A plasma membrane template for macropinocytic cups

1Macropinocytosis provides cells with an efficient way of taking up large volumes of medium into intracellular vesicles, from which they can extract nutrients, antigens and other useful molecules. It is an ancient process, used for feeding by amoebae, but one that is important for a wide spectrum of human biology, including uptake of drugs, and large-scale sampling of extracellular medium for antigens by immune cells. It has also been hijacked by pathogens as a major route of entry. Recent data suggest that macropinocytosis is a principal and widely used method for sustaining the excessive metabolic demands of cancer cells and may be implicated in the spread of neurodegenerative disease within the brain.

2Considering its biological importance, macropinocytosis is not well understood. Macropinosomes form from cup-shaped extensions of the plasma membrane, often known as circular ruffles, which are extended by actin polymerisation. The leading rims of these ruffles must be driven outwards to enclose liquid – often for a very significant fraction of the cell’s diameter – but the base must be held static. The resulting cups can be several microns in diameter, and eventually close by constriction of their rim, with membrane fusion producing an endocytic vesicle. Here we address a critical and mysterious question about this process - how do cells organize actin to polymerize in a ring and so form the walls of the cup?

3In the closely related process of phagocytosis, in which solid particles are taken up, it is proposed that cup formation is guided by engaging receptors with the particle to be engulfed in a zippering process. However, macropinosomes take up fluid and so cannot use a particle as a template in this way. Nor is there any known equivalent in macropinocytosis of the coat that organises clathrin-mediated endocytosis. Thus it appears that macropinocytic cups must form by a self-organizing process within the actin polymerization machinery and its regulators.

4The dynamic actin that polymerises around macropinocytic cups is probably initiated by a number of nucleators, including both formins, such as ForG, which is needed for the basal part of phagocytic cups in Dictyostelium, and the Arp2/3 complex, which produces dendritic structures, like the actin that drives pseudopods. Assembly of Arp2/3 based actin is controlled by the WASP family of nucleation promoting factors; the two family members that act at the plasma membrane are WASP and SCAR/WAVE (hereafter called SCAR). WASP is important for actin polymerisation during clathrin-mediated endocytosis, and SCAR, acting in a five-membered complex, for the formation of pseudopods. It is not known which is responsible for macropinocytosis.

5Ras and phosphoinositide signalling help organize the cytoskeleton for macropinocytosis and phagocytosis. There is evidence that Ras activity stimulates macropinocytosis in both mammalian and Dictyostelium cells, and macropinocytic cups are associated with an intense domain, or ‘patch’, of PIP3, which is essential for their function.

6In macrophages, which often evolve macropinocytic cups from linear ruffles, it has been suggested that ruffle circularisation creates a diffusion barrier in the membrane leading to intensified PIP3 signalling and a domain of PIP3 in the centre of the circular ruffle. This domain then drives the further progression of the macropinocytic cup.

7In axenic strains of Dictyostelium, which grow efficiently in liquid medium, macropinocytosis is massively up-regulated due to mutation of the RasGAP neurofibromatosis-1 (NF1). These strains are thus an excellent starting point for research into the organising principles behind macropinocytic cup formation. Here we examine macropinocytosis with unprecedented 3D detail using lattice light sheet microscopy, and map the spatial and temporal control of actin regulators such as SCAR and WASP with respect to signalling molecules including PIP3 and active Ras. This leads us to propose a new and general hypothesis for the formation of cups from the plasma membrane.

8To determine whether macropinosomes form in Dictyostelium by circularization of linear ruffles, as reported for macrophages, we used lattice light sheet microscopy, which allows unparalleled high-resolution imaging of light-sensitive and dynamic cells over prolonged periods. In axenic cells expressing an F-actin reporter, three types of large F-actin structure are routinely detected: macropinocytic cups, which predominate, pseudopods and basal waves.

93D movies show that the majority of macropinocytic cups initiate by splitting of existing ones (62%; n = 152, Figure 1E). Splitting occurs by a variety of routes, including: simple division in the middle; detachment of a small ruffle that grows into a new macropinocytic cup (Figure 1A, Video 1); and abortive fragmentation of a parental macropinocytic cup into multiple daughter cups. We examined a reporter for active Rac in some of these movies and found that it is spatiotemporally associated tightly with F-actin in this morphological process (Figure 1B, Video 2).

10The remaining macropinocytic cups form de novo, expanding from places where no previous F-actin activity was detected. In more than 90% of cases the initiation is close to the base of the cell, even though most mature macropinocytic cups are present on the top, and are commonly described as 'crowns'. In the example illustrated in Figure 1C (Video 3), the parental ruffle first emerges close to the substratum (t = 0, white arrow) and cannot unequivocally be classified as either pseudopod or circular ruffle. The ruffle then quickly sweeps to the top of the cell, during which time it grows in size and splits several times, to produce multiple full-grown macropinocytic cups. As with splitting macropinocytic cups, the F-actin in de novo macropinosome cups is closely associated with signalling molecules, as illustrated by active Ras in Figure 1D (Video 4) and discussed later. Circular ruffles can persist on the cell surface for prolonged periods before either closing successfully or regressing back into the cell body. Closure of the cup can be quite abrupt and often appears to involve the inward collapse of the rim (Figure 1F, Video 5).

11The other large F-actin projections in growing Dictyostelium cells are pseudopods. These are distinguished from macropinocytic circular ruffles by their shape, which is convex instead of concave. De novo pseudopods also initiate close to the substratum (Figure 1G, white arrow and Video 6) and expand steadily to their full size. Pseudopods are surprisingly rare in growing axenic cells, accounting for less than 5% of all large F-actin structures.

12Finally, we could follow the enigmatic actin waves that move across the basal surface of vegetative cells. These waves also generally originate from existing ruffles by splitting. When the splitting ruffle in Figure 1H (Video 7) contacts the substratum, it initiates an actin wave that spreads across the entire footprint of the cell, becoming so dominant that other large F-actin structures are suppressed and flattening the cell into a smooth bell-shape.

13These observations show that the large macropinocytic cups of axenic Dictyostelium cells generally form by splitting or by expanding de novo from a small focus, as in fibroblasts, rather than by circularization of linear ruffles. The smaller macropinocytic cups of wild-type cells (see later) also more normally form de novo or by splitting, rather than by circularization.

14We tested the involvement of PIP3 and Ras signalling in macropinocytosis using an isogenic set of mutants in which we measured both fluid uptake and growth in liquid medium (Supplementary material, Table 1). Either increased or decreased PIP3 levels (PTEN and PI3-kinase mutants) are deleterious to fluid uptake and growth in liquid medium, as expected from earlier work. Two independent RasG- mutants are substantially impaired in growth in liquid medium, as previously described, but contrary to the earlier report, both are also defective in fluid uptake. Compensation by other Ras proteins and genetic background differences may account for the discrepancy. RasC null cells have no growth defect and a lesser defect in fluid uptake, while RasS (not tested here) may also contribute to macropinocytosis. Notably, we confirm that the Arp2/3 activator, SCAR, is required for efficient fluid uptake.

15The SCAR complex is mostly cytosolic and basally inactive, but when recruited to the plasma membrane it causes actin polymerization through the Arp2/3 complex. A GFP reporter tagged at the HSPC300 subunit accumulates at sites of actin polymerization. To confirm that this accumulation signifies the presence of activated SCAR complex, we correlated the signal with the expansion of pseudopods, using this as a proxy for actin polymerization. The results clearly show that the reporter is recruited during expansion phases but lost in stalls (Figure 2A,B). This correlation holds true globally: SCAR reporter intensity along the membrane correlates well with the local membrane expansion speed (Figure 2C) as all pixels with high SCAR reporter are associated with positive instantaneous membrane speed. Note that the small set of pixels with very high membrane speeds and no SCAR (Figure 2C, red arrow) are due to blebs.

16In images recorded in 3D, the reporter reveals a thin, sometimes broken ring of active SCAR around the lip of macropinocytic cups (Figure 2D and Video 8). The presence of a ring could not be predicted by imaging the actin cytoskeleton itself, as actin filaments are distributed rather uniformly throughout the cup (Figure 2H). These circular SCAR structures are not seen in pseudopods, where 3D images show the same discrete blocks of SCAR as in 2D images (Figure 2E).

17The discovery of these remarkable rings immediately raises the question of how individual SCAR molecules are coordinated to maintain the ring shape. We visualised PIP3 using a double reporter that co-expresses the PH-domain of CRAC fused to RFP. This revealed a second remarkable feature of SCAR rings: they follow the edges of intensely stained domains, or ‘patches’ of PIP3. In all macropinocytic cups examined, of whatever size, the concave cup contains a patch of PIP3 and SCAR is present as a ring around this patch, without detectable recruitment to its centre (Figure 2F and G and Video 9).

18This was confirmed in a larger sample by measuring fluorescence intensity at the centre and rim of 17 macropinocytic cups from nine cells rendered in 3D (Figure 3A–B). All cup centres contained high levels of PIP3, but SCAR consistently followed the rim of the cup with the mean fluorescence of the SCAR reporter significantly higher than cytosolic background (p<0.01), while signal at the centre of the PIP3 patch was not statistically different from the cytosolic background. This is also clear in 2D images, but is easily overlooked as the narrow SCAR ring appears only as tiny puncta in the cross sections obtained from confocal microscopy.

19We further tested the spatial relation between PIP3 and SCAR in two ways not requiring visual recognition of macropinocytic cups. In the first, a number of growing axenic cells was analysed as follows. Membrane areas with fluorescence intensity of the PIP3 reporter greater than cytosolic background plus one standard deviation were defined as PIP3 patches, and the associated SCAR signal was measured. In all cases SCAR is consistently and significantly enriched at patch edges (p<0.01), and never at their centres (Figure 3C–D). In the second test, the fluorescence intensity of the SCAR and PIP3 reporters was extracted from the circumference of a number of growing cells and the results plotted as a 2D histogram (Figure 3E). In pixels with high PIP3 signal, the SCAR signal is low, and conversely in pixels with high SCAR, PIP3 is low. This method cannot show whether high SCAR and PIP3 pixels are adjacent, but it does confirm that SCAR and PIP3 do not co-localise but instead are anti-correlated.

20The complete lifetime of a de novo macropinocytic cup is shown in Figure 4A (Video 10 shows another example). The PIP3 patch first becomes visible at t = 1 and this sub-micron sized patch is already flanked by puncta of SCAR. As the patch of PIP3 grows the SCAR puncta remain dynamically associated with its edge right up to closure of the macropinocytic cup, after which the SCAR signal quickly disappears and the vesicle is internalised. The SCAR is not detected at the centres of the patches above background.

21This is also shown in a kymograph of the membrane pixels of a single cell as it makes several macropinosomes (Figure 4B). The SCAR signal, though sometimes weak, can be traced along the edge of the PIP3 patches from the start of a patch to its abrupt loss at invagination. Combining the data from several macropinocytosis events confirms this continuous association (Figure 4C–E). Thus despite changes in size and shape of PIP3 patches, SCAR remains associated with their edges, and only their edges, throughout the macropinocytic cup lifetime.

22It seemed possible that as a rule of cytoskeletal organization in Dictyostelium, PIP3 patches always recruit SCAR to their edges. Four other examples of PIP3 patches support this: two from growing cells, and two from starved cells, which are highly migratory, chemotactically sensitive, and morphologically very different from growing cells:

23During phagocytosis Dictyostelium cells make a PIP3 patch where they contact the particle to be ingested. In the yeast case shown in Figure 5A, SCAR is recruited to the edges of the PIP3 patch, and not the centre, while a 3D view reveals a full ring of SCAR around the rim of the phagocytic cup. Indeed, a clear ring of SCAR could be detected in all such cases, provided expression of the SCAR reporter was low enough to avoid excessive background fluorescence.

24Basal waves have a core of PIP3 surrounded by a ring of F-actin, and again, the basal PIP3 patches are invariably surrounded by a ring of SCAR (Figure 5B). Basal waves are favourable for microscopy, and we found that WASP is also excluded from PIP3 patches and forms a ring around them, though weaker and less coherently than SCAR (Figure 5C). Similarly, WASP forms rings at the edges of PIP3 patches of normal macropinocytic cups, again more weakly than SCAR (Figure 5D). The remaining Arp2/3 activator in Dictyostelium, WASH, does not associate with PIP3 patches (Figure 5—figure supplement 1).

25During chemotactic aggregation, developing cells form small chains and streams with strong head-to-tail adhesions between them and PIP3 patches in their front. These patches are invariably surrounded by a ring of SCAR. In the example shown in Figure 5E–G, a cell strongly expressing reporters is situated between two poorly expressing cells. The strongly expressing cell forms a PIP3 contact patch, with SCAR present as a clear ring and excluded from the centre.

26Cells respond to cyclic-AMP by making PIP3, initially homogenously and then, after about a minute, in patches at the membrane. In the low light conditions required for time-lapse imaging, the SCAR signal is weak, but where detected, it is clearly at the edges of the PIP3 patches (Figure 5H–J and Video 11). These patches have sometimes been regarded as new pseudopods (for example), but many become concave and close to engulf a drop of medium, indicating that the cell is performing macropinocytosis, not a chemotactic response.

27PIP3 is largely made by Ras-activated PI3-kinases. We confirmed that a patch of activated Ras exactly coincides with each PIP3 patch (Figure 6A). Similarly, plots of intensity, pixel-by-pixel, show exceptional correlation between the Ras and PIP3 signals (Figure 6B). Thus PIP3 patches have a matching patch of activated Ras, which could sustain them by activating PI3-kinase.

28Similarly, the Ras/PIP3 patch overlaps a patch of active Rac1, as detected by the CRIB domain (Figure 6C). Rac1 is an upstream regulator of SCAR, and has been implicated in macropinocytosis. However, its broad distribution cannot simply account for the much narrower SCAR ring. Alternatively, Rac1 may define a permissive area where SCAR can be activated or other Rac isoforms may be involved, such as RacB, RacC or RacG. No specific markers exist for their activated state, but the RacG molecule itself is modestly enriched at the rim of phagocytic cups.

29It has been proposed that PIP3 patch formation requires a positive feedback loop where PIP3 activates Ras. We tested this by genetically manipulating PIP3 levels. A mutant without Ras-activated PI3-kinases and producing only 10% of wild-type PIP3 levels still forms patches of activated Ras at a similar frequency to parental cells (Figure 6D). The SCAR signal in confocal cross sections of macropinocytic cups is too small for an accurate comparison of SCAR ring formation between mutants and therefore we used basal waves as a proxy for macropinocytic cups. Ras patches on the basal surface of PI3-kinase null cells still exclude SCAR from their centre and recruit a peripheral ring of SCAR as normal, albeit more weakly than in parental cells (Figure 6E–F). Conversely, when PIP3 levels are increased 10-fold by eliminating the PTEN phosphatase, the activated Ras domains do not expand correspondingly (Figure 6G) and remain associated with rings of SCAR (Figure 6H). Thus Ras, rather than PIP3, is the primary determinant of patches and SCAR rings.

30It has also been proposed that PIP3 patch formation requires an enclosing circular ruffle to act as a diffusion trap. We tested this by controlled use of the actin inhibitor latrunculin-A to inhibit ruffle formation. Latrunculin-A at 1 µM leaves some actin polymerisation intact, and at 5 µM abolishes all visible actin filaments, resulting in spherical cells (Figure 7A). Neither treatment abolishes the patches of PIP3, which become larger but less numerous, with a fluorescence intensity not significantly different from control cells (Figure 7B–E). We tested whether the sharp boundaries of patches are affected by latrunculin-A by measuring the intensity across the edges of more than 30 patches for each condition, (Figure 7F–H). It is clear that latrunculin-A has little effect on the sharpness of the patch, suggesting that a diffusion barrier is not required to maintain its strong spatial coherence.

31In summary, a circular ruffle is not essential to create signalling patches, which appear to largely depend on Ras, with PIP3 playing a secondary though still important role.

32To test whether Ras plays an instructive role in macropinocytic cup morphogenesis, directly regulating their formation and size rather than acting as a remote trigger or passive participant, we examined the effect of genetically increasing Ras activity. The RasGAP NF1, encoded by the Dictyostelium axeB gene, is present in the wild-isolate NC4 but inactivated in its axenic derivatives, including the standard Ax2 used here. We found that macropinocytic cups in NC4 maintain exactly the same organization as in Ax2, with a central patch of PIP3 surround by a ring of SCAR, but are much smaller and shorter-lived and often arise de novo (Figure 7I,J and Videos 12, 13 and 14). To confirm that macropinocytic cup size is controlled by NF1 we compared an isogenic NF1 knock-out with its parent (DdB; also derived from NC4;). Cells from each strain were cultivated for 48 hr in axenic medium to maximally induce the rate of macropinocytosis. Under these conditions the axeB-null cells that have lost NF1 make significantly larger macropinocytic patches compared to cells from the parental strain (p<0.01, Figure 7K–L).

33The effect of the loss of NF1 on basal PIP3 patches (basal actin waves) is equally striking. Basal PIP3 patches are prevalent in axenic laboratory strains, especially during early starvation, but absent from all wild-type strains tested (Figure 7—figure supplement 1 and compare Videos 15 and 16). axeB knockout cells that have lost NF1 form abundant basal PIP3 patches, but their wild-type parent does not (Figure 7—figure supplement 1D). Thus the intensity of Ras signalling governs the size and frequency of SCAR rings in macropinocytic cups and basal waves, showing that Ras must play an instructive role.

34We therefore propose that Ras patches, assisted by PI3-kinase and Rac, cause macropinocytic cup formation by recruiting rings of SCAR/WAVE complex to their edge.

35Macropinosomes develop from cup-shaped projections of the plasma membrane, whose walls are driven outwards by actin polymerization. They contain a central patch of activated Ras and PIP3 throughout their life and we find that in Dictyostelium, this patch is invariably associated with a ring of active SCAR at its edge. We propose that this ring of active SCAR is recruited by the signalling patch and drives a hollow ring of F-actin to extend the walls of the macropinocytic cup.

36A possible alternative mechanism comes from immune cells, which make abundant linear ruffles. These occasionally fold back to form circular ruffles, which have been described as diffusion traps that can intensify signalling within them, leading to the formation of a patch of active Ras and PIP3,. In this model, PIP3 patches form as a consequence of circular ruffle formation, rather than as a cause of it. Despite the evidence that sharply curved membrane areas such as those present at the leading edge of lamellipods can act as a diffusion barrier, this idea does not easily extend to Dictyostelium, where linear ruffles are much less common, and the central PIP3 patch of macropinocytic cups can still form when ruffle formation is inhibited. However it remains possible that a diffusion barrier forms by a ruffle-independent mechanism, for example by septin-like molecules, or perhaps by cross-linking components within the patch. Further, Dictyostelium patches become larger when Ras signalling is increased by NF1 inactivation, showing that Ras plays an instructive part in their formation.

37PIP3 patches are coincident with patches of activated Ras, which presumably support them by activating PI3-kinase, and also of activated Rac. Previous work suggest that patches are self-organising structures, which can form independently of input from G-protein coupled receptors and are likely dependent on positive feedback loops between their components. Our results argue against an essential role for feedback from PIP3 to Ras, because activated Ras patches can form independently of type-1 PI3-kinases and are still able to recruit SCAR to their edges, albeit less efficiently than when PI3-kinases are present. Thus it appears that the kinetics that lead to patch formation must lie largely within the compass of the GEFs and GAPs activating and inactivating Ras.

38We can only speculate on how SCAR is recruited to the periphery of Ras/PIP3 patches. One possibility is that SCAR and Arp2/3 are preferentially recruited by newly synthesised F-actin produced by formins, which might therefore be the initial actin nucleator to be recruited. However, this seems unlikely in the light of recent work showing that ForG contributes to the base of the macropinocytic cup, but seemingly not to the extending lip. Alternatively, SCAR might be moved to the periphery of Ras/PIP3 patches, perhaps by myosin-1 motors. Early work showed that myosin-1 is genetically important for macropinocytosis in Dictyostelium. In support of the genetic evidence, myosin-1 isoforms are recruited to macropinocytic cups in both Dictyostelium and Acanthamoeba, most likely due to their affinity for PIP3. The PIP3-binding MyoE and MyoF are recruited in the centre and MyoB at the periphery of macropinocytic cups, forming a striking ‘bull’s eye’ pattern. In such a scenario, CARMIL may provide the link between myosin-1 and SCAR.

39Our work also has implications more specific to Dictyostelium biology. First, the basal actin waves, which give a valuable window into actin dynamics, appear to be formed as a consequence of the loss of NF1 in standard laboratory axenic strains. Knowing this should allow for better manipulation of these waves and for modelling to take account of their underlying need for activated Ras. Second, we consider that all patches of PIP3 and activated Ras are related by their common recruitment of SCAR to their periphery and are therefore likely to organise circular rings of actin polymerization, rather than the solid blocks characteristic of pseudopods. Therefore, the proposed role of these patches in chemotaxis, where they have been mistaken for pseudopods, needs to be re-evaluated.

40In summary, our work suggests a general hypothesis for the formation of cupped actin structures: that these structures arise from a ring of actin polymerization formed by recruiting actin nucleators to the periphery, but not the centre, of self-organizing patches of intense Ras and PIP3 signalling. This hypothesis suggests many new lines of experimentation.

41The following Dictyostelium discoideum strains were used: Ax2 (R. Kay lab strain), NC4 (from K. Raper, obtained via P. Schaap), DdB (from M. Sussman, obtained via D. Welker), NC66.2 (from D. Francis), Ax3 (R. Chisholm laboratory strain, obtained via Stock Center) and Ax4 (W. Loomis laboratory strain, obtained via Stock Center). Axenic strains were cultured in Petri dishes under HL5 medium (Formedium, Hunstanton, UK) using standard methods. Non-axenic strains were cultivated on SM agar plates with a lawn of live Klebsiella pneumoniae and where necessary washed free of bacteria by repeated low-speed centrifugation from KK2 (20 mM KH2PO4/K2HPO4, 2 mM MgSO4, 0.1 mM CaCl2, pH 6.2) (for detailed protocols of these standard techniques, see and dictybase.org/techniques/). Mutant strains, all in the parental Ax2 (Kay) background, are listed in Supplementary Material, Table 1.

42Fluid uptake was measured using TRITC-dextran and flow cytometry. Cells were grown on bacterial lawns, washed free of bacteria, resuspended in HL5 with antibiotics, 50 µl aliquots were distributed into 96 plates and allowed to adapt for about 18 hr, until macropinocytosis was maximally up-regulated. TRITC-dextran was added to 0.5 mg/ml in HL5 to the wells and the cells incubated for various times, after which the TRITC dextran was removed, the cells washed once and uptake terminated with ice-cold, 5 mM NaN3, which also detaches the cells. Fluorescence in individual cells was then measured by flow cytometry, and the rate determined while uptake was linear with time (first 45–60 min).

43Single and dual expression vectors were used for all experiments. Specifically, the following vectors were used: plasmid pDM1219 - expression of mCherry-LimEΔcoil (residue 1–145 of Dd LimE), pDM767 - dual expression of HSPC300-GFP and PH-CRAC-mRFPmars (residue 1–126 of Dd DagA), pDM1492 - dual expression of mCherry-RBD-Raf1 (residue 1–134 of Hs Raf1) and PH-PkgE-mCherry (residue 1–100 of Dd PkgE), pDM1383 - dual expression of HSPC300-GFP and mCherry-RBD-Raf1 and pDM1424 - dual expression of HSPC300-GFP and PH-PkgE-mCherry.

44The act6 promoter that drives the resistance marker on the expression vectors is not active when cells are cultivated using bacteria as a food source. Therefore, this promoter was replaced by the coaA promoter (bp −293 to bp −1 relative to the start codon of coaA) for those vectors that were used to transfect non-axenic, wild-type cells.

45Transfection of non-axenic cells was performed as follows: 5 × 106 cells were harvested from the feeding front of an SM agar plate, washed once in H40 buffer (40 mM HEPES/KOH pH 7.0, 1 mM MgCl2), and resuspended in 100 μl H40 buffer. Cells were mixed with 5 μl miniprep DNA (~0.5–1 μg total) and put on ice. Cells were then electroporated with two square waves of 350 V, 8 ms, 1 s apart using a Gene Pulser Xcell (Biorad) and immediately transferred to a Petri dish with SorMC buffer (15 mM KH2PO4, 2 mM Na2HPO4, 50 μM MgCl2, 50 μM CaCl2, pH 6.0) supplemented with live Klebsiella pneumoniae at an OD600 of 2. Selection marker was added after 5 hr (10 μg/ml G418 or 100 μg/ml hygromycin).

46Lattice light sheet microscopy 3D images were acquired as described, using a massively parallel array of coherently interfering beams comprising a non-diffracting 2D optical lattice as light sheet illumination focused by 0.65 NA objective for excitation (Special Optics). This creates a coherent structured light sheet that can be dithered to create uniform excitation in a 400 nm thick plane across the entire field of view determined by the length of the light sheet. In order to obtain the array of lattice light sheet, a binary spatial light modulator (SXGA-3DM, Forth Dimension Displays) is placed conjugate to the sample plane, and a binarized version of the desired structured pattern at the sample is projected on the display. In the time-lapse dithered mode, 3D stacks were acquired either by moving the detection objective (Nikon, CFI Apo LWD 25XW, 1.1 NA, 2 mm WD), which is synchronized with the scanning galvo mirror, or moving the sample by fast piezoelectric flexure stage (Physik Instrumente, P-621.1CD) with 100 ~150 z planes, to have about 20 µm in z axis with respect to the detection objective. Exposure time was 5 or 10 msec per plane, for a total exposure time of ~1 s for one 3D stack and a 1 s pause was added between each time point to have time series data. Raw data was deconvolved via a 3D iterative Lucy-Richardson algorithm in Matlab (The Mathworks, Natick, MA) utilizing an experimentally measured point spread function.

47Spinning disk microscopy was performed on an Andor Revolution system with a Yokogawa CSU spinning disk confocal unit. The microscope was fitted with a 1.49 Plan Apo 100x oil immersion objective and an additional 1.2x magnification lens. GFP and mCherry signals were separated by a Tucam beam splitter and detected using two Andor iXon Ultra backlit EMCCD cameras with 16 µm pixel size. Z-scans were performed with the 1.5x optovar in place using 70 ms exposure per frame and a Z-spacing of 0.19 µm. Typically, 80 frames were collected from each camera in a total of 8 s.

48TIRF microscopy was performed using a Nikon N-STORM system fitted with a 1.49 Plan Apo 100x oil immersion objective and the 1.5x optovar in place. GFP and mCherry fluorescence signals were recorded sequentially on an Andor iXon Ultra backlit EMCCD camera.

49Confocal microscopy was performed on a Leica SP8 system using a 1.4 NA plan apo oil immersion objective and GFP/mCherry fluorescence was detected using two HyD detectors. All microscopy was performed at room temperature.

50General image handling, such as brightness/contrast adjustments and generation of kymographs was done using ImageJ (NIH). 3D cellular fluorescence images were generated as follows. A Z-stack was recorded on a spinning disk microscope using previously indicated settings. The dataset was deconvolved with Huygens Professional software (Scientific Volume Imaging) using a calculated point spread function. The images presented are maximum intensity projections of the deconvolved dataset.

51Correlation between speed and membrane fluorescence intensity was analysed using Quimp11 (www.warwick.ac.uk/quimp). Identification of membrane pixels and measuring their fluorescence intensity was done using a custom-written MATLAB (The MathWorks) script (Supplementary file 1 and Source code 1).

52Image sets that were used for quantification were taken from at least two independent transfections. Only those cells with very low HSPC300-GFP expression were included for analysis, as overexpression dramatically reduces image contrast. For the quantification of SCAR fluorescence on the edge and centre of macropinocytic cups a paired 2-tailed T-test was used in Figure 3B and D and a 2-tailed T-test was used in Figure 6F.

53All lattice light sheet microscopy movies (1–7 and 12–13) show a maximum intensity projection of the fluorescence intensity. The F-actin marker LimEΔcoil is used in all images unless otherwise specified. Images were deconvolved using a custom-written Richardson-Lucy algorithm. The maximum intensity projection was generated using Huygens software. Indicated time is in the min:sec format.