Background

Gradients of bone morphogenetic proteins (BMPs) act as mesenchymal guidance cues during development, disease and tissue repair by molecular mechanisms that remain poorly defined [1]. In particular, the directional migration (chemotaxis) of neural crest cells, bone marrow stromal cells and endothelial cells along gradients of BMP2 has been reported [2–5]. BMPs signal through binding to cell surface hetero-oligomeric receptor complexes comprising type I (BMPRI) and type II (BMPRII) receptors [6]. Activated BMP receptor complexes induce canonical-Smad and non-Smad signalling cascades [7]. Activation of the type I receptor kinase by the type II receptor kinase induces phosphorylation and thus nuclear translocation of Smad1/5/8, leading to transcription of Smad-dependent target genes [8].

Whereas the molecular basis of canonical Smad signalling and its role in gene transcription is well explored, the molecular activation mechanism and the cellular functions of the non-Smad pathways, which rather act directly and independently of gene transcription, are poorly understood. In particular, the molecular mechanism of BMP-induced phosphatidylinositol 3-kinase (PI3K) activation, its signalling route and cellular function are poorly characterised. In recent years, several studies unveiled a requirement of PI3K for BMP2-induced migration of various cell types with mesenchymal origin by yet unknown mechanisms [9–11].

Here, for the first time, we addressed the molecular activation mechanism of BMP2-induced PI3K signalling in undifferentiated mesenchymal progenitor cells and the role of the lipid-product of PI3K, the membrane-bound second messenger PtdIns-3, 4, 5-triphosphate (PI (3, 4, 5) P3; hereafter referred to as PIP3) in BMP2-induced actin reorganisation.

Class Ia PI3Ks are dimeric lipid kinases composed of one out of five possible regulatory subunits encoded by Pik3r1 (encoding splice isoforms p85α, p55α and p50α), Pik3r2 (p85β) or Pik3r3 (p55γ) [12, 13]. The regulatory subunit is bound by one of three catalytic subunits, termed p110, encoded by Pik3ca (p110α), Pik3cb (p110β) or Pik3cd (p110δ) [14]. Catalytic activity is initiated upon regulatory subunit Src homology 2 (SH2) domain binding to phospho-tyrosine (pTyr) residues within a specific peptide context [15]. Thereafter, activated PI3K phosphorylates the 3-hydroxyl group of PtdIns-4, 5-bisphosphate (PIP2) to produce the second messenger PIP3. PIP3 recruits Pleckstrin homology (PH) domain-containing regulators to the inner plasma membrane. One main PI3K effector is protein kinase B (PKB/Akt) [16]. Besides Akt, PH-domain-containing cytoskeletal regulators sense PIP3 and mediate cortical actin dynamics at the so-called leading edge cytocortex. As such, the PH-like domain family B member 2 (Phldb2, hereafter referred to as LL5β) acts as a sensitive PIP3 effector during the establishment of planar cell polarity (PCP), lamellipodia formation, protrusion and subsequent chemotaxis [46–48], suggests that the p55γ/p110α complex positively regulates BMP2-induced motility, chemotaxis, and invasion of endothelial and cancer cells [9, 49, 50]. Whether the PI3K p55γ/p110α dimer indeed represents an attractive molecular target to interfere with BMP2-related cancers will require intense investigations in future.

BMP2-induced PIP3 acts as a cellular compass at the leading edge and recruits LL5β

Numerous cellular activities have been reported to depend on BMP2-induced PI3K signalling [9–11, 51–56]. Most previous studies focused on the role of PI3K-induced Akt activity with Akt being the major PI3K effector. In the present study, we investigated the role and function of PIP3 beyond Akt activation and focused on PIP3 localisation and recruitment of cytoskeletal regulators. We visualised BMP2-dependent PIP3 production in a spatiotemporal manner to gain further insight into its function. We found PIP3 became quickly enriched in BMP2-induced lamellipodia at the cytocortex, especially in cells that displayed strong PCP, suggesting that PIP3 acts as a cellular compass at the leading edge of migrating cells. PIP3 recruits PH-domain-containing proteins that facilitate rearrangements of the actin cytoskeleton [57]. With this knowledge, we aimed to identify PH-domain proteins that link BMP2-induced PIP3 to actin regulators. The BMP2-induced lamellipodia are tightly cross-linked F-actin networks located at the cytocortex of the leading edge. During maturation and protrusion, these actin-rich lamellipodia form broad lamella that allow for the formation of new adhesion sites [58]. In agreement with our observations, we identified a specific interaction of PH-domain protein LL5β with PIP3. LL5β acts as a highly sensitive PIP3 effector during epidermal growth factor-induced chemotaxis and lamellipodia formation [17]. It regulates the actin cytoskeleton through interaction with and co-recruitment of filamin C [19] and filamin A [17]. Filamins orchestrate cortical actin into three-dimensional structures by cross-linking of F-actin filaments [59]. Interestingly, besides tethering filamins, LL5β also tethers Cytoplasmic linker associated proteins (CLASPs) to the leading edge [17, 18]. CLASPs attach microtubule tips to the cell cortex, which is important for microtubule stabilisation and thus PCP. Therefore, our findings provide evidence that LL5β acts as a BMP2-dependent multifunctional PIP3-sensing scaffold that eventually also orchestrates microtubule stabilisation at the cytocortex and thus links BMP2-dependent actin rearrangements to microtubule stabilisation.

p55γ and LL5β are required for BMP2-induced migration and chemotaxis

The potency of BMP2 in stimulating migration of cells with mesenchymal origin is well known. Here, we raised the question of whether our findings contribute in particular to BMP2-induced cortical actin rearrangements, PCP and chemotaxis. We demonstrated that loss of p55γ prevents cells from efficient PCP establishment during wound healing and that knock-down of p55γ or LL5β resulted in impaired BMP2-induced chemotaxis. We therefore conclude that the pro-migratory effects of BMP2 are driven by PI3K signalling leading to PIP3-dependent cytoskeletal actin rearrangements, and result mainly in directional migration explained by the ‘compass’ function of PIP3.

Conclusions

Our molecular findings provide a basis for explaining the important mechanism of BMP2-induced cortical actin rearrangements and chemotaxis, which we have graphically summarised (Figure 8). The novel in vitro data presented here close gaps in our current understanding of how BMP2 gradients influence the cellular cytoskeleton and hence mesenchymal progenitor cell chemotaxis. Interestingly, PIP3 production increases the efficacy of cells in detecting and processing shallow chemokine gradients [60]. This suggests that the molecular mechanism identified here is important for mesenchymal progenitor cells that respond to BMP2 gradients in vivo where they might originate from distant locations. To visualise this in vivo in the context of our novel molecular findings will be the future goal and a translation of this knowledge towards the fields of developmental biology and regenerative medicine is expected.

Methods

Chemicals, recombinant growth factors and inhibitors

All chemicals were purchased from Sigma Aldrich unless stated otherwise. Recombinant human BMP2 was kindly provided by Walter Sebald (University of Würzburg, Würzburg, Germany). The inhibitor LDN-193189 was a kind gift from Paul Yu (Harvard Medical School, Boston, MA, USA) and described elsewhere [61]. LY294002 was purchased from Cell Signaling Technology (Cell Signaling Technology Inc., Danvers, MA, USA) and PI103 was purchased from Echelon Bioscience (Echelon Bioscience Inc., Salt Lake City, USA).

Antibodies

Phospho-specific antibodies, as well as protein- and tag-specific antibodies, were used and applied as recommended by the manufacturer. A detailed list of all antibodies used in this study is provided in Additional file 7.

Cell culture

C2C12 cells and HEK293T cells (both from American Type Culture Collection) were cultivated in Dulbecco’s modified Eagle’s Medium (DMEM) (Biochrom GmbH, Berlin, Germany) supplemented with 10% (v/v) foetal calf serum and 100 U/ml penicillin/streptomycin. To maintain highest plasticity, C2C12 cells were kept undifferentiated and competent for BMP-induced signalling by subculture conditions that maintained a low density corresponding to approximately 150,000 cells per 182 cm2. Cells were split every other day when reaching 30% to 40% confluency and not used at passages higher than 20. Seeding in higher densities such as required for scratch wound healing was performed 12 hours prior to the experiment. C2C12 cells were transfected 48 hours prior to seeding in six-well plates with 0.5 to 3 μg plasmid DNA or 50nM siRNA (Dharmacon, GE Healthcare, Lafayette, CO, USA) (see Additional file 8: Table T1) using Lipofectamine2000 and Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA, USA) according to manufacturer’s instructions. HEK293T cells were transfected using polyethyleneimine and expanded in high glucose (4,500 mg/l glucose) DMEM, 48 hours prior to experiment. All experiments requiring BMP2 stimulation were conducted after 6 hours starvation in DMEM without serum. Cells were grown on uncoated cell culture plastic unless stated otherwise.

Expression plasmids

The plasmids encoding human BMPRII-LF-HA or mouse BMPRIb-HA were described previously [20, 62, 63]. Single point mutations used to generate kinase dead receptors were generated by cyclic mutagenesis PCR as described in [64]. The construct encoding N-terminal flag-tagged p55γ was generated by cloning the full-length open reading frame of mouse p55γ into the TOPO-TA vector (Invitrogen, Carlsbad, CA, USA) before ligation via EcoRI/NotI into pcDNA3.1 basic. Cloning primers used in this paper are available upon request. The construct encoding HA-tagged p85α was a kind gift from Bart Vanhaesebroeck (QMUL, London, UK). The construct encoding GFP-tagged PH-domain of Akt was a kind gift from Kerstin Danker (Charité Berlin, Germany). All constructs were verified by DNA sequencing.

Immunoprecipitation assays

Immunoprecipitation of expressed proteins from HEK293T cells was performed using a modified radio-immunoprecipitation assay buffer containing 0.5% (w/v) sodium dodecyl sulphate and 0.1% Nonidet P-40. Immunoprecipitation from C2C12 cell extracts was performed using a modified radio-immunoprecipitation assay with 0.1% sodium dodecyl sulphate and 0.5% Nonidet P-40. A detailed description of the immunoprecipitation and immunoblotting procedures can be found in Additional file 7. PIP bead assay was purchased from Echelon Bioscience and precipitation was performed according to manufacturer’s instructions.

Mass spectrometry

Identification of p55γ binding to GST-BMPRII was performed as described in [20]. PIP bead-binding proteins were identified by matrix-assisted laser desorption ionisation-time of flight mass spectrometry-based peptide mass fingerprinting as described previously [65].

Scratch wound healing

The scratch wound healing assay was performed using cell culture inserts (ibidi GmbH) according to the manufacturer’s instructions on uncoated tissue culture plastic. A detailed description of the procedure can be found in Additional file 7. The rate of cell migration was measured by quantifying the intensity translocation values for three independent biological replicates per condition using a selective mask filter (Slidebook).

Boyden chamber assay

The assay was performed in a similar manner to [10] with a detailed description of the procedure in Additional file 7.

Chemotaxis assays

Two-dimensional chemotaxis was assayed using the μ-slide chemotaxis chamber system (ibidi GmbH, Martinsried, Germany) according to accompanying instructions with the following modifications: 1 day prior to seeding, chambers were coated with 0.5% gelatin solution in humidified atmosphere washed for 1 hour and dried at 37°C. Pictures were taken using a 4× objective in bright field modus. Measurements were performed using an automated sample table mounted on an Axiovert 200 M (Carl Zeiss, Jena, Germany) in combination with Axiovision Mark&Find tool. Manual cell tracking was performed using the open source ImageJ plugin Manual tracking v2.0.

Immunofluorescence and live cell imaging

For detection of fluorescent signals, we used the Alexa-conjugated secondary antibody system (Invitrogen, Carlsbad, CA, USA) and an inverted fluorescence Axiovert 200 microscope (Carl Zeiss, Jena, Germany) equipped with a live cell imaging heating and CO2 chamber mounted to a CoolSnapHQ CCD camera (Roper Scientific, Martinsried, Germany). Confocal images were taken using a Zeiss LSM519 laser scanning confocal using 63× magnification Plan Apochromat objective. A detailed description is provided in Additional file 7.

Statistics and bioinformatics

Detailed information and description of statistical analysis on co-localisation studies, intensity translocation values, western blot quantification, used databases and artwork programmes is provided in Additional file 7.

We provide an inventory of supplemental information, supplemental experimental procedures, supplemental information and supplemental references (Additional file 7).