Rab5 regulates macropinocytosis by recruiting the inositol 5-phosphatases OCRL and Inpp5b that hydrolyse PtdIns(4,5)P₂

Macropinocytosis is an actin-driven process that involves the formation and extension of plasma membrane ruffles and the eventual closure of large (≥0.2 to 5.0 µm) endocytic vacuoles (Swanson, 2008). Dendritic cells and macrophages perform macropinocytosis constitutively in order to sample their environment for immune surveillance (Bohdanowicz et al., 2013). However, most cell types require stimulation by growth promoters to initiate macropinocytosis (Haigler et al., 1979; Mellström et al., 1988; Racoosin and Swanson, 1989). In these cases macropinocytosis can play an important role in the delivery of extracellular nutrients, including proteins and amino acids, to lysosomes (Bloomfield and Kay, 2016; Racoosin and Swanson, 1992). The significance of macropinocytosis to cellular physiology is illustrated by its common dysregulation in a variety of diseased states; the manipulation of macropinocytosis is a hallmark of cancer metabolism (O'Donnell et al., 2018) and provides an entry gateway for a variety of viruses, including vaccinia and Ebola (Mercer and Helenius, 2012).

The formation of macropinosomes depends on the integration of receptor signalling, phosphoinositide metabolism, activation of small GTPases and the remodelling of the actin cytoskeleton (Buckley and King, 2017; Levin et al., 2014; Marques et al., 2017). Growth factors, such as macrophage colony stimulating factor (M-CSF, also known as CSF1) and epithelial growth factor (EGF), bind to cognate tyrosine kinase receptors and activate signalling cascades that culminate in the recruitment and activation of phosphatidylinositol 3-kinase (PI3K) (Abella et al., 2010; Araki et al., 1996; Cheng et al., 2015; Wang et al., 2001). The resulting generation of phosphatidylinositol-3,4,5-trisphosphate [PtdIns(3,4,5)P₃] is critical for macropinocytosis in a variety of cell types (King and Kay, 2019; Veltman et al., 2016; Welliver and Swanson, 2012). Together with its precursor, phosphatidylinositol-4,5-bisphosphate [PtdIns(4,5)P₂], PtdIns(3,4,5)P₃ promotes the recruitment and activation of small GTPases of the Ras and Rho families that induce the formation of the circular ruffles that precede macropinocytosis (Bar-Sagi and Feramisco, 1986; Fujii et al., 2013; Kay et al., 2018; Porat-Shliom et al., 2008; Welliver et al., 2011). Rho-family GTPases support the recruitment of Arp2/3 and formins that mediate the actin polymerization observed in macropinocytic cups (Junemann et al., 2016; Veltman et al., 2016).

Completion of macropinosome formation requires membrane fusion and is followed by maturation of the resulting vacuole, which proceeds to merge with endosomes and, ultimately, lysosomes. Fusion and maturation are accompanied by rapid disassembly of the rich actin meshwork that initiated ruffling and macropinosome formation. The disassembly is due, at least in part, to the breakdown of PtdIns(3,4,5)P₃ and PtdIns(4,5)P₂ that accompanies macropinosome closure (Araki et al., 2007; Hasegawa et al., 2011; Maekawa et al., 2014; Welliver and Swanson, 2012; Yoshida et al., 2009). Closure is also coincident with the acquisition of small GTPases of the Rab family, notably Rab5 (Araki et al., 2006; Feliciano et al., 2011; Porat-Shliom et al., 2008; Roberts et al., 2000; Welliver and Swanson, 2012). Rab5, a prototypical component of early endosomes, is thought to drive the initial stages of macropinosome maturation. However, expression of dominant-negative Rab5 was found to block macropinocytosis (Feliciano et al., 2011; Roberts et al., 2000), suggesting an additional early role for Rab5 in macropinosome formation. Accordingly, active Rab5 has been localized to the plasma membrane during macropinocytosis (Bucci et al., 1992; Chavrier et al., 1990; Lanzetti et al., 2004). However, at this time, the manner whereby Rab5 affects macropinosome formation remains unclear. We therefore revisited the role of Rab5 in macropinocytosis.

Serum-starved A431 cells respond quickly to the addition of EGF with plasmalemmal ruffling and the formation of 1.5-2 μm-sized macropinosomes (Fabricant et al., 1977; Haigler et al., 1979). We chose this well-established model system to investigate the role of Rab5 in macropinocytosis. As expected, in the absence of EGF, A431 cells internalized little tetramethylrhodamine-conjugated 70 kDa dextran (TMR-D), a prototypical marker of macropinocytic uptake, and uptake was unchanged by expression of wild-type Rab5A (Fig. 1A, left panel). After incubation with EGF, Rab5A⁺ cells internalized TMR-D robustly (Fig. 1A, right panel), as did non-transfected neighbouring cells (not shown). In contrast, EGF-treated A431 cells expressing a dominant-negative version of Rab5A (Rab5A_S34N) failed to internalize TMR-D, whereas the non-transfected neighbouring cells did (Fig. 1B). Expression of Rab5A_S34N decreased macropinocytosis by 86.8% (Fig. 1C, black bars), in agreement with earlier findings (Feliciano et al., 2011; Roberts et al., 2000). Because expression of dominant-negative alleles can have pleiotropic effects, e.g. by inhibiting exchange factors shared by multiple GTPases, we validated the requirement for Rab5 by silencing the three Rab5 isoforms; the combined use of siRNA to the A, B and C isoforms decreased the overall expression of Rab5 by 88.1±4.3% (n=4), as assessed by immunoblotting (not shown). This collective silencing depressed TMR-D uptake by 73% (Fig. 1C, grey bars), validating the specific involvement of Rab5 in macropinosome formation.

Macropinocytosis entails the development of membrane ruffles and their closure to form sealed vacuoles. We therefore sought to define whether Rab5 is required for the extension of ruffles or, instead, for their sealing. To this end, the dynamics of macropinosome formation was monitored by time-lapse video microscopy, using the N-terminal domain of Lyn (Lyn₁₁) tagged with RFP to selectively visualize the plasma membrane. This region of Lyn is myristoylated and palmitoylated, effectively targeting the Lyn₁₁-RFP construct to the plasma membrane and nascent macropinosomes. Temporal projections of A431 cells expressing Lyn₁₁-RFP revealed the extensive ruffling elicited by treatment with EGF (e.g. Fig. 1D). Importantly ruffling was qualitatively indistinguishable in cells that were co-transfected with Lyn₁₁-RFP and either wild-type Rab5A or Rab5A_S34N (Fig. 1D,F; Movies 1, 2). A more rigorous comparison of the effects of the wild-type and dominant-negative forms of Rab5 was made by measuring the sum of the differences in fluorescence intensity (ΣΔ_pixel) associated with membrane displacement (i.e. ruffling). The differences were squared (ΣΔ_pixel²) to eliminate the cancelling effects of positive versus negative changes in fluorescence intensity. As shown in Fig. 1E,G, where the behaviour of multiple individual cells is compared, membrane positioning changed comparatively little over time before the addition of EGF, but marked oscillations were apparent upon addition of the growth factor. Comparison of the s.d. of the ΣΔ_pixel² in the period before and after the addition of EGF confirmed that the ruffling triggered by the growth factor was similar in cells expressing Rab5A or Rab5A_S34N (Fig. 1H). We therefore concluded that functional Rab5 is not required for membrane ruffling and that the GTPase is instead involved in another downstream step in vacuole generation.

Because Rab5 is seemingly not required for ruffling, we considered the possibility that it is instead involved in macropinosome sealing. This would require the engagement of the GTPase before closure of the macropinocytic vacuole. Rab5 has been reported to be present in macropinosome membranes shortly after closure (Feliciano et al., 2011; Roberts et al., 2000), but the precise timing of its acquisition has not been established. It is possible that Rab5 is involved in the sealing of the macropinosome; alternatively, Rab5 may play a role in stabilizing the nascent macropinosome, preventing its back-fusion with the plasma membrane. In accordance with the earlier reports, we routinely observed Rab5 merging with nascent macropinosomes within ∼4 min of EGF addition (Fig. 2A). While examining the macropinocytic process by time-lapse microscopy we observed the striking recruitment and tight apposition of GFP-Rab5A⁺ endocytic vesicles to emerging circular ruffles/early macropinosomes as early as ∼1 min after EGF addition (Fig. 2A,B; Movie 3).

When acquiring optical slices by confocal microscopy it is not possible to conclusively differentiate between open circular ruffles and sealed vacuoles. This strategy alone was therefore insufficient to establish whether the tight apposition and fusion of Rab5 vesicles occur before macropinosome sealing. To more definitively establish the timing of Rab5 recruitment, we induced macropinocytosis with EGF for 2 min in the presence of TMR-D, then rapidly chilled the cells to arrest the process and stained them with the membrane-impermeant dye FM4-64. FM4-64, which is brightly fluorescent only when intercalated into membranes, was viewed immediately (< 5 min) to prevent its endocytic uptake. Thus, only the plasma membrane, which is continuously exposed to the extracellular milieu, as well as open (but not sealed) macropinosomes, was expected to show dye fluorescence. Conversely, only macropinosomes were expected to retain TMR-D. Strikingly, we found that Rab5A⁺ vesicles were distinctly recruited to circular ruffles that had not closed, identified as round structures that were nearly continuous yet stained clearly with FM4-64; multiple examples are illustrated in transverse (X versus Y; bottom) and in orthogonal reconstructions (X versus Z; top) in Fig. 2C. These structures did not retain TMR-D (Fig. 2C, left), which was instead retained by FM4-64⁻ macropinosomes (Fig. 2C, right). To ensure that these observations were not an artifactual result of Rab5A overexpression, we also examined the distribution of the early endocytic compartment in EGF-treated cells by immunostaining endogenous EEA1, a Rab5 ligand and effector (Simonsen et al., 1998). As described for ectopically expressed Rab5A, EEA1⁺ vesicles localized to membrane ruffles and open circular ruffles (visualized using plasmalemmal PLCδ-PH as a marker) before macropinosome closure (Fig. 2D). Although we cannot ascertain that the recruited endosomes had fused with open circular ruffles, these data indicate that tight apposition of Rab5⁺ organelles precedes macropinosome closure.

Because Rab5-containing vesicles were recruited to circular ruffles at early stages of macropinosome formation (Fig. 2), and genetic inhibition of Rab5 blocked macropinocytosis (Fig. 1B,C), we investigated whether their attachment and/or fusion with the plasma membrane is required to promote macropinosome formation. To this end, we expressed plasma membrane targeted chimeras of wild-type Rab5A, Rab5A_Q67L – a constitutively active variant – and the dominant-negative Rab5A_S34N fused to the 20 C-terminal amino acids of K-Ras4B. This sequence includes a CAAX motif and a polylysine sequence that jointly target the chimeric construct to the plasma membrane (Galperin and Sorkin, 2003). Expression of these constructs did not have a noticeable effect on endocytosis, as measured by internalization of transferrin (Fig. 3A,B, insets). Interestingly, expression of wild-type or constitutively active Rab5A at the plasma membrane was neither sufficient to induce macropinocytosis – measured as TMR-D internalization – in the absence of EGF (Fig. 3A), nor did it prevent macropinocytosis induced by EGF (Fig. 3B). In contrast, plasmalemmal Rab5A_S34N-CAAX robustly inhibited macropinocytosis (Fig. 3B,C), in a manner indistinguishable from that observed for Rab5A_S34N expressed in the cytosol, which partitions into the membrane by virtue of its geranylgeranylation (Fig. S1B; Fig.3B,C).

Additionally, we recapitulated these results using an inducible rapamycin heterodimerization system (iRAP) to recruit CFP-FKBP-Rab5A or Rab5A_S34N to the plasma membrane (using Lyn₁₁-FKB). Rapamycin-induced recruitment of CFP-FKBP-Rab5A was not sufficient to induce macropinocytosis in the absence of EGF, but recruitment of CFP-FKBP-Rab5A_S34N blocked macropinocytosis in EGF-stimulated cells (Fig. S1), consistent with the results obtained with CAAX-tagged versions of the GTPase. Jointly, these observations suggest that the dominant-negative effect of Rab5A_S34N on macropinocytosis is exerted at the plasma membrane.

Endogenous inactive GDP-bound Rab5 is thought to reside in the cytosol, complexed with Rab GDI, and the active GTP-bound form associates with early endocytic membranes (Zerial and McBride, 2001). During macropinocytosis, it was conceivable that Rab5 could be activated/recruited to the plasma membrane from the cytosol, or be acquired through fusion of endocytic vesicles containing active Rab5, as suggested by the observations in Fig. 2. To distinguish between these mechanisms, we adapted a reverse dimerization system (Rivera et al., 2000) to acutely block soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE)-mediated membrane fusion with dominant-negative N-ethylmaleimide-sensitive factor (NSF_E329Q). Expression of this reverse dimerization construct (EGFP-F_M4-NSF_E329Q) results in the formation of multimeric aggregates in the cytosol, due to the presence of four tandem mutant FKBP (F_M4) domains in the construct, which can be solubilized promptly by the addition of a cell-permeant rapamycin analog (Fig. 3D). Due to the acute release of NSF_E329Q, the cells do not experience the toxicity normally seen when expressing soluble dominant-negative NSF for extended periods (Coppolino et al., 2001).

As expected, when expressed in A431 cells, EGFP-F_M4-NSF_E329Q formed large cytosolic aggregates (Fig. 3E, left) that were functionally inconsequential. This was verified by monitoring the appearance of the Golgi complex, which was unaffected by the aggregated form of the dominant-negative NSF (Fig. S2A). As expected, aggregated EGFP-F_M4-NSF_E329Q also did not affect macropinocytosis (Fig. 3E,F). The EGFP-F_M4-NSF_E329Q aggregates dissociated gradually upon addition of the D/D solubilizer (Fig. S3E). The appearance of soluble EGFP-F_M4-NSF_E329Q was accompanied by inhibition of endomembrane fusion, as evinced by the fragmentation and dispersal of the Golgi complex (Fig. S2A, right). Importantly, solubilization of the dominant-negative NSF caused a profound (92%) inhibition of macropinocytosis (Fig. 3E,F). Of note, the membrane ruffling induced by EGF was unaffected by the soluble NSF_E329Q (Fig. S2B) and Rab5⁺ endosomes, seemingly coated with NSF_E329Q, were found in the vicinity of actin-rich circular ruffles (Fig. 3G). We therefore concluded that a SNARE-mediated fusion event, possibly that of Rab5-containing vesicles with the plasma membrane, is required for macropinocytosis. However, an indirect effect of NSF_E329Q cannot be ruled out.

Loss of PtdIns(4,5)P₂, which is abundant in the plasma membrane, is one of the early events that accompany macropinosome formation (Porat-Shliom et al., 2008; Welliver and Swanson, 2012). While examining EGF-induced macropinocytosis by time-lapse microscopy, we observed that the docking of Rab5A⁺ endocytic vesicles onto circular ruffles preceded the loss of PtdIns(4,5)P₂ (Fig. 4A). Strikingly, expression of dominant-negative Rab5A_S34N prolonged the presence of PtdIns(4,5)P₂ on circular ruffles (Fig. 4B), suggesting that active Rab5 is normally required for the depletion of the inositide. These observations suggested that PtdIns(4,5)P₂ depletion may be required for the completion of macropinocytosis.

This notion was tested by expressing the genetically encoded PtdIns(4,5)P₂ biosensor PLCδ-PH-GFP in cells also expressing Lyn₁₁-RFP, used to visualize the membrane and nascent macropinosomes. As expected, PtdIns(4,5)P₂ was present in the plasma membrane and was clearly visible in the ruffles induced by EGF (Fig. 4C; Movie 4). As reported previously (Porat-Shliom et al., 2008; Welliver and Swanson, 2012), PtdIns(4,5)P₂ was lost from the macropinosome membrane as it internalized (Fig. 4C, right; Movie 4). The loss of PtdIns(4,5)P₂ from the macropinosome membrane was associated with the retention of TMR-D, the hallmark of closed macropinosomes (Fig. 4D). This inverse correlation was analysed in detail by measuring the abundance of PLCδ-PH-GFP on Lyn₁₁-RFP⁺ cups/vesicles over time and comparing it to the retention of TMR-D. Numerous circular ruffles containing high PtdIns(4,5)P₂ and little TMR-D were observed 5 min after stimulation with EGF (Fig. 4E). Between 10 and 20 min the number of vacuoles retaining TMR-D increased progressively as the PtdIns(4,5)P₂ levels decreased (Fig. 4E; Fig. S3A). Analysis of these populations (Fig. S3B) showed the vesicles fell into well-defined groups: those that were open (i.e. devoid of TMR-D) and contained PtdIns(4,5)P₂, and those that were seemingly closed (i.e. able to retain TMR-D) and depleted of PtdIns(4,5)P₂. There were virtually no sealed vesicles that retained PtdIns(4,5)P₂. Taken together, these data confirmed that PtdIns(4,5)P₂ localized predominantly to open macropinocytic cups, and was depleted quickly at the time of macropinosome sealing or shortly thereafter.

These observations raised the possibility that PtdIns(4,5)P₂ depletion is in fact required for the completion of macropinocytosis or to prevent macropinosome back-fusion with the plasmalemma. To examine this possibility, we sought to prevent this depletion by overexpressing the type I PtdIns 4-phosphate 5-kinase (PIP5K), which catalyzes the conversion of plasmalemmal PtdIns(4)P to PtdIns(4,5)P₂. When expressed heterologously in A431 cells, the GFP-tagged α and β isoforms of PIP5K localized to the plasma membrane, in which they effectively blocked macropinocytic retention of TMR-D by 76.2% and 76.3%, respectively, compared to cells expressing GFP only (Fig. 4F,G). Notably, expression of a kinase-dead version of PIP5Kβ, which is normally targeted to the membrane, did not block macropinocytosis, indicating that the catalytic activity of PIP5K, and thus the overproduction of PtdIns(4,5)P₂, is responsible for blocking the completion of macropinocytosis. Temporal projections of A431 cells expressing Lyn₁₁-RFP to visualize the plasma membrane confirmed that the plasma membrane underwent ruffling upon EGF addition in cells expressing PIP5Kβ-GFP or PIP5Kβ_KD-GFP; this ruffling was indistinguishable from that of EGF-stimulated cells co-expressing GFP (Fig. S4), suggesting that expression of the kinases did not interfere grossly with actin dynamics. Instead, we interpreted these data to mean that the sustained presence of PtdIns(4,5)P₂ in the circular ruffle blocks macropinosome sealing.

The timing of Rab5 acquisition and loss of PtdIns(4,5)P₂ from forming phagosomes led us to question whether the two events are causally related. The mammalian 5-phosphatases Inpp5b and OCRL, which can degrade PtdIns(4,5)P₂, are both Rab5-associating effectors implicated in endocytosis and macropinocytosis (Fukuda et al., 2008; Hyvola et al., 2006; Loovers et al., 2007; Mao et al., 2009; Shin et al., 2005; Swan et al., 2010; Williams et al., 2007). Therefore, these phosphatases were likely candidates to catalyze the observed PtdIns(4,5)P₂ depletion. Inpp5b and OCRL can interact with Rab5 directly, or interact with endomembranes by way of another Rab5 effector, APPL1 (Bohdanowicz et al., 2011; Erdmann et al., 2007; Swan et al., 2010; Zhu et al., 2007). We observed that soon after stimulation by EGF, punctate structures containing Inpp5b and OCRL – likely endomembrane vesicles – were recruited to PtdIns(4,5)P₂⁺ plasma membrane ruffles (Fig. 5A) in a manner similar to that observed for Rab5. Notably, APPL1 was also recruited in this fashion (Fig. 5A, right). Time-lapse imaging revealed that the recruitment of Inpp5b, OCRL and APPL1 preceded the loss of PtdIns(4,5)P₂ from macropinocytic cups (Fig. S5A-C). Like the analogous early recruitment of Rab5⁺ vesicles, we interpreted this as evidence that the docking of Inpp5b-, OCRL- and APPL1-bearing vesicles occurred before macropinosomes sealed (Fig. S5D-F). Extensive acquisition of Inpp5b, OCRL and APPL1 was obvious on macropinosomes that retained TMR-D, and were therefore closed (Fig. 5B), consistent with earlier findings (Swan et al., 2010; Wall et al., 2018; Zoncu et al., 2009). The requirement for APPL1 for the recruitment of the 5-phosphatases was tested using siRNA-mediated silencing. When APPL1 expression was suppressed by 85.9% (Fig. S6A), APPL1-depleted cells recruited considerably less OCRL to Rab5-containing endosomes: the Manders' (M) coefficient was M_red=0.280, compared to M_red=0.785 in cells treated with scrambled RNA (Fig. 5C,D). This was consistent with previous findings that APPL1 cooperates with Rab5 to recruit 5-phosphatases to endosomes and phagosomes (Bohdanowicz et al., 2011; Erdmann et al., 2007).

The functional requirement of these Rab5 effectors for macropinocytosis was then assessed by RNA-mediated silencing. The 5-phosphatases were silenced individually or together, and APPL1 was silenced as well. The effectiveness of the silencing protocol was assessed: the expression of Inpp5b was reduced by 80.9% and that of OCRL by 77.8%, when silenced separately (Fig. S6B). When silenced jointly, expression was suppressed by 82.1% and 77.7% for Inpp5b and OCRL, respectively. APPL1 silencing was the most effective, reaching 95.9% (Fig. S6B). The effects of these treatments on macropinocytosis were assessed next. Inhibiting the expression of either Inpp5b or OCRL caused a marked reduction in macropinocytic efficiency, and simultaneous silencing of both phosphatases resulted in an even greater reduction (∼85%; Fig. 5E). The extent of the latter was comparable to the one attained by silencing APPL1 (Fig. 5E).

These findings are consistent with the notion that APPL1 cooperates with Rab5 in the recruitment of OCRL and Inpp5b. We therefore used APPL1-silenced cells to further assess the role of PtdIns(4,5)P₂ breakdown by the phosphatases in the macropinocytic process. The effect of downregulating the adaptor on membrane ruffling was investigated using the same approaches introduced earlier. Visualization (Fig. 5F versus H; compare also Movie 5 versus Movie 6) and quantitative analysis (Fig. 5G,I,J) of ruffling in cells transfected with Lyn₁₁-RFP showed no significant differences between cells treated with scrambled siRNA and those silenced using APPL1-directed siRNA, which responded normally to stimulation by EGF. These findings reinforce the conclusion that depletion of PtdIns(4,5)P₂ occurs after plasma membrane ruffling and is likely involved in macropinosome sealing.

Jointly, the preceding observations suggest that Rab5 may participate in macropinosome sealing by recruiting Inpp5b and OCRL, which facilitate sealing by dephosphorylation of PtdIns(4,5)P₂. If this model were true, we hypothesized that it would be possible to bypass the inhibitory effect of Rab5A_S34N by recruiting to the plasma membrane an alternative heterologous 5-phosphatase capable of hydrolysing PtdIns(4,5)P₂ while the cells are undergoing EGF-induced ruffling. We chose the yeast 5-phosphatase Inp54 as it is exquisitely specific for PtdIns(4,5)P₂ (Raucher et al., 2000; Stolz et al., 1998). As before, we used an iRAP system to recruit Inp54 to the plasma membrane of cells expressing either Rab5A or Rab5A_S34N. These cells were co-transfected with YFP-FKBP-Inp54 plus the plasma membrane-localizing Lyn₁₁-FKB construct. The addition of rapamycin resulted in the rapid recruitment of Inp54 to the plasma membrane (Fig. 6A, insets). As expected, macropinocytosis proceeded normally in cells expressing Rab5A, regardless of whether Inp54 was recruited or not to the membrane (Fig. 6A, top panels). However, although cytosolic Inp54 was not able to overcome the effect of dominant-negative Rab5A on macropinocytosis, the phosphatase enabled TMR-D uptake when recruited to the membrane at the time of EGF stimulation (Fig. 6A, bottom panels). Recruitment of Inp54 to the plasma membrane restored macropinocytosis to levels similar to cells expressing wild-type Rab5A (Fig. 6B). Taken together, these findings confirm that the breakdown of PtdIns(4,5)P₂ to PtdIns(4)P in the macropinocytic cup is critical for the completion of macropinocytosis. Active Rab5 plays an essential role in this process by recruiting to the macropinocytic cup 5-phosphatase activity required for sealing, exerted primarily by OCRL and Inpp5b, which are recruited through the adaptor APPL1.

The involvement of Rab5 in macropinocytosis has been known for decades, but its mode of action remained enigmatic and controversial. Some reports concluded that Rab5 operated at the plasma membrane, promoting ruffling (Barbieri et al., 1998; Lanzetti et al., 2004; Spaargaren and Bos, 1999), whereas others disputed this claim and proposed instead a role for Rab5 only after vacuole internalization (Feliciano et al., 2011; Roberts et al., 2000). Here, we demonstrate that breakdown of PtdIns(4,5)P₂ is a central requirement for the formation and/or persistence of sealed macropinosomes, a mechanism dependent on Rab5 and its effectors APPL1, Inpp5b and OCRL. These effectors may enable sealing and detachment of the nascent vacuole, or may prevent its back-fusion with the plasma membrane, which would negate the internalization event.

Our findings suggest that Rab5 exerts its effects before the scission of the nascent macropinocytic vacuole from the surface membrane. Accordingly, dominant-negative Rab5 prevents formation of the vacuole while not affecting the occurrence of membrane ruffling. The docking and likely fusion of Rab5-decorated vesicles with circular ruffles before sealing is consistent with this conclusion, as is the observation that the dominant-negative form of the GTPase is inhibitory when recruited to the plasmalemma constitutively or through heterodimerization. Using current technology it is impossible to unambiguously ascertain that Rab5⁺ vesicles in fact fuse with the surface membrane before macropinosome scission, but experiments in which fusion was precluded support this notion. Briefly, we implemented a new system whereby a dominant-negative form of NSF was acutely released into the cytosol of the test cells, obviating the detrimental non-specific effects associated with long-term expression of such inhibitory constructs. Blocking membrane fusion using this comparatively rapid manoeuvre caused a profound inhibition of macropinocytosis. Although the resolution limits of light microscopy do not yet allow for dynamic visualization of vesicle-circular ruffle fusion, these data are highly suggestive of the requirement for focal fusion of endomembranes with the plasma membrane during macropinocytosis, analogous to what has been reported for phagocytosis (Bajno et al., 2000; Braun et al., 2004). The sealing of macropinosomes is accompanied by loss of PtdIns(4,5)P₂ from their limiting membrane (Porat-Shliom et al., 2008; Welliver and Swanson, 2012), an observation we were able to verify (Fig. 4). Although firmly documented in the past, the functional implications of the disappearance of the inositide were unclear. Our data suggest a correlation between PtdIns(4,5)P₂ depletion and macropinosome sealing, although temporal and resolution limits of live-cell imaging warrant caution when assigning causation. Most pertinently, hydrolysis of the inositide on circular ruffles failed to occur when Rab5 activity was impaired (Fig. 4B), suggesting that Rab5 functions upstream of this event. Moreover, preventing the depletion of PtdIns(4,5)P₂ by enhancing its rate of synthesis, through overexpression of PIP5K, precluded macropinosome formation. These findings led us to postulate that PtdIns(4,5)P₂ hydrolysis from forming macropinosomes is essential for their scission from the membrane and/or to prevent their back-fusion with the surface membrane. Rapid PtdIns(4,5)P₂ hydrolysis accompanies endosome and phagosome formation also, and may be a common feature and requirement of all endocytic pathways.

We noted that docking of Rab5 to circular ruffles preceded PtdIns(4,5)P₂ depletion from the macropinocytic cup (Fig. 4A) and therefore postulated a role for the GTPase in the hydrolysis of the phosphoinositide. Our subsequent experiments provided a plausible mechanism: we propose that through the adaptor APPL1, Rab5 recruits Inpp5b and OCRL to the membrane, two active phosphatases capable of degrading PtdIns(4,5)P₂ to PtdIns(4)P. The adaptor, as well as the phosphatases, is recruited to nascent macropinosomes (Fig. 5A,B) and, more importantly, their depletion greatly inhibits macropinocytosis without altering membrane ruffling (Fig. 5E,I).

Dephosphorylation by Inpp5b and OCRL is not solely responsible for the elimination of PtdIns(4,5)P₂ from macropinosomes. Welliver and Swanson (2012) and Yoshida et al. (2015) had clearly demonstrated earlier that phospholipase C contributes to the process, generating diacylglycerol and Ins(1,4,5)P₃. We believe that the initial hydrolysis of PtdIns(4,5)P₂ is caused by phospholipase C, whereas the final stages of its removal, which occur at the time of sealing, are catalyzed by Inpp5b and OCRL. Alternatively, phospholipase C activity may predominate at the base of the macropinocytic cup, while the phosphatases may be most active at the singular point at which sealing occurs. According to this scenario, restoration of inositol phosphatase activity should bypass the inhibition of macropinocytosis by dominant-negative Rab5. Indeed, using an inducible heterodimerization system we found that macropinocytosis could be restored in cells expressing Rab5A_S34N by recruiting an exogenous phosphatase, Inp54, to the membrane at the time of stimulation with EGF (Fig. 6A,B).

In summary, our experiments have defined the mechanism by which Rab5 and its associated inositol 5-phosphatases OCRL and Inpp5b catalyze macropinosome sealing. The hydrolysis of PtdIns(4,5)P₂ occurs in the macropinocytic cup as a result of the recruitment and eventual fusion of early endosomes containing Rab5 and its effectors. These data serve to highlight the intricate intersection between cytoskeletal dynamics, small GTPase activity and phosphoinositide signalling that occurs during dynamic cellular uptake processes, such as macropinocytosis.

Mammalian expression vectors were obtained from the following sources: GFP-Rab5A (Roberts et al., 2000); GFP-Rab5A_S34N (35141, Addgene); CFP-Rab5A (Heo et al., 2006); CFP-Rab5A_S34N (Heo et al., 2006); RFP-Rab5A (Chen et al., 2009); mCherry-Rab5A_S34N (35139, Addgene); PLCδ-PH-GFP (Stauffer et al., 1998); PLCδ-PH-RFP (Stauffer et al., 1998); Lyn₁₁-RFP (Lee et al., 2007); GFP-PIP5Kα (Fairn et al., 2009); GFP-PIP5Kβ (Fairn et al., 2009); GFP-PIP5Kβ_KD (Fairn et al., 2009); pC₄S₁-EGFP-F_M4-FCS-hGH (Ariad Pharmaceuticals; Rivera et al., 2000); NSF_E329Q (Coppolino et al., 2002); GFP-APPL1 (Zoncu et al., 2009); Inpp5b-GFP (Williams et al., 2007); OCRL-GFP (Swan et al., 2010); Lyn₁₁-FKB (20147, Addgene); CFP-FKBP (20160, Addgene) YF-FKBP-Inp54 (Suh et al., 2006); CFP-Rab5A-CAAX (Galperin and Sorkin, 2003); CFP-Rab5A_Q67L-CAAX (Galperin and Sorkin, 2003); and CFP-Rab5A_S34N-CAAX (Galperin and Sorkin, 2003).

Primary antibodies were purchased from the following vendors: Rab5 (3547, Cell Signaling Technology), α-tubulin (T5168, Sigma-Aldrich), GM130 (610822, BD Pharmingen), EEA1 (sc-6451; Santa Cruz Biotechnology). Horseradish peroxidase (HRP), Alexa-568- and Alexa-647-conjugated secondary antibodies against mouse IgG were from Jackson ImmunoResearch Labs.

The A431 cell line was obtained from and authenticated by the American Type Culture Collection (ATCC). Before experimentation, this cell line was revalidated by assessing the expression of epithelial membrane markers (E-cadherin and β-catenin), and responsiveness to EGF. A431 cells were grown in Dulbecco's modified Eagle Medium (DMEM) containing (MultiCell, Wisent) and 10% heat-inactivated fetal calf serum, at 37°C under 5% CO₂. The cell line tested negative for mycoplasma contamination by DAPI staining.

A431 cells were plated on 18 mm glass coverslips at a concentration of 1.5×10⁵ cells ml⁻¹ in DMEM containing L-glutamine and 10% heat-inactivated FCS at 37°C and 5% CO₂. The next day, the medium was changed to Hank's balanced salt solution (HBSS) (311-515-CL; MultiCell, Wisent), and cells were serum starved for 3 h at 37°C. Cells were treated for 20 min with 200 ng ml⁻¹ EGF (Invitrogen, Thermo Fisher Scientific), in the presence of 100 μg ml⁻¹ TMR-D (Molecular Probes, Thermo Fisher Scientific). Monolayers were rinsed three times with PBS, placed in HBSS and imaged live. The cells were visualized at the indicated intervals after the addition of EGF±TMR-D (5, 10, 15, 20 min).

To distinguish sealed macropinosomes from open circular ruffles, A431 cells were serum starved for 3 h at 37°C and incubated for 2 min with 200 ng ml⁻¹ EGF and 100 μg ml⁻¹ TMR-D. Cells were then washed three times with cold 1× PBS, and stained for 1 min with 20 μM FM4-64 (Invitrogen, Thermo Fisher Scientific) at 10°C. Samples were viewed immediately by confocal microscopy. In some cases, after EGF treatment, monolayers were fixed in 3% paraformaldehyde for 10 min at room temperature. Monolayers were blocked/permeabilized in PBS containing 5% BSA and 0.1% saponin, and stained with antibody to EEA1 (1:100 dilution in blocking buffer), followed by fluorescently conjugated secondary antibody (1:1000 dilution in blocking buffer).

For transient transfection, A431 cells were plated on 18 mm glass coverslips at 1.5×10⁵ cells ml⁻¹, 16-24 h before the experiments. Lipofectamine LTX with PLUS reagent (Invitrogen, Thermo Fisher Scientific) was used according to the manufacturer's instructions. A431 cells were transfected at a 3:1 ratio using 1.5 μl Lipofectamine LTX, 0.5 μg DNA and 0.5 μl of PLUS reagent per well for 4 h in Opti-MEM medium (Gibco, Thermo Fisher Scientific). Following this, medium was changed to DMEM containing L-glutamine and 10% heat-inactivated FCS, and cells were used for experiments 16 h after transfection.

siRNAs for Rab5 isoforms A, B and C were obtained from Sigma-Aldrich, based on previously described sequences (Huang et al., 2004). ON-TARGETplus human siRNA SMARTpools targeting APPL1 (L-005138-00-0005), Inpp5b (L-021811-00-0005) and OCRL (L-01-0026-00-005) were obtained from Dharmacon, along with an ON-TARGETplus non-targeting pool (D-001810-10-20). siRNA delivery was performed using either Nucleofector Kit V (Lonza) or a Neon transfection system (Life Technologies, Thermo Fisher Scientific), according to the manufacturers' protocols. For nucleofection of Rab5a/b/c siRNA, 1×10⁶ A431 cells were resuspended in 100 μl of Nucleofector Solution T, containing 200 pmol of Rab5a/b/c siRNA, and electroporated with program X-01 on the Nucleofector I system (Lonza). Cells were used 48 h after siRNA treatment. For Neon transfection of APPL1/Inpp5b/OCRL siRNA smart-pools, A431 cells were resuspended to 5×10⁶ cells ml⁻¹, and 100 μl of suspension mixed with 200 pmol siRNA. Electroporation was then performed using two 20 ms pulses of 1400 V. After electroporation, cells were immediately transferred to DMEM containing L-glutamine and 10% heat-inactivated FCS, before seeding on coverslips at a concentration of 1.25×10⁵ cells ml⁻¹. siRNA-treated cells were used for macropinocytosis experiments 48 h after electroporation. In some cases, after 24 h, cells were transiently transfected with mammalian expression vectors as described above.

siRNA-mediated knockdown was confirmed at 48 h by either western blotting or qPCR. For Rab5A/B/C knockdown, cells were lysed in Laemmli buffer (Bio-Rad). Lysates were separated by SDS-PAGE, followed by transfer to a polyvinylidene difluoride membrane. The membrane was blocked in TBS containing 5% bovine serum albumin (BSA) and 0.05% Tween 20 for 30 min at room temperature, followed by primary antibody staining for 1 h at room temperature in blocking buffer. Primary antibody dilutions were as follows: Rab5 (1:1000); and α-tubulin (1:1000, loading control). After washing the membrane in TBS containing 0.05% Tween 20, samples were incubated 30 min at room temperature with an HRP-conjugated secondary antibody at a 1:5000 dilution. Blots were visualized using the ECL Prime Western Blot detection reagent (GE Healthcare) on film.

For validation of APPL1, Inpp5b and OCRL knockdown, RNA was isolated from siRNA-treated cells using the GeneJET RNA Purification kit (Thermo Fisher Scientific). An aliquot of 1 µg of RNA was used for single strand cDNA synthesis with the SuperScript VILO cDNA Synthesis kit (Invitrogen, Thermo Fisher Scientific). Quantitative PCR was performed using the TaqMan System (Applied Biosystems, Thermo Fisher Scientific) and TaqMan Fast Advanced Mastermix (Applied Biosystems, Thermo Fisher Scientific) on a Step One Plus Real-Time PCR thermal cycler (Step One software v2.2.2; Applied Biosystems, Thermo Fisher Scientific). The Taqman gene expression assays for the reference gene (GAPDH – ID# Hs02786624_g1) and target genes (APPL1 – ID# Hs00179382_m1; Inpp5b – ID# Hs00299982_m1; OCRL – ID# Hs00240844_m1) were duplexed in triplicate. Target gene expression was determined by quantification relative to the GAPDH reference gene and the non-targeting pool control sample (ΔΔC_T method; Livak and Schmittgen, 2001).

For Rab5 iRAP, full length human Rab5A or Rab5A(S34N) were excised from GFP expression vectors using HindIII/BamHI or XhoI/BamHI sites, respectively. These fragments were cloned downstream of CFP-FKBP using the same sites, and co-expressed in cells with the Lyn₁₁-FRB plasma membrane localizing sequence. Treatment of A431 cells with 2.5 μM rapamycin mediated the rapid translocation of FKBP-CFP-Rab5A/Rab5A(S34N) to the plasma membrane by induced heterologous dimerization of rapamycin-binding FKBP and FKB domains. After 5 min pre-treatment with rapamycin, macropinocytosis assays (TMR-D uptake) were performed as described above in the presence of rapamycin. iRAP experiments using YFP-FKBP-Inp54 and Lyn₁₁-FRB, to recruit Inp54 to the plasma membrane upon addition of rapamycin, were performed similarly.

pC₄S₁-EGFP-F_M4-FCS-hGH is a part of an in vitro reverse dimerization system (Ariad Pharmaceuticals; Rivera et al., 2000). This vector contains a CMV enhancer/promoter (C), the secretion signal sequence of human growth hormone (S), EGFP, four tandem repeats of rapalogue-binding FKBP_36M (F_M4), and the human growth hormone coding sequence (hGH) downstream of a furin cleavage signal (FCS). To create a reverse dimerization construct for NSF_E329Q, the pC₄S₁-EGFP-F_M4-FCS-hGH vector was modified by cloning. The hGH signal sequence was removed using EcoRI/XbaI digestion, and replaced with a synthetic Kozak sequence. Next, the FCS-hGH SpeI/BamHI fragment was replaced with NSF_E329Q. The resultant vector, pC₄-EGFP-F_M4-NSF_E329Q, generates a protein that undergoes spontaneous aggregation that reverses upon the addition of D/D solubilizer (635054; Takara Bio), a cell-permeant rapamycin analog that binds to the F_M4 domains (see Fig. 3E).

A431 cells were transiently transfected with pC₄-EGFP-F_M4-NSF_E329Q, as described above. Transfected monolayers were then serum-starved in HBSS for 2 h, the last hour of which included treatment with 500 nM D/D solubilizer or vehicle (DMSO) control. Macropinocytosis assays were performed as above. The reverse dimerization of NSF_E329Q was confirmed in transfected monolayers by assessment of the effect of pC₄-EGFP-F_M4-NSF_E329Q expression±D/D solubilizer on Golgi morphology, using immunostaining. Briefly, 1 h after treatment with 500 nM D/D solubilizer or vehicle (DMSO) control, transfected monolayers were fixed in 3% paraformaldehyde for 10 min at room temperature. Monolayers were blocked/permeabilized in PBS containing 5% BSA and 0.1% saponin, and stained with antibody to GM130 (1:500 dilution in blocking buffer), followed by fluorescently conjugated secondary antibody (1:1000 dilution in blocking buffer).

Confocal images were acquired using a Yokogawa CSU10 spinning disk system (Quorum Technologies). Images were acquired using a 63×/1.4 NA oil objective or a 25×/0.8 NA water objective (Zeiss), as indicated, with an additional 1.5× magnifying lens. For live experiments, cells were maintained at 37°C using an environmental chamber (Live Cell Instruments). Routine analyses were performed using Volocity software (Perkin Elmer) or Fiji (Schindelin et al., 2012).

For some colocalization analyses, Volocity software was used to calculate positive product of the differences of the mean channels (Li et al., 2004), which were then overlaid on merged images for visualization. Alternatively, Manders' overlap coefficients were calculated in Volocity for individual Rab5A⁺ endosomes.

For fluorescence intensity calculations, background-subtracted intensities per unit area for expressed fluorescent protein constructs or endogenous proteins (immunofluorescence) were measured in Volocity software. Ratios were calculated comparing relative intensities of transfected markers in the macropinosome membrane compared to plasma membrane, as indicated in the text. All statistics were calculated using GraphPad Prism software (GraphPad Software).

For the quantification of ruffling, a Java plug-in (Ruffle_Analysis.java; www.doi.org/10.6084/m9.figshare.12349967) was written for Fiji to analyse background-subtracted ruffling movies of EGF-treated A431 cells expressing Lyn₁₁-RFP. Movies were recorded for 2 min pre-EGF and 8 min post-EGF treatment (200 ng ml⁻¹). For a selected region of the ruffling plasma membrane, the plug-in calculates a summed squared pixel difference within a user defined region of interest (ROI), between frame_n+1 and frame_n, for every frame pair in a background-subtracted time-lapse (background subtracted ΣΔ_pixel²). The code first takes each individual pixel value within a square/rectangular ROI and subtracts it from the corresponding pixel value in the next frame (Δ_pixel). Next, the differences are squared (Δ_pixel²) to eliminate the cancelling effects of positive versus negative changes in fluorescence intensity. Finally, these values are then summed generating a (ΣΔ_pixel²) value for each frame, starting with frame_n+1. For normalization, background-subtracted ΣΔ_pixel² values were normalized to the mean pre-EGF values. For each cell, the s.d. of the background-subtracted ΣΔ_pixel² for the pre-EGF period was compared to that of the post-EGF period, as a metric of the continuous intensity changes/pixel that are seen in an ROI during plasma membrane ruffling. Additionally, temporal projections of A431 cells pre- and post-EGF treatment were generated in Fiji using the included Temporal Color Code plug-in (Daste et al., 2017) and cool LUT. Data calculations and normalizations were performed using Microsoft Excel software. All statistics were calculated using GraphPad Prism software.

Experiments were, for the most part, in vitro imaging determinations of individual cells. As such, samples were assigned to groups according to specific experimental treatments (control versus experimental group) applied by the observer/investigator. The number of individual experiments and the number of determinations per experiment were selected to attain an estimate of the variance compatible with the statistical tests used, primarily Student's t-test. Each type of experiment was performed a minimum of three separate times (biological replicates) and a minimum of ten individual event determinations (equivalent but not identical to technical replicates). Data were tested for normality, and appropriate testing was applied. No data were excluded as outliers.

We thank Dr Philip Ostrowski for writing the ruffle quantification Fiji plug-in, and Dr Michal Bohdanowicz for the construction of the Rab5A and Rab5A_S34N iRAP constructs. The plasmid for the reverse dimerization system was kindly provided by Dr Victor Rivera (Ariad Pharmaceuticals, Cambridge, MA, USA).

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/134/7/jcs252411/