Delivery of CSF-1R to the lumen of macropinosomes promotes its destruction in macrophages

Following stimulation by extracellular ligands, transmembrane receptor tyrosine kinases (RTKs) initiate signaling cascades that ultimately impact upon cellular metabolism and growth (Lemmon and Schlessinger, 2010). Clearance of RTKs from the plasma membrane by small endocytic vesicles, including those involved in clathrin-mediated endocytosis (CME), is a primary mechanism for attenuating RTK mitogenic signals (Polo and Di Fiore, 2006; Sorkin and von Zastrow, 2009; von Zastrow and Sorkin, 2007). RTKs such as the epidermal growth factor receptor (EGFR) are targeted to clathrin-coated pits by a combination of signals that include ubiquitylation (Goh et al., 2010). RTK-containing endosomes mature by fusing sequentially with membranes of the early endosome, late endosome and the lysosome, leading to RTK degradation. Alternatively, the RTK can be delivered to recycling endosomes and returned to the cell surface (Sorkin and Goh, 2009). Ubiquitylated RTKs are generally targeted for degradation through the endosomal sorting complex required for transport (ESCRT) machinery, which packages them into vesicles that bud into the lumen of the late endosome to form a multivesicular body (MVB) (Hurley and Emr, 2006). These two activities of RTK degradation and recycling modulate RTK signals, highlighting the complexity of membrane traffic as route for activating and terminating signaling (Huynh et al., 2012; Polo and Di Fiore, 2006; Porat-Shliom et al., 2008; Sigismund et al., 2008; Sorkin and von Zastrow, 2009; Valdez et al., 2007).

Many RTKs stimulate actin-driven membrane ruffling and the formation of macropinosomes, which mediate the internalization of large quantities of extracellular fluid and plasma membrane (Kerr and Teasdale, 2009; Swanson, 2008). Following membrane scission, different trafficking fates are possible for macropinosomes. In macrophages, CSF-1R-stimulated macropinosomes mature by acquiring Rab5 and Rab7, marking them as early and late endosomes, respectively, prior to fusion with the lysosome (Feliciano et al., 2011; Racoosin and Swanson, 1993). Alternatively, EGF stimulation of A431 cells leads to the formation of macropinosomes that do not mature to form endosomes, but eventually fuse back with the plasma membrane (Hewlett et al., 1994; Kerr and Teasdale, 2009). Although the exact function of macropinocytosis in RTK signaling remains unknown, a number of RTKs, such as tropomyosin receptor kinases (Trks) and platelet-derived growth factor receptors (PDGFRs) (Philippidou et al., 2011; Porat-Shliom et al., 2008; Schmees et al., 2012; Valdez et al., 2005; Valdez et al., 2007) have been reported to localize to macropinosomes, where they are thought to signal. Importantly, macropinosomes have been implicated in the uptake of nutrients such as extracellular protein in Ras-transformed cancer cells (Commisso et al., 2013) and in the axenic mutants of Dictyostelium (Clarke and Kayman, 1987; Veltman et al., 2014). Macropinosomes have also been ascribed other non-RTK functions, such as antigen sampling by dendritic cells (Norbury, 2006) and the internalization of viruses (Mercer et al., 2010). Thus, macropinosomes play diverse and as yet poorly defined roles, some of which are downstream of RTK signaling.

Although the signal transduction downstream of the CSF-1R is well-defined, the mechanisms of its clearance from the cell surface remain poorly characterized. Extensive biochemical characterization has shown that the CSF-1R signals through the Ras–MAPK–Erk, Jak–STAT and Akt–PI3K pathways to control macrophage proliferation, differentiation and chemotaxis (Lee et al., 1999; Pixley and Stanley, 2004; Wang et al., 1999; Xiong et al., 2011; Yeung et al., 1998; Yu et al., 2008; Yu et al., 2012). CSF-1 potently stimulates macropinocytosis in macrophages in a phosphoinositide 3-kinase (PI3K)- and Rac1-dependent manner (Araki et al., 1996; Boocock et al., 1989; Fujii et al., 2013; Racoosin and Swanson, 1992). Additionally, signals downstream of the CSF-1R, including PI3K products, activated Rac1 and Ras, have been reported to localize to forming macropinosomes (Feliciano et al., 2011; Welliver et al., 2011; Yoshida et al., 2009). However, the distribution of receptors relative to this structure has not previously been mapped, nor has the mechanism of CSF-1R internalization been defined. Thus far, CSF-1R has been observed to be internalized by small vesicles, which are thought to fuse to form large endosomal structures (Boocock et al., 1989; Murray et al., 2000) where CSF-1R is ultimately degraded (Chen and Lin, 1984). The importance of understanding the interplay between CSF-1R signaling and traffic was highlighted recently by the finding that endocytosis of the CSF-1R mediates sustained Erk and Akt signaling that is necessary for transcriptional activities associated with macrophage growth (Huynh et al., 2012).

Here, we define the internalization and intracellular transport of the CSF-1R in macrophages. We find that macropinosomes contribute little to the internalization of the CSF-1R. Instead, small endocytic vesicles that are sensitive to clathrin inhibition internalize the CSF-1R, undergo homotypic fusion and then traffic to Rab5-labeled macropinosomes where they mature into Rab7-positive compartments. The receptor is subsequently transported into the macropinosome lumen, consistent with MVB formation, where it is degraded. These data demonstrate that CSF-1R-stimulated macropinosomes regulate the CSF-1R, thereby coupling the capture of extracellular solutes through macropinocytosis with CSF-1R-mediated cell growth.

Exposure of bone-marrow-derived macrophages (BMM) to CSF-1 increases the number and volume of macropinosomes as measured by the uptake of Texas-Red–dextran (TXR–Dex; 70,000 Da) (Fig. 1A,B; supplementary material Movie 1; Racoosin and Swanson, 1992). From these data, we estimate that 10% of the macrophage plasma membrane was internalized (neglecting exocytic events) during the first 5 minutes of exposure to CSF-1. The large surface area internalized by macropinocytosis in response to CSF-1 and the colocalization between CSF-1–CSF-1R and large endocytic compartments (Boocock et al., 1989; Murray et al., 2000) led us to consider the possibility that that macropinosomes facilitate internalization of the activated CSF-1R.

To test this possibility, we made functional Texas-Red–CSF-1 (TXR–CSF-1; supplementary material Fig. S1). Surprisingly, live-cell imaging of CSF-1-starved BMMs showed that TXR–CSF-1 was internalized in small (<500-nm diameter) endosomes (Fig. 1C; supplementary material Movie 2) within 2–5 minutes following stimulation. Macropinosomes (>500-nm diameter) formed during this time; however, when first formed, they did not enrich TXR–CSF-1 above plasma membrane levels (Fig. 1C, arrow at 5 minutes). After ∼5 minutes, the small vesicles containing TXR–CSF-1 were delivered to macropinosomes in a two-step process. First, the small endosomes containing TXR–CSF-1 fused into slightly larger and brighter vesicles that were trafficked to patches on the limiting membrane of the macropinosome (Fig. 1C, arrow at 8 minutes; supplementary material Movie 2). In a second step, the TXR–CSF-1 was transported into the macropinosome lumen, which could then be visualized as a large fluorescent sphere containing TXR–CSF-1 (Fig. 1C, arrow at 11 minutes; supplementary material Movie 2). By exposing cells to a fluid-phase marker (Lucifer Yellow) and TXR–CSF-1 we verified that these were indeed macropinosomes and not pre-existing late endosomes (supplementary material Fig. S2; Movie 3). Thus, during acute CSF-1 exposure, macropinosomes are not a major endocytic route for the CSF-1R; rather, they are a destination for the CSF-1R following internalization of small vesicles.

To confirm that the CSF-1R trafficked to macropinosomes as predicted by watching TXR–CSF-1, we devised a sequential immunofluorescence procedure to visualize the CSF-1R and the macropinosome lumen marked by TXR–Dex. This sequential staining procedure was necessary, because TXR–Dex or lysine-fixable TXR–Dex would leak out from the macropinosome upon permeabilization, leaving only a weak label on the limiting membrane of the macropinosome. Following a 5-minute pulse with CSF-1 and a 10-minute chase, macrophages were fixed and imaged. Macropinosomes appeared as fluorescent spheres (Fig. 1D, left panel). The macropinosome coordinates were recorded using a robotic stage and the cells were then permeabilized, resulting in loss of TXR–Dex fluorescence as the macromolecules leaked out (Fig. 1D, center panel). Subsequent immunodetection of CSF-1R revealed that, on average, three out of four macropinosomes contained the CSF-1R and about half of all new macropinosomes showed punctate CSF-1R structures in their lumen (Fig. 1D, right panel; Fig. 1E). Indeed, 15 minutes following CSF-1 exposure, macropinosomes could be readily identified in macrophages by their accumulation of the CSF-1R (supplementary material Fig. S3).

Macropinosomes have been regarded as possible routes for nutrient uptake for catabolism (Commisso et al., 2013; Veltman et al., 2014), implying that CSF-1 might drive macropinocytosis to promote growth by facilitating nutrient uptake. To determine whether macropinocytosis and cell growth are regulated downstream of the CSF-1R, we measured macropinosome number and BMM growth during continuous exposure to a range of CSF-1 concentrations. Indeed, macropinosome number and cell number increased in an identical CSF-1 dose-dependent manner (Fig. 1F). Moreover, acute exposure to physiological concentrations of ∼5 ng/ml CSF-1 (Scholl et al., 1994) caused the CSF-1R to traffic to macropinosomes (Fig. 1G). These results demonstrate that macropinocytosis is not an artifact of high levels of acute exposure to CSF-1 and that cell growth and macropinocytosis are coupled downstream of CSF-1R at all concentrations that support BMM growth.

The delivery of the CSF-1R to the macropinosome was not unique to murine BMMs. Exposure of human monocyte-derived macrophages (hMDM) to CSF-1 induced macropinocytosis and trafficking of TXR–CSF-1 to macropinosomes. Thus, this mechanism is conserved across human and mouse macrophages (Fig. 1H,I).

To test whether the activated CSF-1R was transported into the macropinosome lumen, we devised a strategy for staining cytosolic and extracellular epitopes on the CSF-1R while specifically permeabilizing either the plasma membrane or all membranes. We found that a 30-second exposure of PFA-fixed BMMs to digitonin permeabilized the plasma membrane while leaving TXR–Dex-labeled macropinosomes intact (Fig. 2A), whereas Triton X-100 permeabilized both the plasma membrane and macropinosomes (supplementary material Movie 4). Staining of CSF-1-starved BMMs showed that in the absence of either permeabilization, the extracellular epitope of the CSF-1R could be seen on the plasma membrane and that digitonin permeabilization was required for staining of the cytosolic epitope (Fig. 2B). Under all conditions, the antibody recognizing the cytosolic tail also gave a weak perinuclear staining pattern that we attribute to a non-specific reaction because this fraction could not be stained by the antibody recognizing the extracellular epitope. After 5 minutes of CSF-1 exposure, only the antibody recognizing the CSF-1R cytosolic epitope could be detected on early endosomes during digitonin permeabilization, consistent with the receptor residing in early endosomes (Fig. 2C). Following permeabilization with Triton X-100, both antibodies stained the CSF-1R (Fig. 2C), indicating that the receptor was intact, but that the extracellular epitope was protected by the early endosomal membrane during digitonin permeabilization. Following 15 minutes of CSF-1, the CSF-1R could only be localized to a few small puncta with either antibody during plasma membrane permeabilization, indicating lipid membranes protected both epitopes. Total membrane permeabilization with Triton X-100 allowed both epitopes to be stained and revealed a strong colocalization on large macropinosome structures matching those observed in our sequential staining experiments (Fig. 2D; Fig. 1D; supplementary material Fig. S3). These results provide strong evidence that the CSF-1R is transported into the lumen of macropinosomes, likely by the conversion of the macropinosome into an MVB.

To determine whether macropinocytosis mediates internalization of the CSF-1R by bulk internalization of plasma membrane, we found an unrelated chemokine, CXCL-12, that stimulated macropinocytosis to the same level as CSF-1 (Fig. 3B). Following CXCL-12 exposure, the CSF-1R was rarely localized to macropinosomes and, when observed, was only present at densities less than those found on the plasma membrane (Fig. 3A,C). This result was easily discernable from BMMs exposed to CSF-1, which showed a dramatic redistribution of the CSF-1R to macropinosomes (Fig. 3A). Furthermore, CXCL-12-induced macropinocytosis did not reduce the cell surface CSF-1R density, other than what could be attributed to cell spreading, whereas ligated CSF-1R was rapidly cleared from the cell surface (Fig. 3D,E). Taken together, these results indicate that macropinocytosis and the bulk internalization of plasma membrane associated with this process is an inefficient endocytic route for resting CSF-1R.

To test the selectivity of this pathway, we stimulated cells with both CSF-1 and CXCL12 and assessed the localization of the CSF-1R and CXCR4 (the CXCL12 receptor). The CSF-1R rapidly accumulated in macropinosomes; however, CXCR4 did not redistribute to macropinosomes at levels above those seen on the plasma membrane (Fig. 3F), indicating that this process was selective for the CSF-1R. Furthermore, we imaged the internalization of fluorescent transferrin as a proxy for the transferrin receptor during exposure of cells to CSF-1. Consistent with the previous observations (Racoosin and Swanson, 1992), macropinocytosis resulted in trace amounts of transferrin uptake, which was subsequently depleted from the macropinosomes, indicating that the transferrin receptor did not accumulate in macropinosomes as was true for the CSF-1R (Fig. 3G). Taken together, these results indicate that the CSF-1R is selectively targeted to the macropinosome when ligated.

Given the small size of the endocytic vesicles carrying CSF-1 (Fig. 1) and the established role of CME in RTK endocytosis (Sorkin and Goh, 2009), we speculated that CME was the predominant mechanism for CSF-1R endocytosis. Indeed, RNA interference (RNAi)-mediated knockdown of clathrin heavy chain 1 (CHC) impaired CSF-1R endocytosis as measured by clearance of CSF-1R from the cell surface (Fig. 4A,B). Near complete knockdown of CHC took 6 days and two rounds of RNAi transfection, during which the cell surface density of the CSF-1R dropped, possibly owing to either failure of transport of the CSF-1R to the cell surface or transcriptional repression. To overcome this limitation, we compared the drop in the CSF-1R cell surface density with the arrest in transferrin uptake during clathrin knockdown. In control BMMs and those exposed to CHC RNAi for 3 days, CSF-1R and transferrin internalization were unaffected (Fig. 4A–C). However, following 6 days of RNAi, CHC was substantially depleted (Fig. 4D), CSF-1R internalization halted (Fig. 4A,B) and transferrin accumulated on the cell surface (Fig. 4C). Consistent with these results, treatment with the dynamin inhibitor dynasore (Macia et al., 2006) and the clathrin inhibitor PitStop2 (von Kleist et al., 2011) impaired clearance of the CSF-1R from the cell surface as well as transferrin uptake (Fig. 4E,F). The correlation of impaired CSF-1R internalization with impaired transferrin internalization under both inhibitor and CHC RNAi is consistent with CME as the route of entry. However, impaired delivery of the CSF-1R to the cell surface during CHC knockdown and possible inhibitor specificity issues (Dutta et al., 2012; Park et al., 2013; Willox et al., 2014) suggest that other small vesicle endocytic routes, such as caveolin-mediated endocytosis might be possible. Nonetheless, small vesicle endocytosis was the predominant route of CSF-1R internalization. This observation was corroborated by the fact that inhibition of macropinocytosis by disruption of the actin cytoskeleton had little or no effect on clearance of activated CSF-1R from the cell (Fig. 6D).

To stage the transport of CSF-1R to nascent macropinosomes, we imaged the arrival of TXR–CSF-1 relative to mCit–Rab5a or YFP–Rab7. Upon exposure of BMMs to CSF-1R, pronounced membrane ruffling and the formation of mCit–Rab5a-positive macropinosomes were observed (supplementary material Movie 5). Other than macropinosomes, few if any pre-existing micrometer-scale Rab5-positive structures were observed in these cells, indicating that the low level expression of mCit–Rab5a resulting from retroviral transduction did not drive the formation of enlarged endosomes (supplementary material Movie 5). Following internalization (∼2 min), small TXR–CSF-1-containing endosomes acquired mCit–Rab5a, during which time they began to fuse into early endosomes, as seen by their increase in brightness and size (Fig. 5A,B; supplementary material Movie 5). Colocalization between TXR–CSF-1 and mCit–Rab5a was maximal around 5 minutes after addition of TXR–CSF-1 (Fig. 5C). At this point, TXR–CSF-1-containing endosomes arrived at macropinosomes decorated with mCit–Rab5a (Fig. 5D, 6.3 minutes). In the majority of cases, TXR–CSF-1-containing endosomes lost mCit–Rab5a prior to their arrival at Rab5-positive macropinosomes (Fig. 5D). We point out, however, that movement of macropinosomes and TXR–CSF-1-containing endosomes towards the Rab5a-rich perinuclear region precluded a definitive conclusion on the timing of endosome–macropinosome fusion and Rab5 loss. By 8 minutes, the TXR–CSF-1 endosomes fused with macropinosomes (Fig. 5D). These macropinosomes then transitioned from mCit–Rab5a positive to YFP–Rab7 positive, and TXR–CSF-1 was delivered to the macropinosome lumen (Fig. 5E,F). Consistent with our CSF-1R staining (Fig. 1D; Fig. 2), we interpret the translocation of TXR–CSF-1 (Fig. 5D,E) as a translocation of the CSF-1R to the lumen during macropinosome maturation into a late endosome.

Delivery of CSF-1 and the CSF-1R to the lumen of macropinosomes suggested that macropinosomes are the site of CSF-1R degradation. Following a 3-minute pulse of 100 ng/ml CSF-1, the CSF-1R was internalized and degraded without recycling to the cell surface (Chen and Lin, 1984; supplementary material Fig. S4). During this time, CSF-1 and the CSF-1R trafficked to macropinosomes (Fig. 1C,D; supplementary material Movies 2, 3). These macropinosomes shrank and underwent fusion with each other as they matured (supplementary material Movies 1, 2), reducing the number and size of macropinosomes in each cell. During this time, CSF-1R remained localized to the shrinking macropinosomes and was completely degraded by 60 minutes (Fig. 6A).

To determine whether macropinosomes mediate the degradation of the CSF-1R, we used a number of previously described inhibitors of macropinocytosis. These included amiloride and its more potent analog 5-(N-ethyl-N-isopropyl)amiloride (EIPA), compound C and expression of dominant-negative Rab5 and Rac1. Although all of these approaches reduced macropinocytosis to about a quarter of its normal levels, none of them eliminated it, and many of these approaches altered receptor traffic and membrane organization (data not shown). We concluded that the best way to block macropinocytosis was to acutely disrupt the actin cytoskeleton using the actin depolymerizing drug latrunculin B and the myosin IIa inhibitor blebbistatin in combination with the F-actin stabilizer and fragmenter jasplakinolide (Mercer et al., 2010). Acute exposure to latrunculin B (5 min) or the blebbistatin/jasplakinolide combination (15 min) completely blocked macropinocytosis as assayed by TXR–Dex uptake (Fig. 6B,C). Importantly, these treatments had little or no impact on CSF-1R endocytosis as assayed by cell surface staining for the CSF-1R (Fig. 6D). Actin-based disruption of macropinocytosis slowed the degradation of the CSF-1R. By using immunofluorescence, we observed that small CSF-1R-containing vesicles persisted in the actin-disrupted BMMs for over an hour, whereas most of the CSF-1R was destroyed between 15 and 30 minutes in unperturbed macrophages (Fig. 6A). Immunoblotting against the CSF-1R confirmed the immunofluorescence results indicating that a blockade of macropinocytosis by actin disruption slowed CSF-1R degradation (Fig. 6E,F). These results indicate that macropinosomes are likely to accelerate the degradation of the CSF-1R, but are not required for this process (Fig. 6). Furthermore, they point towards the need for more selective methods of inhibiting macropinocytosis, as acute actin disruption could have pleiotropic effects on membrane traffic that make it impossible to clearly delineate the degree to which the macropinosome accelerates CSF-1R degradation.

In our search for macropinosome-blocking drugs, we were surprised to find that the commonly used amiloride-based inhibitor EIPA triggered rapid internalization and degradation of the CSF-1R (Fig. 6G). Recently, this inhibitor has been used to study macropinocytosis in the context of the cancer cell growth (Commisso et al., 2013). In the macrophage context, however, reduction in macropinocytosis by EIPA might be due to loss of CSF-1R signaling rather than a direct action on the proposed Na⁺/H⁺ exchanger and alteration of inner leaflet pH (Koivusalo et al., 2010).

Our data define a model in which the CSF-1R is internalized by small vesicle endocytosis and is then transported to nascent macropinosomes and degraded (Fig. 7). Initially, we speculated that the large surface area of plasma membrane consumed during CSF-1-stimulated macropinocytosis would be an efficient route for CSF-1R internalization. This idea was supported by the previous reports that CSF-1–CSF-1R localizes to large endocytic structures (Boocock et al., 1989; Murray et al., 2000) and reports of Rac, PI3K and Ras signaling to CSF-1-stimulated macropinosomes (Welliver et al., 2011; Yoshida et al., 2009). However, our results showed only traces of CSF-1 or CSF-1R on forming macropinosomes. Rather, we found that the CSF-1 and CSF-1R accumulated rapidly at small endocytic sites that were sensitive to disruption of clathrin and dynamin (Fig. 3). Furthermore, inhibition of the actin cytoskeleton blocked macropinocytosis but had little effect on CSF-1R internalization. Thus, the CSF-1R must stimulate macropinocytosis through signaling by, for example, PI3K, Ras and PLC (Amyere et al., 2000; Maekawa et al., 2014; Veltman et al., 2014), which can diffuse away from the CSF-1R during its accumulation at sites of small vesicle endocytosis (Fig. 6). These diffuse signals must remotely self-organize at sites of macropinocytosis through the creation of local diffusion barriers (Welliver et al., 2011; Yoshida et al., 2009).

Our data suggest a new role for macropinosomes in RTK signaling. Multiple RTKs have been observed within macropinosomes (Trk, PDGFR) (Philippidou et al., 2011; Schmees et al., 2012) following exposure to growth factors. For macropinocytosis to mediate the endocytosis of RTKs, internalization of bulk plasma membrane would need to compete with CME or other small vesicle endocytosis of RTKs. Our data indicates that small vesicle endocytosis of the CSF-1R is much faster than macropinocytosis. Thus, our data are consistent with a model in which macropinosomes produced in response to RTK signaling mediate receptor sequestration and degradation but not internalization. Trafficking of RTKs to 500–1000-nm diameter vesicles has been widely reported in the literature; however, these compartments are often thought to arise from the fusion of small endosomes carrying the receptors to create late endosomes and MVBs (Boocock et al., 1989; Raiborg and Stenmark, 2009). Although a number of trafficking mechanisms might exist for different cell types, in primary macrophages at physiological ligand concentrations, macropinocytosis appears to be the primary mechanism leading to the formation of large late endosomes to which CSF-1R selectively traffics. The live-cell imaging of fluorescent CSF-1 traffic allowed us to capture this rapid internalization and transport to the macropinosome. Careful analysis of the trafficking of other RTKs will be needed to distinguish the endocytic roles and late endosome roles of macropinosomes from the conventional models.

What targets CSF-1R vesicles to the macropinosome? Our data indicate that homotypic fusion of small Rab5-positive vesicles (Rink et al., 2005; Sorkin and Goh, 2009) occurs prior to the delivery of the CSF-1R to the macropinosome (Fig. 3D). It is possible that fusion of these Rab5-positive early endosomes with Rab5-positive macropinosomes contributes to delivery. However, within our data, many of the early endosomes arriving at the macropinosome appeared to be Rab5 negative, suggesting that another mechanism might target CSF-1R-containing endosomes to the macropinosome. Additional experiments that manipulate Rabs and other trafficking machinery will be needed to define these mechanisms.

Macropinosomes degrade the CSF-1R in their lumen, indicating that they are precursors to MVBs. During the acquisition of Rab7, the CSF-1R was transported into the lumen of the macropinosome (Figs 2, 5) and degraded by ∼30 minutes post-exposure to CSF-1 (Fig. 6). These results are consistent with the data, obtained by using electron microscopy, that in peritoneal macrophages, I¹²⁵-CSF-1 localizes to large vacuoles that contain debris, some of which resembles lumenal vesicles (Chen and Lin, 1984), and the observation that large vacuoles containing lumenal vesicles were observed by electron microscopy in RAW 264.7 macrophages allowed to internalize horseradish peroxidase by fluid-phase endocytosis and macropinocytosis (Braun et al., 2004). Taken together, these data indicate that macropinosomes can serve as sites for lumenal budding and might be MVB precursors. Electron microscopy studies will be needed to determine the prevalence of MVB formation and additional molecular biology will be needed to determine the machinery responsible for transporting the receptor into the lumen. The robustness of this process (supplementary material Fig. S3) indicates that macropinosomes contribute to CSF-1R signal termination by accelerating the sequestration of the cytoplasmic tail of CSF-1R into macropinosomes (Fig. 2). Furthermore, the observation of this mechanism in both mouse and human macrophages underscores the relevance of this mechanism for controlling macrophage function.

Macropinosomes might couple signal amplification with signal termination. Although CSF-1R signaling originates at the plasma membrane, recent evidence indicates that sustained signaling from internalized CSF-1R through the MAPK–Erk pathway mediates transcriptional changes associated with growth (Huynh et al., 2012). CSF-1R trafficking to the macropinosome is consistent with this observation and with the proposed roles of macropinocytosis in amplifying signal from other RTKs such as the PDGFR and Trk receptors (Philippidou et al., 2011; Schmees et al., 2012). Thus, the macropinosome might be a trafficking destination for the CSF-1R as a possible mechanism for amplifying signaling while also serving as their signal terminators through delivery to the macropinosome lumen.

Traffic of the CSF-1R to the macropinosome might link CSF-1 signaling to nutrient uptake. Specifically, we observed that macrophage proliferation and macropinocytosis are tightly coupled over the range of CSF-1 concentrations that support macrophage growth (Fig. 1). Recent work has shown that in Ras(V12)-transformed cancer cells, macropinocytosis of serum proteins provides an amino acid supply in vitro and in vivo (Commisso et al., 2013). Thus, it is likely that CSF-1 signaling drives macropinocytosis for nutrient uptake in macrophages, which have evolved to be efficient at the uptake and degradation of extracellular protein. Trafficking of the CSF-1R to the macropinosome might couple CSF-1R signaling with nutrient sensing (Laplante and Sabatini, 2012).

Primary antibodies were acquired that recognized an extracellular epitope or intracellular domain for the CSF-1R (clone AFS98, eBioscience; SC692, Santa Cruz Biotechnology), CXCR4 (SC9046, Santa Cruz Biotechnology), Rab7 (clone D95F2, Cell Signaling Technology) and the clathrin heavy chain (CHC) (clone sc-12734, Santa Cruz Biotechnology). Secondary antibodies were conjugated to Alexa Fluor 488 (Life Technologies), Dylight 690, Dylight 780 or Dylight 594 (Thermo Fisher Scientific). CSF-1 was labeled using the succinimidyl ester of Texas Red-X (TXR; catalog number T-6134; Life Technologies) or Dylight 594 (product number 46412, Thermo Fisher Scientific). Macropinosome blocking experiments used several drugs, including latrunculin B (Santa Cruz Biotechnology), 5-(N-ethyl-N-isopropyl)amiloride (EIPA, Sigma), jasplakinolide and blebbistatin (Enzo Life Science). Macropinosomes were specifically labeled using TXR–Dex (70,000 Da, Life Technologies) or Lucifer Yellow (LY, Sigma). Blockade of endocytosis used dynasore (Sigma) or PitStop2 (Abcam), as described previously (Macia et al., 2006; von Kleist et al., 2011). Alternative stimulation of macropinocytosis used the recombinant chemokine CXCL12 (also known as SDF-1a, PeproTech). Mammalian-expressed human recombinant CSF-1 (Biolegend) was used in hMDM culture. Dulbecco's Modified Eagle Medium (DMEM, Hyclone) and fetal bovine serum (Hyclone) were used for cell culture. Ni-NTA affinity resin (Qiagen) was used in CSF-1 purification. Digitonin and Triton X-100 (Sigma) were used to permeabilize cells.

Recombinant CSF-1 (N-terminal 151 amino acids) was purified from Escherichia coli. A sequence for the N-terminal 453 nucleotides of CSF-1 cDNA (Biosystem) was amplified and cloned into the pQE31 plasmid between the restriction sites for BamHI and HindIII using standard digestion and ligation to enable expression of 6×His–CSF-1. Protein expression was carried out in BL21(DE3) E. coli under induction by isopropyl β-D-1-thiogalactopyranoside (IPTG). CSF-1 inclusion bodies were solubilized in urea buffer [8 M urea, 100 mM NaH₂PO₄, 10 mM Tris-HCl pH 8.0, 10 mM dithiothreitol (DTT)] overnight at room temperature, and His–CSF-1 was purified by affinity chromatography according to the manufacturer's instructions. Purified monomeric His–CSF-1 was refolded by dilution into glycine buffer (50 mM glycine pH 9.0) at a concentration of 0.3 mg/ml. Refolded CSF-1 dimers were separated from monomers by gel-filtration chromatography (HiPrep Sephacryl S-100 HR, GE-Healthcare; supplementary material Fig. S1). The biological function of fluorescently labeled and non-labeled CSF-1 was assayed by macrophage differentiation from bone marrow cells (supplementary material Fig. S1).

mCit–Rab5a and YFP–Rab7 were cloned into an MoMLV-based retroviral cassette, PIB2-MCS, between the restriction sites MluI and NsiI. BMMs were differentiated in DMEM supplemented with 20% heat-inactivated fetal bovine serum and 30% L929-cell-conditioned medium from C57BL/6 mice. The PIB2-MCS plasmids were transiently transfected into Plat-E packaging cells (Cell Biolabs, Inc.) using Fugene 6 (Roche Applied Sciences). Viral supernatant was collected 72 hours after transfection. For transduction, BMMs were exposed to viral supernatant for 72 hours in culture medium.

Freshly drawn blood was diluted 1∶1 in Dulbecco's phosphate-buffered saline (DPBS), layered on Ficoll-Paque (GE Healthcare) and centrifuged at 900 g for 20 minutes. Mononuclear cells at the interface were collected and washed three times in DPBS. Cells were plated at a density of 2×10⁶ cells/ml on 10-cm culture dishes in complete growth medium (high-glucose DMEM containing 20% fetal bovine serum). After 4 hours, the medium was removed and adherent cells were washed with DPBS to remove non-adherent mononuclear cells. Remaining adherent mononuclear cells were cultured in growth medium supplemented with 5 ng/ml human CSF-1 for 7 days before imaging (medium was replaced every 48 hours).

Small interfering (si)RNA (Smart Pool, Dharmacon) was used to knock down CHC in mouse BMMs. 1×10⁶ D7 BMM cells were transfected with 100 pmol of siRNA using the mouse macrophage Nucleofector kit (Lonza), and the cells were re-transfected at day 3 after first transfection for double CHC knockdown. A transferrin uptake assay, using Alexa-Fluor-488-conjugated transferrin (Invitrogen), and a CSF-1R internalization assay were conducted at day 6 post-transfection. Western blot analysis was used to verify CHC knockdown.

To measure macropinosome formation in response to CSF-1 stimulation, BMMs were starved of ligand overnight in DMEM supplemented with 10% fetal bovine serum. Cells were exposed to 100 ng/ml recombinant CSF-1 or 400 ng/ml CXCL12 in the presence of 0.5 mg/ml 70,000 Da TXR–Dex. Cells were washed and fixed with 4% paraformaldehyde (PFA) at 5 or 15 minutes, as specified in the figure legends. The macropinosome number per cell was quantified by counting all labeled vesicles in fluorescence images. The diameter for macropinosomes was quantified in ImageJ by line measurements across individual organelles in fluorescence images. Macropinosomes were assumed to be spheres, and volumes were calculated using the formula volume = 4/3πr³.

Acute blockade of macropinocytosis was achieved by pre-incubating BMM with either 0.5 µM latrunculin for 5 minutes or 75 µM blebbistatin for 10 minutes, followed by a 5-minute incubation with 1 µM jasplakinolide. All treatments were performed in a 5% CO₂ incubator at 37°C. The number of macropinosomes in drug-treated cells was quantified as described above.

Following CSF-1 stimulation of starved BMMs for specific times, cells were fixed with 4% PFA for 10 minutes, permeabilized with 1 mg/ml Triton X-100 for 10 minutes at room temperature and blocked with 1% bovine serum albumin (BSA), each in a PBS buffer. Stimulation and chase steps were performed in a 5% CO₂ incubator at 37°C. Subsequent fixation and immunofluorescence steps were conducted at room temperature. Primary and secondary antibody incubations were each performed for 1 hour at room temperature. Surface-localized receptor staining was similar to the immunofluorescence protocol described above, except that cells were fixed with 2% PFA for 5 minutes and the cells were not permeabilized. Cells co-stained for CSF-1R and CXCR4 were fixed and permeabilized in methanol for 10 minutes at −20°C.

Antibodies recognizing CSF-1R extracellular or cytosolic epitopes were used to map the topology of the CSF-1R. Following CSF-1 stimulation for specific periods of time, cells were fixed with 2% PFA for 5 minutes at room temperature. The plasma membrane was selectively permeabilized by exposure of cells to 40 µg/ml digitonin for 30 seconds at room temperature. Following 1% BSA blocking at room temperature for 30 seconds, cells were incubated with two CSF-1R antibodies. Alexa-Fluor-488- and Dylight-594-conjugated secondary antibodies were used to specifically detect each primary antibody.

Sequential immunostaining of CSF-1R on macropinosomes was achieved on the Till iMic using an x-y-z positioning stage. BMMs were starved overnight and were stimulated with L929 cell supernatant plus 0.5 mg/ml TXR–Dex (70 kDa) for 15 minutes, fixed and mounted into an Attofluor cell chamber (Invitrogen). Several cells spanning ten fields of view were imaged in a single experiment by using an automated stage tracking X, Y and Z coordinates. Images were acquired following each stage of processing: fixation, permeabilization and immunostaining against CSF-1R.

Following CSF-1 stimulation, BMMs were lysed with mPer lysis buffer (Thermo Fisher) plus 1∶100 proteinase inhibitor cocktail (Sigma) and 100 µM sodium orthovanadate (Sigma). Protein concentration was determined by using the bicinchoninic acid (BCA) assay. Following protein quantification, protein samples were separated by using the Mini-PROTEAN dual cell (BioRad) on 8% pre-cast Tris-Glycine NB gels (NuSep). Gels were transferred to nitrocellulose membranes using the Mini-PROTEAN dual cell (BioRad). Membranes were blocked for 1 hour at room temperature using 5% BSA and were incubated with primary antibody overnight at 4°C or for 2 hours at room temperature. Membranes were imaged by using the LI-COR Odyssey infrared imaging system. Western blot images were quantified using ImageJ. For all experiments, protein values were normalized against an internal loading control (total Erk). For each experiment, CSF-1R intensity values for each data point were normalized against the intensity of the corresponding unstimulated sample.

Immunofluorescence images were acquired on a Leica CTR4000 inverted microscope equipped with an QICAM 12-bit color camera, 12 V 100 W halogen lamp, QCapture software and 60× oil lens. Exposure times were optimized according to sample brightness. For quantification of immunofluorescence images in MATLAB (MathWorks), imaging parameters were held constant for each experimental condition. A binary mask was created using threshholding. Intensity was calculated on a per cell basis. Only regions of the image in which whole cells were evident were measured.

Live-cell imaging and the sequential dextran immunofluorescence experiments were performed on a custom-built Till iMic microscope (FEI) using a 60× water lens (a complete schematic of the imaging system can be found in Hoppe et al., 2013). Macrophages starved of CSF-1 overnight were pulsed with 100 ng/ml TXR–CSF-1 for 1 minute, washed and chased for 14 minutes. Images were taken before the pulse and chase with CSF-1 on both yellow and red channels at a rate of 20 seconds between frames. Red and yellow channel images were registered and normalized. A colocalization coefficient between yellow and red images was calculated according to the Pearson correlation coefficient in MATLAB.

We thank Lu Huang, Shuangling Zhang and Dakota Weathers for technical assistance and Geoffrey Graff for the drawing in Fig. 7.