ArfGAP1 restricts Mycobacterium tuberculosis entry by controlling the actin cytoskeleton

Abstract

The interaction of Mycobacterium tuberculosis (Mtb) with pulmonary epithelial cells is critical for early stages of bacillus colonization and during the progression of tuberculosis. Entry of Mtb into epithelial cells has been shown to depend on F‐actin polymerization, though the molecular mechanisms are still unclear. Here, we demonstrate that mycobacterial uptake into epithelial cells requires rearrangements of the actin cytoskeleton, which are regulated by ADP‐ribosylation factor 1 (Arf1) and phospholipase D1 (PLD1), and is dependent on the M3 muscarinic receptor (M₃R). We show that this pathway is controlled by Arf GTPase‐activating protein 1 (ArfGAP1), as its silencing has an impact on actin cytoskeleton reorganization leading to uncontrolled uptake and replication of Mtb. Furthermore, we provide evidence that this pathway is critical for mycobacterial entry, while the cellular infection with other pathogens, such as Shigella flexneri and Yersinia pseudotuberculosis, is not affected. Altogether, these results reveal how cortical actin plays the role of a barrier to prevent mycobacterial entry into epithelial cells and indicate a novel role for ArfGAP1 as a restriction factor of host–pathogen interactions.

Introduction

Many pathogenic bacteria are able to interfere with host signaling to enter and replicate within cells during infection 1, 2, 3. Phagocytic cells, such as macrophages and dendritic cells, engulf bacteria upon engagement of phagocytic receptors, while many bacterial virulence factors interfere with the host at later stages of the phagocytic process, for example, by subverting phagosome maturation or to achieve phagosomal escape. In contrast, many pathogens are also able to enter non‐phagocytic cells, such as epithelial cells and fibroblasts, by actively interfering with the actin cytoskeleton to trigger pathogen uptake. For example, Listeria monocytogenes and Yersinia pestis use bacterial surface proteins to induce Rac1/Cdc42‐mediated actin rearrangement to enter the host 4, 5, while Salmonella typhimurium and Shigella flexneri apply a type III secretion system to trigger Cdc42/Rac1/RhoA‐dependent actin reorganization 6, 7.

Mycobacterium tuberculosis (Mtb) is the causative agent of tuberculosis, which remains a leading infectious disease around the world with 1.8 million deaths and more than 10 million new cases each year 8. The pathology of tuberculosis is directly linked to the tight interplay between the host immune response and the persistence of Mtb infection 9. Although mostly alveolar macrophages and dendritic cells are colonized by Mtb, several findings indicate that non‐phagocytic cells are an essential niche for this pathogen to promote its replication and dissemination from the lung to other organs 10, 11, 12. In pulmonary epithelial cells, the redistribution of actin filaments was shown to be altered upon Mtb infection and is important for pathogen entry 13. Although a participation of macropinocytosis in Mtb uptake has been demonstrated 14, further examination showed that Mtb entry relies also on other, yet to be characterized, cellular components to enter into non‐phagocytic cells 15. Furthermore, it has been suggested that the trafficking pattern of Mtb in alveolar epithelial cells differs from the one observed in macrophages, since Mtb‐containing vacuoles show features of mature, late endosomes and were labeled with Rab5 and Rab7 16. In addition, it was reported that mycobacterial heparin‐binding hemagglutinin (HBHA) reduces G‐actin polymerization into actin filaments and blocks autophagy pathways in epithelial cells 17, 18, 19. Recently, it has been shown that HBHA interacts with syndecan 4, a mycobacterial attachment receptor in epithelial cells that promotes mycobacterial entry 20. Furthermore, it also has been shown that the Mtb vaccine strain, Mycobacterium bovis BCG, is able to manipulate G protein‐coupled receptors (GPCRs) in epithelial cells, such as CXCR1 and CXCR2, resulting in Rac1 upregulation and altered actin cytoskeleton signaling 21. However, despite these recent findings, the precise mechanisms regulating Mtb entry and persistence in epithelial cells remain to be elucidated.

Here, we aimed to identify the main players influencing Mtb infection of epithelial cells using a genomewide RNAi screen, similar to the one reported previously in macrophages 22, and investigated the underlying mechanisms that lead to the colonization of this cell type. We identified Arf GTPase‐activating protein 1 (ArfGAP1) as a key host restriction factor of Mtb entry. By dissecting the molecular mechanisms that ArfGAP1 engages to modulate mycobacterial colonization, we characterized a novel signaling pathway involved in regulating the uptake of mycobacteria into epithelial cells. We found that Mtb infection induced ADP‐ribosylation factor 1 (Arf1) and phospholipase D1 (PLD1) activation downstream of the muscarinic receptor 3 (M₃R). M₃R is belonging to the GPCR family and was originally identified as the major receptor for acetylcholine neurotransmitter for regulating physiological activities in various organs 23. Knock‐down (KD) of ArfGAP1 resulted in actin stress fiber formation activating massive colonization of epithelial cells by Mtb. Importantly, we demonstrated that these actin cytoskeleton rearrangements favor specifically mycobacterial entry but not uptake of other bacteria, such as Shigella flexneri and Yersinia pseudotuberculosis, and that this pathway is specific to epithelial cells and not to other cell types. Therefore, our findings demonstrate a novel role for ArfGAP1 and M₃R in the regulation of host–pathogen interactions in epithelial cells by modulating and controlling the actin cytoskeleton.

Results and Discussion

To assess the interactions between epithelial cells and Mtb, we decided to identify host effector proteins that restrict the colonization of the pathogen during infection. We applied a phenotypic assay based on genomewide RNAi screening, which we developed previously 24, 25, to human lung epithelial A549 cells. This system relies on automated confocal microscopy and allowed us to quantify the intracellular localization of the EGFP‐expressing Mtb H37Rv strain (H37Rv‐GFP) inside infected A549 cells. Briefly, 3 days before the infection epithelial cells were transfected with smart pool siRNAs from a genomewide library (targeting 16,532 human genes; Fig 1A). Cells were then infected with H37Rv‐GFP for 5 h, and extracellular bacilli were removed by extensive washing and amikacin treatment. The infected cells were incubated for 5 days, labeled by DAPI followed by image acquisition using an automated confocal microscope 24, 25. Customized image analysis was used for the quantification of relevant parameters, such as percentage of infected cells compared to the total number of cells 26, and non‐targeting siRNA (scramble) was used as control. Classification of hits was based on Z scores taking into account the percentage of infected cells compared to the cell number (Fig 1B). Among the genes that had an impact on Mtb colonization, silencing of 109 genes resulted in increased Mtb infection rates of A549 cells, including the genes encoding the anti‐inflammatory cytokine IL10 and the transcription factor XBP1, which is involved in the cellular stress response. Both molecules were shown previously to be affected by Mtb infection in epithelial cells 27, 28 confirming our findings. Interestingly, among the most significant siRNA hits, we identified ArfGAP1 that dramatically increased mycobacterial infection of epithelial cells upon its silencing (Fig 1B) without affecting the intracellular distribution of ArfGAP1 (Appendix Fig S1). In order to validate our finding of the siRNA screen, we investigated Mtb uptake in epithelial cells 3 days after transfecting them with ArfGAP1 siRNA. Compared to control cells, we found a threefold increase in the percentage of infected ArfGAP1 KD cells already 1 h post‐infection (Fig 1C, left panel). In addition to the overall Mtb infection rate, we also found that ArfGAP1 KD cells were infected by more mycobacteria per cell at all time points (Fig 1C, right panel). These findings are not restricted to the investigated virulent Mtb strain, because a similar increase in mycobacterial colonization was found in ArfGAP1 KD cells upon infection with an attenuated, EGFP‐expressing Mtb strain (H37Ra‐GFP) (Fig EV1A–E). In agreement with this, Mtb colonization was also enhanced in control epithelial cells by adding QS11, a chemical inhibitor of ArfGAP1 function (Fig EV1F). Since ArfGAP1 is well known for its function in the formation of COPI vesicles at the Golgi 29, we addressed the possibility that ArfGAP1 silencing interferes with Golgi‐dependent transport to the cell surface, for example, of an inhibitory factor, which might account for the observed effects on Mtb uptake. Therefore, we analyzed Golgi structure and integrity as well as secretory trafficking upon ArfGAP1 KD by confocal and spinning disk microscopy, respectively. While ArfGAP1 KD did not alter Golgi integrity compared to scramble cells, as shown by the two Golgi markers TGN46 and golgin97, Arf1 KD induced a fragmentation of the trans‐Golgi network (Appendix Fig S2A–D), similar than brefeldin A (BFA) treatment 30. ArfGAP1 silencing also did not interfere with the intracellular distribution and Golgi‐associated accumulation of Rab6, a GTPase that is regulating secretory trafficking from the Golgi to the cell surface 31 (Appendix Fig S2E and F). Moreover, we also did not observe differences between scramble and ArfGAP1 KD cells in the formation and exit of neuropeptide Y, a secretory cargo known to be transported from the Golgi in a Rab6‐dependent manner 31, 32 (Movies EV1 and EV2). These findings exclude the possibility that ArfGAP1 silencing alters Golgi‐dependent transport and suggest that ArfGAP1 function on the cytoskeleton is restricting Mtb uptake into epithelial cells. Altogether, our data demonstrate that ArfGAP1 works as a critical host restriction factor regulating the intracellular colonization of epithelial cells by Mtb.

Figure 1. Genomewide siRNA screening identified ArfGAP1 as a host factor restricting Mtb colonization of epithelial cells.

• A, B
  A549 epithelial cells were transfected with pooled siRNA in 384‐well plates and infected with Mtb H37Rv‐GFP at an MOI of 5. Five days post‐infection, images were acquired by automated confocal microscopy followed by image analysis. Shown are the applied workflow (A) and a Z score scatter plot of the obtained data (B, upper panel). Among the most significant siRNA hits, ArfGAP1 was identified in Mtb‐infected cells. A representative confocal image of scramble and siArfGAP1‐transfected cells is shown below (B, lower panel). Cell nuclei were stained by DAPI (blue), while GFP‐expressing Mtb are shown in green. Scale bars: 50 μm.
• C, D
  The impact of ArfGAP1 silencing on Mtb uptake (C) and intracellular replication of Mtb (D) were analyzed in infected A549 cells and by automated confocal microscopy. Shown are the applied workflow (upper panel) and results of three independent experiments (lower panel). Both overall infection rates (left histogram in C and D) and bacterial load per cell (right histogram in C and D) are indicated. Data analysis was carried out to obtain percentages of infected cells (n ≥ 700) and quantified areas of intracellular bacteria (px: pixels). Data are presented as mean ± SEM. ***P < 0.0005, ns: not significant (Student's t‐test).

Figure EV1. Effect of ArfGAP1 silencing on mycobacterial colonization using an attenuated Mtb strain.

• A, B
  Scramble and ArfGAP1 KD cells were infected with Mtb H37Ra‐GFP at MOI 10 and analyzed 4 days post‐infection by automated confocal microscopy. Shown are representative images (A) and Mtb colonization of cells (B). Cell nuclei were stained by DAPI (blue), while GFP‐expressing Mtb are shown in green. Scale bars: 50 μm.
• C, D
  Similarly, these cells were also lysed and plated at different dilutions (C) to determine colony‐forming units (CFU) (D).
• E
  The impact of ArfGAP1 silencing on Mtb H37Ra uptake at MOI 5 was analyzed by automated confocal microscopy at the indicated time points post‐infection.
• F
  A549 cells were treated with the ArfGAP1 inhibitor QS11 and analyzed for Mtb H37Ra colonization at MOI 5 after 4 h post‐infection.

Data information: Data are presented as mean ± SEM. ***P < 0.0005, **P < 0.005, *P < 0.05, ns: not significant (Student's t‐test). Histograms display at least 1,400 (B), 700 (E), 400 (F) analyzed cells in three replicates (B), in four replicates (E), and in three replicates (F). Shown are representative examples of two (B, D, F) or three (E) independent experiments.

In addition to the findings of Mtb uptake, we also infected cells with Mtb at a higher MOI followed by ArfGAP1 silencing 2 h after the infection to analyze a possible influence on the intracellular replication of Mtb in A549 epithelial cells. While there was no difference 1 day post‐infection, we observed an increased bacterial load in ArfGAP1 KD cells compared to scramble cells 3 and 5 days post‐infection (Fig 1D), though it was less pronounced compared to the impact of ArfGAP1 silencing on Mtb uptake. These results suggest that the absence of ArfGAP1 also has an impact on mycobacterial replication. It is possible that ArfGAP1 is able to regulate processes that have an impact on phagosomal membrane integrity and rupture, which would allow Mtb to get access to the cytosol and to nutrients, which could influence Mtb replication rates. However, this would need to be further investigated in future studies. Here, we decided to focus on the role of ArfGAP1 on Mtb uptake by epithelial cells.

ArfGAP1 is known to control Arf1‐dependent membrane trafficking by facilitating the hydrolysis of the active, GTP‐bound form of Arf1 to its inactive, GDP‐bound conformation 33, 34. Therefore, we investigated next whether mycobacterial entry into epithelial cells involves a signaling complex comprising Arf1 and other associated proteins. We analyzed the amount of activated Arf1‐GTP by pull‐down and Western blotting and found, as expected, that ArfGAP1 KD of cells increased the levels of Arf1‐GTP compared to control cells (Fig 2A). While levels of Arf1‐GTP increase rapidly upon Mtb infection and reach a peak after 1 h post‐infection (Fig 2B), ArfGAP1 expression only starts to increase significantly 1–3 h after Mtb infection (Fig 2C). Furthermore, the KD of Arf1 or of GRP1/ARNO3, the guanine nucleotide exchange factor (GEF) of Arf1 35, both reduced dramatically mycobacterial colonization (Fig 2D). Similarly, also the presence of BFA, an inhibitor of ArfGEF function 30, 36, induced a decrease in mycobacterial colonization (Fig 2E). Of note, already sub‐nanomolar concentrations of BFA were able to impair bacterial uptake, suggesting that this effect is due to the absence of activated Arf1. These findings demonstrate that the initial activation of Arf1 is essential to promote mycobacterial entry, while ArfGAP1 expression increases downstream of these events, suggesting a negative feedback loop that downregulates Arf1 activity. The Arf GTPase activation state depends strongly on the activity of phospholipases (PLD), as it has been demonstrated previously in other cell types 37. In particular, ArfGAP1 expression is known to be induced by PLD1‐mediated biosynthesis of phosphatidylinositol 38. Therefore, we decided to investigate PLD activity in Mtb‐infected, epithelial cells. Similar to Arf1‐GTP function, PLD activity was found to increase during Mtb infection (Fig 2F) and when Arf1GAP1 was silenced (Fig 2G). In contrast, absence of Arf1 in epithelial cells decreased PLD activity significantly. In order to verify which specific phospholipase is interfering with Mtb infection, we knocked down either PLD1 or PLD2 in epithelial cells and found that only the absence of PLD1 increased Mtb infection rate of A549 cells (Fig 2H). Moreover, we also confirmed by Western blotting that PLD1 KD is reducing the cellular expression of ArfGAP1 (Appendix Fig S3). Taken together, these results demonstrate that Mtb entry into A549 cells relies on a pathway that includes Arf1 and PLD1 activity, which is negatively controlled by ArfGAP1.

Figure 2. ArfGAP1, Arf1, and PLD1 signaling modulate Mtb entry into epithelial cells.

• A–C
  A549 cells were transfected with scramble and ArfGAP1 siRNA and analyzed for active Arf1‐GTP by pull‐down and Western blotting (A). Tubulin was used as loading control. Similarly, Arf1‐GTP was analyzed in cells infected with Mtb H37Rv at an MOI of 5 at the indicated times post‐infection (B). Mtb‐infected cells were also analyzed for the expression of ArfGAP1 by Western blotting (C). Shown are representative blots of two independent experiments.
• D
  A549 cells were transfected with scramble, Arf1, and GRP1 siRNA, and Mtb colonization at MOI 20 was analyzed after 4 h.
• E
  A549 cells were treated with the indicated concentrations of BFA for 18 h prior infection at MOI 20, and Mtb colonization was analyzed after 4 h by automated confocal microscopy.
• F, G
  PLD activity was measured enzymatically using the Amplex Red reagent in Mtb‐infected A549 cells at MOI 5 (F) and in cells that were transfected with scramble, Arf1, and ArfGAP1 siRNA (G) (n = 3).
• H
  Mtb colonization at MOI 20 was also analyzed after 4 h by automated confocal microscopy in cells transfected with scramble, PLD1‐specific siRNA, and PLD2‐specific siRNA.

Data information: Data are presented as mean ± SEM. ***P < 0.0005, **P < 0.005, *P < 0.05, ns: not significant (Student's t‐test). Histograms display at least 200 analyzed cells in seven replicates (D), in three replicates (E), and six replicates (H). Shown are representative examples of two (D, E, H) or three (F, G) independent experiments. Source data are available online for this figure.

The correlation between Mtb colonization and Arf1 activity suggests a regulation of mycobacterial uptake by cytoskeletal organization, which is controlled by PLD1‐dependent signaling. Many GPCRs at the cell surface were shown to induce PLD activity and to facilitate Arf‐associated signaling pathways 39. Interestingly, the muscarinic receptor M₃R has been shown to induce the BFA‐sensitive activation of PLD 40 and in particular the Arf1‐dependent route of PLD1 activation 41. Furthermore, we tested the impact of a selective KD of different receptors, which are known to interact with Arf1 (Fig EV2A), on Mtb colonization of epithelial cells during additional ArfGAP1 KD. The increased Mtb infection rate induced upon ArfGAP1 silencing (Fig 1C) was found for all tested receptors except for M₃R (Fig EV2B), which suggests that M₃R is a putative candidate receptor that is involved in Mtb uptake. Interestingly, silencing of the alpha‐IIb/beta‐3 integrin ITGA2B in addition to the simultaneous KD of ArfGAP1 increased Mtb uptake compared to control conditions. We did not find any reported association between ITGA2B and ArfGAP1. It is possible that the known function of ITGA2B in cell adhesion 42 in addition to the simultaneous KD of ArfGAP1 leads to local actin rearrangements that further enhance Mtb entry, which would need to be investigated in future. The KD of M₃R in A549 cells led to a reduction in Mtb colonization (Fig 3A and B, open bars), and importantly, was sufficient to inhibit the massive colonization of Mtb induced upon ArfGAP1 KD (Fig 3B, black bars; Fig EV2B). Moreover, the increased uptake of Mtb induced upon ArfGAP1 KD was also reduced by the addition of the M₃R antagonist Zamifenacin, which decreased Mtb colonization at increasing concentration (Fig 3C). These findings indicate that ArfGAP1 functions downstream of M₃R. In addition, M₃R as well as Arf1 silencing induced rearrangements in the intracellular distribution of ArfGAP1 (Appendix Fig S4), further emphasizing the functional association between ArfGAP1 and these proteins. Knock‐down of M₃R also led to reduced levels of Arf1‐GTP, as shown by pull‐down and Western blotting of Mtb‐infected cells (Fig 3D), demonstrating that Mtb internalization is facilitated by M₃R, which leads to the activation of Arf1. In turn, Arf1‐GTP triggers PLD1 activity recruiting ArfGAP1, which negatively controls this signaling pathway by hydrolyzing Arf1‐GTP into Arf1‐GDP and decreasing PLD1 activity. As mentioned before, some pathogens induce actin rearrangements by Rac1, RhoA, and Cdc42 to enter cells 2. We tested this possibility and investigated Mtb colonization in epithelial cells upon KD of these proteins in the absence or presence of additional ArfGAP1 KD. In the absence of additional ArfGAP1 KD, only Arf1 KD efficiently reduced the Mtb infection rate compared to control cells, while KD of Rac1, RhoA, and Cdc42 did not or only slightly decreased Mtb colonization of A549 cells (Fig EV3A). The presence of additional ArfGAP1 KD did not result in increased Mtb uptake when Arf1 was knocked down, demonstrating that ArfGAP1 acts as a restriction factor of Mtb colonization in an Arf1‐dependent manner (Fig EV3B). In contrast, KD of Rac1, RhoA, Cdc42, and scramble siRNA in the presence of additional ArfGAP1 KD resulted in increased Mtb uptake or had no influence on Mtb colonization. These findings show that the observed entry of Mtb into epithelial cells occurs independent of the mechanisms of actin rearrangements that are induced by other pathogens.

Figure EV2. Impact of the silencing of receptors known to interact with Arf1 in presence and absence of ArfGAP1 silencing.

1. Known relationships of Arf1 with different receptors obtained by Ingenuity Pathway Analysis (IPA).
2. ArfGAP1 KD cells were transfected with scramble siRNA and the indicated targeting siRNA and analyzed for Mtb H37Ra colonization at MOI 20 4 h post‐infection. Data are presented as mean ± SEM. ***P < 0.0005, *P < 0.05, ns: not significant (Student's t‐test). Histograms display at least 400 analyzed cells in six replicates. Shown are representative examples of two independent experiments.

Figure 3. Mtb entry is facilitated by the muscarinic receptor M₃R.

• A, B
  Impact of M₃R silencing on Mtb H37Rv colonization in A549 cells at MOI 20 was analyzed after 4 h by automated confocal microscopy. Shown are representative images (A), overall infection rates (left histogram in B), and bacterial load per cell (right histogram in B) in the absence and presence of additional ArfGAP1 silencing. Scale bar: 25 μm.
• C
  Mtb infection at MOI 20 was also analyzed after 4 h in ArfGAP1 KD cells using the M₃R inhibitor Zamifenacin.
• D
  Active Arf1‐GTP was measured by pull‐down and Western blotting in A549 cells transfected with scramble and M₃R siRNA. Shown are a representative WB image (left panel) and the quantification of two independent experiments (right panel).

Data information: Data are presented as mean ± SEM. ***P < 0.0005, **P < 0.005, *P < 0.05, ns: not significant (Student's t‐test). Data are derived from at least 500 analyzed cells of six replicates (B) and four replicates (C), respectively. Source data are available online for this figure.

Figure EV3. Involvement of regulators of actin cytoskeleton rearrangements in Mtb colonization.

• A, B
  A549 cells were transfected with scramble siRNA and the indicated targeting siRNA in the absence (A) and presence of additional ArfGAP1 silencing (B). Mtb H37Rv colonization of transfected cells at MOI 20 was analyzed by automated confocal microscopy 4 h post‐infection. Data are presented as mean ± SEM. ***P < 0.0005, **P < 0.005, ns: not significant (Student's t‐test). Histograms display at least 400 analyzed cells in six replicates. Shown are representative examples of two independent experiments.

Next, we asked how F‐actin reorganization could influence Mtb infection of epithelial cells. A549 cells were transfected with GFP‐actin before Mtb infection and analyzed by automated confocal microscopy. We found that in infected control cells, single mycobacteria induced local actin rearrangements in close vicinity of their contact sites, which we detected as spots and patches (Fig 4A, upper panel) that increased constantly over time (Fig 4B). In contrast, ArfGAP1 KD not only increased Mtb colonization but also actin stress fiber formation, which mycobacteria seemed to use for their entry process (Fig 4A, lower panel). The formation of actin patches during mycobacterial infection has been found previously in macrophages 43, 44.To further analyze this aspect, we applied structured illumination microscopy (SIM) on phalloidin‐labeled epithelial cells to investigate F‐actin distribution in detail. While F‐actin in control cells was rather concentrated in cortical areas, where it is involved in homeostatic turnover as reported previously 45, the KD of ArfGAP1 led to a drastic accumulation of actin throughout the entire cell body and prominent formation of actin stress fibers (Fig 4C). In contrast, KD of M₃R and Arf1 did not drastically affect the actin cytoskeleton distribution compared to control cells and only led to more prominent cortical F‐actin accumulation. We quantified these two phenotypes, cells presenting cortical actin (Fig 4D) and those displaying actin stress fibers (Fig 4E), and confirmed our observations. Actin rearrangements have a direct impact on the stiffness of cells and can be measured by atomic force microscopy 46. Therefore, we applied this approach to monitor cortical elasticity properties in control cells and ArfGAP1 KD cells. Using this technique, we observed that ArfGAP1 KD cells were more stiff than control cells (Fig 4F), in good correlation with the differences we observed for the actin distribution between these cell types and reflecting a more prominent appearance of rigid cytoskeleton elements, such as actin stress fibers, in ArfGAP1 KD cells.

Figure 4. ArfGAP1 silencing leads to actin cytoskeleton rearrangements.

• A, B
  Scramble and ArfGAP1 KD cells were transfected with GFP‐actin, and images were recorded by automated confocal microscopy at the indicated time points during Mtb H37Rv infection at MOI 5 (A). Insets of detailed areas (white frames) are shown on the right. Scale bar: 20 μm. The average number of actin patches per cell was quantified in scramble siRNA cells over time (n = 7) (B).
• C
  Cellular F‐actin distribution of scramble, ArfGAP1 KD, Arf1 KD, and M₃R KD cells was analyzed by SIM microscopy. F‐actin was labeled by phalloidin, and nuclei were stained by DAPI. Scale bars: 5 μm.
• D, E
  Phalloidin‐labeled cells were also analyzed by automated confocal microscopy (n ≥ 1,000) and quantified for the presence of cortical F‐actin (D) and actin stress fibers (E). Shown are representative examples of two independent experiments.
• F
  Cellular elasticity of scramble and ArfGAP1 KD cells was measured by atomic force microscopy. Histograms display 42 analyzed cells. Shown are representative examples of three independent experiments.
• G
  Scramble and ArfGAP1 KD cells were also treated with 0.8 μM cytochalasin D (CytoD) and Mtb colonization at MOI 20 was analyzed after 4 h by automated confocal microscopy.

Data information: Data are presented as mean ± SEM. ***P < 0.0005, **P < 0.005, *P < 0.05 (Student's t‐test).

Previous findings have shown that Mtb entry into non‐phagocytic cells also partially relies on macropinocytic uptake 14, 15. We tested this possibility and analyzed siRNA‐transfected A549 cells for their macropinocytosis and endocytosis efficiency using a previously developed fluid‐phase uptake assay 47. While the silencing of M₃R and Arf1 led to a reduction in macropinocytosis (Appendix Fig S5A and B) and PLD1 KD had no influence (Appendix Fig S5C), induced the knock‐down of ArfGAP1 an increased macropinocytic activity (Appendix Fig S5D). In contrast, the efficiency of endocytic uptake was not altered in ArfGAP1 KD cells compared to scramble siRNA cells (Appendix Fig S5E). In addition, we also interfered with F‐actin dynamics by treating control and ArfGAP1 KD cells with the actin depolymerization agent cytochalasin D. Indeed, Mtb colonization efficiency was significantly reduced in both cytochalasin D‐treated ArfGAP1 KD and control cells, compared to mock‐treated (DMSO) cells (Fig 4G). Remarkably, cytochalasin D treatment of ArfGAP1 KD cells efficiently decreased mycobacterial entry to a similar level than the one observed in control cells, supporting the idea that the ArfGAP1 effect on Mtb colonization is directly linked to actin cytoskeleton organization. Collectively, these findings show that Mtb entry relies on actin cytoskeleton rearrangements in epithelial cells that are influenced by M₃R, Arf1, and PLD1, which are controlled and restricted by ArfGAP1.

Finally, we aimed to investigate whether this restriction mechanism of mycobacterial entry is specific for epithelial cells or can also be found in other human cell types. In addition to A549 cells, we also knocked down ArfGAP1 in epithelial HeLa cells, in the monocyte cell line THP‐1, and in primary human blood‐derived macrophages (Fig 5A). Interestingly, only in human epithelial cells (A549 and HeLa) ArfGAP1 KD induced an increased Mtb colonization (Fig 5B), while monocytes and macrophages were unaffected by ArfGAP1 KD, suggesting that different uptake mechanisms, such as phagocytosis, are utilized predominantly by Mtb to enter those cell types. Moreover, in HeLa cells we also observed similar F‐actin rearrangements upon ArfGAP1 KD when we analyzed them by confocal microscopy (Fig EV4A) demonstrated by reduced levels of cortical F‐actin (Fig EV4B) and increased formation by actin stress fibers (Fig EV4C). We also questioned whether the difference of Mtb colonization upon ArfGAP1 KD of epithelial cells could be observed upon infection of this cell type by other pathogens. We infected HeLa cells with Shigella flexneri (Fig 5C) and Yersinia pseudotuberculosis (Fig 5D) at MOI 5 and MOI 50 (or MOI 20, respectively), but at all conditions, we did not observe any differences in bacterial colonization between ArfGAP1 KD and control cells. In addition, we also analyzed whether the ArfGAP1‐dependent entry process is engaged by other, attenuated or non‐pathogenic mycobacteria. We observed an increased colonization of epithelial cells by Mycobacterium bovis BCG (Fig EV5A) and by Mtb H37Ra (Fig EV1B), ruling out a possible contribution of the main virulence factors encoded by the ESX1 region in this process 48. Finally, also a M. bovis BCG mutant lacking HBHA 18 led to a similar increased infection rate upon ArfGAP1 silencing (Fig EV5B) excluding the dependency of this process on HBHA.

Figure 5. ArfGAP1‐controlled entry is specific to mycobacteria and epithelial cells.

• A
  Scramble and ArfGAP1 KD epithelial cells (HeLa and A549), differentiated monocytes (THP‐1), and primary macrophages (HuMac) were analyzed by Western blotting for the expression of ArfGAP1. β‐actin and tubulin were used as loading control.
• B
  siRNA‐transfected A549 cells were infected with Mtb H37Ra and analyzed by automated confocal microscopy. siRNA‐transfected HeLa and A549 cells were infected at an MOI of 20, while THP‐1 and HuMac were infected at MOI 2. Mtb colonization was analyzed after 4 h by automated microscopy.
• C, D
  Transfected HeLa cells were analyzed for colonization by Shigella flexneri (C) and Yersinia pseudotuberculosis (D) at two different MOI.
• E
  Proposed working model of actin rearrangements and Mtb entry in the presence (left panel) and absence of ArfGAP1 (right panel). See text for further details.

Data information: Data are presented as mean ± SEM. ***P < 0.0005, **P < 0.005, ns: not significant (Student's t‐test). Histograms display at least 200 analyzed cells in four replicates (B) and five replicates (C, D). Shown are representative examples of two independent experiments. Source data are available online for this figure.

Figure EV4. F‐actin rearrangements in HeLa cells induced by ArfGAP1 silencing.

• A–C
  Scramble and ArfGAP1 KD HeLa cells were labeled by phalloidin and analyzed by automated confocal microscopy. Shown are representative images (A) and cells (n ≥ 700) quantified for the presence of cortical F‐actin (B) and actin stress fibers (C).
• D, E
  Actin expression was measured by Western blotting in A549 cells transfected with scramble and ArfGAP1 siRNA. Tubulin was used as loading control. Shown are one representative WB image (D) and the quantification of actin expression normalized to the expression of tubulin of three independent experiments (E).

Data information: Data are presented as mean ± SEM. ***P < 0.0005, ns; non‐significant (Student's t‐test). Source data are available online for this figure.

Figure EV5. Increased colonization of A549 cells by Mycobacterium bovis BCG upon ArfGAP1 silencing.

• A, B
  Mycobacterial colonization of scramble and ArfGAP1 KD A549 cells infected with Mycobacterium bovis BCG (A) and M. bovis BCG ∆hbhA (B) at MOI 1 4 h post‐infection analyzed by automated confocal microscopy. Data are presented as mean ± SEM. ***P < 0.0005, **P < 0.005 (Student's t‐test). Histograms display at least 400 analyzed cells in six replicates. Shown are representative examples of two independent experiments.

In this study, we report a novel mechanism of mycobacterial entry into epithelial cells engaging M₃R‐ and Arf1‐mediated actin cytoskeleton rearrangements, which are restricted and controlled by ArfGAP1. Importantly, the expressed level of cellular actin does not change upon silencing of ArfGAP1 (Fig EV4D and E). The engagement of M₃R upon Mtb entry leads to the activation of Arf1 (Arf1‐GTP) triggering PLD1 activity and the biosynthesis of phosphatidylinositol recruiting the negative Arf1 regulator, ArfGAP1. ArfGAP1 then exerts control over Arf1 signaling by hydrolyzing Arf1‐GTP into Arf1‐GDP. This negative control leads to a decrease in PLD1 activity and no further actin remodeling preserving F‐actin in cortical areas of the cell, which prevents massive mycobacterial entry (Fig 5E, left panel). Upon ArfGAP1 silencing (Fig 5E, right panel), Arf1‐GTP is not hydrolyzed and both, Arf1 and PLD1 activity, do not decrease leading to F‐actin rearrangements throughout the cell body and prominent actin stress fiber formation, which both favor massive mycobacterial colonization of epithelial cells. A similar mechanism has been reported recently during Salmonella invasion, where Arf GEFs and GAPs collaborate to induce Arf1‐dependent pathogen entry 49. In contrast, we did not find evidence that ArfGAP1 silencing influences the overall infection rates of epithelial cells during invasion by Shigella flexneri and Yersinia pseudotuberculosis, suggesting a different control of entry. In addition to the reported contribution of macropinocytosis 14 and phagocytosis 50, the mechanism we propose here demonstrates how ArfGAP1 engages actin cytoskeleton rearrangements to restrict mycobacterial invasion of epithelial cells and provides a novel function of ArfGAP1 in host–pathogen interactions.

Materials and Methods

siRNAs, plasmids, antibodies, and reagents

The SiGENOME SMARTpool siRNA library (targeting 16,532 genes of the whole human genome), pooled single siRNAs targeting four different positions to each gene (M ₃ R, Arf1, RHOA, RAC1, CDC42, GRP1, ArfGAP1, PLD1, PLD2, VCAM1, ICAM1, ITGA2B, HMMR; Table EV1), and non‐targeting siRNA (scramble) were purchased from GE Healthcare Dharmacon. The actin‐GFP plasmid was a kind gift by Michael Way (Francis Crick Institute, London, UK), and the neuropeptide Y (NPY)‐venus plasmid 31 was kindly provided by Cédric Delevoye (Institute Curie, Paris, France). The following reagents were purchased from ThermoFisher Scientific: Lipofectamine RNAiMax, active Arf1 Pull‐down and Detection Kit, Amplex Red Phospholipase D Assay Kit, anti‐ArfGAP1 (#PA5‐12109), anti‐Golgin‐97 (#A21270), Alexa Fluor 647 phalloidin (#A22287), Alexa Fluor 488 phalloidin (#A12379), DAPI (#D1306), RPMI1640 + Glutamax‐I medium, MEM medium, fetal bovine serum (FBS), 200 mM L‐glutamine, DPBS, versene, hygromycin B, streptomycin, and DQ‐ovalbumin (#D12053). The HiPerFect Transfection Reagent was purchased from Qiagen. The anti‐M₃R (#126168) was purchased from Abcam. The anti‐α/β‐tubulin (#2148S) and the anti‐β‐actin (#4967S) were purchased from Cell Signaling. Anti‐TGN46 (#AHP1568) was purchased from AbD Serotec and anti‐Rab6 (#SC‐310) from Santa Cruz Biotechnology. Both, HRP anti‐rabbit and anti‐mouse, were purchased from Jackson Immune Research. Genejuice transfection reagent was from Novagen. RIPA lysis buffer and QS11 were purchased from Calbiochem. Brefeldin A, cytochalasin D, DMSO, tyloxapol, amikacin, ampicillin, kanamycin, IPTG, CongoRed, protease inhibitor cocktail, 10% formalin solution, paraformaldehyde, fluorescein isothiocyanate–dextran (#46945), and anti‐ARF1 (#SAB2100143) were purchased from Sigma‐Aldrich. Fluorescent mounting medium was from Dako. Zamifenacin fumarate was purchased from R&D Systems. Middlebrook 7H9 broth, Middlebrook 7H11 agar, OADC Enrichment, TSA medium, TSB medium, and LB broth were purchased from BD Biosciences. Tween‐80 and glycerol were purchased from Euromedex.

Cells

A549 cells (ATCC, CCL‐185) were cultured in RPMI 1640 + Glutamax‐I medium supplemented with 10% FBS. HeLa cells (ATCC, CCL‐2) were cultured in MEM supplemented with 2 mM L‐glutamine and 10% FBS. THP‐1 monocytes were differentiated in RPMI 1640 + Glutamax‐I medium supplemented with 10% FBS and 50 ng/ml phorbol 12‐myristate 13‐acetate (Sigma‐Aldrich) for 72 h. Human CD14⁺ peripheral blood mononuclear cells (hPBMCs) were isolated from whole blood of healthy donors and differentiated into macrophages by incubation in RPMI 1640 + Glutamax‐I medium supplemented with 10% FBS and 50 ng/ml of recombinant human M‐CSF (Miltenyi).

Ethics statement

Human CD14⁺ peripheral blood mononuclear cells were purified from blood samples obtained from healthy donors under strict anonymity (Etablissement Français du Sang “Nord de France”, EFS, Lille). Written informed consents were obtained from the donors under EFS contract no. NT/18/2016/200 with respect to Decree no. 2007‐1220 (articles L1243‐4, R1243‐61 and following) dated August 10, 2007, of the French Public Health Code. The use of human samples was approved by the French Ministry of Education and Research under the agreement DC 2015‐2575.

Bacteria

Mtb H37Rv‐GFP and H37Ra‐GFP 24, 25 were cultured at 37°C for 14 days in 7H9 broth supplemented with 10% OADC enrichment, 0.5% glycerol, and 50 μg/ml hygromycin B. Tween‐80 (0.05%) or 0.002% tyloxapol were added to the Mtb H37Rv‐GFP strain or Mtb H37Ra‐GFP strain, respectively. Mycobacterium bovis BCG‐GFP and M. bovis BCG ∆hbhA‐GFP (kind gift by C. Verwaerde and C. Locht) were grown in Sauton medium containing 25 μg/ml kanamycin or a combination of 25 μg/ml kanamycin and 25 μg/ml streptomycin, respectively. Prior to infection of cells, mycobacteria were washed three times with DPBS followed by sonication at 1‐s on/off pulse for 3 min or by centrifugation at 120 g for 2 min to minimize bacterial aggregation. Mtb was quantified by fluorescence measurement using the Victor 3 plate reader (Perkin Elmer) at an excitation wavelength of 488 nm, as previously described 25. Shigella flexneri serotype 5a strain M90T (kind gift by L. Ligeon and F. Lafont) was cultured overnight at 37°C in TSB medium. Prior to cell infection, S. flexneri was inoculated at 1:50 dilution and incubated at 37°C for around 3 h to reach, and optical density of 0.5. GFP‐expressing Yersinia pseudotuberculosis 2777 (kind gift by L. Ligeon and F. Lafont) was cultured overnight at 28°C in LB broth supplemented with 100 μg/ml ampicillin. Prior to cell infection, the bacteria were inoculated at a dilution of 1:50 and incubated at 28°C for 2 h. Then, 100 μM IPTG was added to induce GFP expression in bacteria, and bacteria were cultured around 1 h to reach an optical density of 0.5. Prior to infection, S. flexneri and Y. pseudotuberculosis were diluted in cell growth medium. After 1‐h incubation, extracellular bacteria were washed out with DPBS twice, and cells were fixed in 10% formalin for 30 min.

Genomewide high‐throughput siRNA screening

2 μl of each pooled siRNA (0.5 μM) of the genomewide SiGENOME SMARTpool siRNA library was first transferred to 384‐well microtiter plates using an automated liquid handling (Velocity 11, Agilent). 8 μl of DPBS containing 0.1 μl of lipofectamine RNAiMax was then distributed to each well using the WellMate dispenser (Agilent). Plates were incubated at room temperature (RT) for 30 min to allow transfectant complex reaction (Fig 1A (1) and (2)). Next, 40 μl of A549 cells was distributed to the well plates using the WellMate dispenser to obtain a final density of 1,500 cells per well and incubated at 37°C (Fig 1A (3)). After 72 h of incubation, Mtb H37Rv‐GFP in cell medium was added to the cells at an MOI of 5. At 5 h post‐infection, cells were washed with DPBS and treated with 50 μg/ml of amikacin for 1 h in order to remove extracellular Mtb. Then, cells were washed twice with fresh cell medium and incubated at 37°C with 5% CO₂ (v/v) (Fig 1A (4)). After 5 days post‐infection, 10% formalin containing 5 μg/ml of DAPI solution was replaced to each well, and plates were incubated at RT for 30 min allowing staining and cell fixation. Cells were stored in DPBS supplemented with 1% FBS. Images of cells and Mtb H37Rv‐GFP were acquired by automated confocal microscopy (Opera, Perkin Elmer) using a 20× water immersion objective with a numeric aperture of 0.7. DAPI‐stained nuclei were detected using excitation (Ex) at 405 nm and an emission filter (Em) at 450 nm. M. tuberculosis H37Rv‐GFP was detected using Ex at 488 nm and Em at 535 nm (Fig 1A (5)). Images were analyzed using Columbus image analysis software (Perkin Elmer, Fig 1A (6)), as described previously 26. The number of cells and the percentage of infected cells were normalized using Z score versus non‐targeting siRNA scramble. Data resulting in Z scores for number of cells less than −3 were excluded, suggesting siRNA‐induced cytotoxicity. Z scores higher than 5 for the percentage of infected cells were applied as a criterion for hit selection. The RNAi screening data were deposited in the GenomeRNAi public database (v.16.0, DKFZ Heidelberg, Germany) 51 under stable ID: GR00402‐S.

siRNA transfection and plasmid overexpression

20–50 nM of pooled‐siRNAs targeting a single gene or two different genes (co‐silencing) was transfected to cells for 72 h using RNAiMax in A549, HeLa and THP‐1 cells or using HiPerFect transfection reagent for PBMC, as described previously 25, 26, 52. Plasmid constructs were transfected to cells 2 days after siRNA transfection using Genejuice transfection reagent according to the manufacturer's instructions.

Chemical inhibitor test

Brefeldin A (BFA) and cytochalasin D were prepared in cell medium supplemented with 10% FBS, whereas QS11 and Zamifenacin were diluted in cell medium with 1% FBS. Prior to Mtb infection, BFA and QS11 were incubated with cells for 18 h, while Zamifenacin was incubated with cells for 2 h. Cytochalasin D treatment (0.8 μM) was started 30 min prior to Mtb infection and maintained during the 4‐h infection period.

Infection assay

Mycobacteria diluted in cell medium were added to siRNA‐transfected cells at the indicated MOI for the mycobacterial uptake assay. After 4 h post‐infection, extracellular bacteria were removed by extensive washing of cells (by pipetting DPBS three times). The mycobacterial replication assay was performed as described previously 25, 26, 52. Briefly, the A549 cell suspension adjusted to 1.5 × 10⁵ cells/ml was infected with mycobacteria at an MOI of 20 in cell medium under mild stirring (100 rpm). At 2 h post‐infection, cells were washed with cell medium by centrifugation at 300 g for 5 min and treated with 50 μg/ml amikacin for 1 h under mild stirring. Again, cells were washed twice with cell medium by centrifugation. Subsequently, 40 μl cell suspension was distributed to the siRNA‐transfectant complex, and cells were incubated at 37°C for 1, 3, or 5 days. Cells were fixed with 10% formalin solution for 30 min followed by DNA staining using 5 μg/ml DAPI for 5 min. Confocal images were acquired using an automated confocal microscope (Opera, Perkin Elmer).

Image acquisition and analysis

DAPI‐stained nuclei were detected using the 405‐nm laser with a 450/50‐nm emission filter. Independent of nucleus detection, green, red and far‐red signals were recorded using 488‐, 561‐, and 640‐nm laser lines with 540/75‐, 600/40‐, and 690/20‐nm emission filters, respectively. Images were analyzed by a multi‐parameters approach using the Columbus image analysis software (Perkin Elmer, version 2.3.1). Briefly, the host cell segmentation was performed using two different DAPI signal intensities—a strong intensity corresponding to the nucleus and a weak intensity in cytoplasm—with the algorithm “Find Nuclei” and “Find Cytoplasm”, as described previously 26. GFP or RFP signal intensities in a cell were used for the intracellular bacterial segmentation with the algorithm “Find Spots”. The identified intracellular bacteria were quantified as intracellular Mtb area with number of pixels. Subsequently, two populations—infected cells and non‐infected cells—were determined, and the percentage of infected cells was calculated. The SER texture algorithm (mainly SER Edge and SER Ridge, normalization by Kernel) with the phalloidin signal in the far‐red channel (Ex 630 nm/Em 690 nm) was used to define cortical F‐actin in the cell periphery and stress fibers in cytoplasmic regions.

Immunofluorescence and confocal microscopy

siRNA‐transfected cells grown on coverslips in 24‐well plates were washed with DPBS and fixed in DPBS containing 4% paraformaldehyde and 4% sucrose for 20 min at RT. Subsequently, cells were quenched in DPBS + 50 mM NH₄Cl for 10 min at RT. After washing samples three times in DPBS, cells were permeabilized using DPBS + 0.1% Triton X‐100 for 5 min at RT. Cells were then incubated with blocking solution (5% FBS and 1% BSA in DPBS) at RT for 60 min followed by antibody labeling using anti‐ArfGAP1, anti‐Arf1, anti‐TGN46, anti‐golgin97, or anti‐Rab6 at 4°C overnight. Cells were washed three times in DPBS for 10 min and incubated with fluorescently labeled secondary antibodies, 5 μg/ml DAPI and Alexa Fluor 488 or 647 phalloidin for 60 min at RT. After three washes in DPBS, coverslips were placed in mounting medium. Images were acquired using a confocal microscope (Zeiss LSM880) equipped with a 63× objective (NA 1.4) and Zen imaging software (Zeiss). Image analysis and generation of maximum projections of z‐stacks were carried out by Fiji software 53.

Analysis of Golgi fragmentation

Three‐dimensional analysis of Golgi integrity was done with Imaris software (Bitplane Inc.). Images of scramble, siArf1, and siArfGAP1 KD were acquired by confocal microscopy (z‐stack interval of 420 nm), and KD cells were selected based on simultaneous labeling of Arf1 and ArfGAP1, respectively. Trans‐Golgi network (TGN) membranes were visualized by TGN46 antibody labeling. TGN membranes were visualized using the IsoSurface mode of the Surpass module. Subsequently, the volume and number of individual TGN elements was determined. Objects smaller than 500 pixels were excluded from the calculations. The mean volume, SEM, and average numbers of TGN fragments were calculated from 50 cells for each condition using GraphPad Prism (version 5.03).

Live cell imaging and spinning disk microscopy

siRNA‐transfected cells grown in LabTek #1.0 borosilicate chambered coverglasses (ThermoFisher Scientific) were transfected with NPY‐venus 48 h before imaging. Live cell imaging was performed by spinning disk microscopy using a CSU‐W1 microscope (Nikon) equipped with a 63× objective (NA 1.4) and controlled by MetaMorph software (Molecular Devices). Images were acquired every second at 37°C and 5% CO₂. Image analysis and generation of movies were carried out by Icy (http://icy.bioimageanalysis.org/) and Fiji software.

Super‐resolution structured illumination microscopy (SIM)

Images were acquired on an ElyraPS1 microscope system (Zeiss) using a 100× oil immersion lens (PlanApo, NA 1.46). The illumination patterns were generated by laser light passing through an optical gating (sinusoidal pattern of high spatial frequency). The SIM system achieved a resolution of 120 nm along the x–y axis and 500 nm along the z‐axis (measured by 100‐nm beads; sampling voxel size: 0.05 × 0.05 × 0.15 μm). Laser lines at 405 and 488 nm were used for excitation of DAPI and GFP, respectively. Fluorescence emission was filtered with band‐pass filters (BP 420–480 nm; BP 495–575 nm). To obtain one super‐resolution SIM image, 15 images (five different phases of three different angular orientations of the illumination pattern with a SIM gating period of 51 μm for the blue channel and 42 μm for the green channel) were collected on an EMCCD camera (Andor Technology Ltd) and were processed using Zen software (Zeiss). Three‐dimensional reconstructions were carried with Imaris software (Bitplane).

Atomic force microscopy

Atomic force microscopy experiments were carried out on a BioScope Catalyst atomic force microscope (Bruker) mounted on an inverted Zeiss ElyraPS1 microscope system. The reverse‐transfection method was applied for siRNA delivery to A549 cells in dishes with glass bottom (WillCo). After 72 h, cells were washed with DPBS, and the medium was changed to RPMI without phenol red, supplemented with 25 mM Hepes, 2 mM glutamine, and 10% FBS. Experiments were carried out at 37°C. Pyramidal tips (Bruker DNP 0.06 N/m) were used. The deflection sensitivity was determined with the help of a force curve on the glass substrate. The spring constants were calibrated with the classical thermal noise fitting procedure. For each condition, between 10 and 42 cells were scanned (one scan per cell, in areas between the nucleus and the cell periphery). Force volume maps were at least 4 × 4 μm big and had at least 32 × 32 pixels, resulting in at least 43,008 force curves per condition. The same cantilever was used to perform measurements on control and corresponding test samples. All experiments were performed in triplicates. The following scanning parameters were applied: force threshold 2 nN, ramp size 2.5 μm (1,024 points), and scan speed 27.9 μm/s. Data were analyzed with in‐house software written in Python. Each force curve was fitted with a line in the non‐interaction part to determine the point of contact. The Young's modulus (E) was computed with the Sneddon model for pyramidal tips1. The sample's Poisson ratio was defined as 0.5, and the tip angle was set to 18°. The reported values are given for an indentation depth of 200 nm.

CFU determination

At 4 days post‐infection, cells were lysed in DPBS + 0.1% Triton X‐100. 5‐fold serial dilutions for six different points were performed in DPBS and plated onto 7H11 agar plates supplemented with 10% OADC. After 2 weeks, CFUs were calculated for all plated samples.

Protein and RNA extraction

Cells were harvested from 6‐well plates and washed twice with ice‐cold DPBS. 300 μl of RIPA lysis buffer containing protease inhibitor cocktail was used for protein extraction. The cell suspension was transferred to Eppendorf tubes and incubated under constant shaking at 700 rpm for 15 min at 4°C followed by centrifugation at 15,800 g at 4°C for another 15 min. The RNeasy mini kit (Qiagen) was used for the isolation of total RNA from cells following the manufacturer's instructions. RNA quantification and quality were measured on a SimpliNano Spectrophotometer (GE Healthcare). RNA and protein samples were stored at −80°C until further use.

Reverse transcription and quantitative PCR

Reverse transcription was used for cDNA synthesis from total RNA. 1 μg of total RNA was added to the master mix containing random primers. The solution was incubated at room temperature for 10 min, and cDNA synthesis was performed at 42°C for 20 min. The reaction was terminated at 95°C for 5 min, and quantitative PCR was performed with the multi‐well plate 384 LightCycler 480 (Roche). 4 μl of cDNA was amplified with 10 pmol of forward and reverse primers (Appendix Table S1) on the LightCycler 480 probes master. The PCR was performed for 40 cycles with denaturation at 95°C for 15 s, annealing at 60°C for 30 s and extension at 72°C for 30 s. All experiments were run in triplicates, and the Livak method 54 was applied for relative quantification using β‐actin as a reference gene. KD efficiency of all used siRNA conditions is shown in Appendix Table S2.

Western blotting

Equal amounts of protein (10–30 μg) were loaded on 4–12% gradient acrylamide gels (Mini‐Protean TGX, Bio‐Rad) and migrated with 120 V for 80 min. Proteins were transferred to 0.2 μm PVDF membranes using the Trans‐blot Turbo device (Bio‐Rad). PBS/Tween was used to wash membranes. PBS/Tween + 5% skim milk was used for 1 h to block non‐specific binding sites. Membranes were incubated with antibodies overnight at 4°C followed by three washes for 10 min with PBS/Tween. HRP‐coupled secondary antibodies were used for 1 h at RT. Chemiluminescence was revealed using Immobilon HRP substrate (Millipore).

Arf1 pull‐down assay

1 mg of cell lysate was incubated with 100 μg GST‐GGA3‐PBD and glutathione resin in columns at 4°C for 1 h with gentle rotation. Unbound proteins were removed by washout, and 2× SDS sample buffer containing beta‐mercaptoethanol was used to elute proteins from the glutathione resin. Eluted samples were analyzed by Western blotting using an antibody against Arf1 at 1/1,000 dilution.

Measurement of PLD activity

The Amplex Red phospholipase D assay kit was used for the measurement of PLD activity following the manufacturer's instructions. PLD activity was recorded after addition of 20 μg of cell lysates. Briefly, PLD in the sample mix cleaved phosphatidylcholine to choline, which is then oxidized by choline oxidase to betaïne and H₂O₂, the latter reacting with Amplex Red reagent in the presence of HRP to produce resorufin. The fluorescence of resorufin was measured using the Victor 3 plate reader (Perkin Elmer) at an excitation of 530/10 nm and an emission of 595/60 nm.

Fluid‐phase uptake assays to measure macropinocytosis and endocytosis

The fluid‐phase uptake assays were performed on siRNA‐transfected A549 cells. At 2 days post‐transfection, serum starvation (culture medium + 1% FBS) was performed overnight prior to experiments. 500 μg/ml of fluorescein isothiocyanate‐dextran or 50 μg/ml DQ‐ovalbumin diluted in RPMI1640 medium supplemented with 1% FBS was added to the cells to analyze macropinocytosis 47 and endocytosis, respectively. Cells were incubated at 37°C and 5% CO₂ for the indicated time periods. Macropinosome formation was analyzed by confocal microscopy using a 60× objective (InCell 6000, GE Healthcare) equipped with Ex 405 nm and Em 450/50 nm for the detection of DAPI‐stained nuclei and Ex 488 nm and Em 540/75 nm for fluorescently labeled dextran over a time course of 30 min at 37°C. As negative control of macropinocytic uptake, samples were incubated at 4°C for the same time periods. The analysis of endocytosis was analyzed over a time period of 120 min with cells incubated at 37°C by flow cytometry using the FACSCANTO II instrument (BD Biosciences). Mean fluorescence intensities (MFI) of 10,000 DQ‐ovalbumin labeled cells per time point were analyzed. MFI values were normalized to those of control cells that were incubated for the same time periods at 4°C.

Statistics

Statistical significance (P value) was calculated with GraphPad Prism software (version 5.03) using the two‐tailed unpaired t‐test, with the exception of the experiments of Fig. 4E, where the Wilcoxon Mann–Whitney test was used.

Author contributions

O‐RS, SM, EW, MP, and EH performed experiments. O‐RS, CJQ, EH, and PB conceived and designed the experiments and performed data analysis. O‐RS, CJQ, RI, VD, RV‐C, SM, EW, IR, SJ, ND, FL, AB, EY, LM, and PB contributed reagents, materials, and analysis tools. O‐RS, SM, EY, LM, EH, and PB wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Table EV1

Movie EV1

Movie EV2

Source Data for Expanded View and Appendix

Review Process File

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 5

EMBO Reports (2018) 19: 29–42