SLIT2/ROBO1-signaling inhibits macropinocytosis by opposing cortical cytoskeletal remodeling

Macropinocytosis is essential for myeloid cells to survey their environment and for growth of RAS-transformed cancer cells. Several growth factors and inflammatory stimuli are known to induce macropinocytosis, but its endogenous inhibitors have remained elusive. Stimulation of Roundabout receptors by Slit ligands inhibits directional migration of many cell types, including immune cells and cancer cells. We report that SLIT2 inhibits macropinocytosis in vitro and in vivo by inducing cytoskeletal changes in macrophages. In mice, SLIT2 attenuates the uptake of muramyl dipeptide, thereby preventing NOD2-dependent activation of NF-κB and consequent secretion of pro-inflammatory chemokine, CXCL1. Conversely, blocking the action of endogenous SLIT2 enhances CXCL1 secretion. SLIT2 also inhibits macropinocytosis in RAS-transformed cancer cells, thereby decreasing their survival in nutrient-deficient conditions which resemble tumor microenvironment. Our results identify SLIT2 as a physiological inhibitor of macropinocytosis and challenge the conventional notion that signals that enhance macropinocytosis negatively regulate cell migration, and vice versa.

Macrophages, which are specialized cells of the innate immune system, perform diverse functions, including host defense against pathogens1. Pro-inflammatory macrophages use various tools to combat invading microorganisms, including secretion of inflammatory mediators, phagocytosis of the pathogen, and production of reactive oxygen species. These cellular processes are carefully regulated by dynamic changes in the macrophage cytoskeleton2. During immune surveillance, macrophages, and immature dendritic cells (iDCs) constitutively sample their extracellular surroundings via macropinocytosis, a phenomenon brought about by active plasma membrane ruffling induced by remodeling of the cortical cytoskeleton3,4. These cells also internalize bacterial pathogen-associated molecular patterns (PAMPs), such as muramyl dipeptide (MDP), via macropinocytosis, thereby initiating an immune response5. Intracellular pathogens, including some viruses, exploit macropinocytosis to facilitate their entry into the host cells6,7. Several growth factors, including colony stimulating factor 1 (CSF1), secreted by the host, as well as some pathogen-derived proteins such as IpaC from Shigella are known to induce macropinocytosis3,8,9. It is estimated that macrophages can consume liquid equal to their cell volume every hour10. This would leave them particularly vulnerable to the entry of soluble PAMPs and viruses, via macropinocytosis, under specific conditions.

Macropinocytosis is a near-universal feature of cancers driven by mutations in the RAS family of oncogenes11,12,13. RAS-transformed tumors account for approximately one third of all malignancies in humans14. Among the RAS family members, KRAS mutations are the most common in cancer, especially in adenocarcinomas such as pancreatic adenocarcinoma (PDAC)15, and they are associated with poorer response to the conventional therapies and worse prognosis16. KRAS-transformed cancer cells use macropinocytosis to internalize large proteins, such as albumin, from their extracellular environment, to subsequently break them down into amino acids, which enter the cell metabolism to promote cancer cell survival11. This might be particularly relevant in the tumor core, which contains low concentrations of amino acids, including glutamine (Glut)17. Inhibition of macropinocytosis could be of therapeutic benefit in this context11. However, no physiological factor that can inhibit macropinocytosis has been identified.

Spatiotemporal regulation of activity of Rho-family of small GTPases is essential to bring about localized and reversible changes in the cellular actin cytoskeleton18. Accordingly, multiple steps during macropinocytosis are regulated by small Rho GTPases, Rac1 and Cdc423,19. The conserved Slit family of neuronal guidance cues, together with their transmembrane Roundabout (Robo) receptors, act as repellents during development of the central nervous system by regulating actin cytoskeletal rearrangements in migrating neurons and projecting axons20. Slit proteins (SLIT1-3) undergo proteolytic cleavage in vivo, with N-terminal fragments binding to Robo receptors to induce signaling21. A growing number of recent studies suggest that Slit-Robo signaling also has potent, localized, tissue-specific effects outside the nervous system22,23,24,25. Of the four mammalian Robo receptors (ROBO1-4), ROBO3 does not bind to Slit proteins26,27, and ROBO4 is exclusively expressed in endothelial cells28. Among Slit proteins, SLIT1 is predominantly expressed in the nervous system, while SLIT2 and to a lesser extent SLIT3 are also found in peripheral tissues24,29. We and others have shown that SLIT2, together with ROBO1, inactivates Rac1/2 to regulate actin networks in immune cells, and as a result, inhibits the directed migration of neutrophils29,30,31, monocytes32, dendritic cells33, and T lymphocytes34 toward chemotactic stimuli in vitro and in vivo31,32,33,35. Although primary murine macrophages were recently shown to express ROBO1 and ROBO324, the actions of SLIT2 on macrophage functions have been largely unexplored.

The role played by Slit-Robo signaling in tumor progression vs. suppression also remains a topic of active investigation and appears to be context dependent36,37. Nonetheless, SLIT2 has been shown to be silenced in several invasive tumors and in cancer cell lines, including PDAC cells, and conversely, high levels of SLIT2 mRNA are associated with suppressed tumor growth in vivo38,39. Although SLIT2-induced tumor suppression has been presumed to be due inhibition of cancer cell migration, its effects on cancer cell growth have not been carefully elucidated.

We report here that SLIT2 prevents macrophage spreading not by inhibiting Rac activation, but rather, by activating RhoA. We further demonstrate that SLIT2 can inhibit macropinocytic uptake of bacterial PAMPs and subsequent upregulation of inflammatory chemokines in vivo, thereby modulating macrophage immune responses. We show that SLIT2, in a ROBO1-dependent manner, negatively impacts the survival of KRAS-transformed cancer cells by inhibiting macropinocytosis, thus limiting protein uptake in a Glut-deprived state similar to the tumor microenvironment found in vivo. Our work identifies an endogenous inhibitor of macropinocytosis with potent effects both in vitro and in vivo.

NSlit2 induces actin remodeling in macrophages in a ROBO1-dependent manner

We first investigated which Slit-binding Robo receptors are expressed in macrophages and found that Robo1, but not Robo2, messenger RNA (mRNA) is expressed in the RAW264.7 macrophage cell line and in primary murine bone marrow-derived macrophages (BMDM) (Fig. 1a). Using immunoblotting, ROBO1 protein was detected in RAW264.7 cells as well as in primary macrophages derived from human peripheral blood mononuclear (MDM) cells (Supplementary Fig. 1a). To investigate the effects of SLIT2 on macrophage cytoskeleton, we incubated cells with control vehicle, or with recombinant bioactive NSlit2 (the amino-terminal fragment of SLIT2), or with other recombinant SLIT2 proteins, CSlit2 and Slit2ΔD2, which do not bind to the Robo receptors (Fig. 1b)22,30,32. The endotoxin levels in all recombinant Slit protein preparations were below 0.05 EU/ml (Supplementary Fig. 1b; Supplementary Table 1). Treatment with NSlit2 significantly reduced macrophage spreading, as seen by reduction in their total three-dimensional surface area (Fig. 1c, d and Supplementary Fig. 1c). To quantify the circularity of cells, we calculated the ‘shape factor’ using Volocity 6.3 software40. Shape factor values range from 0 to 1, the latter defining a perfectly circular shape (see “Methods” for mathematical calculation). Only NSlit2 treatment significantly increased rounding in human and murine macrophages (Fig. 1c, e; Supplementary Fig. 1d). To determine whether these effects of NSlit2 were dependent on ROBO1, we used a specific siRNA to knockdown its expression in RAW264.7 cells (Fig. 1f; Supplementary Fig. 1e). NSlit2-induced inhibition of cell spreading was abolished by ROBO1 knockdown (Fig. 1g). Furthermore, in ROBO1-deficient macrophages, NSlit2 failed to induce the rounding (Fig. 1h). To validate this finding using an independent experimental approach, we pre-incubated NSlit2 with the soluble N-terminal fragment of the human ROBO1 receptor (Robo1N) containing the NSlit2-binding Ig1 domain41. Robo1N has previously been shown to act as an NSlit2 antagonist22,42. Treatment with Robo1N alone did not have any effect on cell spreading or rounding (Supplementary Fig. 1f, g). However, Robo1N-treated NSlit2 failed to reduce cell spreading (Supplementary Fig. 1f) as well as to increase cell rounding (Supplementary Fig. 1g). These findings suggest that actions of NSlit2 on the macrophage cytoskeleton are mediated by its binding to the ROBO1 receptor.

a Total RNA was isolated from murine brain and kidney, as well as from RAW264.7 cells and primary murine bone marrow-derived macrophages (BMDM). Robo1 and Robo2 mRNA expression were investigated by RT-PCR. Gapdh was used as a loading control. b A schematic representation of full-length human SLIT2 isoform 1 (accession number- NP_004778.1) protein, and recombinant NSlit2, CSlit2, and Slit2ΔD2 proteins used in this study. c RAW264.7 cells were incubated with vehicle (phosphate-buffered saline), NSlit2, CSlit2, or Slit2ΔD2 for 15 min at 37 °C and allowed to spread on poly-D-lysine-coated coverslips for 1 h. Cells were fixed, permeabilized, and incubated with AF-488-conjugated phalloidin (green) to label polymerized actin. Scale bar, 25 μm. d, e, g, h All data are presented as mean ± standard error of mean (SEM). Comparisons between the groups were made by one-way analysis of variance (ANOVA), followed by post hoc Tukey’s multiple comparison test. n = 50 cells per treatment group per experiment over three independent experiments. d Experiments were performed as in (c) and cell surface area was measured using Volocity 6.3 software. ****p < 0.0001, NSlit2 vs vehicle, CSlit2, or Slit2ΔD2. e Shape factor for cells in (c) using Volocity 6.3 software. ****p < 0.0001, NSlit2 vs vehicle, CSlit2, or Slit2ΔD2. f ROBO1 levels were knocked down in RAW264.7 cells using specific siRNA. Protein levels (band intensity) were quantified using ImageJ software, version 1.51v and normalized to corresponding β-actin levels. n = 3 independent experiments (Supplementary Fig. 1e). Comparison between the two groups was made by unpaired, two-tailed t-test. ***p = 0.0005, scrambled vs Robo1 siRNA. g ROBO1 protein levels were knocked down in RAW264.7 cells. After 72 h, experiments were performed as in (c). Cell surface area was measured as in (d). ****p < 0.0001 for the indicated comparisons and p = 0.9887, NSlit2 vs Slit2ΔD2 in ROBO1 knockdown conditions. (h) Shape factor for cells in (g) was determined as in (e). ****p < 0.0001 for the indicated comparisons and p = 0.4268, NSlit2 vs Slit2ΔD2 in ROBO1 knockdown conditions. Source data for (a, d, e, f–h) are provided as a Source Data file.

Full size image

SLIT2 has previously been shown to inhibit human and murine monocyte chemotaxis by inactivating Rac132. In addition, it has been reported that NSlit2 can inactivate Rac1 and Cdc42 GTPases in RAW264.7 cells43,44. To determine whether inactivation of Rac1 is responsible for NSlit2-induced inhibition of macrophage spreading, we transiently expressed a constitutively active form of RAC1(Q61L)45 in RAW264.7 cells. In line with previous reports46,47, the Q61L mutant increased macrophage spreading (Supplementary Fig. 2a). The Q61L mutant, however, failed to prevent the decrease in cell spreading following exposure to NSlit2 (Supplementary Fig. 2a). To further validate this finding, we studied the effects of NSlit2 on primary BMDM from Rac1/2 double knockout (Rac1/2 DKO) mice. Rac1/2 DKO macrophages were smaller than the wild-type (WT) counterparts (Fig. 2a). However, exposure of Rac1/2 DKO macrophages to NSlit2 further reduced their spreading (Fig. 2a). The Arp2/3 protein complex is an important downstream effector of both Rac1 and Cdc42 GTPases mediating actin polymerization48. Exposure to CK-666, a potent Arp2/3 inhibitor49, could not reverse the NSlit2-induced cell spreading phenotype (Supplementary Fig. 2b). The results indicate that NSlit2-induced inhibition of macrophage spreading is independent of its actions on Rac1/2 and Cdc42 activities.

(a, d, e, f) Data are presented as boxplots where middle line is the median; lower and upper hinges correspond to the first and third quartiles; whiskers represent minimum and maximum values. n = 50 cells per treatment group per experiment over three independent experiments. Comparisons between the groups were made by one-way ANOVA in (b, e, f) and two-way ANOVA in (a, d), followed by post hoc Tukey’s multiple comparison test. a Primary BMDM from wild-type (WT) or Rac1/2 double knockout (DKO) mice were cultured for 10 days, and cell spreading assays were performed as described in Fig. 1c. Cell surface area was measured as in Fig. 1d. ****p < 0.0001, WT vs DKO BMDM and p < 0.0001, NSlit2 vs vehicle, CSlit2, or Slit2ΔD2. b RAW264.7 macrophages were incubated with vehicle, NSlit2, CSlit2 or Slit2ΔD2 for 15 min at 37 °C and active and total RhoA levels were measured using a G-LISA assay. **p = 0.0063, 0.0093, and 0.0053 for NSlit2 vs vehicle, CSlit2, Slit2ΔD2, respectively. c RAW264.7 cells were incubated with the RhoA/B/C inhibitor, TAT-C3, for 4 h followed by vehicle, NSlit2, CSlit2, or Slit2ΔD2 treatment for 15 min and cell spreading was performed as described in Fig.1c. Scale bar, 20 μm. d Surface area for cells in (c) was measured as in Fig. 1d. ****p < 0.0001, vehicle vs TAT-C3 and p = 0.8877, 0.9470, and 0.9952 for NSlit2 vs vehicle, CSlit2, Slit2ΔD2 in TAT-C3-treated conditions, respectively. e, f Experiments were performed as in Fig. 1c but cells were first incubated with a formin inhibitor, SMIFH2, for 30 min. e Cell surface area was measured as in Fig. 1d. p = 0.9220, 0.8356, and 0.7215 for NSlit2 vs vehicle, CSlit2, Slit2ΔD2, respectively. f Shape factor was measured as in Fig. 1e. p = 0.7557, 0.6594, and 0.9950 for NSlit2 vs vehicle, CSlit2, Slit2ΔD2 respectively. Source data for (a, b, d, e, f) are provided as a Source Data file.

Full size image

NSlit2 activates RhoA and results in cytoskeletal rearrangement in macrophages

We asked whether NSlit2 inhibited cell spreading through its effects on RhoA, which is the most abundantly expressed Rho-family GTPase member in monocytes/macrophages50. We used absorbance-based G-LISA assays to measure total and active RhoA levels in RAW264.7 cells. NSlit2 treatment significantly increased the active RhoA levels (Fig. 2b). We next tested the effects of TAT-C3 exoenzyme, a cell-permeable inhibitor of RhoA, RhoB, and RhoC, but not Rac1 nor Cdc4251, on macrophage spreading. Cells incubated with TAT-C3 were significantly larger than the control counterparts (Fig. 2c, d). In macrophages treated with TAT-C3, NSlit2 no longer inhibited cell spreading (Fig. 2c, d). To verify this, we reduced the expression of RhoA using a siRNA (Supplementary Fig. 2c). Exposure to NSlit2 failed to attenuate spreading in RhoA-deficient cells (Supplementary Fig. 2d). Overall, these data indicate that NSlit2 inhibits the spreading of macrophages primarily through the activation of RhoA.

Several downstream effectors of RhoA, such as Rho-associated coiled-coil containing protein kinases (ROCK1/2) and diaphanous-related formins, are known to regulate actin polymerization via distinct mechanisms52. To determine which of these two pathways mediates NSlit2-induced cytoskeletal changes, we selectively blocked their actions. In the presence of Y-27632, a ROCK1/2 inhibitor53, NSlit2 still inhibited macrophage spreading (Supplementary Fig. 2e). In contrast, SMIFH2, an inhibitor of formin homology 2 domains54, completely abrogated the NSlit2-induced reduction in cell surface area (Fig. 2e; Supplementary Fig. 2f) and increase in the cell rounding (Fig. 2f; Supplementary Fig. 2f). These findings suggest that NSlit2 inhibits macrophage spreading primarily by augmenting RhoA-mediated formin activation.

NSlit2-induced RhoA activation in macrophages is mediated by inactivation of MYO9B

Backer S. et al. recently reported that NSlit2 activates TRIO, a dual Rho/Rac guanine nucleotide exchange factor (GEF), in murine embryonic fibroblasts and neuronal cells55. We found that Trio mRNA is expressed at low levels in RAW264.7 cells but not in primary murine BMDM (Supplementary Fig. 3a). Liu et al. have shown that NSlit2 mediates RhoA activation, via inactivation of the Fyn kinase, a member of the Src-family tyrosine kinases, in rat oligodendrocyte precursor cells, but not in mature oligodendrocytes56. SLIT2-ROBO1-signaling has also been shown to mediate pan-Src activation in cancer cells