PDGFR dimer-specific activation, trafficking and downstream signaling dynamics

The platelet-derived growth factor receptors (PDGFRs) are a family of receptor tyrosine kinases (RTKs) known to direct multiple cellular processes during development, such as migration, proliferation and differentiation (Heldin and Westermark, 1999; Hoch and Soriano, 2003). In mammals, this family is composed of four dimeric growth factor ligands – PDGF-AA, PDGF-BB, PDGF-CC and PDGF-DD – which variously signal through two receptors, PDGFRα and PDGFRβ (Bostrom et al., 1996; Ding et al., 2004; Leveen et al., 1994; Soriano, 1994, 1997; Williams, 1989). Each receptor consists of an extracellular region harboring five immunoglobulin-like loops, a single transmembrane domain and an intracellular domain containing a split, catalytic tyrosine kinase (Williams, 1989). Ligand binding results in the dimerization of two PDGFRs to form PDGFRα homodimers, PDGFRα/β heterodimers or PDGFRβ homodimers (Fantauzzo and Soriano, 2016; Herren et al., 1993; Matsui et al., 1989; Seifert et al., 1989; Shim et al., 2010), followed by activation of the tyrosine kinase domains and trans-autophosphorylation of intracellular tyrosine residues (Herren et al., 1993; Kelly et al., 1991). Signaling molecules possessing Src homology 2 phosphotyrosine recognition motifs subsequently bind to phosphorylated residues in the intracellular domains of the receptors and initiate downstream signaling cascades (Heldin and Westermark, 1999; Williams, 1989). When expressed in the same cell type, signaling through the two receptors can result in different cell activities. For example, in the mouse cranial neural crest lineage, PDGFRα plays a predominant role in cranial neural crest cell migration, whereas PDGFRβ primarily contributes to proliferation of the cranial neural crest-derived facial mesenchyme (Mo et al., 2020). Given the structural similarities of PDGFRs and the fact that they interact with similar subsets of signaling molecules (Heldin and Westermark, 1999), a critical question is how this biological specificity is introduced to generate unique responses downstream of receptor engagement.

PDGFR internalization into the cell following ligand binding and dimerization also contributes to the regulation of PDGFR signaling (Bonifacino and Traub, 2003). Once internalized, PDGFRs can continue to bind adaptor proteins and/or signaling molecules within endosomes to activate downstream signaling pathways (Jastrzebski et al., 2017; Teis et al., 2002; Wang et al., 2004). Endosomal trafficking of the receptors commonly results in receptor degradation (Miyake et al., 1998, 1999; Mori et al., 1995). However, two studies have demonstrated that PDGFRβ homodimers and PDGFRα/β heterodimers can also be recycled to the cell membrane (Hellberg et al., 2009; Karlsson et al., 2006), indicating potential dimer-specific differences in PDGFR trafficking following activation. Despite these findings, the dimer-specific dynamics of PDGFR activation, trafficking and signal attenuation, as well as the roles of these processes in mediating signal transduction and cellular activity, have not been studied in detail.

Here, we implemented bimolecular fluorescence complementation (BiFC) (Hu and Kerppola, 2003; Magliery et al., 2005) to probe for PDGFR dimer-specific dynamics. The BiFC technique adapted for this study employs a split Venus fluorescent protein (Croucher et al., 2016) fused to the C-terminus of individual PDGFRs to investigate receptor dimerization. This method overcomes limitations with prior approaches to examine receptor expression and/or dimerization using antibody-based methods that cannot distinguish between PDGFRs present as monomers or engaged in homodimers versus heterodimers. Here, we generated two stable cell lines – one for each PDGFR homodimer pair – expressing BiFC-tagged PDGFRs. The formation of a functional Venus fluorescent protein upon PDGFR dimerization allows for the visualization and purification of individual PDGFR dimers. Utilizing this technique, we demonstrated significant differences in the timing and extent of PDGFRα versus PDGFRβ homodimer dimerization, activation and trafficking, which lead to changes in downstream signaling and cellular activity. Taken together, these data shed considerable light on the mechanisms by which biological specificity is introduced downstream of PDGFR activation.

To investigate PDGFR dimer-specific dynamics, we generated stable cell lines expressing BiFC-tagged PDGFRs. We cloned plasmids expressing C-terminal protein fusions of each PDGFR with BiFC fragments corresponding to the N-terminal (V1) or C-terminal (V2) regions of the Venus fluorescent protein (Fig. S1A). Using lentiviral transduction of the human-derived HCC15 cell line (Fig. S1B–C′), we stably integrated PDGFR-BiFC sequences [PDGFRα-V1 with PDGFRα-V2 (denoted PDGFRαV1/V2), and PDGFRβ-V1 with PDGFRβ-V2 (denoted PDGFRβV1/V2)] to generate two cell lines – a PDGFRα homodimer cell line and a PDGFRβ homodimer cell line, respectively. HCC15 cells were selected based on the lack of PDGFR expression, as the presence of endogenous PDGFRs would confound the interpretation of BiFC events. Furthermore, these cells have little to no PDGF ligand expression, such that the BiFC-tagged PDGFRs would only be activated upon exogenous ligand stimulation. Finally, HCC15 cells express all the relevant adaptor proteins and signaling molecules known to function downstream of PDGFR activation (Barretina et al., 2012; Cancer Cell Line Encyclopedia and Genomics of Drug Sensitivity in Cancer, 2015; Heldin and Westermark, 1999). We confirmed that the relevant sequences were inserted into the genome via PCR amplification and Sanger sequencing. Next, we examined expression of transcripts encoding each receptor using quantitative real-time PCR (qRT-PCR), revealing similar expression of PDGFRA (6.49±1.92% of B2M expression; mean±s.e.m.) and PDGFRB (14.8±2.26% of B2M expression) in their respective cell lines (Fig. S1D). Interestingly, this ratio of PDGFRA to PDGFRB expression is comparable to the ratio of endogenous receptor expression previously reported in mouse embryonic fibroblasts, in which all three PDGFR dimers are active (Fantauzzo and Soriano, 2016). Finally, we established that the relevant PDGFR protein was expressed in each cell line and migrated at the expected mass on SDS-PAGE gels (Fig. S1E,F). Within these cell lines, the individual N- and C-terminal fragments of the Venus fluorescent protein are expected to be non-fluorescent when bound to monomeric receptors. However, upon receptor dimerization, the N- and C-terminal fragments should colocalize, resulting in a functional Venus fluorescent protein. This BiFC event can be visualized through fluorescence microscopy or purified biochemically utilizing a GFP-Trap nanobody, which has an epitope that spans V1 and V2 (Croucher et al., 2016).

We initially starved the cells for 24 h in HITES medium, which lacks any growth factors, and performed fluorescence microscopy to determine baseline expression of Venus. For the PDGFRα homodimer cell line, the majority of Venus expression localized just outside of the nucleus (54%) (Fig. S1G,G′), with additional expression detected in the cytoplasm (21%), cell membrane (17%) and nucleus (8%). For the PDGFRβ homodimer cell line, Venus expression equally localized just outside of the nucleus (33%) (Fig. S1H,H′) and within the nucleus (33%) (Fig. S1I,I′), with additional expression in the cell membrane (19%) and cytoplasm (14%). To confirm the BiFC event and assess Venus expression upon exogenous ligand treatment, we starved the cells as above, photobleached the coverslips to ensure that all subsequent imaging captured newly formed BiFC events and stimulated with PDGF ligand for 5 min. We stimulated PDGFRα homodimer cells with PDGF-AA ligand and PDGFRβ homodimer cells with PDGF-BB ligand, as previous data indicated that PDGFRα homodimers are primarily responsive to activation by PDGF-AA, while PDGFRβ homodimers are primarily activated by PDGF-BB (Bostrom et al., 1996; Ding et al., 2004; Fantauzzo and Soriano, 2016; Leveen et al., 1994; Soriano, 1994, 1997). An important caveat to keep in mind, however, is that it is unclear whether the ligands used in this study have the same affinities for their respective BiFC-tagged receptors. Amongst images with an average number of 15.4 cells for the PDGFRα homodimer cell line and 19.5 cells for the PDGFRβ homodimer cell line, we detected 1.45 (9.40%) and 1.53 (7.83%) Venus-positive cells, respectively. At this timepoint, Venus expression in the PDGFRα homodimer cell line was typically observed as small, internalized puncta, whereas in the PDGFRβ homodimer cell line, Venus expression was at both the cell membrane and inside the cell (Fig. 1A–D′). Quantification of fluorescence intensity revealed a significant increase in Venus intensity upon 5 min of ligand treatment for the two cell lines (PDGFRαV1/V2, P=0.0232; PDGFRβV1/V2, P=0.0123; Fig. 1E,F), with comparable fluorescence intensity between the two cell lines [PDGFRαV1/V2, 21.3±2.79; PDGFRβV1/V2, 18.2±1.12; mean±s.e.m., arbitrary units (A.U.); Fig. 1E,F].

Next, we investigated colocalization of the Venus signal with signal from anti-PDGFR antibodies. These data revealed that Venus signal colocalized with a subset of PDGFR expression at 5 min of ligand stimulation (Fig. 1G–J″). The Pearson's correlation coefficient (PCC) for the Venus signal and PDGFR expression was similar between the PDGFRα homodimer (0.324±0.0547 PCC; mean±s.e.m.) and the PDGFRβ homodimer cell lines (0.358±0.0793 PCC; Fig. 1G,H). The regions of receptor expression that were not Venus-positive likely represent monomeric receptors and/or potential V1/V1 or V2/V2 dimers, which would not be expected to fluoresce (Fig. 1I–J″). The fact that we observed less than complete positive colocalization (PCC values <1) indicated that the presence of the BiFC fragments alone did not drive dimerization.

We further confirmed the BiFC event upon ligand treatment for both cell lines biochemically. Cells for both PDGFR homodimers were stimulated with PDGF ligand for 2, 5 and 15 min. Subsequent immunoprecipitation with the GFP-Trap nanobody was followed by western blotting for each receptor, normalizing the immunoprecipitation signal to the whole cell lysate signal. This revealed an increase in dimerized receptors following ligand treatment of both cell lines, with peak dimerization occurring at 5 min for both PDGFRα homodimers [1.25±0.0361 relative induction (R.I.); mean±s.e.m.] and PDGFRβ homodimers (1.49±0.361 R.I.; Fig. 2A–C). Furthermore, PDGFRβ homodimers appeared to dimerize more quickly than PDGFRα homodimers, reaching near peak dimerization at 2 min of ligand treatment (1.47±0.268 R.I.; Fig. 2C).

Next, we investigated the activation of both PDGFR homodimers by assessing receptor autophosphorylation. Each cell line was stimulated with PDGF ligand from 2 min to 4 h. Western blotting of whole-cell lysates using an anti-phospho-PDGFR antibody revealed that both PDGFR homodimers were significantly and robustly autophosphorylated upon ligand treatment. Peak activation occurred at 5 min of ligand stimulation for PDGFRα homodimers (11.2±2.44 R.I.) and at 15 min for PDGFRβ homodimers (28.2±3.89 R.I.; Fig. 2D,E). Critically, there was very little to no activation of the PDGFRs in the absence of ligand (Fig. 2D), consistent with previous findings showing that both ligand binding and dimerization are required for receptor activation (Kelly et al., 1991; Yang et al., 2010). Furthermore, PDGFRβ homodimers exhibited more robust autophosphorylation than PDGFRα homodimers from 5 min to 4 h of PDGF ligand stimulation with a significantly greater area under the curve (P=0.0003; Fig. 2D,E). Together, these results suggest that PDGFRβ receptors homodimerize more quickly and have increased levels of autophosphorylation than PDGFRα homodimers.

Interestingly, each of the blots for total PDGFR protein levels revealed a doublet with a lower band at 170 kDa (PDGFRα) or 180 kDa (PDGFRβ) and an upper band at 190 kDa (Fig. S1E,F; Fig. 2A,B,D). To determine whether the bands correspond to V1- and/or V2-tagged PDGFRs, we transiently transfected HEK 293T/17 cells with PDGFR-BiFC expression constructs. Each western blot lane had two bands (Fig. S2), likely representing the non-glycosylated and glycosylated versions of the PDGFRs (Shim et al., 2010). These bands ran slightly higher in the individual V1 lanes than the same bands in the individual V2 lanes (Fig. S2), which would be expected given the sizes of V1 (156 amino acids) and V2 (78 amino acids). However, the anti-phospho-PDGFR blots revealed a single, upper band of 190 kDa for both cell lines (Fig. 2D), indicating that only the glycosylated BiFC-tagged PDGFRs are phosphorylated upon PDGF ligand stimulation.

It is well established that PDGFRs dimerize at the cell membrane upon ligand binding (Herren et al., 1993; Shim et al., 2010) and become internalized and trafficked in the cell following activation (De Donatis et al., 2008; Disanza et al., 2009; Miaczynska, 2013; Pahara et al., 2010; Rogers and Fantauzzo, 2020). To further investigate the different subcellular localization of Venus in the two cell lines following PDGF ligand treatment, we evaluated the membrane localization of the PDGFR dimers at early timepoints of PDGF ligand treatment by measuring colocalization of the Venus signal with signal from an antibody recognizing the membrane marker Na⁺/K⁺-ATPase (Skou, 1989). We observed relatively similar levels of colocalization for PDGFRα homodimers and PDGFRβ homodimers at 1 and 2 min following PDGF ligand stimulation (Fig. 3A). At 5 min of ligand treatment, however, PDGFRα homodimer colocalization with Na⁺/K⁺-ATPase had decreased from a peak at 1 min, whereas PDGFRβ homodimers reached peak colocalization with the membrane marker (Fig. 3A–C″). To assess whether PDGFR dimer internalization could account for this comparative decrease in membrane localization for PDGFRα homodimers, we next measured the colocalization of the Venus signal with signal from an antibody recognizing the early endosome marker Rab5 (all isoforms) (Zerial and McBride, 2001). These analyses revealed that PDGFRα homodimers colocalized significantly more with Rab5 at 2 min (0.250±0.0316 PCC; mean±s.e.m.) and 5 min (0.215±0.0310 PCC) of PDGF ligand treatment than PDGFRβ homodimers (0.123±0.0192 PCC; P=0.0359; 0.100±0.0198 PCC; P=0.0440, respectively), indicating a greater likelihood for PDGFRα homodimers to be trafficked at these timepoints (Fig. 3D–F″). Collectively, these data suggest that PDGFRα homodimers are trafficked more quickly, whereas PDGFRβ homodimers dwell at the membrane longer.

We next assessed trafficking of the PDGFR homodimers through late endosomes, which precedes trafficking to the lysosome for degradation (Mellman, 1996), by examining colocalization of the Venus signal with signal from an antibody recognizing the late endosome marker Rab7 (all isoforms) (Zerial and McBride, 2001). These data revealed that PDGFRα homodimers colocalized significantly more with Rab7 (0.280±0.0111 PCC) than PDGFRβ homodimers (0.237±0.00404 PCC; P=0.0474) at 1 h of ligand treatment (Fig. 4A–C″). This finding suggests that PDGFRα homodimers might be degraded by lysosomes more quickly and/or to a greater extent than PDGFRβ homodimers. To investigate this further, we biochemically evaluated receptor degradation following PDGF ligand stimulation. We performed a 30 min pretreatment with cycloheximide to inhibit translation and then stimulated the cells with PDGF ligand from 2 min to 4 h. We collected cell lysates and performed western blotting for PDGFR levels in each condition. Following 4 h of ligand treatment, we observed 80% and 78% decreases from initial receptor levels for PDGFRα homodimers (P=0.0316) and PDGFRβ homodimers, respectively, indicating a similar extent of degradation by the end of the time course (Fig. 4D,E). Importantly, we observed that PDGFRα homodimers were degraded at a quicker rate, particularly between 1 and 2 h (Fig. 4E), consistent with the Rab7 colocalization findings (Fig. 4A–C″). Taken together, these data suggest that PDGFRα homodimers are degraded more quickly than PDGFRβ homodimers.

To test whether the above receptor degradation is due to trafficking to the lysosome, we performed a 1 h pretreatment with the lysosome inhibitor chloroquine (Wiesmann et al., 1975) followed by a 30 min pretreatment with cycloheximide. Cells were stimulated with PDGF ligand and receptor levels were assessed via western blotting as described above. Although the PDGFRα homodimer cell line showed no significant difference in receptor levels in the absence versus presence of chloroquine, there was a trend of increased receptor levels upon lysosome inhibition at 1, 2 and 4 h. In particular, the 1 h timepoint revealed a 19% increase in PDGFRα receptor levels in the presence of chloroquine (Fig. 4F,G). The PDGFRβ homodimer cell line exhibited a noticeable (42%) increase in PDGFRβ receptor levels at 1 h in the presence of chloroquine, as well as a significant difference in receptor levels in the absence [0.224±0.0812 relative receptor levels (R.R.L.)] versus presence of chloroquine (0.534±0.0734 R.R.L.; P=0.0479) at 4 h (Fig. 4F,G). Collectively, these data suggest that trafficking to the lysosome contributes to PDGFR degradation following ∼1 h of ligand treatment. Furthermore, although PDGFRα was degraded to a similar extent in the absence versus presence of chloroquine, indicating that it might primarily undergo proteasomal degradation in response to sustained ligand treatment, the second wave of PDGFRβ degradation appeared to be almost exclusively lysosomal.

In contrast to being degraded, PDGFRs can alternatively be recycled to the membrane for continued signaling. To examine receptor recycling, we first quantified the colocalization of the Venus signal with signal from an antibody recognizing the rapid recycling endosome marker Rab4 (all isoforms) (Zerial and McBride, 2001). Quantification indicated that PDGFRβ homodimers colocalized significantly more with Rab4 (0.224±0.0131 PCC) than PDGFRα homodimers (0.167±0.00765 PCC; P=0.0286) at 15 min of ligand treatment (Fig. 5A–C″). Next, we examined colocalization of the Venus signal with signal from an antibody recognizing the slow recycling endosome marker Rab11 (all isoforms) (Zerial and McBride, 2001). This analysis revealed that PDGFRβ homodimers colocalized significantly more with Rab11 (0.203±0.0122 PCC) than PDGFRα homodimers (0.124±0.0116 PCC; P=0.0094) at 90 min of PDGF ligand treatment (Fig. 5D–F″). Consistent with this finding, at 90 min of PDGF ligand stimulation, PDGFRβ homodimers colocalized significantly more with the membrane marker Na⁺/K⁺-ATPase (0.150±0.00680 PCC) than PDGFRα homodimers (0.101±0.00722 PCC; P=0.0077; Fig. S3A–C″). Taken together, these data demonstrate that PDGFRβ homodimers are more likely to be recycled back to the cell membrane through these rapid and slow recycling routes than PDGFRα homodimers.

We postulated that the differences in dimerization, activation and trafficking between PDGFR homodimers might affect downstream cell signaling. We performed a time course of ligand stimulation for each PDGFR homodimer cell line from 2 min to 4 h followed by western blotting for two well-studied effector molecules downstream of PDGFR activation – ERK1 and ERK2 (ERK1/2, also known as MAPK3 and MAPK1, respectively) and AKT (all isoforms) (Fantauzzo and Soriano, 2016; Lemmon and Schlessinger, 2010; Vasudevan et al., 2015). Phosphorylation of ERK1/2 in response to PDGFRα homodimer signaling peaked at 15 min following PDGF ligand stimulation (1.50±0.236 R.I.; mean±s.e.m.) and remained relatively stable throughout the time course (1.33±0.0693 R.I. at 4 h; Fig. 6A,B). Phosphorylation of ERK1/2 in response to PDGFRβ homodimer signaling displayed a similar curve to that of PDGFRα homodimers but had an initial peak later, at 30 min of PDGF stimulation (2.32±0.581 R.I.), with a higher amplitude than the PDGFRα homodimers (Fig. 6A,B). Phosphorylation of AKT in response to PDGFRα homodimer activation had an initial peak at 15 min of ligand treatment (2.04±0.284 R.I.; P=0.0382) and, comparable to the phosphorylation dynamics of ERK1/2, remained relatively stable (2.30±0.556 R.I. at 4 h; Fig. 6C,D). Phosphorylation of AKT in response to PDGFRβ homodimer activation also peaked at 15 min of ligand stimulation (5.85±1.25 R.I.; P=0.0165) but had a greater amplitude than PDGFRα homodimers and appeared more dynamic, sharply declining after the peak at 15 min (Fig. 6C,D). These signaling data suggest that PDGFR homodimers differ in downstream signaling dynamics and that PDGFRβ homodimers generate a greater signaling response to PDGF ligand stimulation than do PDGFRα homodimers.

An important question is whether these differences in signaling downstream of the PDGFR homodimers were due to innate differences in the clones that we chose for each of these cell lines, rather than differences in signaling activity downstream of PDGFRα versus PDGFRβ homodimers. We tested this question by measuring activation of an alternative RTK, EGFR. We stimulated each cell line with EGF for 10 min and performed western blotting for the phosphorylation of the same effector molecules, revealing no difference in phosphorylation of ERK1/2 and AKT between the PDGFR homodimer cell lines (Fig. S4A–D). These results indicate that the differences that we observed upon PDGF ligand stimulation of our two PDGFR homodimer cell lines reflect true differences in signaling response downstream of these receptors.

To test the requirement for PDGFR endocytosis in downstream cell signaling, we performed a 1 h pretreatment with Dyngo-4a, which inhibits the large GTPase dynamin from pinching off clathrin-coated vesicles from the cell membrane (McCluskey et al., 2013; Sadowski et al., 2013). In this case, PDGFRα homodimer colocalization with Rab5 at 2 min of PDGF ligand treatment (0.121±0.0175 PCC; mean±s.e.m.; Fig. 7A–B″) was significantly reduced from its previous value (P=0.0350) to a level similar to that of PDGFRβ homodimers (0.137±0.0135 PCC; Fig. 7A,C–C″). Furthermore, treatment with Dyngo-4a resulted in distinct changes in signaling downstream of each PDGFR homodimer. PDGFRα homodimers exhibited delayed phosphorylation of ERK1/2 in response to PDGF ligand following dynamin inhibition, with relative induction remaining below baseline levels until 1 h or more of ligand stimulation and peaking at 4 h (1.37±0.518 R.I.; Fig. 7D,E). By contrast, the early phosphorylation dynamics of ERK1/2 in response to PDGFRβ homodimer activation were not affected by dynamin inhibition, peaking at 30 min of ligand stimulation (2.47±0.442 R.I.; Fig. 7D,E) as in non-Dyngo-4a experiments. Similar to the phosphorylation of ERK1/2, phosphorylation of AKT in response to PDGFRα homodimer activation exhibited significantly reduced induction at 15 min (P=0.0311) and 30 min (P=0.0194) of PDGF ligand stimulation following dynamin inhibition compared to non-Dyngo-4a experiments, with relative induction remaining below baseline levels until 1 h or more of ligand stimulation and peaking at 4 h (1.60±0.0529 R.I.; Fig. 7F,G). Comparably, phosphorylation of AKT in response to PDGFRβ homodimer activation also peaked at a later timepoint of 1 h of ligand stimulation (4.22±1.10 R.I.). Taken together, these data demonstrate that inhibition of clathrin-mediated endocytosis results in alterations in cellular trafficking and downstream signaling dynamics for PDGFR homodimers, with a greater impact on PDGFRα than PDGFRβ.

As with many RTKs, signaling through the PDGFRs has been shown to direct a range of cellular processes (Fantauzzo and Soriano, 2015; Schlessinger, 2000). To determine how the above trafficking and signaling dynamics of the PDGFR homodimers affect cellular activity, we first measured proliferation via a soft agar anchorage-independent growth assay in growth medium with 10% fetal bovine serum (FBS). In the absence of ligand, the cells expressing PDGFRβ proliferated significantly more than those expressing PDGFRα, as measured by colony area (7.80×10⁵±4.72×10⁴ versus 3.32×10⁵±4.81×10⁴ A.U.; P=0.0027; mean±s.e.m.; Fig. 8A,B,D). Upon exogenous PDGF ligand stimulation, proliferation, as measured by colony count increased for the PDGFRα homodimer cell line (8.60×10³±8.77×10² versus 7.43×10³±8.59×10² colonies; P=0.0074; Fig. 8A,A′,C) and proliferation as measured by colony area increased for the PDGFRβ homodimer cell line (8.92×10⁵±4.44×10⁴ versus 7.80×10⁵±4.72×10⁴ A.U.; P=0.0325; Fig. 8B,B′,D). Finally, the cells expressing PDGFRβ proliferated significantly more than those expressing PDGFRα upon exogenous ligand treatment, as measured by colony area (8.92×10⁵±4.44×10⁴ versus 3.81×10⁵±6.63×10⁴ A.U.; P=0.0048; Fig. 8A′,B′,D). These effects were confirmed via immunofluorescence analysis using the proliferation marker Ki67 (Gerdes et al., 1984) (Fig. S5A–B′). Here, both cell lines exhibited trends for increased proliferation upon exogenous PDGF ligand stimulation, with cells expressing PDGFRβ proliferating more than those expressing PDGFRα in the absence and presence of ligand (Fig. S5C). We next performed 24 h transwell migration assays in growth medium with 10% FBS to assess cell migration for the two PDGFR homodimer cell lines. Only the PDGFRβ homodimer cell line exhibited increased migration towards growth medium containing PDGF ligand (3.14×10⁸±3.22×10⁷ versus 1.24×10⁸±1.59×10⁷ A.U.; P=0.0022; Fig. 8F,F′,G). Furthermore, addition of exogenous ligand resulted in increased migration for cells expressing PDGFRβ over those expressing PDGFRα (3.14×10⁸±3.22×10⁷ versus 2.07×10⁸±2.14×10⁷ A.U.; P=0.0218; Fig. 8E′,F′,G).

To confirm these results in a physiologically relevant setting in which both PDGFRα and PDGFRβ are endogenously expressed, we cultured primary mouse embryonic palatal mesenchyme (MEPM) cells derived from the palatal shelves of embryonic day 13.5 Pdgfra^fl/fl;Wnt1-Cre^+/Tg or Pdgfrb^fl/fl;Wnt1-Cre^+/Tg embryos (Fantauzzo and Soriano, 2016). We reasoned that PDGF-AA ligand treatment of Pdgfrb^fl/fl;Wnt1-Cre^+/Tg cells could only activate PDGFRα homodimers, while PDGF-BB ligand treatment of Pdgfra^fl/fl;Wnt1-Cre^+/Tg cells could only activate PDGFRβ homodimers. Confirming the results from the HCC15-based stable cell lines, upon exogenous PDGF ligand stimulation, the cells expressing PDGFRβ proliferated significantly more than those expressing PDGFRα in both starvation medium containing 0.1% FBS (0.200±0.0148 versus 0.0778±0.0115; P=0.0006; Fig. S6A) and growth medium containing 10% FBS (0.317±0.0318 versus 0.137±0.0326; P=0.0056; Fig. S6A). Similarly, the cells expressing PDGFRβ migrated significantly more than those expressing PDGFRα in the presence of exogenous PDGF ligand (4.89×10⁴±1.59×10³ vs 2.64×10⁴±4.02×10³; P=0.0193; Fig. S6B–D). Collectively, these cellular activity assays provide evidence that activated PDGFRβ homodimers generate stronger pro-proliferation and pro-migration signals than PDGFRα homodimers.

It has previously been impossible to investigate PDGFR dimer-specific dynamics. Former studies utilized antibody- and ligand propensity-based approaches to study PDGFR expression and/or activity, yet both approaches have considerable caveats. Antibodies cannot distinguish whether a receptor is present as a monomer or engaged in a homodimeric or heterodimeric complex. Proximity ligation techniques are also insufficient, as they detect protein–protein interactions within a relatively large range of 40 nm (Bagchi et al., 2015), while dimerized PDGFRs might have considerably closer interactions (Chen et al., 2015). Moreover, evidence suggests that some PDGF ligands are promiscuous and can result in the formation of multiple PDGFR dimers (Fantauzzo and Soriano, 2016). Therefore, in studies where conclusions were drawn about a particular PDGFR dimer based on ligand treatment alone, it is often unclear which dimer was assayed. Our aim was to implement an optimized method, BiFC, to both visualize and purify individual PDGFR dimers to more precisely investigate homodimer-specific subcellular distribution and signaling dynamics. We observed less than complete colocalization between Venus and PDGFR antibody signals, demonstrating that PDGFR expression is not predictive of receptor dimerization and activation, and further indicating that the presence of the BiFC fragments alone does not drive dimerization in our cell lines.

Interestingly, we took multiple approaches to generate a stable PDGFRαV1/βV2 cell line, with limited success. A single clone expressing both receptors was obtained among 188 clones screened. However, despite two rounds of subcloning, the frequency of BiFC events remained surprisingly low. Related to this finding, qRT-PCR analyses revealed that expression of PDGFRA and PDGFRB was ∼4-fold and ∼13-fold lower, respectively, in this clone compared to our PDGFRα and PDGFRβ homodimer cell lines. Moreover, the cells grew very slowly compared to parental HCC15 cells and both PDGFRα and PDGFRβ homodimer cell lines. Taken together, we conclude that HCC15 cells do not tolerate combined expression of PDGFRα and PDGFRβ, and/or formation of PDGFRα/β heterodimers.

Our initial characterization of Venus expression in the absence of photobleaching and PDGF ligand treatment for the PDGFR homodimer cell lines indicated some baseline BiFC expression. This Venus expression is likely due to low levels of PDGF ligand expressed by HCC15 cells and/or PDGFR homodimer formation in the absence of ligand stimulation. Similarly, our GFP-Trap purification of the PDGFR homodimers revealed increased dimerization for both upon PDGF ligand stimulation, as well as, surprisingly, the presence of dimerized receptors in the absence of ligand stimulation. As recently discussed, an in silico study has implied that there might be an inactive dimerization state for the PDGFRs (Polyansky et al., 2019; Rogers and Fantauzzo, 2020). Interestingly, other RTKs, most notably EGFR, have been shown to exist in an inactive dimerized state in the membrane prior to a ligand-induced conformational change (Bae and Schlessinger, 2010; Zhang et al., 2006). Importantly, even though we detected dimerized PDGFRs in the absence of ligand stimulation, we revealed that PDGFR homodimers were only activated upon exogenous ligand treatment.

In studying these receptor dimerization and activation dynamics, we observed that PDGFRβ receptors dimerized more quickly than PDGFRα receptors. On the other hand, PDGFRα homodimers reached their peak autophosphorylation faster than PDGFRβ homodimers, though the PDGFRα peak was significantly lower than that for PDGFRβ. It is possible that these trends in dimerization and activation might be due to different conformations of the receptors prior to and during activation. There is precedent for this concept in the RTK field, as the extracellular regions of ErbB family members EGFR, ErbB3 and ErbB4 possess an intramolecular tether in the absence of ligand binding that buries the dimerization arm of the cysteine-rich 1 subdomain (Bouyain et al., 2005; Cho and Leahy, 2002; Ferguson et al., 2003), whereas ErbB2 exists in an extended conformation that exposes both cysteine-rich subdomains (Cho et al., 2003; Garrett et al., 2003). A similar explanation for the differences in autophosphorylation activity between the PDGFR homodimers may be the auto-inhibitory function of a glutamic acid/proline repeat motif in the C-terminal tail of PDGFRβ (Chiara et al., 2004), which is alleviated upon ligand-induced dimerization and conformational change of the receptor. As this motif has not been detected in PDGFRα, it could create a relative delay in activation for PDGFRβ compared to PDGFRα homodimers.

Our PDGFR homodimer dimerization and activation findings are noteworthy in the context of the Na⁺/K⁺-ATPase and Rab5 colocalization data, from which we determined that PDGFRα homodimers are trafficked from the cell membrane more quickly than PDGFRβ homodimers. Additionally, as discussed above, we demonstrated that PDGFRβ homodimers are activated to a greater extent than PDGFRα homodimers. It is possible that the extended time at the cell membrane for the PDGFRβ homodimers allows for increased autophosphorylation. Conversely, earlier peak activation of the PDGFRα homodimers might result in faster recruitment of trafficking machinery and, thus, quicker intracellular trafficking. It is also possible that dimer conformation might play a role in these internalization and trafficking dynamics, though the complete conformations of the PDGFR homodimers remain unknown. Molecules that post-translationally modify both PDGFRα and PDGFRβ to recruit endosomal machinery, such as the RING-finger E3 ubiquitin ligase Cbl, likely finetune these dynamics (Reddi et al., 2007; Rorsman et al., 2016; Schmid et al., 2018). Furthermore, PDGFRβ has a known internalization motif located in the C-terminal domain of the receptor (Mori et al., 1991; Pahara et al., 2010), whereas such a motif has yet to be uncovered for PDGFRα, potentially imparting unique internalization and trafficking mechanisms for the various PDGFR dimers.

In addition to differences in the timing of initial PDGFR trafficking into early endosomes, our data also revealed varied degradation dynamics and recycling between the PDGFR dimers. We found that PDGFRα homodimers colocalized more with Rab7 following 1 h of PDGF ligand stimulation and were degraded faster than PDGFRβ homodimers shortly afterwards. However, both PDGFR homodimers experienced similar levels of receptor degradation following 4 h of ligand treatment. We further demonstrated that lysosomal inhibition partially alleviated both PDGFRα and PDGFRβ degradation following 1 h of PDGF ligand treatment, indicating an early wave of lysosomal degradation for both homodimers. Notably, although lysosomal inhibition had little effect on PDGFRα levels after 4 h of ligand treatment, it significantly increased PDGFRβ levels, suggesting later waves of proteasomal and lysosomal degradation for PDGFRα and PDGFRβ, respectively. These findings align with prior results demonstrating the relevance of both degradation pathways for PDGFRs (Li et al., 2017; Mori et al., 1995; Reddi et al., 2007; Rorsman et al., 2016). We additionally found that PDGFRβ homodimers were more likely to be recycled than PDGFRα homodimers. However, the fact that only 20–22% of initial receptor levels for both cell lines were detected following 4 h of PDGF ligand treatment indicates that receptor degradation is the predominant result of intracellular trafficking, consistent with previous findings for PDGFRβ (Hellberg et al., 2009; Sadowski et al., 2013).

We next sought to determine how these dimerization and activation dynamics, together with PDGFR dimer-specific subcellular distribution, affected downstream signaling. We found that both PDGFR homodimers induced phospho-ERK1/2 peaks that remained relatively stable over time. PDGFRα homodimers drove an early peak of ERK1/2 phosphorylation, whereas the peak driven by PDGFRβ homodimer activation occurred slightly later and had a higher amplitude. Interestingly, it has previously been shown that various subcellular localizations of Erk1/2, or its upstream activator RAS, result in different timings of the Erk1/2 signaling response (Bruggemann et al., 2021; Herrero et al., 2016; Keyes et al., 2020). For example, following EGF ligand stimulation, the phosphorylation of Erk1/2 at the membrane results in sustained signaling, whereas its phosphorylation in the cytoplasm and nucleus results in transient signaling, ultimately leading to differences in cellular activity (Keyes et al., 2020). Given that PDGFRs have been shown to signal within various cellular compartments, including endosomes (Wang et al., 2004), engagement of signaling molecules at different subcellular locations might explain, at least in part, the observed differences in ERK1/2 signaling dynamics downstream of PDGFRα versus PDGFRβ homodimer activation. Correspondingly, the trends of AKT phosphorylation were also different between the two PDGFR homodimers. Although initial phospho-AKT peaks occurred at a similar timepoint for both PDGFR homodimers, the PDGFRα homodimer-driven peak remained relatively stable, whereas PDGFRβ homodimer-stimulated phosphorylation of AKT was much more transient. In line with the phospho-ERK1/2 results, PDGFRβ homodimer activation induced a greater level of phospho-AKT than PDGFRα homodimers, suggesting that signaling through PDGFRβ homodimers results in greater activation of downstream signaling cascades.

Interestingly, two previous studies have suggested that activation of PDGFRα/β heterodimers leads to a greater MAPK downstream signaling response than that of either PDGFR homodimer. In the first, a porcine aortic endothelial cell line stably expressing exogenous PDGFRα and PDGFRβ was treated with PDGF-AB and PDGF-BB ligand. Stimulation with PDGF-AB ligand, which was proposed to generate PDGFRα/β heterodimers, led to reduced phosphorylation at Y771 in the cytoplasmic domain of PDGFRβ in comparison to PDGF-BB ligand treatment, which was proposed to more efficiently activate PDGFRβ homodimers. This reduced phosphorylation led to decreased association of PDGFRβ with a negative MAPK regulator, RasGAP, and increased downstream Ras and ERK2 activation (Ekman et al., 1999). In a second study comparing PDGF-AA, PDGF-BB and PDGF-DD ligand treatment of mouse embryonic fibroblasts, PDGF-BB ligand led to the greatest physical association of PDGFRα and PDGFRβ receptors and induced the highest peak of phospho-Erk1/2 signal (Fantauzzo and Soriano, 2016). However, in each case, it cannot be ruled out that multiple PDGFR dimers formed in response to ligand treatment and contributed to the downstream signaling response. Thus, the signaling dynamics downstream of PDGFRα/β heterodimer activation, and how these dynamics compare to those detected upon PDGFR homodimer activation, remain to be determined.

Importantly, inhibition of clathrin-mediated endocytosis indicated a particular requirement for rapid internalization and trafficking of PDGFRα homodimers in the propagation of downstream signaling. The observed difference in AKT phosphorylation following PDGFRα homodimer activation between untreated versus Dyngo-4a-treated conditions is particularly noteworthy, as PDGFRα homodimers have been shown to predominantly signal through the phosphoinositide 3-kinase (PI3K)/AKT pathway during development (Klinghoffer et al., 2002). These findings support a model in which differences in the timing and extent of PDGFR homodimer internalization and trafficking finetune downstream signaling responses. An important consideration, however, is that we evaluated whole-cell lysates reflecting population-level dynamics, and it is likely that these dynamics differ between individual cells (Sparta et al., 2015).

Finally, cell activity assays revealed that PDGF ligand stimulation of PDGFRβ homodimers resulted in increased levels of both proliferation and migration from what was observed upon PDGFRα homodimer activation. The duration of Erk1/2 signaling downstream of RTK activation has been shown to determine cellular responses in various settings, with transient signaling resulting in proliferation and sustained signaling leading to differentiation or migration (Bruggemann et al., 2021; Freed et al., 2017; Marshall, 1995). Given the relatively sustained ERK1/2 signaling downstream of PDGFRβ homodimer activation, it is interesting that both proliferation and migration were stimulated. Whether the differences in cellular responses that we observed downstream of PDGFRα versus PDGFRβ homodimers stem from varying responses to a relatively low PDGF ligand concentration and/or inherent differences in the enzymatic activity of the homodimers remains to be determined. In support of the former hypothesis, varying concentrations of PDGF ligand have been shown to result in differences in both PDGFR trafficking and downstream proliferation and migration (De Donatis et al., 2008). Future signaling and cell activity studies will incorporate a range of PDGF ligand concentrations. Alternatively, activation of PDGFRβ homodimers might generate stronger kinase activity than that of PDGFRα homodimers. In fact, analysis of chimeric PDGFRs in vivo revealed that the intracellular domain of PDGFRβ can compensate for the loss of PDGFRα signaling, while the inverse is not true (Klinghoffer et al., 2001). Both hypotheses considered, it remains that PDGFRβ homodimers stimulated increased cell activity upon exogenous PDGF ligand stimulation in the context of both HCC15 cells and primary MEPM cells. However, we cannot rule out that the higher expression of PDGFRB compared to that of PDGFRA in the PDGFR-BiFC-HCC15 cell lines contributed, at least in part, to the differences in signaling and cell activity dynamics that we observed between the two homodimer cell lines.

Overall, we have identified PDGFR homodimer-specific differences in dimerization, activation, trafficking, downstream signaling and cellular activity by utilizing a novel BiFC system. Our findings thus provide significant insight into how similar receptors within an RTK family utilize a combination of mechanisms to differentially propagate downstream signaling. This approach will be invaluable in future studies characterizing PDGFR dimer-specific interactomes to further delineate the means by which these receptors generate distinct cellular outputs.

Using Gateway cloning, we cloned pDONOR223-PDGFRA (#23892; Addgene, Watertown, MA, USA) and pDONOR223-PDGFRB (#23893; Addgene) each into pDEST-ORF-V1 (#73637; Addgene) and, separately, into pDEST-ORF-V2 (#73638; Addgene). PDGFRα-V1, PDGFRα-V2, PDGFRβ-V1 and PDGFRβ-V2 sequences were amplified by PCR using the following primers (Integrated DNA Technologies, Inc., Coralville, IA, USA): 5′-TTACTCGAGATGGGGACTTCCCATCCGGC-3′ and 5′-TTACCCGGGTTACTGCTTGTCGGCGGTGA-3′, 5′-TTACTCGAGATGGGGACTTCCCATCCGGC-3′ and 5′-TTACCCGGGTTACTTGTACAGCTCGTCCA-3′, 5′-TTAGAATTCATGCGGCTTCCGGGTGCGAT-3′ and 5′-CTGTCTAGATTACTGCTTGTCGGCGGTGA-3′, 5′-TTAGAATTCATGCGGCTTCCGGGTGCGAT-3′ and 5′-CGGTCTAGATTACTTGTACAGCTCGTCCA-3′, respectively. The sequences were cloned into the pLVX-Puro vector using XhoI and XmaI sites for PDGFRα-V1 and PDGFRα-V2 and EcoRI and XbaI sites for PDGFRβ-V1 and PDGFRβ-V2. The above lentiviral constructs (10 mg) and packaging vectors pCMV-VSV-G (Stewart et al., 2003) and pCMV-dR8.91 (Zufferey et al., 1997) (5 mg each) were transfected into HEK 293T/17 cells (American Type Culture Collection) using Lipofectamine LTX (Thermo Fisher Scientific, Waltham, MA, USA). Medium containing lentivirus was collected 48 h and 72 h following transfection and filtered using a 13 mm syringe filter with a 0.45 µm PVDF membrane (Thermo Fisher Scientific) following addition of 4 mg/ml polybrene (Sigma-Aldrich, St Louis, MO, USA). Lentiviral-containing medium was added to HCC15 lung cancer cells (see below) every 24 h for 2 days, and cells were subsequently grown in the presence of 2 µg/ml puromycin for 10 days. Individual Venus-positive cells were isolated on a Moflo XDP 100 cell sorter (Beckman Coulter Inc., Brea, CA, USA) following 5 min of stimulation with 10 ng/ml PDGF-AA or PDGF-BB (R&D Systems, Minneapolis, MN, USA) for the PDGFRα homodimer and PDGFRβ homodimer cell lines, respectively, and expanded to generate clonal cell lines. Final clones chosen for PDGFRα and PDGFRβ homodimer cell lines were confirmed by PCR amplification of the inserted sequences from genomic DNA using primers found in Table S1. These PCR products were subcloned into the Zero Blunt TOPO vector (Thermo Fisher Scientific) and Sanger sequenced.

The HCC15 lung cancer cell line was obtained from the laboratory of Dr Lynn Heasley (University of Colorado Anschutz Medical Campus). The cell line was authenticated through short tandem repeat analysis and tested for mycoplasma contamination every 10 passages using the MycoAlert Mycoplasma Detection Kit (Lonza Group Ltd, Basel, Switzerland). PDGFR-BiFC stable cells were cultured in RPMI growth medium [RPMI (Gibco, Thermo Fisher Scientific) supplemented with 100 U/ml penicillin (Gibco), 100 µg/ml streptomycin (Gibco) containing 10% FBS (Hyclone Laboratories Inc., Logan, UT, USA)] at 37°C in 5% CO₂. Once the stable cell lines were established, they were split at a ratio of 1:5 for maintenance. PDGFRα homodimer cells were used for experiments at passages 6–19, and PDGFRβ homodimer cells were used for experiments at passages 10–23. When serum starved, cells were grown in HITES medium [DMEM/F12 (Corning, Corning, NY, USA) supplemented with 0.1% bovine serum albumin (Fisher Scientific, Thermo Fisher Scientific), 10 mM β-estradiol (Sigma-Aldrich), 10 mM hydrocortisone (Sigma-Aldrich), 5 µg/ml insulin (Sigma-Aldrich), 100 U/ml penicillin (Gibco), 100 µg/ml streptomycin (Gibco), 1.2 mg/ml NaHCO₃ (Santa Cruz Biotechnology, Inc., Dallas, TX, USA), 30 nM Na₃SeO₃ (Sigma-Aldrich) and 10 µg/ml apo-transferrin (Sigma-Aldrich)].

Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Germantown, MD, USA) according to the manufacturer's instructions. First-strand cDNA was synthesized using a ratio of 2:1 random primers:oligo (dT) primer and SuperScript II RT (Invitrogen, Thermo Fisher Scientific) according to the manufacturer's instructions. qRT-PCR was performed on a CFX Connect Real-Time PCR Detection System and analyzed with CFX Manager software (version 3.1; Bio-Rad Laboratories, Inc., Hercules, CA, USA). All reactions were performed with SYBR Select Master Mix (Applied Biosystems, Thermo Fisher Scientific), 300 nM primers (Integrated DNA Technologies, Inc.) and cDNA in a 20 µl reaction volume. PCR primers for qRT-PCR analyses can be found in Table S2. The following PCR protocol was used: step 1, 2 min at 50°C; step 2, 2 min at 95°C; step 3, 15 s at 95°C; step 4, 1 min at 60°C; repeat steps 3 and 4 for 39 cycles; step 5 (melting curve), 5 s per 0.5°C increment from 65°C to 95°C. All samples were run in triplicate and normalized against an endogenous internal control, B2M. qRT-PCR experiments were performed across three independent experiments, each using a separate passage of cells.

PDGFR-BiFC-HCC15 cells were cultured as described above. To induce PDGFRα homodimer or PDGFRβ homodimer signaling, PDGFRα homodimer cells and PDGFRβ homodimer cells at ∼60–70% confluence were serum starved for 24 h in HITES medium and stimulated with 10 ng/ml PDGF-AA or PDGF-BB ligand (R&D Systems), respectively, for the indicated length of time. To induce EGFR signaling, cells were similarly stimulated with 10 ng/ml EGF (PeproTech, Thermo Fisher Scientific). When applicable, cells were pretreated with 10 µg/ml cycloheximide (Sigma-Aldrich) in DMSO (or an equivalent volume of DMSO) for 30 min before PDGF ligand stimulation and/or 25 μM chloroquine (Sigma-Aldrich) in water (or an equivalent volume of water) for 1 h before PDGF ligand stimulation; or 30 µM Dyngo-4a (Abcam, Waltham, MA, USA) in DMSO (or an equivalent volume of DMSO) for 1 h before PDGF ligand stimulation. Protein lysates for immunoprecipitation were generated by resuspending cells in ice-cold GFP-Trap lysis buffer [20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM EDTA, 1× complete Mini protease inhibitor cocktail (Roche, MilliporeSigma, Burlington, MA, USA), 1 mM PMSF] and collecting cleared lysates by centrifugation at 13,400 g at 4°C for 20 min. For immunoprecipitations, cell lysates (500 µg) were incubated with GFP-Trap agarose beads (Bulldog Bio, Inc., Portsmouth, NH, USA) for 1 h at 4°C. Beads were washed three times with ice-cold GFP-Trap wash/dilution buffer [10 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5 M EDTA] and the precipitated proteins were eluted with Laemmli buffer containing 10% β-mercaptoethanol, heated for 10 min at 100°C and separated by SDS-PAGE. For western blotting analysis of whole-cell lysates, protein lysates were generated by resuspending cells in ice-cold NP-40 lysis buffer [20 mM Tris-HCl pH 8, 150 mM NaCl, 10% glycerol, 1% Nonidet P-40, 2 mM EDTA, 1× complete Mini protease inhibitor cocktail (Roche), 1 mM PMSF, 10 mM NaF, 1 mM Na₃VO₄, 25 mM β-glycerophosphate] and collecting cleared lysates by centrifugation at 13,400 g at 4°C for 20 min. Laemmli buffer containing 10% β-mercaptoethanol was added to the lysates, which were heated for 5 min at 100°C. Proteins were subsequently separated by SDS-PAGE. Western blot analysis was performed according to standard protocols using horseradish peroxidase-conjugated secondary antibodies. Full, uncropped western blots are provided in Fig. S7. The following antibodies were used for western blotting: PDGFRα (1:200; C-20; sc338; Santa Cruz Biotechnology, Inc.) (validation, dimerization and receptor degradation experiments); PDGFRβ (1:200; 958; sc432; Santa Cruz Biotechnology, Inc.) (validation, dimerization and receptor degradation experiments); β-tubulin (1:1000; E7; E7; Developmental Studies Hybridoma Bank, Iowa City, IA, USA); phospho-PDGFRα (Tyr 849)/PDGFRβ (Tyr857) (1:1000; C43E9; 3170; Cell Signaling Technology, Inc., Danvers, MA, USA); PDGFRα (1:1000; D13C6; 5241; Cell Signaling Technology, Inc.) (autophosphorylation and chloroquine experiments); PDGFRβ (1:1000; 28E1; 3169; Cell Signaling Technology, Inc.) (autophosphorylation and chloroquine experiments); phospho-Erk1/2 (1:1000; Thr202/Thr204; 9101; Cell Signaling Technology, Inc.); Erk1/2 (1:1000; 9102; Cell Signaling Technology, Inc.); phospho-Akt (1:1000; Ser473; 9271; Cell Signaling Technology Inc.); Akt (1:1000; 9272; Cell Signaling Technology Inc.); horseradish peroxidase-conjugated goat anti-rabbit IgG (1:20,000; 111035003; Jackson ImmunoResearch Inc., West Grove, PA, USA); and horseradish peroxidase-conjugated goat anti-mouse IgG (1:20,000; 115035003; Jackson ImmunoResearch Inc.). Quantifications of signal intensity were performed with ImageJ software (version 1.52a, National Institutes of Health, Bethesda, MD, USA). Relative dimerized PDGFR levels were determined by normalizing GFP-Trap immunoprecipitated PDGFR levels to total PDGFR levels. Relative phospho-PDGFR levels were determined by normalizing to total PDGFR levels. Relative degraded PDGFR levels were determined by normalizing cycloheximide-treated or chloroquine- and cycloheximide-treated PDGFR levels to DMSO-treated or water- and DMSO-treated PDGFR levels. Relative phospho-ERK1/2 levels were determined by normalizing to total ERK1/2 levels. Relative phospho-AKT levels were determined by normalizing to total AKT levels. When applicable, statistical analyses were performed with Prism 9 (GraphPad Software Inc., San Diego, CA, USA) using a two-tailed, ratio paired t-test within each cell line (comparing individual ligand treatment timepoint values to the no ligand 0 min timepoint value) and a two-tailed, unpaired t-test with Welch's correction between each cell line or within each cell line comparing treatments. Immunoprecipitation and western blotting experiments were performed across at least three independent experiments.

For fluorescence intensity and marker colocalization experiments, cells were seeded onto glass coverslips coated with 5 µg/ml human plasma fibronectin purified protein (MilliporeSigma) at a density of 80,000 cells and 40,000 cells for the PDGFRα homodimer cell line and PDGFRβ homodimer cell line, respectively, in RPMI growth medium. After 24 h, cells were washed with 1× phosphate buffered saline (PBS) and serum starved in HITES medium. HITES medium was replaced 23 h later. After 54 min, coverslips were photobleached for 1 min with an Axio Observer 7 fluorescence microscope (Carl Zeiss Microscopy LLC, White Plains, NY, USA) using the 2.5× objective and 488 nm laser. Cells were allowed to recover for 5 min and were treated with 10 ng/ml PDGF-AA or PDGF-BB ligand (R&D Systems) for the indicated amount of time. When applicable, cells were pretreated with 30 µM Dyngo-4a (Abcam) in DMSO (or an equivalent volume of DMSO) for 1 h before PDGF ligand stimulation. Cells were fixed in 4% paraformaldehyde (PFA) in PBS with 0.1% Triton X-100 for 10 min and washed in PBS. Cells were blocked for 1 h in 5% normal donkey serum (Jackson ImmunoResearch Inc.) in PBS and incubated overnight at 4°C in primary antibody diluted in 1% normal donkey serum in PBS. After washing in PBS, cells were incubated in Alexa Fluor 546-conjugated donkey anti-rabbit secondary antibody (1:1000; A21206; Invitrogen) or Alexa Fluor 546-conjugated donkey anti-mouse secondary antibody (1:1000; A10036; Invitrogen) diluted in 1% normal donkey serum in PBS with 2 µg/ml DAPI (Sigma-Aldrich) for 1 h. Cells were mounted in either VECTASHIELD HardSet Antifade Mounting Medium or VECTASHIELD Vibrance Antifade Mounting Medium (Vector Laboratories, Inc., Burlingame, CA, USA) and photographed using an Axiocam 506 mono digital camera (Carl Zeiss Microscopy LLC) fitted onto an Axio Observer 7 fluorescence microscope (Carl Zeiss Microscopy LLC) with the 63× oil objective with a numerical aperture of 1.4 at room temperature. The following antibodies were used for immunofluorescence analysis: PDGFRα (1:100; RB-1691; Thermo Fisher Scientific); PDGFRβ (1:25; 28D4; 558820; BD, Franklin Lakes, NJ, USA); Na⁺/K⁺-ATPase (1:500; EP1845Y; ab76020; Abcam); Rab5 (1:200; C8B1; 3547; Cell Signaling Technology Inc.); Rab7 (1:100; D95F2; 9367; Cell Signaling Technology Inc.); Rab4 (1:200; ab13252; Abcam); and Rab11 (1:100; D4F5; 5589; Cell Signaling Technology Inc.). For characterization of Venus expression in the absence of photobleaching, one independent trial, or biological replicate, was performed for each cell line, with at least 20 technical replicates consisting of individual cells imaged with Z-stacks (0.24 µm between Z-stacks with a range of 3–14 Z-stacks). For fluorescence intensity measurements, three independent trials, or biological replicates, were performed for each cell line. For each biological replicate, at least 40 technical replicates consisting of individual cells were imaged with Z-stacks (0.24 µm between Z-stacks with a range of 3–15 Z-stacks). For marker colocalization experiments, three independent trials, or biological replicates, were performed for each cell line. For each biological replicate, at least 20 technical replicates consisting of individual cells were imaged with Z-stacks (0.24 µm between Z-stacks with a range of 1–15 Z-stacks) per timepoint. Images were deconvoluted using ZEN Blue software (Carl Zeiss Microscopy LLC) using the ‘Better, fast (Regularized Inverse Filter)’ setting. For all images, extended depth of focus was applied to Z-stacks using ZEN Blue software (Carl Zeiss Microscopy LLC) to generate images with the maximum depth of field. For fluorescence intensity measurements, background was subtracted using rolling background subtraction with a radius of 30 pixels using Fiji software (version 2.1.0/1.53c; National Institutes of Health). A region of interest (ROI) was drawn around each Venus-positive cell, and integrated density was measured and recorded as the fluorescence intensity. For marker colocalization measurements, an ROI was drawn around each Venus-positive cell in the corresponding Cy3 (marker) channel using Fiji software (version 2.1.0/1.53c). For each image with a given ROI, the Cy3 channel and the EGFP channel were converted to 8-bit images. Colocalization was measured using the Colocalization Threshold function, where the rcoloc value [Pearson's correlation coefficient (PCC)] was used in statistical analysis. Statistical analyses were performed on the average values from each biological replicate with Prism 9 (GraphPad Software Inc.) using a two-tailed, unpaired t-test with Welch's correction.

For Ki67 immunofluorescence experiments, cells were seeded onto glass coverslips coated with 5 µg/ml human plasma fibronectin purified protein (MilliporeSigma) at a density of 40,000 cells in RPMI growth medium. After 24 h, cells were washed with 1× PBS and serum starved in HITES medium. HITES medium was replaced 23 h later and 1 h later cells were treated with 10 ng/ml PDGF-AA or PDGF-BB ligand (R&D Systems). After 24 h, cells were fixed in 4% PFA in PBS with 0.1% Triton X-100 for 10 min and washed in PBS with 0.1% Triton X-100. Cells were blocked for 1 h in 5% normal donkey serum (Jackson ImmunoResearch Inc.) in PBS and incubated overnight at 4°C in Ki67 primary antibody (PA1-38032; Invitrogen) diluted 1:300 in 1% normal donkey serum in PBS. After washing in PBS, cells were incubated in Alexa Fluor 546-conjugated donkey anti-rabbit secondary antibody (1:1000; A21206; Invitrogen) diluted in 1% normal donkey serum in PBS with 2 µg/ml DAPI (Sigma-Aldrich) for 1 h. Cells were mounted in VECTASHIELD HardSet Antifade Mounting Medium (Vector Laboratories, Inc.) and photographed using an Axiocam 506 mono digital camera (Carl Zeiss Microscopy LLC) fitted onto an Axio Observer 7 fluorescence microscope (Carl Zeiss Microscopy LLC) with the 20× air objective with a numerical aperture of 0.8 at room temperature. Three independent trials, or biological replicates, were performed for each cell line. For each biological replicate, 10 technical replicates consisting of individual fields of cells were imaged. All Ki67-positive signals were confirmed by DAPI staining. Statistical analyses were performed on the average values from each biological replicate with Prism 9 (GraphPad Software Inc.) using a paired t-test within each cell line and a two-tailed, unpaired t-test with Welch's correction between each cell line.

HEK 293T/17 cells were cultured in DMEM growth medium [DMEM (Gibco) supplemented with 100 U/ml penicillin (Gibco), 100 µg/ml streptomycin (Gibco) containing 10% FBS (Hyclone Laboratories Inc.)] at 37°C in 5% CO₂. Passage 20 cells at 70% confluence were transfected with pDEST-PDGFR-V1 and/or pDEST-PDGFR-V2 constructs (2.5 µg each) using Lipofectamine LTX (Thermo Fisher Scientific). Whole-cell protein lysates were collected 24 h following transfection by resuspending cells in ice-cold NP-40 lysis buffer as detailed above.

For measurement of anchorage-independent cell growth, 25,000 cells were suspended in 1.5 ml RPMI containing 100 U/ml penicillin (Gibco), 100 µg/ml streptomycin (Gibco), 10% FBS (Hyclone Laboratories Inc.) and 0.35% Difco Agar Noble (BD) and overlaid on a base layer containing 1.5 ml RPMI containing 100 U/ml penicillin (Gibco), 100 µg/ml streptomycin (Gibco), 10% FBS (Hyclone Laboratories Inc.) and 0.5% Difco Agar Noble (BD) in six-well plate wells. A feeding layer of 2 ml RPMI growth medium supplemented with 10 ng/ml PDGF-AA or PDGF-BB ligand (R&D Systems) was added on top of the agar and replaced every day. Plates were incubated at 37°C in 5% CO₂ for 10 days, and viable colonies were stained overnight with 1 mg/ml Nitrotetrazoleum Blue (Sigma-Aldrich) in PBS at 37°C. Following a second overnight incubation at 4°C, duplicate wells from each independent trial were photographed using a COOLPIX S600 digital camera (Nikon Inc., Melville, NY, USA). Images were made binary using Adobe Photoshop (version 21.0.3; Adobe Inc., San Jose, CA, USA), and colony number and area were quantified using MetaMorph imaging software (Molecular Devices, LLC, San Jose, CA, USA). Statistical analyses were performed on average values from each of three biological replicates with Prism 9 (GraphPad Software Inc.) using a paired t-test within each cell line and a two-tailed, unpaired t-test with Welch's correction between each cell line.

HCC15 cells were serum starved for 24 h in HITES medium. Cell culture inserts for 24-well plates containing polyethylene terephthalate membranes with 8 µm pores (Corning) were coated with 5 µg/ml human plasma fibronectin purified protein (MilliporeSigma). HCC15 cells were seeded at a density of 315,000 cells per insert in HITES medium and primary mouse embryonic palatal mesenchyme (MEPM) cells (see below) were seeded at a density of 250,000 cells per insert in DMEM starvation medium [DMEM (Gibco) supplemented with 100 U/ml penicillin (Gibco), 100 µg/ml streptomycin (Gibco) containing 0.1% FBS (Hyclone Laboratories Inc.)]. Inserts with HCC15 cells were immersed in 500 µl RPMI growth medium and inserts with MEPM cells were immersed in 500 µl DMEM growth medium in the absence or presence of 10 ng/ml PDGF-AA or PDGF-BB ligand (R&D Systems) for 24 h. Migrated cells were subsequently fixed in 4% PFA in PBS for 10 min and stained in 0.1% Crystal Violet in 10% ethanol for 10 min. Dried inserts were photographed using an Axiocam 105 color camera fitted onto a Stemi 508 stereo microscope at 2.5× magnification at room temperature (Carl Zeiss Microscopy LLC). Five fields of cells from each independent trial were photographed and quantified by measuring integrated density with ImageJ software (version 1.52a). For MEPM cells, independent trials consisted of cells derived from a single conditional knockout embryo or pooled cells from two conditional knockout littermates. Statistical analyses were performed on average values from each of six (HCC15 cells) or three (MEPM cells) biological replicates with Prism 9 (GraphPad Software Inc.) using a paired t-test within each cell line and a two-tailed, unpaired t-test with Welch's correction between each cell line.

All animal experimentation was approved by the Institutional Animal Care and Use Committee of the University of Colorado Anschutz Medical Campus. Pdgfra^tm8Sor mice (Tallquist and Soriano, 2003), referred to in the text as Pdgfra^fl; Pdgfrb^tm11Sor mice (Schmahl et al., 2008), referred to in the text as Pdgfrb^fl; and H2az2^{Tg(Wnt1-cre)11Rth} mice (Danielian et al., 1998), referred to in the text as Wnt1-Cre^Tg, were maintained on a 129S4 coisogenic genetic background and housed at a sub-thermoneutral temperature of 21–23°C. Mice were euthanized through inhalation of CO₂ from compressed gas. Cervical dislocation was used as a secondary verification of death. Both male and female embryos were analyzed. Primary mouse embryonic palatal mesenchyme (MEPM) cells were isolated from the palatal shelves of embryos dissected at embryonic day 13.5 in PBS and cultured in DMEM growth medium at 37°C in 5% CO₂.

11,500 passage 2 MEPM cells were seeded into wells of a 24-well plate and cultured in DMEM growth medium. After 24 h, medium was aspirated and replaced with fresh DMEM growth medium or DMEM starvation medium. After 24 h, cells were stimulated with 10 ng/ml PDGF-AA or PDGF-BB ligand (R&D Systems). Cells were fixed after 24 h in 4% PFA in PBS, stained with 0.1% Crystal Violet in 10% ethanol, extracted with 10% acetic acid, and the absorbance measured at 590 nm. Three independent trials were performed, each consisting of MEPM cells derived from a single conditional knockout embryo or non-pooled cells from two conditional knockout littermates. Statistical analyses were performed on individual values with Prism 9 (GraphPad Software Inc.) using a two-tailed, unpaired t-test with Welch's correction.

We thank Julia Mo, Damian Garno and Jessica Johnston for technical assistance. Cell sorting was performed at the University of Colorado Cancer Center Flow Cytometry Shared Resource (Cancer Center Support Grant P30CA046934). We are grateful to the laboratory of Dr Rytis Prekeris for lentiviral and packaging plasmids and the laboratory of Dr Lynn Heasley for HCC15 cells. We thank members of the Fantauzzo laboratory and Dr Rytis Prekeris for their critical comments on the manuscript.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.259686.reviewer-comments.pdf.