TGF-β/BMP superfamily ligands require heteromeric complexes of type 1 and 2 receptors for ligand-dependent downstream signaling. Activin A, a TGF-β superfamily member, inhibits growth of multiple myeloma cells, but the mechanism for this is unknown. We therefore aimed to clarify how activins affect myeloma cell survival. Activin A activates the transcription factors SMAD2/3 through the ALK4 type 1 receptor, but may also activate SMAD1/5/8 through mutated variants of the type 1 receptor ALK2 (also known as ACVR1). We demonstrate that activin A and B activate SMAD1/5/8 in myeloma cells through endogenous wild-type ALK2. Knockdown of the type 2 receptor BMPR2 strongly potentiated activin A- and activin B-induced activation of SMAD1/5/8 and subsequent cell death. Furthermore, activity of BMP6, BMP7 or BMP9, which may also signal via ALK2, was potentiated by knockdown of BMPR2. Similar results were seen in HepG2 liver carcinoma cells. We propose that BMPR2 inhibits ALK2-mediated signaling by preventing ALK2 from oligomerizing with the type 2 receptors ACVR2A and ACVR2B, which are necessary for activation of ALK2 by activins and several BMPs. In conclusion, BMPR2 could be explored as a possible target for therapy in patients with multiple myeloma.
This article has an associated First Person interview with the first author of the paper.
The transforming growth factor (TGF)-β superfamily of ligands consists of over 30 human members, including bone morphogenetic proteins (BMPs), growth differentiation factors (GDFs), activins and TGF-β isoforms (Wakefield and Hill, 2013). These structurally related ligands are involved in a plethora of processes in the developing and adult organism. The ligands signal by binding to a heteromeric complex composed of type 1 and type 2 receptors, causing phosphorylation of receptor-activated SMAD transcription factors. In the case of activins, two type 1 receptors are known to relay signals between ligands and SMADs, namely ALK4 (ACVR1B) and ALK7 (ACVR1C) (ten Dijke et al., 1994; Tsuchida et al., 2004). For BMPs, four different type 1 signaling receptors have been described: ALK1 (ACVRL1), which is primarily expressed by endothelial cells, and the more ubiquitously expressed ALK2 (ACVR1), ALK3 (BMPR1A) and ALK6 (BMPR1B) (for review on BMP receptors, see Yadin et al., 2016). The activin type 1 receptors activate the SMAD2/3 pathway, as TGF-β does through ALK5 (TGFBR1), whereas the BMP type 1 receptors activate the SMAD1/5/8 pathway. Two type 2 receptors, ACVR2A and ACVR2B, are shared between activins, BMPs and GDFs, whereas only the type 2 receptor BMPR2 is used by BMPs and GDFs (Yadin et al., 2016).
Activin A has been shown to inhibit growth of normal B cells as well as murine myeloma and plasmacytoma cells, however the mechanism behind was not clear (Brosh et al., 1995; Hashimoto et al., 1998; Nishihara et al., 1993; Zipori and Barda-Saad, 2001). We have shown earlier that activation of SMAD1, SMAD5 or SMAD8 (SMAD1/5/8) is necessary and sufficient to induce apoptosis in myeloma cells (Holien et al., 2012a). BMP-induced activation of SMAD1/5/8 causes apoptosis in multiple myeloma cells through downregulation of the oncogene MYC, which is important for myeloma cell survival (Holien and Sundan, 2014; Holien et al., 2012a,b). Ligands, such as TGF-β, that activate the SMAD2/3 pathway, but not the SMAD1/5/8 pathway do not induce apoptosis in myeloma cells (Baughn et al., 2009; Olsen et al., 2015). We and others have shown that activin A may compete with BMPs for binding to ACVR2A, ACVR2B and ALK2 (Aykul and Martinez-Hackert, 2016; Hatsell et al., 2015; Hino et al., 2015; Olsen et al., 2015; Piek et al., 1999; Seher et al., 2017). Activin A could also induce weak phosphorylation of SMAD1/5/8 in addition to the commonly described activation of SMAD2 (Olsen et al., 2015). This coincided with a modest reduction in myeloma cell viability. Interestingly, it was shown that activin A binds and signals through a hypersensitive R206H mutant variant of ALK2 (Hatsell et al., 2015; Hino et al., 2015).
In this study, we wanted to clarify how activins can affect myeloma cell survival. We found that both activin A and activin B signal through endogenous wild-type ALK2 to phosphorylate SMAD1 or SMAD5, and consequently induce myeloma cell death. Surprisingly, depletion of BMPR2 strongly augmented ALK2-mediated, but not ALK4-mediated signaling, induced by either activins or BMPs. In conclusion, we propose that BMPR2 prevents ALK2 from forming active receptor signaling complexes with ACVR2A and ACVR2B, thus inhibiting ALK2-mediated SMAD1/5/8 activation and myeloma cell death.
Activin A and activin B inhibit myeloma cells through ALK2 and SMAD1/5/8
Five different myeloma cell lines were treated with increasing doses of activin A or activin B for 3 days. A dose-dependent decrease in relative viable cell numbers was seen in IH-1 cells and to a lesser extent in the INA-6 cell line (Fig. 1A,B). The other three cell lines were for a large part refractory to activin A- or activin B-induced growth inhibition in the doses used here. Notably, both activins induced activation of SMAD2 in all cell lines, whereas a clear activation of SMAD1/5/8 was also observed in the IH-1, INA-6 and KJON cell lines (Fig. 1C and Fig. S1A,B). Contamination by other ligands is a potential issue when using recombinant proteins (Carthy et al., 2016; Olsen et al., 2017). Thus, to confirm that the observed effects were actually caused by activin A and activin B, we performed experiments on SMAD activity in the presence of neutralizing antibodies (Fig. S2). As shown, antibodies towards activin A and activin B reduced activin A- and activin B-induced SMAD1/5/8 phosphorylation to background levels, indicating that the observed SMAD1/5/8 phosphorylation was indeed due to activin A and activin B and not to contaminants. We and others have previously shown that activation of SMAD1/5/8, but not SMAD2, induced growth arrest or apoptosis in human myeloma cells (Baughn et al., 2009; Holien et al., 2012a; Olsen et al., 2015). In line with this, we also show here that activin A and activin B induced apoptosis in IH-1 cells as measured by annexin-V labeling and cleavage of caspase 3 (Fig. 1D,E and Fig. S1C). Activin A has previously been shown to signal through ALK4 whereas activin B binds and signals through either ALK4 or ALK7 (Wakefield and Hill, 2013). Both these receptors activate SMAD2/3, but not SMAD1/5/8. Thus, the results above suggest that activin A and activin B also may bind and signal through complexes containing BMP-type 1 receptors.
Activins signal through endogenous wild-type ALK2 in myeloma cells
To clarify which BMP-type 1 receptor could be responsible for activin-induced activation of SMAD1/5/8, we first measured the mRNA levels of the potential TGF-β superfamily receptors (Fig. 2A). ALK1 (ACVRL1) and ALK6 (BMPR1B) were also measured, but omitted from the figure since none of the cell lines showed expression of these two receptors. The only BMP-type 1 receptor expressed in INA-6 cells is ALK2, whereas IH-1 cells express both ALK2 and ALK3. This suggested that ALK2 is the crucial type 1 receptor responsible for the effects measured. We then employed DMH1 and K02288, small molecule BMP type 1 receptor kinase inhibitors, to evaluate their effects on apoptosis induced by activin A and activin B. DMH1 targets BMP-activated type 1 receptors, whereas K02288 has been described as an inhibitor that targets ALK2 more potently than other BMP-activated ALKs (Hao et al., 2010; Sanvitale et al., 2013). Both compounds significantly inhibited the reduction in cell viability caused by activin A and activin B (Fig. 2B,C). We then compared these results with inhibition of ALK4/5/7 by using the SB431542 inhibitor (Inman et al., 2002). This inhibitor showed no significant effects on activin A- or activin B-induced apoptosis (Fig. 2D). To target ALK2 even more specifically, we utilized a novel, monoclonal ALK2-neutralizing antibody termed DaVinci (Katagiri et al., 2017). The antibody blunted apoptosis induced by activin A, activin B or BMP9 (Fig. 2E), the latter ligand we have previously shown to activate SMAD1/5/8 via ALK2 in these cells (Olsen et al., 2014). Taken together, the results support that ALK2 is the BMP-type 1 receptor for activin A and activin B in myeloma cells. Of note, whole exome sequencing of 69 human myeloma cell lines, including INA-6, did not reveal any mutations in ALK2 (data available from the Keats lab website; http://www.keatslab.org/data-repository). We also sequenced RNA from our own batch of INA-6 cells. Two single nucleotide variations (SNVs) were detected in the ALK2 gene (ACVR1), denoted rs2227861 and rs1146031, neither of which causes a change in the amino acid sequence (data not shown). Thus, we report here that both activin A and activin B signal through wild-type ALK2 in myeloma cells to induce apoptosis.
Loss of BMPR2 potentiated activin-induced SMAD1/5/8 activity
ACVR2A and ACVR2B are known type 2 receptors for activins and BMPs. Also, BMPR2 is an important BMP type 2 receptor, but the role for BMPR2 in activin-mediated signaling is less clear. We thus generated cells with stable expression of shRNA targeting BMPR2 (shBMPR2), using as controls non-targeting shRNA (shCTR) or shRNA targeting ACVRL1 (shALK1), which is not expressed by myeloma cell lines. Expression of BMPR2 mRNA was measured in shRNA-expressing cells and indicated a reduction to ∼10% and 40% compared with the levels in control-transduced INA-6 and IH-1 cells, respectively (Fig. 3A,B). BMPR2 protein levels could be detected by immunoblotting in INA-6 cells and there was a clear downregulation of BMPR2 in the shBMPR2-cells compared with control cells (Fig. S3). In IH-1 cells we were not able to detect BMPR2 protein clearly, and thus, we could not determine the degree of downregulation at the protein level. Importantly, activin A activates SMAD1/5/8 more strongly in cells where BMPR2 was downregulated than in control cells. Activin A-induced SMAD2 phosphorylation in the same cells however, was less affected by BMPR2 knockdown (Fig. 3C,D and Fig. S1D-G) and mRNA levels of ID1, a known SMAD1/5/8 target gene (Chen et al., 2006) were also more potently induced by activin A in shBMPR2 cells (Fig. 3E,F). Finally, activin A- and activin B-induced growth inhibition was potentiated by knockdown of BMPR2 (Fig. 3G-J). To exclude possible compensatory upregulation of other receptors, we measured mRNA levels of ACVR1, ACVR2A and ACVR2B in stable knockdown cells. No changes were measured that could explain the potentiating effects in the shBMPR2 cells (Fig. S4). Taken together, we show here that activin A- and activin B-induced signaling through ALK2 is potentiated by stable knockdown of BMPR2 in both short- and long-term experiments. The results indicate that presence of BMPR2 inhibits activin-induced ALK2 signaling.
Loss of BMPR2 potentiates the effect of some BMPs
Based on the effects obtained with activins, we wondered whether BMP signaling could also be potentiated in BMPR2-knockdown cells. The INA-6 cell line expressed the type 2 receptors BMPR2, ACVR2A and ACVR2B, but only ALK2 as a BMP type 1 receptor (Fig. 2A). BMP6, BMP7 and BMP9 are the only BMPs currently known to activate SMAD1/5/8 and induce apoptosis in this cell line. IH-1 cells express in addition to ALK2 the ALK3 receptor and thereby responds to all BMPs tested (BMP2, BMP4, BMP5, BMP6, BMP7 and BMP9) (Hjertner et al., 2001; Olsen et al., 2014; Ro et al., 2004). Expression of shBMPR2 in either INA-6 (Fig. 4A-C) or IH-1 (Fig. 4D,E) cells resulted in a more pronounced, dose-dependent decrease in viability after treatment with BMP6, BMP7 (only shown for INA-6) or BMP9 than in control cells (Fig. 4A-E). In contrast, treatment with BMP2, BMP4 or BMP10 was not affected by shBMPR2, indicating that these ligands depend on other receptor combinations to produce a signal (Fig. 4F-H).
Effect of BMPR2 knockdown in HepG2 liver carcinoma cells
To investigate if the effects of BMPR2 expression levels on BMP- and activin-induced activation of SMAD1/5/8 signaling also applies for non-myeloma cells, we transiently transfected HepG2 liver carcinoma cells with siRNA targeting BMPR2 (siBMPR2) or non-targeting siRNA (siCTR). The cells were treated for 4 h with BMP6, activin A or activin B (all at a concentration of 10 ng/ml) and activation of SMADs was determined by immunoblotting. Interestingly, activation of SMAD1/5/8 by BMP6 or activin B was also potentiated by BMPR2 knockdown in HepG2 cells, whereas activation of SMAD2 remained unaffected for all three ligands (Fig. 5A and Fig. S1H,I). The mRNA expression of the SMAD target gene ID1 in activin A-treated HepG2 cells after BMPR2 knockdown was also augmented (Fig. 5B). BMPR2 mRNA in these experiments was reduced to <20% when compared with levels in control cells (Fig. 5C). In summary, ligand-induced activation of SMAD1/5/8 is potentiated by lowering the expression of BMPR2 in HepG2 cells as well as in myeloma cells.
Re-expressing long or short forms of BMPR2 inhibited activin-induced SMAD1/5/8 activity
We wanted to further elucidate the role of BMPR2 in regulating ALK2-mediated signaling in myeloma cells. Therefore, we took advantage of INA-6 shBMPR2 cells, which have low levels of BMPR2, and re-expressed the long (LF) and the short form (SF) of BMPR2. The plasmids used encoded BMPR2 variants with N-terminal MYC tags between the signal peptide (SP) and the extracellular domain (ECD), and have been described earlier (Fig. 6A) (Amsalem et al., 2016). The levels of BMPR2-LF and BMPR2-SF were highly increased compared with levels in control cells (Fig. 6B,C). BMPR2 mRNA was not detected in cells overexpressing BMPR2-LF or BMPR2-SF by PCR spanning exons 1 and 2 (Fig. 6D). Interestingly, both the long and the short form counteracted activin B-induced SMAD1/5/8 activation (P<0.05 and P<0.01, respectively), whereas the effect on activin A-induced SMAD1/5/8 activation was not significant (Fig. 6E and Fig. S1J). Taken together, we propose that BMPR2 counteracts signaling via ALK2 in multiple myeloma cells.
Multiple myeloma is regarded as incurable and there is a need for improved therapy options. BMPs potently induce growth arrest or apoptosis in myeloma cells and the BMP signaling pathway could therefore be a potential target for therapy. The aim of this study was to clarify how activins affected myeloma cell survival. We previously found that activin A could inhibit BMP6 and BMP9 signaling through ALK2 (Olsen et al., 2015). However, other studies have reported that activin A could induce apoptosis in B cells and murine myeloma and plasmacytoma cells (Brosh et al., 1995; Hashimoto et al., 1998; Nishihara et al., 1993; Zipori and Barda-Saad, 2001). In myeloma cells, activation of the SMAD2 pathway has not been found to induce apoptosis, whereas we previously found that activation of the SMAD1/5/8 pathway was necessary and sufficient for induction of apoptosis (Holien et al., 2012a). The most important finding presented here is that activin A and activin B activated the SMAD1/5/8 pathway through ALK2, thus inducing apoptosis in multiple myeloma cells. Moreover, reducing the levels of the type 2 receptor BMPR2 potentiates ALK2-SMAD1/5/8 signaling by activins and those BMPs that act through ALK2.
ALK2 was initially described as a type 1 receptor that could bind activin A, but no evidence was found for activin A-induced activation of the SMAD1/5/8 pathway (Attisano et al., 1993; Ebner et al., 1993; ten Dijke et al., 1994). However, activin A was shown to induce activation of SMAD1/5/8 by signaling through an R206H mutated variant of ALK2 (Hatsell et al., 2015; Hino et al., 2015). Just before submission of this manuscript, there was also a report describing that activin A could induce SMAD1/5/8 activity through overexpressed wild-type ALK2, strongly supporting our findings (Haupt et al., 2018). Moreover, activin A or BMP-4 synergistically activate SMAD1/5/8 in connective tissue progenitor cells from patients with the rare inherited disease fibrodysplasia ossificans progressiva (FOP) as well as from healthy individuals (Wang et al., 2018). Activin B has been shown to activate SMAD1/5/8 in hepatocytes through ALK2, ACVR2A and ACVR2B (Besson-Fournier et al., 2012; Canali et al., 2016). In line with these results, we show here that both activin A and activin B are able to activate SMAD1/5/8 through the endogenous wild-type ALK2 receptor in myeloma cells, leading to cell death.
The variation in response to activin A and activin B between the different myeloma cell lines was puzzling. The KJON cells expressed the same levels of ALK2 as INA-6 cells did, and lower levels of BMPR2, but still responded more poorly to activins. One explanation could be that KJON cells express or lack expression of other co-factors that regulate activin-induced SMAD1/5/8 signaling. The JJN-3 cell line did not respond significantly to activins either, but this could possibly be explained by lower levels of ALK2.
Activin A-induced SMAD1/5/8 activation and consequent myeloma cell death was strongly augmented when levels of BMPR2 were reduced. Furthermore, the effects of activin B, BMP6, BMP7 and BMP9, which may also utilize ALK2 as a type 1 receptor, were potentiated in the same way. Our results resemble previously described effects in human mesenchymal stem cells, where BMP2 and BMP4 preferred BMPR2 over ACVR2, whereas BMP6 and BMP7 preferred ACVR2 over BMPR2 for signaling (Lavery et al., 2008). Also, in hepatocytes, knockdown of BMPR2 potentiated BMP6- and activin B-induced SMAD1/5/8 activity (Canali et al., 2016). In pulmonary artery smooth muscle cells (PASMCs) loss of BMPR2 augmented signaling by BMP6 and BMP7, whereas BMP2 and BMP4 signaling was diminished (Yu et al., 2005). In contrast to these reports in PASMCs, we observed no change in the activity of BMP2 or BMP4. We speculate that this may be due to compensatory utilization of the type 2 receptors ACVR2A and ACVR2B by BMP2 and BMP4 in the cell types used here. The same might apply for BMP10-induced growth inhibition, which was also not changed in cells with lower levels of BMPR2. Notably, we cannot rule out that some of the ligands used in this study could signal via heterodimeric type 1 receptors instead of just ALK2, at least in the IH-1 cell line, which expresses both ALK2 and ALK3. On the other hand, another study found that deletion of Bmpr2 in mouse skeletal progenitor cells selectively impaired activin A-induced SMAD2/3 activation, but had no effect on BMP signaling, causing increased bone formation and bone mass (Lowery et al., 2015). We did not see much difference in activin A- or activin B-induced activation of SMAD2 in myeloma cells by lowering BMPR2 levels. The differences between our study and the previously reported results may be due to variations in cellular receptor expression.
Based on our results, one might speculate that increasing the ability of BMPR2 to interact with ALK2 could dampen ALK2 signaling. Mutations found in ALK2 in the rare inherited disease FOP increase the ability of activin A to signal through ALK2 (Hatsell et al., 2015; Hino et al., 2015). It has been suggested that the effect of activin A signaling through ACVR1 R206H is cell type specific, i.e. activin A-induced SMAD1/5/8 phosphorylation was induced in human induced pluripotent stem cells (hiPSCs), but not in hiPSC-derived endothelial cells (iECs) from FOP patients (Barruet et al., 2016). Interestingly, the expression levels of BMPR2 in hiPSCs were lower than in iECs, and in light of our findings, this could possibly explain some of the cell-specific observations. It was shown that type 2 receptors were absolutely required for signaling via the hypersensitive ALK2 in FOP patients, with ACVR2A being more efficient than BMPR2 (Bagarova et al., 2013). The requirement for the type 2 receptor was not dependent on ligand binding, type 2 receptor kinase activity or the BMPR2 tail domain, but the presence of the cytoplasmic domain of this receptor was essential.
Many ligands bind ACVR2A and ACVR2B more strongly than BMPR2 (Aykul and Martinez-Hackert, 2016), thus we cannot rule out that some of the observed effects are due to differences in type 2 receptor affinity and/or avidity. On the other hand, BMP10 binds more strongly to ACVR2A and ACVR2B than BMP6 does, but is not potentiated by BMPR2 knockdown, suggesting that this could not entirely explain our results. Another possible mechanism for the potentiating effects of activin A seen after knocking down BMPR2, could be that activin A binds equally well to the ALK2-BMPR2 and ALK2-ACVR2 complexes, but induces a weaker signal through ALK2-BMPR2. This hypothesis could be supported by a report showing that the tail domain of BMPR2 could inhibit ALK2-mediated BMP7 signaling, but did not affect BMP4 signaling, in PASMCs (Leyton et al., 2013). In our experiments, ALK2-mediated activation of SMAD1/5/8 was inhibited by both the long and short forms of BMPR2, indicating that the inhibitory effect we see is independent of the tail domain. Crosslinking experiments with radioactive activin A showed binding of activin A to ALK2 when the receptor was co-expressed with ACVR2A or ACVR2B, but not with BMPR2 (Hino et al., 2015). Another study found that overexpression of Müllerian inhibiting substance receptor II (MISRII) could sequester ALK2. Activin A signaling through ALK4 was thereby potentiated, suggesting that the presence of ALK2 inhibited activin A signaling by blocking ACVR2 receptor binding to ALK4 (Renlund et al., 2007). The latter study also showed that ALK2 could bind ACVR2A in the absence of ligand. Both studies support our hypothesis that BMPR2 might sequester ALK2 and prevent complex formation between ALK2 and ACVR2A or ACVR2B. A proposed model of our hypothesis is provided (Fig. 7).
This is the first report to show that the TGF-β family ligands activin B and BMP10 can induce apoptosis in myeloma cells. Activin B has been shown to bind ALK2 and activate SMAD1/5/8 in hepatocytes (Canali et al., 2016). Both activin B and BMP10 have been shown to bind with strong affinity to the type 2 receptors ACVR2A and ACVR2B (Aykul and Martinez-Hackert, 2016; Koncarevic et al., 2012). We also found that in myeloma cells, activin B could signal via ALK2 to activate SMAD1/5/8, thus inducing apoptosis. BMP10 and BMP9 are members of the same subfamily based on their amino acid sequence homologies (Mazerbourg et al., 2005). Multiple myeloma cells expressing both ALK2 and ALK3 (IH-1 cells) respond to BMP10 by inducing apoptosis, while cells only expressing ALK2 (INA-6 cells) are not affected by BMP10. Further studies are needed to determine the exact involvement of type 1 receptors for BMP10 in myeloma cells.
BMPs and activins can be targeted clinically by the use of soluble decoy receptors (for a recent review, see Lowery et al., 2016). This approach is based on the affinities of the ligands to the receptor and may not be a specific way of targeting a particular signaling pathway. In attempts to treat multiple myeloma, it may be beneficial to specifically increase BMP activity in cancerous cells. Our results clearly indicate that one promising way of doing this is by reducing the levels of BMPR2. To conclude, we hypothesize that reducing the levels of BMPR2 enables increased formation of endogenous receptor complexes of ALK2 with ACVR2A or ACVR2B. The BMPR2 levels could thus dictate the degree of SMAD1/5/8 activation by activin A, activin B and several BMPs. Increased BMPR2 expression in myeloma cells could therefore be one way of escaping the tumor suppressive effects of activins and BMPs. Thus, lowering BMPR2 levels in myeloma cells may be beneficial, but this issue needs to be investigated further.
MATERIALS AND METHODS
Cells and reagents
The human multiple myeloma cell lines INA-6 and JJN-3 were kind gifts from Dr Martin Gramatzki (University of Erlangen-Nurnberg, Erlangen, Germany) and Dr Jennifer Ball (University of Birmingham, UK), respectively. RPMI-8226 cells were from American Type Culture Collection (ATCC; Rockville, MD, USA). IH-1 and KJON were established in our laboratory (Hjertner et al., 2001; Våtsveen et al., 2016). The myeloma cell lines were grown as described previously (Holien et al., 2015). The hepatocyte carcinoma cell line HepG2 was from European Collection of Authenticated Cell Cultures (ECACC; Salisbury, UK). HepG2 cells were grown in 10% fetal calf serum (FCS) in Eagle's minimum essential medium supplemented with 2 mM glutamine and non-essential amino acids (Sigma-Aldrich Norway, Oslo, Norway). All cell lines were cultured at 37°C in a humidified atmosphere containing 5% CO2 and tested regularly for mycoplasma. Fingerprinting was used to confirm the authenticity of the myeloma cell lines. For experiments, 2% human serum (HS) (Department of Immunology and Transfusion Medicine, St. Olav's University Hospital, Trondheim, Norway) in RPMI was used as medium, with IL-6 (1 ng/ml) (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) added for INA-6 and KJON cells. All other recombinant human proteins were from R&D Systems (Bio-Techne, Abingdon, UK). The neutralizing rat monoclonal ALK2 antibody, DaVinci, was a kind gift from Saitama Medical University (Saitama, Japan).
Cells treated as indicated were used for immunoblotting as previously described (Olsen et al., 2015). Primary antibodies used were against: phospho-SMAD1/5, phospho-SMAD1/5/9, phospho-SMAD2, and cleaved caspase-3 (Cell Signaling Technology, Medprobe, Oslo, Norway), and GAPDH (Abcam, Cambridge, UK). Details of all primary antibodies are listed in Table S1. Blots were incubated with horseradish peroxidase-conjugated secondary antibodies (Dako Cytomation, Glostrup, Denmark) and detected using SuperSignal West Femto (Thermo Fisher Scientific) as a luminescence substrate and an Odyssey Fc Imager with Image Studio software (LI-COR Biosciences, Cambridge, UK).
Relative viable cell numbers were determined by the CellTiter-Glo assay (Promega, Madison, WI, USA), which measures ATP levels by luminescence, as described (Olsen et al., 2015). Luminescence was detected using a Victor 1420 multilabel counter and Wallac software (PerkinElmer, Waltham, MA, USA).
To measure changes in cell viability, cells were stained using Apotest FITC kit (Nexins Research, Kattendijke, The Netherlands). In brief, cells were incubated with annexin-V–FITC (0.2 µg/ml in 1× binding buffer) for 1 h on ice. Propidium iodide (PI) (1.4 µg/ml) was added 5 min prior to data acquisition using an LSRII flow cytometer (BD Biosciences, San Jose, CA). Cells negative for both annexin-V and PI staining were considered viable.
Generation of shRNA-expressing cell lines
ShRNA constructs were purchased as ready-made lentiviral particles and consisted of pools of three different sequences targeting BMPR2 (sc-40220-V) or ALK1 (sc-40212-V), and non-targeting control shRNA (sc-108080) (Santa Cruz Biotechnology, Heidelberg, Germany). Briefly, cells were treated with lentiviral particles at a multiplicity of infection (MOI) of 15 in the presence of polybrene (8 µg/ml). Fresh medium was added after 24 h. Puromycin was added after 48 h to select for cells expressing the shRNA and the recovered viable cells were used for further experiments. Aliquots of the stable cells were also frozen on liquid nitrogen and thawed for later experiments. Knockdown of BMPR2 was controlled regularly.
RNA isolation, complementary DNA (cDNA) synthesis and PCR were performed using StepOne Real-Time PCR System and Taqman Gene Expression Assays (Applied Biosystems, Thermo Fisher Scientific) as described previously (Olsen et al., 2015). The Taqman assays used were: ACVR1 (Hs00153836_m1), ACVR2A (Hs00155658_m1), ACVR2B (Hs00609603_m1), BMPR2 (Hs00176148_m1, Hs01556134_m1 and Hs01574531_m1), GAPDH (Hs99999905_m1) and ID1 (Hs00357821_g1). The comparative Ct method was used to calculate relative changes in gene expression with GAPDH as housekeeping gene.
Nanostring gene expression analysis
For mRNA transcript counting, we used nCounter Technology (Nanostring Technologies, Seattle, WA, USA) and a custom-made TGF-β nCounter Kit. The standard mRNA gene expression experiment protocol provided by Nanostring was used. Briefly, 100 ng total RNA from myeloma cell lines was hybridized with reporter probes overnight at 65°C. The nSolver Analysis Software (Nanostring) was used for calculations of transcript numbers. Sample data was normalized against internal kit positive controls and the following housekeeping genes: ABCF1, EEF1G, GAPDH, OAZ1, POLR2A, RPL19 and TUBB.
Transient knockdown in HepG2
HepG2 cells were seeded in 6-well plates and left overnight to adhere. Cells at a confluence of about 50% were transfected with siRNA using Lipofectamine RNAiMAX (Invitrogen) according to the protocol. For BMPR2 knockdown, a mix of two different Silencer Select siRNAs was used (siRNA IDs s2045 and s2046) and compared with a mix of Silencer Select Negative Control No. 1 and No. 2 (Ambion, Thermo Fisher Scientific). On the day after transfection, the cells were treated with the indicated ligands for 4 h and harvested for immunoblotting or PCR.
Transfection of INA-6
INA-6 shBMPR2 cells were transfected using the Nucleofector device (Amaxa Biosystems, Cologne, Germany) and Amaxa Cell Line Nucleofector Kit R (Lonza, Basel, Switzerland), as previously described (Fagerli et al., 2008). The plasmids encoding the long and short isoforms of BMPR2, BMPR2-LF and BMPR2-SF, were generated earlier as described (Amsalem et al., 2016). A plasmid encoding CD4Δ was used as a negative control (kind gift from Dr Martin Janz, Charité-University Medicine Berlin, Germany). On the day after transfection, dead cells were removed by use of Optiprep density gradient medium (Axis-Shield, Oslo, Norway) and the remaining viable cells were used directly in experiments.
GraphPad Prism 7 (GraphPad Software, San Diego, LA) was used to analyze statistical significance. The tests used for each particular experiment are described in the figure legends.
The authors are grateful for skilled technical help from Berit Størdal. We also thank Anders Sundan for helpful comments on the manuscript.
Conceptualization: O.E.O., P.K., T.H.; Methodology: O.E.O., T.H.; Investigation: O.E.O., M.S., S.E., H.H., G.B., S.R.D., K.M., T.H.; Resources: T.K., P.K.; Writing - original draft: O.E.O., T.H.; Writing - review & editing: O.E.O., M.S., S.E., H.H., G.B., S.R.D., K.M., T.K., P.K., T.H.; Supervision: T.H.; Project administration: T.H.; Funding acquisition: T.H.
The work was supported by a grant from Kreftforeningen (Norwegian Cancer Society; grant 5793765) (T.H.), the Helse Midt-Norge (O.E.O., T.H.), the Joint Research Committee between St. Olav's Hospital and Faculty of Medicine and Health Science, Norges Teknisk-Naturvitenskapelige Universitet (T.H.), Daiichi-Sankyo (T.K.) and the Deutsche Forschungsgemeinschaft (SFB958) (P.K.).
The authors declare no competing or financial interests.