S1. We also detected the endogenous TGFb-RII not associated with microparticle-derived Alk5 (white squares). Importantly, even if the merged image does not reveal colocalization, the immunoreactivity for the microparticle-derived pre-labeled Alk5 is always in close association with the TGFb-RII immunoreactivity (Fig. 6D, white arrowheads, d.1). At all time points at 37˚C (20 min is shown), the microparticle-derived, Alexa-Fluor-594 pre-labeled Alk5 was detected in the cytosol, consistent with transfer and incorporation of Alk5 from the MP KD-ITSN to EC KD-ITSN (Fig. 6F). Cells were counterstained with ubiquitin antibody (Fig. 6G), followed by the appropriate secondary antibody, for easier identification. Worth mentioning is the significant colocalization between Alk5 and ubiquitin ( Fig. 6G, inset g.1), consistent with our hypothesis that, in EC KD-ITSN , Alk5 undergoes increased ubiquitylation. To rule out the possibility of non-specific attachment of anti-Alk5 and Alexa-Fluor-594-conjugated IgG aggregates to the MP KD-ITSN and, thus, their endocytic internalization, control experiments were performed using acidwashed microparticles, as described in Materials and Methods. Representative results are shown in supplementary material Fig.  S1B.
The intercellular microparticle-mediated transfer of Alk5, the downstream signaling molecules of which include Erk1/2 MAPK (Derynck and Zhang, 2003), raised the question of whether the survival of EC KD-ITSN in vivo might be a consequence of an interaction between microparticles and endothelial cells. Cultured EC KD-ITSN , 48 h post siRNA transfection, were exposed to 12.5 mg/ml, 25 mg/ml and 50 mg/ml MP KD-ITSN , for 24 h (Fig. 7A,f-h). EC Ctrl (Fig. 7A,a) and EC KD-ITSN (Fig. 7A,b) not exposed to MP KD-ITSN were used for comparison. After 3 days, the cells were counted; despite ITSN-1s deficiency, exposure to 12.5 mg/ml MP KD-ITSN doubled the survival rate of EC KD-ITSN without microparticle exposure (f versus b) and reached 83% of the EC Ctrl number (f versus a); exposure to 25 mg/ml or 50 mg/ml MP KD-ITSN increased more than twofold the survival rate compared to EC KD-ITSN without microparticle exposure (g, h versus b), and reached 98.8% and 95%, respectively, of the EC Ctrl number (g, h versus a). Moreover, exposure of EC KD-ITSN to MP Ctrl (Fig. 7A,ce) showed only 6%, 26% and 32.5% improvement of survival rate. A ratio of 1:2 between MP Ctrl :MP KD-ITSN was used, to approximate their distribution in the murine systemic circulation.
An enzyme-linked immunosorbent assay (ELISA)-based BrdU cell proliferation assay indicated that EC KD-ITSN exposed for 2 days to 25 mg/ml and 50 mg/ml MP KD-ITSN showed BrdU incorporation similar to that of EC Ctrl ; however, when compared to EC KD-ITSN without MP KD-ITSN exposure, the BrdU incorporation showed a greater than 2.5-fold increase (Fig. 7B). These data indicate that the intercellular transfer of Alk5 to EC KD-ITSN might rescue EC KD-ITSN from apoptosis. Thus, we next evaluated the effects of the microparticle-EC KD-ITSN interaction on Erk1/2 phosphorylation by western blotting with a phospho-Erk1/2 specific antibody. Exposure of EC KD-ITSN to 12.5 mg/ml or  (Fig. 7C,d versus a) and was significantly higher when EC KD-ITSN were exposed to 25 mg/ml MP KD-ITSN (Fig. 7C,e versus a). It appears that the interaction between MP KD-ITSN and endothelial cells and a MP KD-ITSN basal threshold are mandatory for Erk1/2 activation and endothelial cell survival following KD-ITSN. EC KD-ITSN without MP KD-ITSN exposure showed 50% lower Erk1/2 phosphorylation compared with that of EC Ctrl (Fig. 7C,b versus a). The blockade of Alk5 by pre-incubation of MP KD-ITSN with 10 mM/l SB525334 (Fig. 7C,g), a selective Alk5 inhibitor, or pre-incubation of MP KD-ITSN with 10 mM diannexin (Fig. 7C,h), an annexin V homodimer known to block the microparticle uptake (del Conde et al., 2005), notably reduced Erk1/2 activation. MP Ctrl did not significantly activate Erk1/2 (Fig. 7C,c). The Alk5 inhibitor affected both the microparticle-derived Alk5 and the endogenous Alk5. EC KD-ITSN not exposed to microparticles (lane b) displayed a low level of Erk1/2 phosphorylation that could be inhibited by 10 mM/l SB525334; the observation is consistent with the idea that the low Erk1/2 activation in EC KD-ITSN is due, at least in part, to the endogenous Alk5 signaling. Exposure of EC KD-ITSN to MP KD-ITSN in the presence of 10 ng/ml TGFb (Fig. 7C,f), revealed less than a 30% decrease in Erk1/2 phosphorylation compared with EC KD-ITSN exposed to MP KD-ITSN in the absence of TGFb. However, the degree of Erk1/2 phosphorylation in the presence of TGFb is still above control levels, consistent with activation of pro-survival signaling and rescue of EC KD-ITSN from apoptotic death caused by ITSN-1s deficiency.
ITSN-1s deficiency alters the Smad2/3-Erk1/2 MAPK signaling balance towards persistent Ras-Erk1/2 MAPK activation ITSN-1s and TGFb-Alk5 induce Erk1/2 MAPK signaling by sharing the same Ras-Raf-MEK cascade (Derynck and Zhang, 2003;Patel et al., 2013;Tong et al., 2000); moreover, ITSN-1s associates with mSos (mammalian Son of sevenless, also known as SOS1) in a protein complex that excludes Grb2 (Tong et al., 2000), raising the possibility that ITSN knockdown might increase mSos availability for Grb2 interaction and, thus, lead to preferential formation of the Alk5-mSos-Grb2 complex and activation of Erk1/2 MAPK signaling. Ras-Erk1/2 MAPK activation might result in ineffective assembly of Alk5-Smad2-SARA complex and subsequent alteration of the Smad2/3-Erk1/2 signaling balance. To address this possibility, control and KD-ITSN mouse lung lysates were subjected to immunoprecipitation with antibodies against mSos, Smad2/3 and SARA, followed by western blot analyses for Alk5 (Fig. 7D). KD-ITSN mouse lungs showed increased Alk5 association with mSos and decreased association with Smad2/3 and SARA, consistent with non-productive assembly of the Alk5-Smad2/3-SARA complex; no changes in the amounts of mSos, Smad2/3 and SARA immunoprecipitated from EC Ctrl and EC KD-ITSN lysates were detected. Apparently, ITSN-1s deficiency steers Alk5 away from its canonical Smad2/3 signaling and preferentially   .1-b.6). Light microscopy (b.1) and confocal images (b.2-b.6) illustrate Alk5 immunoreactivity associated with microparticles and the capability of microparticles to fuse with each other (B, arrow; b.4-b.6); the arrow in B indicates a 3-4 mm diameter particle, most likely generated by fusion of individual microparticles due to the relatively long time of processing before 1% paraformaldehyde fixation. (C) Morphometric analyses of Alk5-positive MP KD-ITSN . Data are shown as a percentage relative to the control and show the mean6s.e.m.; *P,0.05; n53. (D) EC KD-ITSN exposed to anti-Alk5-Alexa-Fluor-594 pre-labeled microparticles (MPs) on ice for 1 h were co-stained with anti-TGFb-RII and an Alexa-Fluor-488-conjugated secondary. Frequent colocalization is suggestive of their heterodimerization and residence on the same microparticle (yellow circles). The endogenous TGFb-RII is not always associated with microparticle-derived Alk5 (white squares). Even when colocalization of Alk5 and TGFb-RII is not obvious, Alk5 and TGFb-RII immunoreactive puncta are found in very close proximity (arrowheads and inset d. stimulates the less common Erk1/2 MAPK pathway. Based on densitometric analyses of Alk5-Smad2/3, Alk5-Sos and Alk5-SARA interactions and on the finding that Alk5 level in KD-ITSN mouse lung lysates is 40% lower compared to that of controls (Fig. 1A), we determined that in control mouse lungs 75% of Alk5 associates with Smad2/3 and only 25% with mSos. In KD-ITSN mouse lungs, only ,8% signals through SARA-Smad2/3 and 52% associates with mSos (Fig. 7E). Taken together, the findings are consistent with ineffective assembly of the Alk5-Smad2/3-SARA complex in favor of the Alk5-Sos-Grb2 signaling complex and persistent Ras-Erk1/2 MAPK activation with protective effects on lung endothelium.
in altered Alk5 intracellular trafficking and enhanced degradation with detrimental consequences for endothelial cell function. Accumulating evidence indicates that perturbation of clathrinand caveolae-mediated endocytosis functionally upregulates alternative pathways that are either underrepresented or even non-existent under normal conditions (Doherty and McMahon, 2009;Predescu et al., 2012). Recent studies have demonstrated that endocytosis into cav1-dependent tubulovesicular structures, in addition to vesicles, is a common event in mammalian cells (Kirkham et al., 2005;Knezevic et al., 2011;Marbet et al., 2006;Predescu et al., 2012). Moreover, EM studies of the cav1-null mouse revealed the presence of cav1-independent vesicles and vesiculo-vacuolar-like organelles able to mediate transendothelial transport (Predescu et al., 2007b). Our in vivo studies indicated that the endocytic deficit generated by modulation of ITSN expression in lung endothelial cells is rescued by upregulation of alternative endocytic pathways and their morphological intermediates (i.e. tubulovesicular and tubular ring-like structures) involved in tracer uptake and transport across endothelium (Knezevic et al., 2011;Predescu et al., 2012). The properties of the cav1-associated tubulovesicular endocytic pathway or clathrin-and cav1independent pathway and their molecular characteristics are not fully understood; most of the knowledge has been derived mainly from EM studies of the morphology of the endocytic structures in different cell types using different tracers, sensitivity to drugs and their dependence on dynamin (Doherty and McMahon, 2009). Caveolae have been shown to be capable of fusing with the early endosomes in Rab5-dependent processes (Pelkmans et al., 2004). Cav1 is palmitoylated, and it binds cholesterol and fatty acids that might be important in ordering local lipids into invaginationcompetent compositions (Doherty and McMahon, 2009). Most endocytic pathways, especially cav1-dependent pathways, are sensitive to cholesterol perturbation and are inhibited by the removal of cholesterol (Mayor and Pagano, 2007). Dynamin dependence of the cav1-dependent tubulovesicular structures is unclear; in lung endothelial cells, overexpression of ITSN (Predescu et al., 2003) or of the SH3A domain of ITSN (Knezevic et al., 2011) inhibits the GTPase activity and oligomerization properties of dynamin, resulting in impairment of membrane fission and, thus, formation of membranous tubules, frequently associated with caveolae-like vesicles. Nonetheless, a dynamin pool might escape ITSN-mediated inhibition and, thus, can still mediate the fission of few vesicles and of the tubulovesicular and tubular ring-like structures from the endothelial plasma membrane. In contrast, in EC KD-ITSN characterized by a significant shortage of ITSN scaffold, and thus inefficient dynamin recruitment to the endocytic site, upregulation of the tubulovesicular and tubular ring-like structures is significant. However, other endocytic accessory proteins might partially compensate for dynamin recruitment and membrane invagination and scission, leading to the formation of a few discrete vesicles and release from the plasma membrane of tubulovesicular and tubular ring-like structures.
The endocytic mechanism is vital for many processes including nutrient uptake, membrane recycling, signal transduction and recycling or degradation of cellular receptors (Doherty and McMahon, 2009). In normal endothelial cells, Alk5 is internalized by (1) CCVs, leading to TGFb-induced Smad2/3 activation, transcriptional responses and recycling to the plasma membrane, and (2) caveolae, which direct Alk5 to the ubiquitinproteasome and turn off TGFb signaling (DiGuglielmo et al., 2003). Our studies demonstrate that internalization of Alk5 through the cav1-associated tubulovesicular structures results in enhanced ubiquitylation and degradation, and thereby, decreased Alk5 protein expression. Interestingly, ubiquitylation of some receptor tyrosine kinases promotes association with caveolae and endocytic internalization through a filipin-and nystatin-sensitive, clathrin-independent pathway; in the case of the epidermal growth factor receptor, at high epidermal growth factor concentration, a switch in the endocytic mechanism resulting in receptor ubiquitylation and degradation when endocytosed through caveolae has been reported (Sigismund et al., 2005).
An intriguing observation made during our studies of knocking down ITSN-1s in cultured endothelial cells and mouse lungs relates to the fact that ITSN-1s deficiency in cultured cells triggers mitochondrial apoptosis (Predescu et al., 2007a), whereas, in vivo, the peak in endothelial cell apoptosis is followed by survival and alterations of endothelial phenotype towards hyperproliferation and apoptosis resistance . The observation raised the possibility that activated or apoptotic cells of KD-ITSN mice generate microparticles that are able to activate pro-survival signaling and modify the endothelial cell phenotype; this is consistent with our results that demonstrate the presence in the systemic circulation of KD-ITSN mice of microparticles bearing the widely expressed Alk5. We have also found that MP KD-ITSN harbor 2.5-fold more Alk5 compared with the MP Ctrl . It is of note that microparticles released by cultured EC KD-ITSN do not contain Alk5, and this observation might explain, at least in part, the apoptotic death of cultured endothelial cells. MP KD-ITSN also contain TGFb-RII, but in smaller amounts compared with MP Ctrl . Apparently, Alk5-TGFb-RII heterodimers are already formed on the microparticles, most likely owing to the TGFb present in the systemic circulation. Even if our studies do not allow us to draw conclusions on the Alk5 phosphorylation and activation status, the signaling will occur only after Alk5 transfer to EC KD-ITSN , which enables Alk5 interaction with downstream partners. Recent studies have shown that receptor endocytosis is not essential for TGFb signaling (Chen et al., 2009;Lu et al., 2002). Consistent with this, TGFb-Alk5-mediated Erk1/2 activation can take place on the plasma membrane, without Alk5 endocytic internalization. In addition to Alk5, endothelial cells express Alk1 (also known as SKR3), both involved in TGFb-induced transcriptional responses, with opposite effects on the activation state of the endothelium; whereas activated Alk5 induces the phosphorylation of Smad2/3, activated Alk1 has been shown to induce the phosphorylation of Smad1/5 (Goumans et al., 2002). TGFb-Alk1 and TGFb-Alk5 signaling might be modulated by two accessory TGFb-R type III receptorsbetaglycan (also known as TGFBR3) and endoglin (Lebrin et al., 2005). Because previous reports indicated that TGFb-RI/ Alk5 signaling might be regulated in a ligand-dependent manner by TGFb co-receptors (Bizet et al., 2012), it is likely these TGFb co-receptors and accessory proteins account for modulation of Erk1/2 phosphorylation in EC KD-ITSN exposed to microparticles in the presence of TGFb.
Moreover, MP KD-ITSN readily interact with EC KD-ITSN that contain less Alk5, and they transfer a functional Alk5 receptor, suggesting in vivo mechanisms of replenishing EC KD-ITSN with functional Alk5. The functionality of Alk5 is supported by signaling events leading to Erk1/2 kinase phosphorylation and endothelial cell survival; Erk1/2 phosphorylation can be prevented by pre-incubation of MP KD-ITSN with SB-525334, a specific Alk5 inhibitor. The event involves phosphatidyl serine residues of the Alk5-containing microparticles, given that use of diannexin blocks Erk1/2 phosphorylation. Our findings are similar to the recently described transfer of the oncogenic epidermal growth factor receptor present on tumor-derived microparticles to endothelial cells or of the tissue factor present on macrophage-derived microparticles to platelets (Al-Nedawi et al., 2008;del Conde et al., 2005). Circulating microparticles can transfer genetic material and proteins from the donor cells (cells generating the microparticles) to a wide range of target cells, by several mechanisms: internalization and lysosomal processing of microparticles, fusion-mediated transfer of surface receptors, proteins, and lipids, outside-in signaling through ligand-receptor internalization and temporary fusion with the target cell, followed by complete or selective transfer of microparticle content (McVey et al., 2012). Extensive endothelial cell apoptosis caused by KD-ITSN might induce a macrophage phenotype that favors tissue repair and suppression of inflammation (McCubbrey and Curtis, 2013), and as part of this process release microparticles comprising Alk5 into the systemic circulation of the KD-ITSN mice. The ability of apoptotic cells to signal for their non-inflammatory and nonimmunogenic removal in vivo is crucial for normal tissue homeostasis and for resolution of inflammation (Xiao et al., 2008). Macrophage interaction with apoptotic cells increases the production of TGFb, which is known to inhibit inflammatory cytokine production through the crosstalk between MAPKs, specifically Erk-dependent inhibition of p38 MAPK (Xiao et al., 2002). In addition, experimental and clinical data indicate that platelets are necessarily involved in repair and regeneration of damaged tissues and preservation of organ function (Gawaz and Vogel, 2013). Platelet-derived microparticles constitute the majority of the pool of microparticles circulating in the blood; they express and might transfer functional receptors, stimulate the release of cytokines, activate signaling pathways, promote angiogenesis and participate in tissue regeneration (Varon et al., 2012). Although an increase in TGFb production by platelets and macrophages as a result of interaction with apoptotic cells has been reported (Dean et al., 2009;Xiao et al., 2002), the release of microparticles enriched in the ubiquitously expressed TGFb-RI is a novelty of our studies. It is well documented that TGFb signaling has crucial functional roles in lung development, injury and repair (Warburton et al., 2013). However, it seems that the activated pathways and the end effects of TGFb signaling are highly dependent on the cellular context and are disease specific. Therefore, it will be of considerable interest to examine whether in human ALI/ARDS patients the number of Alk5-harboring microparticles is increased and whether these particles interact with endothelial cells and impact on the lung vasculature. In most cell types, endothelial cells included, TGFb signals through TGFb-RI/Alk5 (Goumans et al., 2002;Lebrin et al., 2005). Although Smad2/3 have been identified as pivotal intracellular effectors of TGFb-Alk5, there is growing evidence that Ras-Erk1/2 MAPK is another major signaling pathway for TGFb (Derynck and Zhang, 2003). TGFb induces modest Ras activation consistent with low level Erk1/2 kinase induction (Mulder, 2000). TGFb-mediated Erk1/2 activation is necessary for TGFb-induced epithelial-to-mesenchymal transformation (Davies et al., 2005), for regulation of Smad nuclear translocation (Kretzschmar et al., 1999) and for Smaddependent gene expression (Mucsi et al., 1996). The mechanism by which TGFb activates Erk1/2 MAP kinases is poorly understood. Our study provides a mechanism whereby MP KD-ITSN -derived Alk5 re-wires dysfunctional endothelial cells to activate pro-survival signaling through Erk1/2 kinase and to become hyperproliferative. TGFb induces Ras-Erk1/2 signaling through phosphorylation of the adaptor protein ShcA (also known as SHC1; Lee et al., 2007), leading to its association with mSos, a Ras GTP/GDP exchange factor, and Grb2 (van der Geer et al., 1996). It appears that ITSN-1s deficiency increases mSos availability for Grb2, favoring the formation of the Alk5-mSos-Grb2 signaling complex. As a result, the assembly of the Alk5-Smad2-SARA signaling complex is unproductive. SARA is a Smad2/3-interacting protein and a control point for Smad2 subcellular localization and TGFb-dependent transcriptional responses (Tsukazaki et al., 1998). Thus, ITSN deficiency by disturbing the SARA-Smad2 interaction might cause Smad2 subcellular mislocalization. ITSN deficiency also decreases the levels of Smad2/3 phosphorylation . Smad2 phosphorylation is required for its association with Smad4, and for the formation and nuclear translocation of the heterotrimeric Smad2/3/4 complex, leading to activation of TGFb target genes and inhibition of cell proliferation (Goumans et al., 2002;Tsukazaki et al., 1998;Xie et al., 2011). Thus, ITSN deficiency suppresses the Alk5-Smad2/3 pathway, leading to inhibition of the anti-proliferative action of TGFb. In addition, the TGFb-Alk5 signaling is switched from the canonical Smad2/3 to the less common Erk1/2 MAPK pathway, with protective effects on endothelial cells and lung vasculature. Given that decreased expression of ITSN-1s favors the assembly of Alk5-mSos-Grb2 signaling complexes resulting in downstream Erk1/2 activation, endothelial cells are rescued from apoptotic death caused by ITSN-1s deficiency. The effects induced by MP KD-ITSN on Erk1/2 activation and cell survival are dependent on membrane fusion and Alk5 transfer from microparticles to endothelial cells. Erk1/2 activation is dependent on microparticle number as well, consistent with previous reports that threshold concentrations of biological effectors are important for microparticle-induced physiological effects (Freyssinet, 2003). Although Alk5 transfer might play an important role in rescuing endothelial cells, a possible involvement of other biological effectors that make up microparticles cannot be ruled out. However, this finding might potentially apply also to other cell surface receptors (Predescu et al., 2012), altering their fate, sorting and the functional consequences for proteins involved (Di Fiore and De Camilli, 2001;Le Roy and Wrana, 2005).
In summary, our studies demonstrate a functional relationship between the intercellular transfer of Alk5 by microparticles and endothelial cell survival and proliferation, and define a novel molecular mechanism for TGFb-Alk5-dependent Erk1/2 MAPK signaling that is significant for the abnormal proliferation of pulmonary endothelial cells.
Specific antibodies were against the following proteins (the relevant suppliers are also indicated): Smad-7 (R&D Systems, Minneapolis, MN); Alk5 N-terminal extracellular epitope, Smurf-1, Smad2/3, cav1, SARA, ubiquitin and mSOS (Santa Cruz Biotechnology, Santa Cruz, CA); ITSN-1 (BD Biosciences, San Jose, CA); actin (Sigma-Aldrich, St Louis, MO); Alk5-APC (e-Bioscience, San Diego, CA) and phospho-Erk1/2 MAPK (Cell Signaling, Beverly, MA). EM reagents were from EM Sciences (Hatfield, PA). Biotin was from ThermoFisher Scientific (Rockford, IL). All fluorophore-conjugated antibodies and the Prolong Antifade reagent were from Molecular Probes (Eugene, OR). Spherotech nano fluorescent beads were from Spherotech, Inc. (Lake Forest, IL). Flow cytometry reagents were from e-Bioscience (San Diego, CA). SB-525334 and diannexin were from Sigma-Aldrich (St Louis, MO) and human TGFb1 was from R&D Systems (Minneapolis, MN). Protein-A/G-agarose beads were from Santa Cruz Biotechnology (Santa Cruz, CA). Animals CD1 male mice, 6-8 weeks old, 20-25 g weight, from Jackson Laboratory (Bar Harbor, ME), kept under standardized housing and feeding conditions were used. The experiments were done under anesthesia, using ketamine (60 mg/kg), acepromazine (2.5 mg/kg) and xylazine (2.5 mg/kg) in 0.1-0.2 ml PBS. A specific ITSN-1 siRNA sequence (100 mg siRNA/mouse) was delivered by using cationic liposomes, by retro-orbital injection, into mouse lungs as described previously Predescu et al., 2012). The siRNA sense sequence -59-GAGAGAGCCAAGCCGGAAAUU-39 -(Dharmacon, Lafayette, CO) was used for knocking down mouse ITSN-1s. Chronic inhibition of ITSN-1s was achieved by repeated retro-orbital delivery of the siRNA ITSN -liposome complexes every 72 h for 24 days as described previously (Predescu et al., 2012). Mice were killed at day 3, day 10, day 15 and day 24; three to four mice per experimental condition [controls (wild-type mice, vehicle-and non-specific siRNA-treated mice) and siRNA ITSN -treated mice] were used; all experiments were repeated at least three times. No mouse mortality was recorded during the 24 days of the study. All experiments were approved and performed in accordance with the guidelines of Rush University Institutional Animal Care and Use Committee.

Isolation of microparticles
Blood of fully anesthetized wild-type and KD ITSN mice was drawn by cardiac puncture and using 3.8% sodium citrate as an anticoagulant. Platelet-free plasma was centrifuged at 80,000 g for 2 h at 4˚C to obtain the microparticle pellets; microparticles were either lysed or used intact for morphological approaches. All morphological approaches were performed with freshly isolated microparticles.

Fluorescent labeling of microparticles and immunofluorescent staining
Microparticles were incubated with 1 mg/ml biotin in PBS containing 0.1 M CaCl 2 and 0.1 M MgCl 2 for 20 min on ice, followed by incubation with neutrAvidin-Alexa-Fluor-594 diluted in 0.1% BSA in PBS for 1 h. The unbound biotin and neutrAvidin-Alexa-Fluor-594 were removed by three successive washings in PBS followed by centrifugation (Beckman centrifuge,TLA-55 rotor) at 80,000 g for 1 h at 4˚C. For double labeling with biotin and Alk5 antibody, microparticles were sequentially incubated with: (1) Alk5 goat primary antibody diluted in 0.1% BSA in PBS, overnight at 4˚C, followed by (2) biotin, as above and (3) a mixture of neutrAvidin-Alexa-Fluor-594 and anti-goat-IgG conjugated to Alexa Fluor 488, for 1 h at room temperature. A blocking step using 1% BSA in PBS preceded incubation with the Alk5 antibody. Successive washings in 0.1% BSA in PBS followed by centrifugation were used to remove excess biotin or antibodies. Final pellets were resuspended in PBS and fixed in 1% paraformaldehyde, and aliquots were mounted on glass slides with Prolong Antifade reagent. Isotype-matched IgG was used as a control. Microparticles were examined and photographed using a Zeiss AxioImager M1 microscope or Zeiss Laser Scanning Microscope LSM 700.
Immunofluorescent staining of endothelial cells grown on coverslips was performed as described previously (Predescu et al., 2005). Incubation of endothelial cells with isotype-matched IgGs or omission of the primary antibody were used as controls for antibody specificity. Each set of experiments was performed in triplicate.

Morphometric analysis
The degree of cav1 and Alk5 colocalization was determined by counting the cav1-positive large endocytic structures in 50 endothelial cells per coverslip, in three different experiments performed in triplicate. All images used for quantification of the degree of colocalization were acquired using identical parameters per experiment.

Microparticle-endothelial cell interaction
Endothelial cells were grown on coverslips and exposed for 1 h on ice to 12.5 mg/ml MP KD-ITSN pre-labeled with an anti-Alk5 antibody and an Alexa-Fluor-594-conjugated secondary antibody, to allow binding of the microparticles to the endothelial plasma membrane; then, cells were transferred to 37˚C for 15 min and 30 min, to allow internalization. Cells were washed, fixed in methanol for 7 min at 220˚C, quenched in 1% BSA/PBS for 1 h at room temperature and counterstained by incubation with TGFb-RII and ubiquitin antibodies, followed by their specific secondary antibodies, as above. Cells were examined and photographed using a Zeiss AxioImager M1 microscope.
Control experiments to rule out the endocytic internalization by EC KD-ITSN of non-specifically attached anti-Alk5-Alexa-Fluor-594 IgG aggregates to the MP KD-ITSN were performed. Briefly, anti-Alk5-Alexa-Fluor-594 pre-labeled MP KD-ITSN were resuspended in ice-cold acid wash buffer (DMEM/10 mM HEPES pH 5.0, 10 mM MES, 120 mM NaCl, 0.5 mM MgCl 2 and 0.9 mM CaCl 2 ) for 30 min as described previously (Koenig et al., 1997;Smalley et al., 2001). At this pH, the MP KD-ITSN are not stripped of their pre-labeled Alk5. The acidwash buffer was removed by ultra-centrifugation and the MP KD-ITSN pellet was resuspended in DMEM containing 0.1% BSA. EC KD-ITSN grown on coverslips were exposed to a mixture of 12.5 mg/ml anti-Alk5-Alexa-Fluor-594 pre-labeled MP KD-ITSN and unlabeled Alk5 antibody (dilution 1: 1000; for 1 h on ice, to allow binding of anti-Alk5-Alexa-Fluor-594 pre-labeled MP KD-ITSN to the endothelial plasma membrane and to block the endogenous Alk5 receptor, respectively. The cells were transferred to 37˚C, for 20 min to allow internalization as above.

Flow cytometry
Microparticles were isolated from wild-type and KD-ITSN mice and labeled with an APC-conjugated anti-Alk5 antibody diluted in flow cytometry staining buffer, according to the manufacturer's indications. Samples were incubated for 1 h at 4˚C in the dark, and then centrifuged for 1 h at 80,000 g. Pellets were resuspended in 1 ml of staining buffer and centrifuged, with this procedure being repeated three times to remove excess antibody. The final pellet was resuspended in 50 ml and analyzed in a LSR Fortessa flow cytometer with Diva software. Control experiments included incubation with isotype control mouse IgG. Microparticle gating was accomplished by preliminary standardization experiments using Spherotech nano fluorescent size standard beads (0.45 mm-1.35 mm). Data are presented as dot plots and the results of data analysis are presented as the mean percentage of total gated events (at least 10,000 events/sample) 6s.e.m.
Negative staining and pre-embedding immuno-EM Microparticles were fixed in 2.5% glutaraldehyde for 30 min at room temperature, absorbed onto formvar-coated nickel grids recently exposed to glow discharge and negatively stained as described previously (Predescu et al., 2001). EM grids were analyzed in a JEOL JEM-2000FX TEM. For Alk5 pre-embedding immuno-EM, thick cryostat sections of polyvinylpyrrolidone-fixed tissue were incubated with anti-Alk5 antibody followed by goat anti-rat-IgG conjugated to 8-nm gold and processed by standard EM procedure as described previously (Predescu et al., 1996).

Statistical analysis
All findings were confirmed in three to five different experiments and data are expressed as the mean6s.e.m. Stimulated samples were compared to controls by using unpaired Student's t-tests. Differences with values of P,0.05 were considered to be statistically significant.