Lipid-gated monovalent ion fluxes regulate endocytic traffic and support immune surveillance

See allHide authors and affiliations

Science  17 Jan 2020:
Vol. 367, Issue 6475, pp. 301-305
DOI: 10.1126/science.aaw9544

Ion fluxes resolve organellar volume

Animal cells continuously sample the surrounding medium, a feature accentuated in immune cells. Sampling is accomplished by trapping external medium into membrane-bound vesicles or vacuoles. These structures are promptly resolved, thus avoiding accumulation of endomembranes and volume expansion. In a variety of cultured cells, Freeman et al. found that this resolution entails conversion of spherical vacuoles into thin tubules, a process that involves marked changes in surface-to-volume ratio (see the Perspective by King and Smythe). Shrinkage of membrane-bound structures is driven by ion fluxes and subsequent osmotic transfer of water. Shriveled vacuoles attract curvature-sensing proteins that promote the extension of fine tubules. Ion channels thereby control membrane remodeling, enabling receptor recycling and proper routing of cellular cargo.

Science, this issue p. 301; see also p. 246


Despite ongoing (macro)pinocytosis of extracellular fluid, the volume of the endocytic pathway remains unchanged. To investigate the underlying mechanism, we used high-resolution video imaging to analyze the fate of macropinosomes formed by macrophages in vitro and in situ. Na+, the primary cationic osmolyte internalized, exited endocytic vacuoles via two-pore channels, accompanied by parallel efflux of Cl and osmotically coupled water. The resulting shrinkage caused crenation of the membrane, which fostered recruitment of curvature-sensing proteins. These proteins stabilized tubules and promoted their elongation, driving vacuolar remodeling, receptor recycling, and resolution of the organelles. Failure to resolve internalized fluid impairs the tissue surveillance activity of resident macrophages. Thus, osmotically driven increases in the surface-to-volume ratio of endomembranes promote traffic between compartments and help to ensure tissue homeostasis.

During endocytosis, cells internalize membrane along with extracellular fluid (pinocytosis). The amount of fluid ingested can be substantial: dendritic cells and macrophages take up the equivalent of their entire cellular volume every 4 hours (1). Despite continuous uptake of large amounts of water and osmolytes, the endocytic pathway and the cells as a whole retain their volume and ionic composition over extended periods (1). To investigate endomembrane volume and ionic regulation, we chose macropinosomes, large (up to 5 μm) vacuoles formed by specialized cell types (2, 3) (Fig. 1 and fig. S1, A and B). Unlike smaller endocytic vesicles, macropinosomes can be resolved by diffraction-limited microscopy (4), enabling detailed assessment of their volume as they mature. Moreover, the volume of medium entrapped by macropinosomes is sufficiently large to alter the overall ionic composition of the cells (fig. S1, C and D). When stimulated by macrophage colony-stimulating factor (M-CSF), bone marrow–derived macrophages (BMDM) underwent a large burst of macropinocytosis. Multiple large vacuoles (10 to 15 per cell; mean volume: 7 μm3) formed within 5 min (Fig. 1A and fig. S1, A and B). The volume of fluid internalized was equivalent to ≈25% of the cell volume, an increase detectable by electronic cell sizing (Fig. 1B). When visualized using rhodamine-dextran (70 kDa), the macropinosomes of BMDM (Fig. 1A), and those formed by peritoneal macrophages and human monocyte–derived macrophages (fig. S1E), resorbed within 30 min, and cell volume returned to basal levels (Fig. 1B). Shrinkage of macropinocytic vacuoles was also observed in vivo. Two-photon imaging of live mice revealed that resident tissue macrophages (RTM) of the peritoneal serosa—interstitial, nonmigratory cells that constitutively sample the surrounding milieu (5) (movie S1)—formed large macropinosomes that subsequently contracted (Fig. 1, C and D, and movie S2).

Fig. 1 Vacuolar shrinkage requires monovalent ion efflux.

(A) Volume (vol) and 70 kDa rhodamine-dextran fluorescence intensity changes of macropinosomes induced in bone marrow–derived macrophages (BMDM) by macrophage colony-stimulating factor (M-CSF); data are means ± SEM of >100 vacuoles from three independent experiments (i.e., n = 3). Measurement of vacuole resolution was initiated after a 5-min stimulation with M-CSF in medium containing dextran, followed by an immediate wash. (B) Cell (BMDM) volume was measured electronically before and at the indicated times after M-CSF stimulation; >104 cells per point, n = 3. P values determined by unpaired, two-sided t tests. Here and elsewhere, ***P < 0.001, **P < 0.01, and *P < 0.05. (C) Intravital observation of td-Tomato–labeled resident tissue macrophages (RTM; pseudocolored yellow and red) of the peritoneal serosa and second harmonic imaging (SHG) of collagen (blue). See also movies S1 and S2. (D) Visualization and volume quantification of M-CSF–induced macropinosomes in RTM in vivo; means, upper and lower quartiles (boxes), and distribution (whiskers) are graphed. >50 vacuoles, n = 3. P values determined by Mann-Whitney U test. (E) Macropinosomes of M-CSF–stimulated BMDM in media containing indicated solutes and dextran. Representative images acquired at 5 min. See also movie S3. Bottom row: mean ± SEM macropinosomal volume and dextran intensity from three independent video recordings representing >150 macropinosomes. See also fig. S1.

Macropinosome shrinkage was accompanied by an increase in the intensity of the luminal dextran fluorescence (Fig. 1A), implying that fluid was extracted from the vacuoles. This suggested that the volume loss of the vacuoles is caused by osmotically driven solvent loss. Na+ and Cl constitute the majority of the osmolytes in the fluid engulfed during macropinocytosis. Accordingly, inducing macropinocytosis with M-CSF resulted in a fourfold increase in the total cellular Na+ concentration (fig. S1D). Thus, loss of Na+ and Cl along with osmotically coupled water may underlie the rapid shrinkage of macropinosomes. This was validated by ion substitution experiments: replacing Na+ for the impermeant cation N-methyl-d-glucamine+ (NMG+) virtually eliminated macropinosome resolution (Fig. 1E; fig. S1, E and F; and movie S3). Similarly, shrinkage was precluded when substituting the impermeant anion gluconate for Cl, implying that electroneutrality needs to be maintained during solute export (Fig. 1E and fig. S1, E and F). Preventing monovalent ion efflux from macropinosomes also prevented restoration of the cell volume (fig. S1G). The absence of luminal Ca2+ did not prevent macropinosome resolution (Fig. 1E).

We tested a series of ion transport inhibitors to gain insight into the pathways involved in macropinosome shrinkage. Tetrandrine, a potent inhibitor of two-pore channels (TPC) (6), impaired volume loss (Fig. 2A; fig. S2, A and B; and movie S4). Interestingly, the endomembrane isoforms TPC1 and TPC2 are expressed at particularly high levels in myeloid cells, including BMDM (fig. S6I) (7). Moreover, TPC1 is highly expressed in the macropinocytic interstitial (Ccr2, CD169+) RTM, compared with neighboring stromal or migratory (Ccr2+, CD169) myeloid cells that are nonmacropinocytic (Fig. 2, D and E) (5). Although undetectable at the plasma membrane, TPC1 was rapidly (in ≤1 min) acquired by nascent macropinosomes (Fig. 2F), whereas TPC2 was recruited later (fig. S3D). BMDM from Tpc1;Tpc2 double-knockout mice formed large macropinosomes when stimulated by M-CSF, but these did not shrink and resolve during our analyses (Fig. 2, B and C). Using single knockout mice and RNA interference, we discerned this effect to be attributable primarily to TPC1 (Fig. 2C and fig. S3, F and G).

Fig. 2 Monovalent ion efflux mechanisms.

(A) Macropinosome volume changes in presence of 5 μM tetrandrine, measured in BMDM. Measurement of vacuole resolution was initiated once cells were washed after a 5-min stimulation with M-CSF in medium containing dextran; tetrandrine or vehicle were present throughout. Means ± SEM, n = 3. See also fig. S2 and movie S4. (B and C) Macropinosome volume changes after stimulation with M-CSF in WT, Tpc1 and Tpc2 single and double knockout (KO), and Trpml1 KO BMDM. In (C), means, upper and lower quartiles (boxes), distribution (whiskers), and observations from fields containing three to five cells (dots) each, measured 10 min after macropinosome formation; n = 3. (D) Staining of the peritoneal serosa. Outline of CD169 signal (left) overlaid on TPC1 signal (right). (E) RNA sequencing. Resident tissue macrophages were Cx3cr1/Ccr2. Migratory cells were Cx3cr1/Ccr2+. (F) BMDM expressing TPC1-tomato or 2xfyve-GFP to detect PtdIns(3)P. Dextran shown in cyan. (G) BMDM stimulated with M-CSF in the presence of 70 kDa rhodamine-dextran and, where indicated, the PIKfyve inhibitors YM201636, apilimod, or WX8 (all used at 500 nM). Resolution was recorded as in (A). 5 min after isosmotic recording, a hyperosmotic solution [final 500 milliosmolar (mOsm)] was added to verify the osmotic responsiveness of the vacuoles. (H and I) Visualization and volume quantification of macropinosomes in RTM treated in situ with YM201636 (500 nM) or tetrandrine (5 μM); >15 cells, n = 3. All P values determined by Mann-Whitney U tests.

Certain ion channels, including TPCs and TRPMLs (mucolipin transient receptor potential channels), require phosphatidylinositol 3,5-bisphosphate [PtdIns(3,5)P2] for activation (8, 9). This phosphoinositide is generated by phosphorylation of PtdIns(3)P by the phosphoinositide kinase PIKfyve (10). PtdIns(3)P and PIKfyve itself were readily detectable on the cytosolic leaflet of nascent macropinosomes (Fig. 2F and fig. S3E), which is consistent with this sequence. 2xMLN–green fluorescent protein (GFP), a putative probe for PtdIns(3,5)P2 (11) was also found in macropinosomes (fig. S3E). Macropinosome shrinkage—whether measured directly or assessed indirectly from the overall cell volume gain—was blocked by PIKfyve antagonists (Fig. 2, G to I, and fig. S1, L to N). Inhibiting PIKfyve did not alter the water permeability or pliability of the membrane, as indicated by the acute volume loss induced by water abstraction caused by hypertonic medium (fig. S1, K and L). A similar response to hypertonicity was observed in macropinosomes formed in Na+-free solution (fig. S1J). Although TRPML1 channels are recruited to maturing macropinosomes, deletion of the Trpml1 gene did not affect macropinosome resorption (Fig. 2C and fig. S3D).

The area of the vacuolar membrane was reduced during shrinkage by emission and severing of tubules and vesicles, which were visualized using FM 1-43 (Fig. 3A) or sulforhodamine B (Fig. 3D) dyes. Tubule extension accompanies and requires the volume loss that is driven by the export of ions and osmotically coupled water. Accordingly, substitution of Na+ with NMG+, or blockade of TPC channels by tetrandrine or their inactivation by depleting PtdIns(3,5)P2, impaired tubulation and vesiculation (Fig. 3, A and B). Genetic deletion of TPC channels also precluded tubulation (Fig. 3C). Applying a hypertonic solution to macropinosomes that initially failed to shrink because they were loaded with NMG+ or formed in cells treated with a PIKfyve inhibitor revealed that tubulation is a consequence, not a cause, of volume loss (Fig. 3, D and E, and movie S5). The tubules emanating from early macropinosomes are very thin, with a modal diameter of ≈30 nm (Fig. 3F and fig. S4C), which likely explains the preferential retention and progressive concentration of large (70-kDa) dextran in the vacuolar lumen. The diameter of the tubules generated during macropinosome resorption makes them well suited for associating with proteins containing BAR domains, concave structures that preferentially bind and stabilize curved membranes of ≈22-nm diameter (12). Indeed, the BAR domain–containing proteins SNX1, SNX2, and SNX5 decorated the tubules emanating from resorbing macropinosomes (fig. S4A) (13). Thus, membrane crenation caused by the volume loss potentially generated the necessary curvature to stabilize BAR domains on the membrane and thereby fostered tubulation (Fig. 3I). This notion was tested by generating liposomes and monitoring the effects of hydrostatic tension on the ability of a recombinant BAR-domain protein, BIN1, to induce tubulation. These experiments revealed a distinctive relationship between volume loss and BAR-mediated tubulation: the relief of hydrostatic tension greatly amplified tubulation by BIN1, whereas swelling the liposomes counteracted it (Fig. 3G and fig. S4D). Given their functional redundancy and ability to form interchangeable heterodimeric complexes, the loss of any one BAR protein is unlikely to prevent tubulation. Because many BAR-domain proteins (including various SNX isoforms) require PtdIns(3)P for optimal binding, we interfered with their association by scavenging the available head groups of the phosphoinositide by expressing tandem FYVE domains. Indeed, high-affinity (multicopy) FYVE-domain tandems prevented SNX recruitment and precluded tubulation and resolution of the macropinosomes (Fig. 3H and fig. S4B).

Fig. 3 Osmotically driven shrinkage induces tubulation.

(A) BMDM were stimulated with M-CSF and the distribution of 70 kDa rhodamine-dextran and FM 1-43 imaged at the indicated times after removal of the stimulus. (B and C) Mean number of tubules (exceeding 1 μm in length) measured 5 min after stimulation with M-CSF; >100 macropinosomes (n = 3) for each condition. (D and E) Macropinosomes containing sulforhodamine B (SRB) and N-methylglucamine chloride or formed in cells treated with YM201636 (500 nM) were recorded 10 min after formation, before and after being subjected to hypertonic solution. See also movie S5. >100 vacuoles, n = 6. (F) Transmission electron microscopy was used to measure the diameter of tubules emerging from macropinosomes; 85 tubules were quantified. (G) Liposomes formed of whole brain lipid, rhodamine-labeled phosphatidylethanolamine, and PtdIns(4,5)P2 in 20 mOsm solution. The mean aspect ratio for three to five fields of liposomes incubated with or without recombinant human BIN1 was quantified by imaging, n = 3. (H) HT1080 cells expressing GFP, 2xfyve-GFP, or 5xfyve-GFP were pulsed with SRB for 10 min. Mean number of tubules (exceeding 1 μm in length); >100 macropinosomes (n = 3) for each condition. All P values determined by unpaired, two-sided t tests. (I) Model of mechanism proposed to underlie macropinosomal tubulation.

The significance of endomembrane shrinkage driven by efflux of monovalent ions is far-reaching. The resulting tubulation mediates the recycling of plasmalemmal components that are internalized in the course of macropinocytosis and endocytosis. One such example is Mac-1 (αMβ2), an integrin that is key to macrophage adherence, migration, and phagocytosis (14). Blocking TPC channels by inhibiting PtdIns(3,5)P2 formation led to a pronounced depletion of plasmalemmal Mac-1, which was instead trapped in endomembrane vacuoles (Fig. 4A and fig. S5C). The effect was phenocopied by simply substituting extracellular Na+ by K+, demonstrating that a Na+ gradient is exploited by the endocytic pathway to direct membrane traffic (fig. S5, A to C). Moreover, nonmacropinocytic cells (e.g., fibroblasts) also required Na+ efflux to execute canonical receptor recycling (fig. S6, A to D, G, and H) (15, 16). Defective plasmalemmal protein recycling caused by ion substitution had acute functional consequences: the ability of BMDM to bind and ingest complement-coated targets and of fibroblasts to form focal adhesions—processes mediated by integrins—were severely depressed (figs. S5, D and E, and S6, E and F).

Fig. 4 Vacuolar resolution maintains cellular responsiveness and tissue surveillance.

(A) BMDM were stimulated for 30 min with M-CSF, with or without YM201636, fixed and immunostained for Mac-1. (B to E) Resident tissue macrophages (LysM-tdTomato) with or without YM201636 or tetrandrine were stimulated with M-CSF for 10 min followed by removal of the stimulus for 30 min. Cells were then imaged for 30 min. (C) surveillance area measured in the absence (left) or presence (right) of YM201636 (YM) over time, n = 3. In (D and E), 30 min after M-CSF removal, laser-induced microlesions were generated, marked by the resultant autofluorescence [orange in (E)]. Mean squared displacement of the macrophages is graphed in (D); means ± SD, n = 3. Representative images in (E) taken at 15 min after injury. (F and G) Experiments performed as in (E). Representative images denoting the presence of neutrophils are shown. In (G), the percentage of lesions with neutrophil swarming was quantified for six microlesions per animal, n = 3. (H) HT1080 cells expressing TPC1-tomato were incubated with 70 kDa dextran for 10 min before imaging. (I and J) WT and TPC1 KO HT1080 cell growth measured in the absence or presence of YM201636 or tetrandrine (TTD) by cell counting. Means ± SD, n = 3. All P values determined by Mann-Whitney U tests.

The need for ion efflux from endocytic compartments for normal cell function was also documented in situ. The ability of interstitial RTM to survey their environment was impaired when Na+ efflux from the endocytic pathway was prevented in vivo (Fig. 4, B to E, and movie S6). Blocking β1 and β2 integrins similarly inhibited surveillance (fig. S7). When microlesions are made by targeted laser ablation to adjacent fibroblasts, RTM normally emit processes to contain the damage (5), preventing the recruitment and activation of neutrophils. When PIKfyve or TPCs were inhibited, the cells failed to resorb vacuoles and were unable to respond to the damage (Fig. 4, D and E). As a result, neutrophil swarming ensued (Fig. 4, F and G). Thus, vacuole resolution, mediated by lipid-gated Na+ efflux, underpins membrane traffic necessary to maintain cellular responsiveness.

Because macropinocytosis is not restricted to phagocytes, inhibition of vacuolar shrinkage also affects other cell types (17). HT1080 fibrosarcoma cells display vigorous constitutive macropinocytosis (Fig. 4H), express TPC1 at comparatively high levels (fig. S8A), and localize the channel to macropinosomes (Fig. 4H). HT1080 cells require growth factors for survival and are highly responsive to epidermal growth factor (EGF). Because the EGF receptor is internalized along with the macropinosomal membrane, effective recycling to the cell surface is key. Preventing vacuole shrinkage resulted in the depletion of EGF receptors from the plasmalemma (fig. S8B), which was associated with reduced responsiveness to EGF and delayed growth (Fig. 4, I and J, and fig. S8C).

Solute transport may be involved in the shrinkage and tubulation of other organelles. In this regard, lysosomes are known to undergo swelling in cells treated with PIKfyve antagonists (18). Impaired solute extrusion could account for the volume gain, which is compounded by ongoing membrane fusion that is not compensated for by shrinkage-dependent tubulation and/or vesiculation and scission. Indeed, lysosomes swollen by inhibition of PIKfyve recovered their ability to tubulate when exposed to hypertonic solution (movie S7) and on removal of the PIKfyve inhibitor but failed to do so when tetrandrine was present (fig. S9). Overexpression of TPC2 alone caused lysosomes to become more tubular, suggesting that the channel may be involved in this process (fig. S10 and movie S8).

We propose a role for ion fluxes in the endocytic pathway: Extrusion of osmotically abundant ions and solutes, accompanied by water, cause vacuolar and vesicular shrinkage leading to crenation of the membrane, forming convex protrusions that stabilize proteins with BAR or similar curvature-sensing domains. These proteins, in turn, foster tubulation and eventual scission that is critical for interorganellar traffic. In macrophages that continuously survey the extracellular space, an ongoing ion efflux from the endocytic pathway drives the resolution of fluid that is taken up during this process, supporting recycling of receptors to maintain their function. Thus, the resolution of organelles formed during surveillance is necessary to preserve tissue integrity.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

References (1928)

Movies S1 to S8

References and Notes

Acknowledgments: We thank K. Aranda, Z. Liu, and M. Capurro for tissue preparation and genotyping. We are grateful to the Transgenic and Chimeric Mouse Core of the University of Pennsylvania for the generation of the TPC1 line. Funding: This work was supported by grant FDN-143202 from CIHR to S.G., the Intramural Research Program of the NIH to R.N.G. and S.U., and by NIH grants R01 HL147379 and R01 GM133172 to D.R. C.M.B. was supported by the Wellcome Trust Sir Henry Wellcome Postdoctoral Fellowship. Author contributions: S.A.F. and S.G. conceptualized the project. S.A.F., S.U., A.S., and S.G. designed the research and wrote the paper. S.A.F., S.U., A.S., R.F.C., C.M.B., S.M., P.B., and J.P. performed the experiments and analyzed the results. R.N.G. and D.R. devised the methodology, provided resources, and edited the manuscript. Competing interests: None declared. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.

Stay Connected to Science

Navigate This Article