The physiological regulation of macropinocytosis during Dictyostelium growth and development

Macropinocytosis, first described in the 1930s (Lewis, 1931), is a process of large-scale, non-specific fluid uptake carried out by a wide variety of cells. Actin-driven protrusions from the plasma membrane form cup-shaped circular ruffles that can be several microns in diameter. When a ruffle closes, it engulfs and delivers extracellular material to the cell interior in macropinosomes. Macropinosomes proceed through the endocytic system where their contents can be broken down by digestive enzymes and useful metabolites extracted (Buckley and King, 2017; Bloomfield and Kay, 2016; Swanson, 2008).

In the immune system, dendritic cells and macrophages use macropinocytosis to sample environmental antigens for presentation to B and T cells (Sallusto et al., 1995; Norbury et al., 1995). Certain bacteria and viruses can utilise macropinocytosis to invade host cells (Marechal et al., 2001; Nanbo et al., 2010; Hardt et al., 1998), while other bacteria stimulate macropinocytosis to promote toxin internalisation (Lukyanenko et al., 2011). Prions and neurodegenerative protein deposits also exploit macropinocytosis to invade new host cells (Magzoub et al., 2006; Fevrier et al., 2004; Münch et al., 2011; Falcon et al., 2015). Tumour cells can maintain a high rate of macropinocytosis (Lewis, 1937), with Ras-activated cancer cells obtaining a substantial part of their nutrition in this way (Commisso et al., 2013).

Considering its widespread importance, the basic biology of macropinocytosis is poorly understood. It has been studied most intensively in tissue culture cells, particularly macrophages, although genetic screens have also been performed in Caenorhabditis elegans (Fares and Greenwald, 2001) and Dictyostelium discoideum (Bacon et al., 1994). Dictyostelium in particular has great potential as a model because of the high constitutive rate of macropinocytosis maintained by cells in the right circumstances, and because the evolutionary distance from mammalian cells should allow conserved core features to be discerned.

The high rate of macropinocytosis by standard axenic strains of Dictyostelium used in the laboratory is due to deletion of the RasGAP NF1 (Bloomfield et al., 2015). This mutation allows cells to grow in nutrient-containing media without a bacterial food source (hence axenic). Wild isolates also perform macropinocytosis, although the rate of fluid uptake is too low to allow growth in the standard media used with laboratory-adapted axenic strains. These strains can, however, grow in medium supplemented with additional nutrients (Maeda, 1983; Bloomfield et al., 2015).

Axenic strains form frequent large macropinosomes, which shrink and concentrate their contents once they have been internalised by the cell. The macropinocytic cups are organised around intense patches of active Ras, Rac and plasmanylinositol (3,4,5)-trisphosphate (PIP3) (Hoeller et al., 2013; Parent et al., 1998; Veltman et al., 2016) [note that in Dictyostelium PIP3 is a plasmanylinositide, rather than a phosphatidylinositide (Clark et al., 2014)], with SCAR/WAVE and WASP localised to their periphery (Veltman et al., 2016). SCAR/WAVE and WASP activate the Arp2/3 complex to polymerise actin and form the walls of the macropinocytic cup, which is also known as a crown or circular ruffle. The base of the cup appears to be supported by actin polymerisation driven by a Ras-activated formin (Junemann et al., 2016).

The rate of fluid uptake through macropinocytosis by axenic cells is regulated by environmental factors, principally whether the nutrient source for the cells is growth media or bacteria (Kayman and Clarke, 1983; Aguado-Velasco and Bretscher, 1999), and their developmental state (Maeda, 1983; Katoh et al., 2007). Macropinocytosis is additionally affected by the stage of the cell cycle and the concentration of bacterial peptone in the medium (Maeda, 1988), as well as the incubation temperature and the pH (Maeda and Kawamoto, 1986). For certain mutants, fluid uptake is dependent upon whether cells are attached to a surface or in shaking suspension (Novak et al., 1995).

Fluid uptake by standard axenic strains of Dictyostelium, such as Ax2, is almost entirely due to macropinocytosis (see Discussion) and can be accurately measured by following the uptake of fluorescent dextran as a fluid-phase marker (Kayman and Clarke, 1983; Thilo and Vogel, 1980; Hacker et al., 1997). We have developed a high-throughput assay to measure macropinocytosis in Dictyostelium, identified useful inhibitors and sought to better understand how macropinocytosis is physiologically regulated during the switch between macropinocytic and phagocytic feeding and the growth-to-development transition.

Macropinocytosis accounts for more than 90% of fluid uptake by axenic strains of Dictyostelium, and can therefore be followed by measuring fluid uptake (Hacker et al., 1997). However, existing methods based on processing individual cell pellets after uptake of fluorescent dextran are relatively low throughput. We therefore developed a high-throughput assay that used flow cytometry to measure TRITC–dextran uptake. The assay is performed in 96-well plates and, after loading with TRITC–dextran, the cells are washed in situ by ‘dunk-banging’ and detached using sodium azide (Glynn and Clarke, 1984) (Fig. 1A), which also prevents exocytosis of internalised dextran (Fig. 1B). Plates are analysed by flow cytometry using a high-throughput sampling attachment to load the flow cytometer, and subsequent analysis is performed with Flowjo, which easily distinguishes Dictyostelium cells from beads and bacteria, but not yeast (Fig. 1C). An advantage of flow cytometry is that the fluorescence of internalised TRITC–dextran, a pH-insensitive fluid-phase marker, can be determined for single cells (Fig. 1D). The accumulation of TRITC–dextran proceeds in a uniform fashion across the population over time, with an extended lagging edge of cells with lower uptake. The median fluid internalisation over time by Ax2 cells is quantified in Fig. 1E, while a comparison of uptake rates with previous work (Kayman and Clarke, 1983; Thilo and Vogel, 1980; Aguado-Velasco and Bretscher, 1999; Pintsch et al., 2001; Traynor and Kay, 2007) is shown in Fig. 1F.

Controls show the efficiency of the wash step (Fig. S1A), that the Ax2 cells take up similar volumes of liquid whether in suspension or attached to a surface (as in the assay; Fig. S1B) [although this is not true for all strains (Novak et al., 1995)], and demonstrate the range of cell numbers that can be accommodated per well (Fig. S1C). The assay is calibrated in terms of volume taken up per cell by reference to measurements of uptake by the same cell population undertaken using a fluorimeter (Fig. S1D) and standardised over time by using Flow-Set fluorosphere calibration beads (Beckman Coulter).

Phagocytosis of beads (Fig. S1E) and bacteria (Fig. S1F), as well as membrane uptake (Fig. S1G), can also be measured in our high-throughput assay. However, as the cells are not shaken, larger beads in particular will settle during the assay, increasing their local concentration and making comparisons between differently sized particles problematic. In addition, all particles tended to make cells detach, limiting the concentration that can be used, and giving uptake rates that are sub-maximal (Sattler et al., 2013). Because of these limitations, particle assays are most suited to making comparisons in standardised conditions, and not for measuring maximal rates. These problems can be circumvented by performing the uptake in shaken suspension, followed by analysis by flow cytometry, but at the cost of throughput.

We standardly use dextran of 155,000 Da for the uptake assays, but since cells may show selectivity in uptake or trafficking of differently sized dextrans, we also tested smaller and larger dextrans (4400 and 500,000 Da), which had similar sensitivity to actin and PIP3 inhibitors (Fig. S1H), consistent with uptake by the same mechanism.

Inhibitors are powerful tools for acutely interfering with biological processes, but relatively few that affect macropinocytosis are currently known. We therefore tested a number of inhibitors affecting both the cytoskeleton and cellular signalling. These were added to Ax2 cells growing in HL5 medium at the start of the uptake assay along with the TRITC–dextran. The internalised fluorescence was measured 1 h later (Table 1). A number of inhibitors were without effect, although whether this was due to lack of inhibitor uptake, target interaction or the target not functioning in macropinocytosis is unknown.

As macropinocytosis is an actin-dependent process, we first tested inhibitors of actin dynamics. Latrunculin B efficiently inhibited macropinocytosis at standard concentrations (Fig. S2A), as expected from its profound effects on the actin cytoskeleton, as did cytochalasin A (Fig. S2B), as previously reported (Hacker et al., 1997). Inhibitors of the Arp2/3 complex (CK666, Fig. S2C), WASP (Wiskostatin, Fig. S2D) and formins (SMIFH2, Fig. S2E) were all potent inhibitors of macropinocytosis, consistent with the localisation of the target proteins to macropinosomes, and genetics showing macropinocytosis defects in WASP⁻ (Veltman et al., 2016) and ForG⁻ (Junemann et al., 2016) mutants and the axenic growth defect of an ArpB mutant (Langridge and Kay, 2007).

The microtubule inhibitors nocodazole (Fig. S2F) and thiabendazole (Fig. S2G) both partially inhibited fluid uptake, indicating a role for microtubules – most likely in macropinosome trafficking (Rai et al., 2016). The myosin II inhibitor blebbistatin had no effect on macropinocytosis up to 100 µM, in contrast to previously published data, which used higher concentrations that precipitated in our conditions (Shu et al., 2005).

Macropinosomes are organised around active Rac, Ras and PIP3-containing patches (Hoeller et al., 2013; Veltman et al., 2016). Accordingly, the phosphoinositide 3-kinase (PI3K) inhibitor LY294002 (Fig. S2H) inhibited fluid uptake, as did TGX221, which targets the mammalian p110β PI3K isoform (Fig. S2I), whereas inhibitors targeting the p110α and p110γ isoforms did not. We found the Rac inhibitor EHT1864 (Shutes et al., 2007) is a potent inhibitor of fluid uptake (Fig. S2J).

Rapamycin, a TORC1 specific inhibitor, did not affect fluid uptake when applied acutely, as found previously, although it does prevent proliferation (Rosel et al., 2012). It has been suggested that mTORC1 has functions that are not inhibited by rapamycin but which can be inhibited by more-potent (but less specific) mTor inhibitors (Thoreen and Sabatini, 2009). We therefore tried alternative Tor inhibitors and observed an inhibition of macropinocytosis in cells treated with torin 1 (Fig. S2K), but not palomid 529 or PP242. Whether this is due to greater inhibition of TORC1, inhibition of TORC2, or both is not clear – TORC2 has previously been described as having no function in Dictyostelium macropinocytosis (Rosel et al., 2012); however, we see a reduction in macropinocytosis when TORC2 components are knocked out in the Ax2 strain used here (T.D.W., unpublished data).

The nearest to diagnostic inhibitors for macropinocytosis in mammalian cells are amiloride and EIPA, which block the plasma membrane Na⁺/H⁺ exchanger, thus affecting sub-membranous pH (Koivusalo et al., 2010). Although Dictyostelium possesses two Na⁺/H⁺ exchangers (Patel and Barber, 2005; Fey et al., 2013), it is not known whether they are sensitive to these drugs and we find that the drugs do not affect macropinocytosis. The removal of extracellular Ca²⁺ by means of EGTA inhibits constitutive macropinocytosis in immune cells (Canton et al., 2016), but had no effect on macropinocytosis by Dictyostelium incubated in a Ca²⁺-free medium (50 mM lysine and 55 mM glucose in 50 mM MES pH 6.5) indicating that extracellular Ca²⁺ is not required. Indeed, high extracellular Ca²⁺ concentrations can inhibit Dictyostelium macropinocytosis (Maeda and Kawamoto, 1986).

These results support previous genetic studies showing that macropinocytosis depends on PI3K, Rac and actin dynamics controlled through SCAR/WAVE, WASP and formins. On the other hand, regulation through extracellular Ca²⁺ is not part of a conserved core mechanism of macropinocytosis across species, while roles for the Na⁺/H⁺ exchanger and Tor have not been confirmed for Dictyostelium in this work.

Ax2 cells grown on bacteria have a low rate of macropinocytosis, which increases greatly when they are switched to HL5 growth medium (a complex medium containing peptone, yeast extract and glucose). A similar, although much reduced, increase is seen in wild-type NC4 cells (which have an intact NF1 gene) in media enriched with protein (Maeda, 1983).

We confirmed the upregulation of macropinocytosis in Ax2 cells switched from growth on bacteria to growth on HL5 medium (Fig. 2A). It is slow, taking ∼10 h (Fig. 2B), similar to what is seen for Ax3 cells (Kayman and Clarke, 1983), and involves both an increased rate of macropinosome formation (Fig. 2C) and increased macropinosome size (Fig. 2D). A 50% increase in diameter, as seen here, would lead to an ∼3.4-fold increase in macropinosome volume.

Wild-type DdB cells (the parent of the standard Ax2, Ax3 and Ax4 strains) with an intact NF1 gene only marginally upregulate macropinocytosis in HL5 medium (Fig. 2A). However, if DdB cells are switched to HL5 medium supplemented with 10% fetal calf serum (FCS; Gibco, providing ∼4 mg ml⁻¹ additional protein), in which they can proliferate (Bloomfield et al., 2015), they substantially upregulate macropinocytosis, although not as much as Ax2 cells. The increased fluid uptake by DdB cells in this case appears to be due only to an increased rate of macropinosome formation (Fig. 2E), with no detectable increase in size (Fig. 2F). Thus, the macropinocytic rate of wild-type cells is also controlled by the availability of environmental nutrients, as in Ax2 cells.

The ability of Dictyostelium cells to ingest large particles, such as yeast, and large volumes of fluid are linked, since both depend on the loss of NF1. Fig. S3A shows the same linkage exists at a physiological level: axenically adapted Ax2 cells (high fluid uptake) phagocytose 2 μm beads better than the same cells grown on bacteria (low fluid uptake). Phagocytosis of smaller 1.75 μm beads is similar between the conditions, while cells grown on bacteria are better at phagocytosis of 1.5 μm beads. Bead uptake by DdB cells, with an intact NF1 gene, is largely unaltered by the nutritional history of the cell (Fig. S3B), consistent with the unaltered macropinosome size of DdB cells observed in Fig. 2F.

Similar trends are apparent when uptake is performed in shaken suspension: axenically adapted Ax2 cells phagocytose yeast better than Ax2 cells grown on bacteria, while DdB is essentially unable to phagocytose yeast in any condition (Fig. S3C). Similarly, axenically adapted Ax2 cells are relatively better at taking up larger 2 µm beads than bacterially grown cells (Fig. S3D).

The macropinocytic and phagocytic states are not mutually exclusive, as we found that Ax2 cells fully adapted to HL5 medium maintain a relatively high rate of phagocytosis of bacteria (Fig. S3E). We therefore asked what happens when Ax2 cells are presented with both bacteria and liquid medium for food. In this case, irrespective of whether the cells had been grown on bacteria or HL5, they adopted a low rate of macropinocytosis (Fig. S3F).

These results show that Dictyostelium has two basic feeding modes: the preferred mode is phagocytosis, which is seen with cells growing on bacteria (although these cells retain a low level of macropinocytosis, as noted in Fig. 2C). Cells in nutrient-containing media without bacteria adopt a second mode where macropinocytosis is upregulated, although the potential for phagocytosis of bacteria is retained.

Ax2 cells do not sustainably increase their rate of macropinocytosis when they are switched from bacteria to KK₂MC buffer (Fig. 2B), but require HL5 medium, or some components of it, to do so. To identify such stimulatory components, we first showed that HL5 medium could be replaced by the defined medium SIH (Fig. 3A) and then dissected this defined medium to find the active components. Leaving out blocks of components showed that vitamins and micro-minerals are not necessary for macropinocytic upregulation and that the effect is accounted for by amino acids and glucose alone (Fig. 3B). Testing amino acids individually showed that only arginine, lysine and glutamate induce macropinocytosis upregulation at the tested concentrations (Table S1). Consistent with this, removal of just these three amino acids from SIH severely impairs the ability of cells to upregulate macropinocytosis, which is restored when the amino acids are returned to the medium (Fig. 3C). Testing different sugars showed that only glucose and other metabolisable sugars that can support cell growth permit macropinocytosis upregulation (Table S2) (Watts and Ashworth, 1970; Ashworth and Watts, 1970).

Based on these results, a simplified medium for macropinocytosis upregulation, simple upregulation medium (SUM) was devised, consisting of KK₂MC buffer plus 55 mM glucose, 4 mM arginine, 3.7 mM glutamate and 8.5 mM lysine (the same concentrations as SIH) at pH 6.5. SUM induces nearly the same level of macropinocytosis as complete SIH, with faster upregulation kinetics (Fig. 3D). Although cells remain healthy in SUM for several days, it does not support long-term growth. SUM has very low background fluorescence, and we have found it very useful for microscopy, particularly for cells with weakly expressed markers, such as cells with fluorescent tags knocked-in to an endogenous gene. Cells can be grown rapidly on bacteria before transfer to SUM a few hours prior to microscopy, during which time macropinocytosis is greatly upregulated.

These results show that macropinocytosis upregulation can be induced by only a handful of the components present in defined medium, while the requirement for the sugar to be metabolisable hints that sugars may be sensed through their effects on metabolism, rather than by dedicated receptors.

We envisioned that nutrients that cause macropinocytosis upregulation might either be sensed by dedicated receptors, such as those for glutamate, or indirectly through their effect on metabolism, or a combination of both. Since nutrients obtained by macropinocytosis can only be utilised after internalisation, this second route implies that macropinocytic upregulation would depend on fluid uptake by macropinocytosis itself. To test this idea, we used inhibitors to block macropinocytosis during upregulation. As this experiment requires prolonged inhibitor treatment, we first tested how well cells recover from the inhibitors. Ax2 cells growing in HL5 medium recover quite well from prolonged treatment with LY294002 and TGX221 (both PI3K), CK666 (Arp2/3 complex), EHT1864 (Rac) and torin 1 (Tor) (Fig. S4A–E). Prolonged incubation with other inhibitors was too deleterious to make them useful for these experiments.

We next used the inhibitors to determine to what extent upregulation of macropinocytosis depends on macropinocytosis (Fig. 4, ‘raw’ curves), also making a correction for the relatively small deleterious effects of long-term exposure of cells to the inhibitors (Fig. 4, ‘corrected’ curves). Although these inhibitors affect macropinocytosis through different targets, they all inhibit upregulation of macropinocytosis (measured after 10 h incubation in HL5 medium) in a dose-dependent manner (Fig. 4A–E). The effect remains even after correcting for the long-term effects of the inhibitors. Upregulation is not completely abolished by the inhibitors, reflecting their incomplete inhibition of macropinocytosis. Thus, these results suggest that the upregulation of macropinocytosis in nutrient-containing media is at least partially dependent on delivery of nutrients into the cell through macropinocytosis.

We considered the possibility that the ingested nutrients delivered by macropinocytosis might be detected through the TORC1 complex, similar to the situation in other organisms. Although rapamycin does not inhibit macropinocytosis acutely, it does somewhat inhibit upregulation (Fig. 4F), with extremely mild effects on control cells (Fig. S4F). Torin 1 has a stronger effect on upregulation (Fig. 4E), but as it is less specific, some of this might be due to inhibition of the TORC2 complex. In summary, these results suggest that nutrients causing cells to increase their rate of macropinocytosis are detected in the macropinocytic pathway, possibly by TORC1.

Bacteria have two distinct effects on the regulation of macropinocytosis. They inhibit upregulation of macropinocytosis in cells that are transferred into HL5 medium after being previously grown on bacteria (Fig. 5A), and promote downregulation of macropinocytosis by cells transferred from HL5 medium to KK₂MC buffer where it otherwise would remain high (Fig. 5B, see later).

Bacteria can be sensed through their release of folate, which is a chemoattractant for Dictyostelium and acts through the G-protein-coupled receptor fAR1 (Pan et al., 2016). We found that folate inhibits the upregulation of macropinocytosis when cells are transferred from bacteria to HL5 medium (Fig. 5C), but has no effect when cells are transferred from HL5 medium to KK₂MC buffer (data not shown). fAR1-null cells are essentially blind to this inhibitory effect of folate (Fig. 5D), as are mutants of the Gβ and Gα4 (Hadwiger and Firtel, 1992), subunits of the cognate heterotrimeric G-protein for fAR1 (Fig. 5E) (Hadwiger and Srinivasan, 1999), and ErkB (Fig. 5F), the downstream MAP kinase. Thus, bacteria can exert some, but clearly not all, of their effects on feeding behaviour through canonical folate signalling.

Development in Dictyostelium is triggered by starvation and, over the first 8–10 h, the cells undertake chemotaxis towards cyclic AMP causing them to aggregate. Macropinocytosis is downregulated during this period (Maeda, 1983; Katoh et al., 2007), and it was therefore surprising that macropinocytosis continues at a high rate for at least 24 h in cells starved by incubation in KK₂MC buffer (Fig. 5B). However, compared to standard developmental conditions, these cells were starved at low density, likely causing attenuation of developmental signalling. This suggests that the downregulation of macropinocytosis during development requires a developmental signal in addition to starvation.

Fig. 6A confirms that macropinocytosis is strongly downregulated by starving Ax2 cells (previously grown in HL5) at high density in shaking suspension and pulsed with cyclic AMP to mimic developmental signalling. By 5 h of development, fluid uptake is negligible. Similar results were obtained when developing cells on non-nutrient agar (Fig. S5A). Similarly, if the cell density in 96-well plates is increased from 5000 to 50,000 cells per well, the cells form visible aggregates and also downregulate their macropinocytosis (Fig. 6B).

Later in development, tight aggregates form, which could distort uptake assays by restricting access of dextran to internal cells. However, the main decline in macropinocytosis occurs before this stage (Fig. S5B,C) and we see no evidence for two populations of cells (inner and outer) in the flow cytometry results, suggesting that restricted access to internal cells does not significantly affect our results, at least in the first 6 h of development.

We tested the effects of known developmental signals on macropinocytosis using cells starving at low cell density. As shown in Table S3, the developmental signals cyclic AMP, ATP (Ludlow et al., 2008; Traynor and Kay, 2017), adenosine and the polyketides DIF-1, DIF-2 and MPBD (Morris et al., 1987, 1988; Saito et al., 2006) were without effect, as was the high-cell density signal polyphosphate (Suess and Gomer, 2016). However, conditioned medium (CM) prepared by shaking starving cells at high density for 8 h was effective at inhibiting macropinocytosis, with the active component(s) being heat-labile and retained by a 30 kDa cut-off membrane and, therefore, likely to be protein(s) (Fig. 6C). Most likely this signal is one of the known proteins controlling early developmental events in Dictyostelium, but unfortunately these were unavailable for testing.

To gain insight into how developmental signals suppress macropinocytosis, we examined possible signal transduction routes, focussing on cyclic AMP-dependent protein kinase (PKA), which is a crucial mediator of both early and late events in development (Mann and Firtel, 1991; Harwood et al., 1992; Kay, 1989). PKA can be directly activated by using the membrane-permeable analogue of cyclic-AMP, 8-bromo-cyclic-AMP (8-Br-cAMP), and we found that this, unlike cyclic AMP, causes up to a 50% downregulation of macropinocytosis in starving cells at low density (Fig. 6D). High concentrations are required, but these are comparable to those used previously (Kay, 1989).

The involvement of PKA is strongly supported by mutants with elevated intracellular cyclic AMP levels, due to defective breakdown. The hybrid cyclic AMP phosphodiesterase RegA is activated by a His/Asp phospho-relay in which RdeA is the essential phosphate carrier protein (Shaulsky et al., 1998; Thomason et al., 1999, 1998; Chang et al., 1998). Elimination of either protein results in strong downregulation of macropinocytosis in starving cells at low density (Fig. 6E).

Conversely, eliminating PKA activity by mutation of the catalytic subunit (pkaC⁻ cells; Primpke et al., 2000) results in cells where macropinocytosis remains high for at least 24 h after starvation, even when they are at high cell density or treated with CM (Fig. 6F). Combined, these results strongly argue that macropinocytosis is downregulated in starving cells because of PKA activation. These results are also relevant to the interpretation of recent work (Scavello et al., 2017) showing that pkaC− cells have a strong defect in chemotaxis towards cyclic AMP (see Discussion).

The great majority of fluid uptake by axenic Dictyostelium cells is by macropinocytosis, rather than, for instance, by clathrin-mediated endocytosis. This has been shown through morphometry (Hacker et al., 1997), the sensitivity of fluid uptake to mutations and inhibitors specifically expected to affect macropinocytosis, such as those affecting PI3K and the RasGAP NF1 (Buczynski et al., 1997; Hoeller et al., 2013; Bloomfield et al., 2015; Veltman et al., 2016; this work), and finally by the lack of correlation between the uptake of membrane (taken up mainly by clathrin-mediated endocytosis) and fluid in various situations, indicating that they are largely separate processes (Aguado-Velasco and Bretscher, 1999). In this context, the reduced fluid uptake seen in clathrin heavy chain mutants is likely an indirect effect, perhaps due to perturbed processing of macropinosomes (O'Halloran and Anderson, 1992). The various lines of evidence suggest that macropinocytosis accounts for more than 90% of fluid uptake by axenic cells fully adapted to growth on liquid medium.

This high rate of macropinocytosis allows fluid uptake to be used as a measure of macropinocytosis, which is an enormous advantage over assays based on counting macropinosomes visualised by microscopy (Commisso et al., 2014). We have adapted the previous macropinocytosis assays based on bulk fluid uptake (Kayman and Clarke, 1983; Thilo and Vogel, 1980; Rivero and Maniak, 2006) to flow cytometry and 96-well plates, thus giving high-throughput and single-cell resolution.

A screen of inhibitors provides new tools for acute inhibition of macropinocytosis and further supports the involvement of PI3K, Rac, WASP, formins and the Arp2/3 complex, as expected from genetic and subcellular localisation studies (Buczynski et al., 1997; Hoeller et al., 2013; Langridge and Kay, 2007; Dumontier et al., 2000; Veltman et al., 2016; Junemann et al., 2016).

Macropinocytosis in Dictyostelium occurs at a high rate in conditions where the cells can proliferate in liquid medium. However, it is under physiological control, with cells slowly transitioning between high and low macropinocytic states according to whether bacteria or soluble nutrients are available. In these transitions, the frequency of macropinosome formation is altered: in axenic cells, where the active Ras patches are unconstrained by NF1, macropinosome size is additionally increased. Wild-type cells with an intact NF1 gene also transition between low and high macropinocytic states according to the nutrients available, showing that this regulation is not just a feature of axenic strains (this work; Maeda, 1983). The presence of a high macropinocytic state in wild-type cells suggests there are ecological circumstances where macropinocytosis is used for feeding, though these are yet to be defined.

Ax2 cells in the low macropinocytic state can sense bacteria through their secretion of folic acid, inhibiting macropinocytic upregulation accordingly. However, due to the relatively modest effects of folate, and the fact that it does not induce downregulation of macropinocytosis, it seems certain that other sensory pathways also play a prominent role. It has recently been reported that certain bacteria secrete cyclic AMP, which functions as a chemoattractant for vegetative Dictyostelium (Meena and Kimmel, 2017); however, cyclic AMP did not affect the upregulation or downregulation of macropinocytosis.

Four nutrients are largely responsible for inducing macropinocytosis upregulation in Ax2 cells: arginine, glutamate, lysine and a metabolisable sugar. None of the other amino acids appears effective individually, and even in combination they have only a modest effect. Arginine and lysine are essential amino acids, but glutamate is not (Marin, 1976; Franke and Kessin, 1977). Dictyostelium has several receptors similar to metabotropic glutamate receptors (Taniura et al., 2006; Fey et al., 2013), but it seems likely that the major route for nutrient sensing is intracellular, with nutrients delivered by macropinocytosis.

In mammalian cells, free amino acids obtained by macropinocytosis are sensed by activation of mTORC1 at the lysosome (Yoshida et al., 2015; Sancak et al., 2010), with one effect being inhibition of autophagy. In Dictyostelium, autophagy is induced within minutes of withdrawing arginine and lysine (King et al., 2011), a step which is necessary to survive prolonged amino acid starvation (Tekinay et al., 2006). Taken together, what is known about mTORC1 and Dictyostelium autophagy alongside our results suggests that Dictyostelium TORC1 may sense arginine and lysine to upregulate macropinocytosis. Our attempts to test this idea using Tor inhibitors are not definitive, but it remains an attractive possibility.

As only metabolisable sugars induce upregulation of macropinocytosis, it is probable that the sensing of these is through a general metabolic readout, such as the ratio of ATP to ADP and AMP. Increased levels of AMP and ADP, as occurs in nutrient-poor conditions (such as without sugar), activate AMP-kinase. Overexpression of a constitutively active AMP-kinase α subunit in Dictyostelium inhibits growth but does not affect macropinocytosis (Bokko et al., 2007), similar to what we observe in low-density starvation conditions. Although an attractive possibility, it remains to be determined whether AMP-kinase has any function in upregulation of macropinocytosis.

Our results show that the cessation of macropinocytosis during early development requires a developmental signal that most likely acts through PKA. Macropinocytosis does not cease immediately when cells are starved, but decreases over several hours and so may occur at reduced levels in cells used for studying chemotaxis to cyclic AMP. This can be a confounding influence since macropinocytosis uses the same actin machinery as pseudopods and thus impairs chemotaxis (Veltman, 2015). In particular, we found that macropinocytosis continues at a high rate in mutants of the PKA catalytic subunit, possibly accounting for the strong chemotactic defect of these strains (Scavello et al., 2017). Continued macropinocytosis could also confound studies on other strains with early developmental defects (Khosla et al., 2005; Wu et al., 1995; Rodriguez et al., 2008; Lee et al., 2005).

Many of the molecular components required for macropinocytosis are the same in both Dictyostelium and mammalian cells, such as actin, Arp2/3, PI3K, SCAR/WAVE, WASP, Rac and Ras proteins. This suggests that macropinocytosis may have first arisen in simple protists as a way of feeding in the absence of bacterial prey. In mammalian cells, there are additional levels of regulation, some of which are cell type specific (such as the Ca²⁺ requirement in immune cells) and others that are more generic (such as growth factor-stimulated macropinocytosis). Dictyostelium with its high intrinsic rate of macropinocytosis in axenic strains, high-throughput assays (this work), and the recent development of transformation techniques that allow easy manipulation of non-axenic strains – and thus of mutants defective in macropinocytosis (P. Paschke, D. A. Knecht, A. Silale, D. Traynor, T.D.W., P. A. Thomason, R. H. Insall, J. R. Chubb, R.R.K., D. M. Veltman; unpublished data) – is now an excellent model for establishing the conserved core elements of macropinocytosis.

Cells were cultivated at 22°C. HL5, SIH (complete, and lacking components) and SM media were from Formedium. Unless otherwise specified, cells were grown on Klebsiella aerogenes (Ka) lawns on SM plates and harvested for experiments from the feeding front, washing three times with KK₂ (16.6 mM KH₂PO₄, 3.8 mM K₂HPO₄, pH 6.1) by centrifugation (280 g, 3 min) to remove the bacteria. Cells were also grown in tissue culture plates with Ka as a food source. In this case, Ka was added to KK₂MC buffer (KK₂ plus 2 mM MgSO₄, 100 μM CaCl₂) to 2 optical density units at 600 nm (2 OD_600nm) units from a 100 OD_600nm stock (the stock bacteria were grown overnight in 2× TY, pelleted by centrifugation, washed twice in KK₂ and stored at 4°C).

Cells were grown axenically in HL5 in conical flasks with shaking at 180 rpm. Media derived from SIH, including SUM, were made in KK₂MC pH 6.5. Conditioned medium was made by washing axenically grown Ax2 cells free of HL5, resuspending them to 10⁷ cells ml⁻¹ in KK₂MC and incubating for 8 h, with 180 rpm shaking, before removing the cells by centrifugation (2400 g, 10 min). Strains are listed in Table S4.

For transformation, cells were harvested from bacteria, resuspended in H40 buffer (40 mM Hepes, 1 mM MgCl₂, pH 7.0), mixed with 500 ng vector for a PIP3 reporter (PkgE-PH–mCherry), electroporated in ice-cold 2 mm cuvettes (Novagen) using a square wave protocol (2×350 volts, 8 ms apart), and then transferred to 2 ml KK₂MC+Ka in a six-well plate to recover for 5 h, before G418 selection was added to 10 μg ml⁻¹ (Paschke et al., in revision).

Chemicals were from Sigma unless otherwise indicated. Polyphosphate was from both Spectrum and Merck.

The method used was based on that from Rivero and Maniak, 2006. Cells at 1×10⁷ ml⁻¹ were shaken at 180 rpm in HL5 with 0.5 mg ml⁻¹ TRITC–dextran (molecular mass of 155,000 Da, unless otherwise stated) and at each time point triplicate 0.8 ml samples were centrifugally washed once in ice-cold KK₂ and resuspended to 1 ml. Fluorescence was measured in a fluorimeter (Perkin-Elmer LS 50 B with excitation at 544 nm, emission at 574 nm, slit width 10 nm). Background ‘0 minute’ fluorescence was subtracted and uptake volume was calculated from standard curves for TRITC–dextran diluted in buffer. Cells loaded in this way were also analysed by flow cytometry (LSR_II flow cytometer, BD Biosciences) to compare the methods.

To measure yeast uptake, cells were resuspended to 5×10⁶ cells ml⁻¹ in KK₂MC in a 5 ml conical flask and shaken at 180 rpm at 22°C. TRITC-labelled yeast (sonicated at level 7.0 for 20 s on a Misonix sonicator 3000) were added to 10⁷ particles ml⁻¹. At 0 and 60 min, duplicate 200 μl samples were added to 20 μl of Trypan Blue quench solution (2 mg ml⁻¹ in 20 mM citrate, 150 mM NaCl, pH 4.5) on ice, shaken for 3 min at 2000 rpm, spun down and washed twice with ice-cold KK₂+10 mM EDTA. The final pellet was resuspended to 1 ml and the fluorescence compared to a standard curve to give the number of yeast per cell.

For high-throughput assays, 50 μl of medium with 10⁵ cells ml⁻¹ was incubated in flat-bottom, 96-well plates at 22°C for the indicated time (usually 24 h). Then 50 μl of 1 mg ml⁻¹ TRITC–dextran in the same medium was added for a final concentration of 0.5 mg ml⁻¹. After 1 h, unless otherwise stated, the medium was thrown off, and the cells washed by ‘dunk-banging’ (the plate was submerged in a container of ice-cold KK₂, which was thrown off and the plate patted dry) before 100 μl KK₂MC containing 5 mM sodium azide was added to each well to detach the cells and stop exocytosis. Cells were analysed by flow cytometry (LSR-II, BD Biosciences) using the high-throughput sampling attachment, which pipetted them up and down twice, before analysing 65 μl per sample at 3 μl s⁻¹. FlowJo software (https://www.flowjo.com) calculated the median fluorescence of cells in each well. The mean was then taken of all biological replicates. To determine volumes taken up, the same population of cells (loaded with TRITC–dextran in suspension, as above) was analysed by both fluorimetry and flow cytometry. The LSR-II flow cytometer was calibrated through all subsequent experiments by using FlowSet fluorospheres calibration beads (Beckman Coulter).

We also used this method to measure uptake of membrane using 10 μM FM1-43 (Invitrogen), phagocytosis of bacteria using 1×10⁸ particles ml⁻¹ Texas Red E. coli bioparticles (Thermo Scientific), or beads of different sizes (YG-beads, Polysciences; 2.0 and 1.75 μm at 5×10⁷ ml⁻¹, 1.0 and 1.5 μm at 10⁸ ml⁻¹). Bead uptake in shaking suspension was measured by using cells at 2×10⁶ cells ml⁻¹ in KK₂MC (prepared as for the yeast uptake assay) with 4×10⁸ beads ml⁻¹. After shaking at 180 rpm for 20 min, samples were diluted into ice-cold KK₂+5 mM NaN₃, spun down (300 g, 3 min), washed once, then resuspended and filtered into tubes for flow cytometry. The number of particles internalised per cell were calculated as described (Sattler et al., 2013). Beads larger than 2 μm were taken up very poorly, so were not used. For time courses, start times were staggered so that all incubations ended concurrently. When inhibitors were used acutely, they were added with the fluorescent medium to the final indicated concentration. Polyketides were synthesised as described (Morris et al., 1987, 1988; Saito et al., 2006).

To initiate development, axenically growing cells were washed twice, resuspended to 10⁷ cells ml⁻¹ in KK₂MC and shaken at 180 rpm for 1 h before delivering pulses of KK₂MC containing cyclic AMP to give a concentration of 100 nM every 6 min using a Watson Marlow 505Di pump. At the indicated times 5×10⁴ cells were diluted into dextran-containing KK₂MC in 24-well plates for 1 h, after which they were washed in situ using ice-cold KK₂+10 mM EDTA and detached with KK₂MC+5 mM sodium azide. 100 μl was transferred to duplicate wells in a 96-well plate for flow cytometry analysis.

Development on non-nutrient agar plates was initiated by settling 1.5 ml of washed, axenically growing cells at 2.5×10⁷ cells ml⁻¹ in KK₂MC onto fresh 1.8% KK₂MC agar (Oxoid L28) in 6-cm plates. After 15 min settling, the medium was aspirated off, and the plates kept on wet tissues at 22°C. At the indicated times, cells were harvested, resuspended in KK₂MC and 10⁵ cells inoculated into KK₂MC in a six-well plate with 0.5 mg ml⁻¹ TRITC-dextran for 1 h. Cells were then washed in situ and resuspended in KK₂+10 mM EDTA before analysis by low-throughput flow cytometry. The zero hour time point was for cells taken immediately after washing.

The rate of macropinosome formation was determined in KK₂MC by loading cells in a two-well microscope slide (Nunc) with 2 mg ml⁻¹ FITC–dextran (molecular mass 70,000 Da) for 1 min, then washing and fixing with 4% paraformaldehyde for 20 min. Fixed cells were washed five times and stored in PBS (pH 5.0) at 4°C for imaging. Z-stacks with 0.1 μm steps were taken using a Zeiss 700 series microscope with 2× averaging to reduce noise. Maximum intensity projections were made using FIJI and FITC-positive endosomes counted by eye. The mean of at least eight cells on a given day was taken as one data point.

To measure macropinosome diameter at closure, cells in KK₂MC expressing a PIP3 reporter (PkgE-PH–mCherry) were filmed in their central section at 1 frame per second for 5 min on a Zeiss 700 series microscope. The maximum diameter of macropinosomes at closure was measured by using the FIJI measure tool. Note that this method will underestimate the diameter of macropinosomes not lying fully within the optical section.

We thank the rest of the Kay laboratory for their assistance in moulding this project, particularly Peggy Paschke. Clelia Amato and Robert Insall (Beatson Institute, Glasgow) alerted us to the Rac inhibitor. Jason King (Sheffield University) provided valuable feedback on the macropinosome formation experiments. Miao Pan and Tian Jin (NIAID, Bethesda) kindly sent us the fAR1⁻ strain. The MRC-LMB flow cytometry facility maintained the flow cytometers and provided technical support.