ABSTRACT
The regulation of macropinocytosis, a specialised endocytosis pathway, is important for immune cell function. However, it is not known whether the biogenesis of macropinosomes involves one or more distinct pathways. We previously identified sorting nexin 5 (SNX5) as a regulator of macropinocytosis in macrophages. Here, we show that bone-marrow-derived macrophages from SNX5-knockout mice had a 60–70% reduction in macropinocytic uptake of dextran or ovalbumin, whereas phagocytosis and retrograde transport from the plasma membrane to the Golgi was unaffected. In contrast, deficiency of SNX5 had no effect on macropinocytosis or antigen presentation by dendritic cells. Activation of macrophages with CSF-1 resulted in a localisation of SNX5 to actin-rich ruffles in a manner dependent on receptor tyrosine kinases. SNX5-deficient macrophages showed a dramatic reduction in ruffling on the dorsal surface following CSF-1 receptor activation, whereas peripheral ruffling and cell migration were unaffected. We demonstrate that SNX5 is acting upstream of actin polymerisation following CSF-1 receptor activation. Overall, our findings reveal the important contribution of dorsal ruffing to receptor-activated macropinocytosis in primary macrophages and show that SNX5 selectively regulates macropinosomes derived from the dorsal ruffles.
INTRODUCTION
Macropinocytosis is a regulated pathway for the non-specific fluid-phase cellular uptake of extracellular fluid, solute molecules and antigens. Macropinocytosis is important in a wide range of cellular events including development, cell motility, tumour progression and metastasis (Lim and Gleeson, 2011; Swanson and Watts, 1995), provides an entry into the cell for a number of pathogens and mediates antigen sampling by antigen-presenting cells of the innate immune system (Mercer and Helenius, 2012; Norbury, 2006).
Macropinocytosis is an actin-dependent process initiated by surface membrane ruffles that give rise to large irregularly shaped endocytic vesicles with a diameter from >0.2 µm to 10 µm called macropinosomes (Hewlett et al., 1994; Swanson and Watts, 1995). A variety of membrane ruffles have been reported to give rise to macropinosomes by fusing with the plasma membrane. Macropinocytosis is a regulated endocytic process mediated by the activation of a variety of cell surface receptors including macrophage colony stimulating factor (CSF-1), epidermal growth factor (EGF) and platelet-derived growth factor (Dharmawardhane et al., 2000; Haigler et al., 1979; Racoosin and Swanson, 1989; Swanson, 1989). Phosphoinositides derived from receptor-mediated signalling events mediate the induction of macropinosomes (Amyere et al., 2000; Anzinger et al., 2012; Araki et al., 1996; Clague et al., 1995; Kerr et al., 2010; Yoshida et al., 2009). An important class of phosphoinositide effectors which regulate a variety of endocytic trafficking events, including macropinocytosis, are the sorting nexins (SNXs) (Cullen, 2008). A number of SNXs have been implicated in the biogenesis of macropinosomes, including SNX1, SNX5, SNX9, SNX18 and SNX33 (Lim et al., 2008; Wang et al., 2010). Our earlier studies have shown that SNX5 in particular is required for the early events in the formation of macropinosomes at the cell surface not only in cultured cells (Lim et al., 2008) but also primary macrophages (Lim et al., 2012). Following EGF stimulation of HEK293 cells, SNX5 is recruited to the plasma membrane and this recruitment is dependent on phosphatidylinositol (3,4)-bisphosphate (PIP2) (Lim et al., 2008). Overexpression of SNX5 results in an increase in macropinocytosis in HEK293 cells (Lim et al., 2008), whereas silencing of SNX5 in bone-marrow-derived macrophages using microRNA results in a substantial decrease in the biogenesis of macropinosomes (Lim et al., 2012). Although SNX5 can form heterodimers with SNX1, we have previously shown that SNX5 can function independently of SNX1 (Lim et al., 2012). Hence, SNX5 appears to play a key role in the biogenesis of macropinosomes.
Macropinosomes can arise from distinct types of membrane ruffles, namely peripheral ruffles, which protrude from the leading edge that fold backwards, and dorsal ruffles, which are located on the dorsal surface of the cell. Dorsal ruffles have a distinctive circular or curved upright architecture, are very dynamic and appear transiently following receptor activation. In some cells, dorsal ruffles are specialised for initiating receptor signalling (Abella et al., 2010; Luo et al., 2014), for example dorsal ruffles are abundant in macrophages following activation (Patel and Harrison, 2008; Yoshida et al., 2009). Although macropinocytosis has been linked to both peripheral ruffles and dorsal ruffles, some reports have suggested that peripheral ruffles might be more relevant than dorsal ruffles for initiating macropinocytosis (Buccione et al., 2004; Hoon et al., 2012; Suetsugu et al., 2003). There remains uncertainty on the functional role of dorsal ruffles in macropinocytosis and the relative contribution of peripheral and dorsal ruffles to the macropinocytic pathway.
Macropinocytosis represents a major pathway by which innate immune cells capture antigens (Norbury et al., 1997, 1995; Sallusto et al., 1995). Professional antigen-presenting cells include dendritic cells and macrophages, and both these immune cells have high levels of macropinocytosis. Following antigen uptake, internalised antigens are processed into antigenic peptides, loaded onto major histocompatibility complex (MHC) molecules and the complex delivered to the cell surface for presentation to T-cells. However, a more detailed understanding of the macropinocytosis pathways in innate immune cells is required to appreciate the mechanism of antigen delivery and the ability to manipulate these pathways in the whole organism for functional analysis. In addition, it is not known whether the same regulators are relevant to macropinocytosis in both dendritic cells and macrophages. Given the importance of SNX5 in regulating the biogenesis of macropinosomes in macrophages, we have established a SNX5-knockout mouse line to investigate these questions. By analysis of macrophages and dendritic cells from SNX5−/− mice, here, we have demonstrated that SNX5 plays a specific role in the biogenesis of macropinosomes from dorsal ruffles and that dorsal ruffles provide a major contribution to macropinocytosis in activated macrophages.
RESULTS
SNX5−/− mouse line
The SNX5 ‘Knock-out First’ mouse strain, was created from an embryonic stem cell (ESC) clone (a2EPD0033_3_D07) generated by the CHORI-Sanger-UC Davis (CSD) consortium for the NIH-funded knockout mouse project (KOMP). The mutant allele, which constitutes a knockout first allele (Skarnes et al., 2011), carries a modifiable gene-trap knockout first, lacZ-tagged insertion between exon 3 and exon 4 of SNX5. Homologous recombination results in the ablation of the SNX5 allele by the insertion a SV40 polyadenylation (pA)-sequence-mediated transcriptional termination as part of the 5.5 kb non-genomic trapping cassette (Fig. 1A). Expression of exon 1 to exon 3 of SNX5 would only result in a small∼9.8 kDa protein representing the N-terminus of SNX5, which does not contain the lipid-binding phox (PX) domain. The SNX5 Knock-out First and control wild-type mice, on the same genetic background, will be referred to throughout the article as SNX5−/− and SNX5+/+ mice, respectively. The SNX5+/+ and SNX5−/− alleles were confirmed by PCR genotyping of mouse tail DNA (Fig. 1A,B).
The 50 kDa SNX5 protein was absent from bone-marrow-derived macrophages (BMMs) (Fig. 1C) and splenic dendritic cells from SNX5−/− mice (Fig. S1A), as determined by immunoblotting using an antibody that recognises the N-terminus of SNX5, and no smaller bands were detected. Whereas SNX5 protein was readily detected associated with endosomal structures in BMMs and dendritic cells from SNX5+/+ mice, immunofluorescence analysis revealed that the punctate staining pattern attributed to endogenous SNX5 was absent in BMMs (Fig. 1D) and dendritic cells (Fig. S1B) from SNX5−/− mice.
There was no detectable abnormality in SNX5−/− mice compared to their wild-type counterparts. Both SNX5−/− and SNX5+/+ mice survived for >52 weeks (n=4 for SNX5+/+ and SNX5−/−) and were fertile with average litter sizes of 5.32 for SNX5+/+ (n=25) and 5.62 for SNX5−/− (n=32). However, both male and female SNX5−/− mice were significantly smaller that SNX5+/+ mice at 9 weeks±3 days (SNX5+/+ male, 24.98±2.95 g; SNX5−/− male, 23.34±2.64 g; SNX5+/+ female, 20.12±1.47; SNX5−/− female, 18.29±2.03; mean±s.d., n=38, 42, 29 and 32, respectively).
Macropinocytosis is substantially reduced in primary macrophages from SNX5−/− mice
Alterations in SNX5 expression either by RNA interference (RNAi)-mediated silencing of endogenous SNX5 (Lim et al., 2012) or elevated expression in HEK293 stable cell lines (Lim et al., 2008) alters macropinocytic activity. As macropinocytosis is considered an essential pathway for antigen sampling by antigen-presenting cells, we had anticipated that SNX5−/− mice might be immune-compromised; however, there was no evidence for an abnormal T cell compartment or viability of mice. Therefore, we investigated macropinocytosis activities in two major classes of antigen-presenting cells, macrophages and immature dendritic cells, from SNX5−/− mice.
BMMs were derived from SNX5+/+ and SNX5−/− mice. Cells that had been rendered quiescent overnight were incubated with lysine-fixable, fluorescein isothiocynate (FITC)-conjugated 70 kDa dextran (FITC–dextran) in the presence of 50 ng/ml of CSF-1 for 15 min at 37°C to induce macropinocytosis. 70 kDa dextran is predominantly internalized by macropinocytosis (Kerr et al., 2006; Lim et al., 2012, 2008). Immunofluorescence analysis showed that endogenous SNX5 in CSF-1 activated BMMs from SNX5+/+ mice is localised to endosomal structures (>500 nm in diameter) labelled with FITC–dextran (Fig. 2A) when BMMs were incubated at 37°C but not at 4°C, characteristics previously identified as macropinosomes (Lim et al., 2012). In contrast, SNX5−/− BMMs had significantly reduced FITC–dextran uptake when incubated with FITC–dextran at 37°C (Fig. 2B). The level of FITC–dextran intensity per µm2 cell area was determined, using Metamorph software for >25 cells, and FITC–dextran uptake in SNX5−/− BMMs was reduced by 61% when compared to the wild-type counterpart (P<0.005) (Fig. 2C). Quantitative analysis by flow cytometry was also performed following fluid-phase FITC–dextran uptake for 15 min (Fig. 2D). A substantial reduction in the total uptake of FITC–dextran by SNX5−/− BMMs was observed compared with SNX5+/+ BMMs, with a 50% reduction in median fluorescence intensity (four independent experiments) by SNX5−/− BMMs compared to SNX5+/+ BMMs (Fig. 2D). The difference between microscopic and flow cytometry analysis is likely due to flow cytometry analysis measuring the total FITC–dextran level associated with cells whereas the microscopic analysis is restricted to intracellular macropinosome-associated FITC–dextran. The data collectively shows that SNX5 is required for efficient uptake of 70 kDa dextran by macropinocytic activity in BMMs.
Macropinocytosis is a major internalisation route for soluble antigen (Lim et al., 2012), and the subsequent processing of protein antigens in acidic intracellular compartments is an essential step in antigen presentation. To study antigen uptake and processing in the late endosomes and lysosomes, we used a self-quenched conjugate of ovalbumin, namely DQ-ovalbumin, which exhibits fluorescence only after proteolytic degradation (Daro et al., 2000; Santambrogio et al., 1999). BMMs from SNX5+/+ and SNX5−/− mice that had been rendered quiescent overnight were pulsed with DQ-ovalbumin for 3 min in the presence of 50 ng/ml CSF-1 and then chased at 37°C for 30 min, the time required for ovalbumin to reach LAMP-1-positive compartments (Lim et al., 2012). Immunofluorescence analysis showed that BMMs from SNX5−/− mice had significantly reduced DQ-ovalbumin fluorescence intensity compared with cells incubated with DQ-ovalbumin at 4°C (Fig. 2E). Quantitative analysis of >25 BMMs showed 67% reduction (P<0.005) in DQ-ovalbumin fluorescent intensity in SNX5−/− BMMs compared with BMMs from SNX5+/+ mice (Fig. 2F). The reduction in DQ-ovalbumin uptake and processing in SNX5−/− BMMs parallels the reduced macropinocytic activity. These data show that SNX5 regulates antigen uptake and processing by primary macrophages.
Macropinocytosis and antigen presentation is not affected in dendritic cells by the absence of SNX5
Given the importance of SNX5 in regulating macropinocytosis in macrophages, we anticipated that immature dendritic cells, which are specialised for antigen capture as they have a high macropinocytic activity (Lim and Gleeson, 2011; Norbury, 2006), might also be compromised in the absence of SNX5. The level of endogenous SNX5 protein in purified, splenic dendritic cells isolated from SNX5+/+ mice is comparable to the level of SNX5 in SNX5+/+ BMMs (Fig. 3A). To investigate whether SNX5 has a role in regulating macropinocytosis in dendritic cells, immature splenic dendritic cells were isolated from SNX5+/+ and SNX5−/− mice as described in the Materials and Methods. Immature dendritic cells are highly endocytic and capture fluid-phase antigens by macropinocytosis and express relatively low levels of CD86 compared with activated mature dendritic cells (Villadangos and Schnorrer, 2007). Enriched populations of immature dendritic cells were incubated with FITC–dextran for 30 min at 37°C, or on ice, before excess FITC–dextran was removed. Dendritic cells were identified by staining for CD11c (encoded by ITGAX) and MHCII, and analysed by immunofluorescence after cytospin collection on glass slides. Dendritic cells from SNX5+/+ and SNX5−/− mice internalised FITC–dextran (Fig. 3B) at 37°C but not on ice (4°C). To quantify the level of FITC–dextran uptake, flow cytometry analysis was performed on dendritic cells suspensions after fluid-phase FITC–dextran uptake. Cells were stained and immature dendritic cells identified using a combination of anti-CD11c, anti-CD86 and anti-MHCII antibodies. Immature dendritic cells (CD11c+, MHCII+, CD86low) from SNX5+/+ and SNX5−/− mice showed a similar intensity of FITC–dextran (Fig. 3B,C). These data indicate that macropinocytosis in immature dendritic cells is unaffected by the absence of SNX5. We also performed T cell proliferation assays to determine the efficacy of antigen presentation by dendritic cells from SNX5−/− mice using soluble ovalbumin as the fluid-phase antigen. Ovalbumin323-339-specific (H-2b) CD4+ T cells (OTII) responded in a similar manner to ovalbumin-loaded splenic dendritic cells from SNX5+/+ and SNX5−/− mice (data not shown), indicating that antigen uptake, processing and loading onto MHCII molecules is unaffected by the lack of SNX5 in dendritic cells. Collectively, and in contrast to primary macrophages, these data indicate that SNX5 does not regulate macropinocytosis in immature dendritic cells.
Lack of SNX5 does not affect phagocytosis or M6PR trafficking in macrophages
Having established that SNX5 plays an important role in macropinocytosis in BMMs, we sought to determine whether the impact of SNX5 is restricted to macropinocytosis or whether SNX5 is also relevant to other endocytosis pathways in macrophages. First, we assessed the phagocytic activity of SNX5-deficient macrophages, as phagocytosis and macropinocytosis share some similarities such actin-dependence, formation of large vascular endocytic structures and common regulators, such as phosphoinositide 3-kinase (PI3K) and Rab5 (Feliciano et al., 2011; Lee and Knecht, 2002; May and Machesky, 2001; Roberts et al., 2000). To monitor phagocytic activity, SNX5+/+ and SNX5−/− BMMs were incubated with Alexa-Fluor-488-labelled zymosan A bioparticles for 60 min at 37°C at two different concentrations. Zymosan particles were internalised efficiently into both SNX5+/+ and SNX5−/− BMMs (Fig. 4A). The average number of zymosan particles per macrophage was similar for SNX5+/+ BMMs [5.9±1.2 zymosan particles; mean±s.d., n=3 (50 cells each)] and SNX5−/− BMMs [6.7±0.8 zymosan particles, n=3 (50 cells for each)] at the higher concentration of 7×105 zymosan particles/ml, and 2.2±0.4 and 2.2±0.5 zymosan particles per cell for SNX5+/+ and SNX5−/− BMMs, respectively, at 1.6×105 zymosan particles/ml (Fig. 4B). Staining of zymosan-loaded SNX5−/− BMMs with antibodies to SNX5 showed that SNX5 was absent from phagosome structures (data not shown). Therefore, these data show that SNX5 does not play a role in macrophage phagocytosis.
SNX5 has been suggested to have a role in retromer-mediated endosomal sorting and transport of cargo to the Golgi (Wassmer et al., 2007, 2009). To determine whether the lack of SNX5 in macrophages has an effect on retrograde transport, we assessed the trafficking of the cation-independent mannose-6-phosphate receptor (M6PR, also known as IGF2R), which is recycled via endosomes and the Golgi. For this experiment, we used a CD8 chimera of M6PR (CD8–M6PR), previously used to analyse the internalisation of M6PR from the plasma membrane to early endosomes, followed by the intracellular transport, in a retromer-dependent manner, to the Golgi (Lieu et al., 2007; Seaman, 2004). The trafficking of the CD8–M6PR was analysed in recombinant adenovirus-transduced BMMs using an internalisation assay by tagging cell surface cargo with antibodies to the CD8 domain. Rapid internalisation of surface antibody-conjugated CD8–M6PR to endosomal structures in both SNX5+/+ and SNX5−/− macrophages was detected within 5 min of incubation at 37°C. By 60 min incubation, the majority of the CD8–M6PR was detected in the Golgi region of both SNX5+/+ and SNX5−/− macrophages (Fig. 4C,D). Hence the deficiency of SNX5 appears to have no effect on clathrin-mediated endocytosis and retrograde endosome to Golgi transport of M6PR, highlighting a specific role for SNX5 in macropinocytosis in BMMs.
SNX5 is located at the cell surface of macrophages following CSF-1 receptor activation
We have previously demonstrated that SNX5 is recruited to membrane ruffles of EGF-activated HEK293 cells expressing GFP–SNX5 (Lim et al., 2008). To determine whether SNX5 behaved in a similar manner in primary macrophages, we investigated the localisation of either endogenous or GFP-tagged SNX5 in CSF-1 activated SNX5+/+ BMMs. In resting macrophages, where there are minimal actin-rich membrane ruffles, SNX5 is located predominantly in early endosomes (Fig. 1D; Lim et al., 2012). In contrast, CSF-1-activated macrophages showed the presence of actin-rich plasma membrane ruffles with strong colocalisation of SNX5 with actin (Fig. 5A) and with CSF-1 receptors on membrane ruffles (Fig. 5B). We assessed the influence of CSF-1 receptor signalling on macropinocytosis using GW2580, a specific CSF-1 receptor tyrosine kinase inhibitor. GW2580 treatment dramatically reduced FITC–dextran uptake of both SNX5+/+ and SNX5−/− BMMs (Fig. 5C). For SNX5+/+ BMMs, there was a 89% reduced uptake of FITC–dextran following GW2580 treatment. As expected, SNX5−/− BMMs had a reduced level of macropinocytosis compared with SNX5+/+ BMMs in the presence of the carrier alone and this residual 30% macropinocytic activity was further reduced in the presence of GW2580 (Fig. 5C) to a similar level observed in SNX5+/+ BMMs in the presence of GW2580. This finding shows that macropinocytosis in CSF-1-activated macrophages is regulated by the tyrosine kinase activity of the CSF-1 receptor and influenced by SNX5. The levels of CSF-1 receptor at the cell surface of SNX5+/+ and SNX5−/− BMMs were similar (Fig. 5D), indicating that SNX5 is acting downstream of receptor activation to promote macropinocytosis.
Deficiency in SNX5 results in reduction of dorsal ruffling
Macropinosomes are derived from plasma membrane ruffles. Treatment of macrophages with CSF-1 has been shown to stimulate actin polymerisation (Sampaio et al., 2011), which is essential for macropinocytosis as well as chemotaxis (Jones, 2000). Given the possible link between macropinocytosis and chemotaxis, we investigated whether the reduced macropinocytic activity in SNX5−/− BMMs affected the chemotactic response of macrophages towards CSF-1 using a transwell migration assay. The results showed that SNX5−/− BMM migration (migration index of 1.55 and 1.55 in two separate experiments) was not significantly perturbed in their ability to migrate through the transwell compared with SNX5+/+ BMMs (migration index of 1.23 and 1.36 in two separate experiments) suggesting that SNX5 is required for macropinocytosis but not for the chemotactic response of macrophages to CSF-1.
Macropinocytosis and chemotaxis require plasma membrane ruffling. There are three types of membrane ruffling: lamellipodia or sheet-like, dorsal ruffles and membrane blebbing (Mercer and Helenius, 2012). Of the three types, only lamellipodia ruffles, which normally occur at the leading edge of cells, and dorsal ruffles have been implicated in fluid-phase uptake and macropinosome formation. Thus, we investigated whether the regulation of macropinocytosis by SNX5 was associated with the initial formation of plasma membrane ruffles. To do this, we monitored the extent of ruffling in SNX5-deficient macrophages following CSF-1 activation. BMMs that had been rendered quiescent overnight were morphologically elongated with small protrusive areas at one or both ends of the cell (Fig. 6A). Stimulation with CSF-1 for 3 min at 37°C resulted in flattened SNX5+/+ and SNX5−/− macrophages and the formation of planar lamellipodia with increased peripheral ruffling as indicated by an increase in F-actin intensity (Fig. 6A).
To investigate the formation of membrane ruffles on the dorsal surface of macrophages, two approaches were employed; an x-z series of confocal images and scanning electron microscopy (SEM). As assessed by confocal microscopy, CSF-1 activation of SNX5+/+ BMMs resulted in considerable actin-stained projections from the dorsal surface of the cell (Fig. 6B; Movies 1 and 2). In contrast, very few actin projections were detected on the dorsal surface of CSF-1 treated SNX5−/− BMMs (Fig. 6B; Movies 3 and 4) and any actin projections detected from the dorsal surface were considerably smaller than those in SNX5+/+ BMMs. Scanning electron microscopy clearly demonstrated the presence of dorsal ruffles on SNX5+/+ BMMs following CSF-1 stimulation (Fig. 7A). In contrast, dorsal ruffles were dramatically reduced on CSF-1-stimulated SNX5−/− BMMs. Quantification of the extent of total membrane ruffling on SEM images was estimated by manually tracing the perimeter of the ruffles and measuring the area (Fig. 7B). There was a 68% and 71% reduction in total membrane ruffling in SNX5−/− BMMs, compared with SNX5+/+ BMMs, after 3 min and 15 min CSF-1 stimulation, respectively (Fig. 7B). The reduction in membrane ruffling correlated with the level of reduction in macropinocytic activity in SNX5−/− BMMs (Fig. 2C). Therefore SNX5−/− BMMs have reduced dorsal membrane ruffling and macropinocytosis in response to CSF-1, whereas peripheral ruffling and cell migration is unaffected by the absence of SNX5.
To confirm that the deficiency of dorsal ruffles of SNX5−/− BMMs is a direct consequence of the lack of SNX5 protein, a rescue experiment was performed to re-introduce full-length SNX5 into SNX5−/− BMMs. BMMs from SNX5−/− mice were transduced with recombinant adenovirus which encodes GFP–SNX5. By performing quantitative immunofluorescence, we observed that transduced GFP-positive BMMs from SNX5−/− mice showed an equivalent or greater number of Texas-Red–dextran-labelled macropinosomes than wild-type BMMs (data not shown). In addition, and more importantly, dorsal ruffles were detected by SEM in SNX5−/− BMMs transduced with adenovirus encoding GFP–SNX5 following CSF-1 activation (Fig. S2). Approximately 52% of transduced SNX5−/− BMMs cells were detected with dorsal ruffles, which is consistent with the transduction efficiency of ∼65% as assessed by GFP fluorescence. In contrast, dorsal ruffles were not detected in CSF-1-stimulated BMMs from SNX5−/− mice transduced with adenovirus encoding GFP alone (Fig. S2). Given that dorsal ruffles are absent in SNX5−/− BMMs, these data show that the re-introduction of full-length SNX5 restores dorsal ruffles and macropinocytosis.
SNX5 and Arp3
Dorsal ruffles require actin branching, which is a coordinated response by a protein complex that includes Arp2/3. We initially assessed the localisation of Arp3 to peripheral membrane ruffles in response to CSF-1 stimulation. After 3 min CSF-1 stimulation at 37°C, Arp3 was localised to actin-rich regions of the peripheral membrane ruffles of SNX5+/+ BMMs as well as SNX5−/− BMMs (Fig. 8A), indicating that the localisation of Arp3 to the peripheral membrane ruffles is not dependent on SNX5. We then investigated the location of Arp3 on dorsal membrane ruffles in relation to SNX5. Here, we generated an orthogonal view of dorsal ruffles formed in SNX5+/+ BMMs after 15 min CSF-1 stimulation (Fig. 8B). Endogenous SNX5 was localised to the base of actin-rich dorsal ruffles, juxtaposing closely with actin (Fig. 8B; actin+SNX5) and Arp3 on the same dorsal ruffles (Fig. 8B; SNX5+Arp3). SNX5 could also be observed in close proximity with Arp3 on actin-rich structures resembling macropinocytic cups 3 min after CSF-1 stimulation (Fig. S3).
In contrast, SNX5−/− BMMs lack dorsal ruffles and associated actin on the dorsal surface (Figs 6B and 7). One possibility is that the distribution of the CSF-1 receptors on the dorsal surface might differ between SNX5+/+ and SNX5−/− BMMs, which could account for differences in spatial actin polymerisation. We assessed the distribution of CSF-1 receptors following activation. Upon 3 min stimulation with CSF-1, receptors on both SNX5+/+ and SNX5−/− BMMs cluster into distinct microdomains on the dorsal surface of the macrophages (Fig. S4), suggesting that the localisation of CSF-1 receptor on the dorsal membrane is not influenced by SNX5 and that SNX5 functions downstream of CSF-1 receptor activation. Taken together, the data suggests that SNX5 can act in concert with the Arp2/3 complex to regulate the formation of CSF-1-stimulated dorsal ruffles on the plasma membrane from which macropinosomes can then be formed.
DISCUSSION
Here, we have identified an essential role for SNX5 in the formation of macropinosomes derived from the dorsal ruffles of macrophages. Using mice deficient in SNX5 our findings show that: (1) macrophages lacking SNX5 have a dramatically reduced (60–70%) macropinocytic activity following CSF-1 activation compared with wild-type macrophages, (2) dendritic cells from SNX5-deficient mice have normal levels of macropinocytosis and antigen presentation, and (3) SNX5 is required for the formation of dorsal ruffles but not planar peripheral ruffles. Hence, our findings demonstrate that there are multiple pathways for macropinocytosis based on the type of membrane ruffling and, furthermore, that there are differences in the regulation of macropinocytosis between different immune cells. Moreover our study also clearly establishes the importance of dorsal ruffles in the generation of macropinosomes.
Mice deficient in SNX5 had normal life spans and were fertile. Other than a smaller size, there were no obvious developmental or physiological phenotype associated with the deletion of SNX5. Another report has described a SNX5-knockout mice which developed respiratory failure early in life (Im et al., 2013); however, these authors did not analyse macropinocytosis or dorsal ruffling in their mice. The phenotype observed by Im et al. (2013) was not observed in our SNX5-knockout mice. A major difference between the two studies is that the study by Im and colleagues (Im et al., 2013) targeted the SNX5 gene in intron 7, which could result in the expression of a 26-kDa truncated protein that has the potential to act as a dominant negative mutant, whereas in our study SNX5 was disrupted in intron 3. SNX5 has been reported to be involved in retromer-mediated endosomal sorting and endosome-to-Golgi retrograde trafficking (Wassmer et al., 2007, 2009). However, SNX5 is clearly not essential for these processes, at least in macrophages, as the retrograde trafficking of M6PR to the Golgi was unaffected by the absence of SNX5.
There are a number of common regulators of both macropinocytosis and phagocytosis, including dependence on PI3K and phosphatidylinositol-dependent recruitment of common effectors and adaptors (Swanson, 2008). However, phagocytosis was not affected by the lack of SNX5 in BMMs and moreover SNX5 was not detected on phagosomes in wild-type BMMs that had internalised insoluble zymogen particles. Hence, in macrophages at least, SNX5 is not recruited to the membranes of phagosomes and is a phosphatidylinositol-binding BAR domain protein that is selective for macropinosomes.
Previously, we noted that macropinocytosis was reduced but not completely blocked after silencing SNX5 in BMMs using microRNA (Lim et al., 2012). It was possible that the residual macropinocytic activity in this earlier study was due to incomplete silencing of SNX5. However, here we also observed a similar partial block in macropinocytosis in the complete absence of SNX5. The block in fluid-phase endocytosis in SNX5-deficient macrophages was specific for the macropinocytosis pathway as phagocytosis and clathrin-mediated endocytosis were not affected by the absence of SNX5. The reduction in uptake of either 70 kDa dextran or ovalbumin by SNX5-deficient macrophages corresponded to the reduction in membrane ruffling of CSF-1-activated SNX5−/− macrophages compared with wild-type macrophages. Hence, the loss of membrane ruffling, and in particular dorsal ruffling, would appear to fully account for the reduced macropinocytic activity as measured by fluid-phase uptake assays.
The contribution of dorsal ruffling to macropinocytosis has been a subject of some debate (Buccione et al., 2004; Hoon et al., 2012; Suetsugu et al., 2003). Although SNX5 is found on macropinosomes derived from peripheral and dorsal ruffles, our findings indicate that SNX5 is essential only for dorsal-ruffle-derived macropinosomes. Only low levels of actin ruffling were detected on the dorsal membrane of SNX5-deficient macrophages as assessed by both light microscopy and SEM analysis. By contrast, microscopic analysis showed that peripheral ruffles were prominent in SNX5-deficient macrophages. Furthermore, the migration index of the SNX5−/− macrophages was similar to wild-type macrophages. As directed migration of cells is mediated from peripheral ruffles rather than dorsal ruffles (Suetsugu et al., 2003), these findings indicate that peripheral ruffling was not affected by the absence of SNX5. Therefore, our findings show that, following CSF-1 receptor activation in macrophages, the majority of macropinocytic activity is derived from dorsal ruffles. This conclusion is consistent with an emerging view that the dorsal surface and ruffles might function as a membrane domain specialised for signalling (Abella et al., 2010; Luo et al., 2014).
What is the role of SNX5 in promoting the generation of dorsal ruffles? Our previous work has shown that SNX5 binds both phosphatidylinositol 3-phosphate (PI3P) and PIP2, the latter of which is derived from cell signalling events (Merino-Trigo et al., 2004) and that SNX5 is localised to newly emerging macropinosomes following receptor activation in HEK293 cells (Lim et al., 2008; Merino-Trigo et al., 2004). In this paper, we show that SNX5 is located at the plasma membrane following receptor activation of macrophages and that the plasma membrane localization of SNX5 is dependent on kinase activity of the CSF-1 receptor. Inhibition of tyrosine kinase activity has previously been shown to reduce macropinocytosis not only in macrophages but also in other cell types (Araki et al., 1996; Lim et al., 2008). Given that dorsal ruffles do not form in the absence of SNX5, our findings suggests that SNX5 is recruited to the plasma membrane, probably through an increase in PIP2 production following receptor tyrosine kinase activation, to promote actin polymerisation and ruffle formation on the dorsal surface, which contributes to macropinocytosis. Our work shows that SNX5 is acting upstream of Arp2/3-mediated regulation of actin polymerisation and dorsal ruffling. Identification of the SNX5-binding partners in macrophages will be important to define the mechanism by which SNX5 co-ordinates dorsal ruffling. Using GFP–SNX5-expressing HeLa cells, SNX5 has previously been reported to interact with DOCK180 (also known as DOCK1), a GEF for the small GTPase Rac1, which is known to link extracellular signals to actin dynamics (Hara et al., 2008). However, we were unable to detect DOCK180 as a binding partner of endogenous SNX5 in activated macrophages from the analysis of pulldown experiments (J.P.L. and P.A.G., unpublished observations). Hence, it is likely that there are other, as yet undefined, binding partners involved in this process in macrophages. Finally, the BAR domain of SNX5 can regulate membrane curvature (Carlton et al., 2004; Peter et al., 2004), and this SNX5 domain might also be an important factor in regulating the membrane curvature required for macropinosome formation.
Immature dendritic cells are very active in macropinocytosis, which is a crucial pathway for their immune surveillance, and uptake and presentation of antigens (Lim and Gleeson, 2011; Mercer and Helenius, 2012; Norbury, 2006). Here, we found that, in contrast to macrophages, dendritic cells were unaffected by the lack of SNX5 and were able to present antigen as effectively to T cells as wild-type dendritic cells. Our study shows that there are differences in the regulation of the biogenesis of macropinocytosis between macrophages and dendritic cells. This difference might reflect the balance between constitutive and receptor-activated macropinocytosis because dendritic cells, and to a lesser extent macrophages, are able to macropinocytose constitutively. Recently constitutive ruffling and macropinocytosis has been shown to be a distinct pathway regulated by the elevated levels of phosphatidic acid in the plasma membrane of these cells (Bohdanowicz et al., 2013). These findings highlight the need to study and compare macropinocytosis in different primary cells.
Thus far, studies to identify the functional importance of macropinocytosis in vivo have used drugs such as amiloride, Sanglifehrin A or rapamycin (Hackstein et al., 2007, 2002; von Delwig et al., 2006). However, these drugs are not specific for macropinocytosis, which represents a major limitation in defining the role of macropinocytosis in a range of physiological processes in vivo. This study, which has identified SNX5 as a regulatory component of macropinocytosis derived from dorsal ruffles, will therefore provide a targeted approach to determine the effect of inhibiting one major pathway of macropinocytosis in vivo. In addition, and more broadly, the functions of dorsal ruffling, which have remained elusive (Hoon et al., 2012), will also be able to be directly investigated.
MATERIALS AND METHODS
Antibodies and reagents
CSF-1 was purchased from Life Technologies and Merck Millipore. Rabbit anti-mouse SNX5 antibody is as described previously (Lim et al., 2012). Rat anti-mouse CD16 and CD32 antibodies (Mouse BD Fc Block™), biotinylated hamster anti-mouse CD11c antibody and rat anti-mouse I-A/I-E MHCII antibody conjugated to PerCP-Cy5.5 were from BD Biosciences (USA). Mouse anti-human CD8a antibody, anti-mouse CD11c antibody conjugated to eFluor® 450 and anti-CD86 antibody conjugated to APC were obtained from eBioscience, USA. Rabbit anti-mouse GRASP65 antibody was from Abcam (UK), and Phalloidin–Tetramethylrhodamine B isothiocyanate (TRITC-Phalloidin) and mouse anti-Arp3 antibody were from Sigma-Aldrich (USA). Horseradish peroxidase (HRP)-conjugated swine anti-rabbit-IgG and rabbit anti-mouse-IgG antibodies were from DAKO Corporation (Denmark). Mouse monoclonal anti-bovine α-tubulin and goat anti-mouse-IgG antibody conjugated to Alexa Fluor 488, goat anti-mouse-IgG antibody conjugated to Alexa Fluor 647, goat anti-rabbit-IgG antibody conjugated to Alexa Fluor 488, goat anti-rabbit-IgG antibody conjugated to Alexa Fluor 568, goat anti-rabbit-IgG antibody conjugated to Alexa Fluor 647, Alexa-Fluor-488-labelled zymosan A bioparticles, lysine-fixable fluorescein isothiocynate (FITC)-conjugated 70 kDa dextran, DQ-ovalbumin and streptavidin conjugated with Alexa Fluor 647 were purchased from Life Technologies (USA). CSF-1 receptor tyrosine kinase inhibitor (GW2580) and rabbit anti-mouse CSF-1 receptor were purchased from Merck Millipore (USA).
Genotyping
The SNX5 ‘Knock-out First’ mouse strain was created from an ESC clone (a2EPD0033_3_D07) and was obtained from the NCRR-NIH-supported KOMP repository and generated by the CSD consortium for the NIH-funded KOMP project. Control (SNX5+/+) mice used in all the experiments have the normal, unmodified SNX5 allele and are on the same C57BL/6 background as the SNX5−/− mice. Litters of weaned pups were genotyped using the following oligonucleotides: primer set 1, 5′-CTTTGCCTCTGATTTGGATCTCC-3′ and 5′-CTAAGACAATAAAACCCACCGGGCG-3′, to identify the SNX5 wild-type allele; and primer set 2, 5′-CCTCTTTAAAGGAATGAAGCCGTGGG-3′ and 5′-GAGATGGCGCAACGCAATTAAT-3′, to identify the SNX5 targeted allele. The annealing temperature for both PCR reactions was 53°C. All animal experiments were conducted according to government and institutional regulations and guidelines (University of Melbourne animal ethics approval 1212502.1).
Generation of BMMs
Eight- to 12-week-old mice were killed by CO2 asphyxiation and BMMs were generated as previously described (Lim et al., 2012). Once generated, BMMs were cultured in RPMI supplemented with 10% (v/v) foetal calf serum, 100 units/ml penicillin, 100 g/ml streptomycin and 2 mM l-glutamine (complete RPMI medium).
Dendritic cell isolation
Dendritic cells were isolated from spleens as previously described (Moffat et al., 2013; Bourges et al., 2014). Dendritic cell isolation yielded preparations with ∼70–85% CD11c+ purity. For antigen presentation assays, dendritic cells were activated by culture overnight in complete RPMI medium supplemented with 10 ng/ml granulocyte macrophage colony stimulating factor (GM-CSF, Peprotech).
Recombinant adenovirus constructs, production and transduction
Mouse SNX5 inserted at the C-terminus of GFP was as previously described (Lim et al., 2012). CD8–M6PR was amplified from pCMU-CD8/M6PR using primers which contain EcoRI and XhoI restrictions sites. The resulting PCR product was cloned into the EcoRI–SalI sites of pDC315 (Microbix Biosystems, Canada) for generating recombinant adenovirus (rADV). All constructs were verified by DNA sequencing. Production and transduction of recombinant adenovirus is as previously described (Lim et al., 2012).
Anti-CD8a internalisation
BMMs expressing CD8–M6PR constructs were chilled on ice for 5 min prior to incubation with pre-chilled Fc Block™ for 10 min. Fc Block™ was then removed and replaced with anti-CD8a antibodies (eBioscience, USA), diluted 1:100 in serum-free RPMI, on ice for 30 min. Unbound antibodies were removed using cold PBS washes and the internalisation of antibody-bound CD8–M6PR was performed in serum-free RPMI for the indicated duration at 37°C. Cells were fixed with 4% (w/v) paraformaldehyde and processed for immunofluorescence.
Phagocytosis
BMMs at density of 2×105 cells/ml were cultured overnight in complete RPMI medium in the absence of CSF-1, washed in serum-free medium, and incubated with 7×105 particles/ml and 1.6×105 particles/ml of Alexa-Fluor-488-labelled zymosan A bioparticles at 37°C for 60 min. Non-phagocytosed zymosan particles were removed with ice-cold PBS washes, cell monolayers fixed and processed for immunofluorescence. Monolayers were counterstained with TRITC–phalloidin and the number of fluorescent particles per cell was then scored for triplicate experiments of 50 cells each.
Fluid-phase uptake assays
BMMs
BMMs at a density of 2×105 cells/ml were cultured overnight in complete RPMI medium in the absence of CSF-1 to render macrophages quiescent. Then, macrophages were incubated with 500 µg/ml fluorescently conjugated 70 kDa dextran in the presence of 50 ng/ml of CSF-1 in serum-free RPMI at 37°C for various times as indicated. Following the incubation, cells were placed on ice for 5 min to inhibit uptake of dextran and cells were then washed with cold complete RPMI and PBS. Control incubations were also performed on ice. Cells were either processed for immunofluorescence or analysed live by flow cytometry. Pulse-chase with DQ-ovalbumin was performed as described (Lim et al., 2012).
For flow cytometry after dextran uptake, BMMs were incubated with 10 mM EDTA on ice for 10–15 min and then gently scraped off Petri dishes. Cells were washed twice in cold PBS containing 0.1% BSA (w/v) and 2 mM EDTA. Cells were then subjected to an acid-salt wash (0.5 M NaCl, 0.2 M acetic acid, pH 3), to remove dextran non-specifically bound to the cell surface, and washed again prior to flow cytometry using a FACSort (Becton-Dickinson, California, USA).
Splenic dendritic cells
1.8×106 Nycodenz-enriched dendritic cells were incubated with 1 mg/ml fluorescently conjugated 70 kDa dextran in 200 µl of serum-free RPMI. Cells were then washed four times in PBS containing 0.1% (w/v) BSA and 2 mM EDTA. For immunofluorescence microscopy, cells were stained with biotinylated hamster anti-mouse CD11c antibody followed by streptavidin-conjugated Alexa Fluor 647, and cell suspensions were pelleted onto microscope slides using a CytoSpin and stained with DAPI. For flow cytometry, live cells were stained with anti-CD11c antibody conjugated to eFluor® 450, anti-CD86 antibody conjugated to APC and anti-MHCII antibody conjugated to PerCP-Cy5.5 prior to flow cytometry analysis on a LSR Fortessa (BD Biosciences).
Antigen presentation assay
Antigen presentation assays with mouse spleen dendritic cells involved dendritic cell isolation and incubation with antigen. Dendritic cells were pulsed with OVA protein (Worthington Biochemicals) in the presence of fully phosphorothioated 0.5 mM CpG 1668, type B (custom made, Geneworks) for 45 min at 37°C. Antigen-pulsed dendritic cells were incubated with Cell Trace Violet (Invitrogen) labelled OT-II transgenic T cells for 64 h in complete RPMI medium supplemented with 10 ng/ml GM-CSF (Peprotech). T cell division was assessed by flow cytometry. Cell number was determined by inclusion of calibration particles (BD Biosciences).
Immunofluorescence and microscopy
Cells were fixed in 4% (w/v) paraformaldehyde, permeabilised and processed for immunofluorescence as previously described (Chia et al., 2011). Unless otherwise indicated, cells were examined using a Leica TCS SP2 system and Leica Confocal Software version 2.61. For multi-colour labelling, images were collected independently. Images were analysed using Metamorph Version 7.5.6.0 software or Volocity 6.3 software (Perkin Elmer).
Surface CSF-1 receptor level
BMMs cultured overnight in complete RPMI medium in the absence of CSF-1 were lifted from Petri dishes by incubating in 10 mM EDTA on ice for 15 min. Cells were then washed in cold PBS containing 0.1% (w/v) BSA and 2 mM EDTA. BMMs was stained with rabbit anti-mouse CSF-1R antibody for 30 min on ice followed by Alexa-Fluor-488-conjugated anti-rabbit-IgG antibody and analysed by flow cytometry.
Inhibition of CSF-1 receptor tyrosine kinase
BMMs that had been cultured overnight in in complete RMPI medium in the absence of CSF-1 to render macrophages quiescent were pre-treated with 5 µM of 5-{3-Methoxy-4-[(4-methoxybenzyl)oxy]benzyl}-pyrimidine-2,4-diamine (GW2580) or DMSO (carrier control) for 1 h at 37°C prior to dextran uptake assays or stimulation with CSF-1.
Transwell migration assay
BMMs were cultured overnight in the absence of CSF-1 to render macrophages quiescent. Cells were then harvested by gentle scraping and resuspended in complete RPMI at 2×106 cells/ml. Cell suspensions (150 µl) in complete RPMI were seeded onto the upper chamber of a transwell insert containing an 8-µm pore size Polyethylene Terephthalate membrane (24-well Millicell Hanging Cell Culture inserts, Millipore, Switzerland). The lower chamber contained either complete RPMI or supplemented with 50 ng/ml CSF-1. After 4 h incubation at 37°C, cells that had not migrated were removed with a cotton swab and the transwell inserts washed four times with PBS. Cells that had migrated to the lower surface of the membrane were fixed in 4% paraformaldehyde and stained with DAPI. The experiments were performed in triplicate and the number of migrating cells in five fields per membrane was counted microscopically at 100× magnification. Data were expressed as the migration index (number of cells migrated to the underside of the membrane in the presence of CSF-1/number of cells migrated in the absence of CSF-1).
Scanning electron microscopy
BMMs grown on 22-mm square glass coverslips were fixed with 0.4% gluteraldehyde and 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, for 16–18 h at 4°C. The coverslips were then rinsed in 0.1 M phosphate buffer followed by three PBS washes. The coverslips were then dehydrated in graded ethanol solutions in water (10%, 30%, 50%, 70%, 90%, 100% for 10 min each); 2×100% ethanol (10 min each). The coverslips were dried using a critical point dryer (Balzers CPD 030). The samples were then mounted on specimen stubs and coated with a gold–palladium alloy using a Dynavac Xenospot gold coater. The presence of membrane ruffling was examined using a Philips XL 30 FEG, field emission SEM with an accelerating voltage of 2 kV. All measurements necessary to quantify the CSF-1-induced responses were performed using ImageJ software. The percentage ruffling for each cell was obtained by dividing the ruffling area by the sum of ruffling area (R) and cell area (C) as (R×100)/(R+C) (Chitu et al., 2005).
Immunoblotting
Immunoblotting performed as previously described (Lim et al., 2012).
Statistical analysis
The unpaired two-tailed Student t-test was used to determine statistical significance. A P-value of 0.05 (*) was considered as significant, a P-value of <0.01(**) was highly significant and a P-value of <0.001 (***) was very highly significant. The absence of a P-value indicates the differences were not significant.
Acknowledgements
We thank Fiona Houghton for expert technical advice and Dr Simon Crawford, School of Botany, University of Melbourne for assistance with SEM sample preparation. The authors acknowledge the assistance of The Biological Optical Microscopy at the University of Melbourne.
Footnotes
Author contributions
Experiments were designed by J.P.L. and P.A.G. BMM experiments were performed by J.P.L., dendritic cell experiments by J.D.M., E.M.R. and J.P.L., SEM by P.G. The paper was written by J.P.L. and P.A.G.
Funding
This work was supported by funding from the National Health and Medical Research Council of Australia [grant number ID566727].
References
Competing interests
The authors declare no competing or financial interests.