ABSTRACT
Migratory macrophages play critical roles in tissue development, homeostasis and disease, so it is important to understand how their migration machinery is regulated. Whole-transcriptome sequencing revealed that CSF-1-stimulated differentiation of bone marrow-derived precursors into mature macrophages is accompanied by widespread, profound changes in the expression of genes regulating adhesion, actin cytoskeletal remodeling and extracellular matrix degradation. Significantly altered expression of almost 40% of adhesion genes, 60–86% of Rho family GTPases, their regulators and effectors and over 70% of extracellular proteases occurred. The gene expression changes were mirrored by changes in macrophage adhesion associated with increases in motility and matrix-degrading capacity. IL-4 further increased motility and matrix-degrading capacity in mature macrophages, with additional changes in migration machinery gene expression. Finally, siRNA-induced reductions in the expression of the core adhesion proteins paxillin and leupaxin decreased macrophage spreading and the number of adhesions, with distinct effects on adhesion and their distribution, and on matrix degradation. Together, the datasets provide an important resource to increase our understanding of the regulation of migration in macrophages and to develop therapies targeting disease-enhancing macrophages.
INTRODUCTION
Tissue macrophages are resident chameleons that finely tune their phenotype to adapt to the local demands of their tissue of residence (Gautier et al., 2012). For example, Kupffer cells in the liver are quite different to peritoneal or splenic macrophages or brain microglia. Further diversity is derived from the developmental origin of the precursor. While most tissue macrophages, including Kupffer cells and microglia, originate from embryonically derived progenitor cells, intestinal macrophages are largely derived from hematopoietic stem cells in the adult bone marrow (Perdiguero and Geissmann, 2015). Nevertheless, these different macrophage populations all perform common duties such as immune surveillance and injury repair along with their specialized tissue-specific functions (Okabe and Medzhitov, 2015). Similarly, most macrophage populations depend on the colony stimulating factor-1 (CSF-1)–CSF-1 receptor (CSF-1R) axis for their establishment and survival in different tissues (Wynn et al., 2013; Sullivan and Pixley, 2014). The reliance of developing and homeostatic macrophages on CSF-1R signaling is reflected in the recommendation that CSF-1-differentiated bone marrow-derived macrophages (BMMs) be considered the in vitro baseline standard for mouse macrophage experimentation (Murray et al., 2014).
The CSF-1R is a class III receptor tyrosine kinase that triggers a range of signaling pathways in BMMs and precursor mononuclear phagocytic cells (Sullivan and Pixley, 2014). It is activated by two ligands, CSF-1 and interleukin (IL)-34, which have different spatiotemporal expression patterns (Wei et al., 2010). However, whereas the CSF-1-deficient osteopetrotic mouse has many developmental defects due to reductions in tissue macrophage populations, the IL-34-deficient mouse is essentially normal, indicating that CSF-1 is the primary macrophage growth factor (Sullivan and Pixley, 2014). CSF-1 is also important for tissue homeostasis in the adult mouse through support of tissue macrophage populations (Wynn et al., 2013). The pleiotropic signaling of the CSF-1R makes it difficult to delineate signals specific to the CSF-1–CSF-1R axis. However, although other cytokines such as granulocyte-macrophage (GM)-CSF (also known as CSF-2) support macrophage survival and proliferation, CSF-1 appears to play a non-redundant role in the differentiation of macrophage precursor cells into mature adherent macrophages capable of interstitial migration (Yu et al., 2008; Pixley and Stanley, 2004; Pixley, 2012).
Cells adhere to underlying extracellular matrix (ECM) proteins through integrin-mediated adhesions, which consist of large numbers of proteins residing in or regulating the assembly and disassembly of these dynamic connections to the intracellular actin cytoskeleton. A bioinformatics approach was used to assemble an adhesome of 232 proteins categorized into adhesion receptors, adaptor proteins, actin regulators and other proteins (Winograd-Katz et al., 2014). A subsequent meta-analysis of fibronectin-induced adhesion proteins increased the list to more than 2400 proteins, from which a ‘consensus adhesome’ of 60 proteins found in at least five different cell types was determined (Horton et al., 2015). Finally, a compartmentalized network of 258 proteins was derived from the consensus and literature-curated adhesomes (Horton et al., 2016). As macrophages form tiny dot-like point contacts and small linear focal complexes rather than larger focal adhesions typical of other mesenchymal cells, the adhesome in mature macrophages contains additional selectively expressed proteins (Pixley, 2012). This is exemplified by their expression of two adhesion kinases, focal adhesion kinase (FAK; also known as PTK2) and Pyk2 (also known as PTK2B), and five Src family kinases (SFKs), Src, Fyn, Hck, Fgr and Lyn, to coordinate phosphorylation on tyrosine of many adhesion proteins during adhesion turnover and signaling (Dwyer et al., 2016; Zaidel-Bar et al., 2007; Owen et al., 2007; Abram and Lowell, 2008). Macrophages also form podosomes, which have both adhesion and matrix-degrading functions, and commonly organize into circular rosettes in mouse macrophages (Dwyer et al., 2016; Linder and Wiesner, 2015).
Cell migration requires the co-regulation of cell adhesion with actin polymerization and acto-myosin contractility (Lauffenburger and Horwitz, 1996; Devreotes and Horwitz, 2015). Rho family GTPases are key regulators of the actin cytoskeleton (Bustelo et al., 2007; Ridley, 2011). Grouped by structural similarities, they are tightly regulated themselves by activators, guanine nucleotide exchange factors (GEFs), and two groups of inactivators, GTPase activating proteins (GAPs) and Rho-GDP dissociation inhibitors (Bustelo et al., 2007; Ridley, 2011). To move through the interstitial space of tissues, macrophages recruit extracellular proteases to podosomes (Linder and Wiesner, 2015). The ECM proteases implicated in macrophage interstitial migration are matrix metalloproteinases (MMPs), cathepsins and urokinase-type plasminogen activator (uPa; also known as Plau) (Vérollet et al., 2011).
Baseline mature macrophages differentiated in the presence of CSF-1 can be further activated by cytokines. Although in vivo cytokines produce a mix of activated macrophage phenotypes, a linear range of activation from interferon (IFN)-γ-induced classically activated ‘M1’ macrophages through to IL-4-induced alternatively activated ‘M2’ macrophages is a commonly cited concept (Xue et al., 2014; Mosser and Edwards, 2008). Macrophage activation is particularly important in the context of disease as macrophages can be subverted to contribute to disease progression (Wynn et al., 2013). Notably, tumor-associated macrophages (TAMs) promote tumor progression in a number of ways (Yang et al., 2018). CSF-1R signaling in TAMs appears to play a critical role in tumor invasion and metastasis, as revealed in a mouse model of breast cancer (Lin et al., 2001). In this model, tumor cells and TAMs participated in a paracrine CSF-1–epidermal growth factor (EGF) loop, leading to their co-migration, invasion and metastasis (Wyckoff et al., 2004). TAMs are thought to predominantly adopt an alternatively activated or M2-like phenotype, and IL-4 has been shown to skew TAMs towards this tumor-promoting state (DeNardo et al., 2009).
As CSF-1 is a critical regulator of macrophage adhesion, migration and invasive capacity in vitro and in vivo (Wyckoff et al., 2004; Webb et al., 1996; Pixley et al., 2001), it is important to delineate the mechanisms by which CSF-1 regulates the development of these functions. Similarly, further characterization of tumor-promoting changes induced by IL-4 in TAMs may help to identify therapeutic targets to inhibit tumor invasion and metastasis. CSF-1 rapidly triggers BMM spreading, adhesion and motility through CSF-1R pY721-activated phosphatidylinositol 3-kinase (PI3K) p110δ (also known as PIK3CD) (Sampaio et al., 2011; Mouchemore et al., 2013), and also regulates macrophage adhesion and migration in the longer term through SFK gene expression changes via CSF-1R pY974 (Dwyer et al., 2016). A stringent ex vivo protocol that produces synchronized differentiation of non-adherent bone marrow-derived mononuclear phagocytic precursor cells into immature, adherent macrophages in the presence of low levels of CSF-1 (12 ng/ml) then more-mature macrophages in the presence of higher levels of CSF-1 (120 ng/ml) was used in Tushinski et al. (1982). To document the CSF-1-induced expression changes during this differentiation process, we used whole-transcriptome sequencing with a focus on expression changes underpinning the development of adhesion, motility and matrix degradation. We then used these baseline mature macrophages to delineate the effect of IL-4 on gene expression and correlate those changes to increased motility and matrix degradation in IL-4-treated macrophages. Finally, we reduced expression of the adhesion proteins paxillin and leupaxin to demonstrate their overlapping and distinct roles in CSF-1-induced spreading, adhesion and matrix degradation.
RESULTS
CSF-1 drives adhesion and spreading in differentiating mononuclear phagocytes and maturing macrophages
Non-adherent bone marrow cells were treated with increasing doses of CSF-1 as described above to collect mononuclear phagocytic precursor cells (C1), immature adherent (C2) and mature adherent (C3) macrophages for transcriptome analysis (Fig. 1A). This stepped increase in CSF-1 concentration stimulated differentiation of mononuclear phagocytic precursors from non-adherent cells into immature adherent macrophages and then into mature larger, better-spread macrophages (Fig. 1B).
Gene induction patterns in CSF-1-induced macrophage differentiation and maturation
Although BMMs cultured in vitro do not fully represent tissue-resident macrophages, they can be used to investigate the process of macrophage differentiation. To characterize gene expression changes underpinning the process of mononuclear phagocytic cell differentiation into adherent, mature macrophages, sequential RNA-sequencing (RNA-Seq) analysis was carried out on precursor cells (C1), immature (C2) and mature (C3) BMMs from a total of nine libraries. The similarity of the biological replicates' transcript expression profiles was demonstrated by SeqMonk principal component analysis (PCA) (Fig. 1C).
The RNA-Seq data (aligned raw reads) were then analyzed by the EdgeR pipeline in SeqMonk (P≤0.01) to identify differentially expressed genes during CSF-1-induced macrophage differentiation and maturation. CSF-1 induced the upregulation of 2407 genes and downregulation of 3355 genes as precursor cells differentiated into fully mature macrophages (C1 to C3) (Fig. 1D, C3 versus C1; Table S1). Most gene expression changes occurred during the differentiation step from non-adherent to adherent cells (C1 to C2) with fewer changes occurring during subsequent maturation (C2 to C3) (Fig. 1D; Table S1). Heatmaps for the top 50 down- and upregulated genes overall and for each transition are shown (Fig. 1E,F; see also Table S2), with a full list of differentially expressed genes in Tables S3 (C1 to C3), S4 (C1 to C2) and S5 (C2 to C3). Notably, the large number of points aligned on the x-axis of the C1-to-C3 scatter plot depicts 1239 genes for which expression was switched off altogether during differentiation into fully mature BMMs. We compared these genes to those listed for specific mouse bone marrow-derived cell populations identified in the Mouse Cell Atlas (Han et al., 2018). Consistent with commitment to macrophage differentiation, highly expressed genes in precursor cells that were switched off during this transition included neutrophil markers calgranulin B (S100a9), neutrophilic granule protein (Ngp) and lactotransferrin (Ltf), B-cell markers Cd79a and Cd19, and dendritic cell markers Flt3 and Siglech (Han et al., 2018) (Tables S1 and S6). The validity of our step-wise C1–C2–C3 differentiation model was supported by evidence that genes downregulated during differentiation of monocytes into colonic macrophages, such as Ly6C (also known as Ly6c2) and selectin-L (Sell) (Schridde et al., 2017), were all but switched off during C1-to-C2 differentiation, while expression of Ccr2, the receptor for the monocyte chemokine Ccl2, was lost in the C2-to-C3 maturation step (Tables S2-S5).
In contrast to the preponderance of hematopoietic lineage markers in the downregulated gene set, the top upregulated genes in differentiating macrophages encoded proteins with disparate functions, including structural (Nes, Fat3 and Emp2), metabolic (Ch25h) and the immune response (Fcna). We compared our list of significantly upregulated genes with markers of four in vivo mouse macrophage clusters (Han et al., 2018). Of the 85 markers found in all in vivo macrophage clusters, 44 were strongly upregulated during in vitro BMM differentiation, indicating that BMMs differentiated in the presence of CSF-1 in vitro are very similar to tissue-resident macrophages (Table S6) (Han et al., 2018). Importantly, CSF-1-differentiated macrophages did not express significant levels of either classically activated (M1) markers Il-6 or Nos2, or alternatively activated (M2) markers Arg1, Chil3 and Retnla.
Gene ontology (GO) functional analysis shows enrichment of adhesion and motility pathways
To identify the major biological processes up- and downregulated during step-wise differentiation of bone marrow precursors into mature macrophages, functional pathway analysis was carried out using the PANTHER classification system (Mi et al., 2016). Pathways important for neutrophil function plus B- and T-cell lineage commitment and associated functions of interleukin production, antigen processing and immune response dominated the downregulated gene patterns (data not shown). Consistent with changes in adhesion and morphology during CSF-1 stimulation of macrophage differentiation, upregulated GO pathways included adhesion and cell migration (Fig. S1, C1–C3). Adhesion and motility pathways also scored highly during adhesion development (C1–C2) and further maturation (C2–C3) (Fig. S1).
Allocation of genes to functional pathways using an unsupervised analysis is impacted by the extent to which those genes have been annotated to pathways in reference databases. Incomplete annotation particularly affects cells with complex transcriptomes such as macrophages (Suzuki et al., 2009). Moreover, the highly specialized motility machinery of macrophages is regulated by selectively expressed proteins and splice variants (Pixley, 2012), many of which are not included in unsupervised functional pathway analyses. To catalogue CSF-1-induced changes correlating with increased adhesion and interstitial migration during macrophage differentiation, we undertook a supervised analysis of molecules important for adhesion, cytoskeleton remodeling and matrix degradation in macrophages.
CSF-1-induced macrophage differentiation and maturation is associated with extensive changes in the expression of adhesion genes and increased organization of adhesion, leading to increased footprint area
The integrated consensus adhesome of 258 proteins was derived from cells that form focal contacts (Horton et al., 2016). As macrophages form smaller adhesions and express a unique set of adhesion-regulating proteins (Pixley, 2012), we added additional genes to the consensus adhesome, including the myeloid-restricted SFKs Hck and Fgr (Dwyer et al., 2016). The curated list was used to identify changes in expression during macrophage differentiation and maturation. Of the 280 adhesion-related genes, 53 were upregulated and 58 downregulated, producing a change in expression (P<0.01) in almost 40% of the macrophage adhesome (Table S7). Looking at the integrin adhesion receptors in detail, two thirds of integrins expressed in either precursor cells or macrophages showed altered expression, seven upregulated (α integrins – α6, α8, α9 and αV; β integrins – β1, β2 and β5) and ten downregulated (α integrins – α1–4, αe, α2b and β integrins – β3, β4, β7 and β8) (Fig. 2A, Table 1). Notably, α9 integrin expression increased 25-fold, whereas integrins α1 and α2b became almost undetectable.
Many adhesion proteins are tyrosine phosphorylated during adhesion-based signaling and turnover. FAK, its homolog Pyk2 and their associated SFKs are the tyrosine kinases responsible for adhesion protein phosphorylation (Pixley, 2012; Zaidel-Bar et al., 2007). Macrophages express both FAK and Pyk2 (Owen et al., 2007), as well as five SFKs, Src, Fyn, Hck, Fgr and Lyn (Dwyer et al., 2016; Abram and Lowell, 2008). Ptk2 (which encodes FAK) was upregulated 7-fold during CSF-1-stimulated BMM differentiation, although final levels remained almost 8-fold lower than the abundantly expressed Pyk2 (Table 1). Of the SFKs, Fyn and Hck increased, whereas Fgr levels fell sharply (Table 1). The multi-adaptor adhesion protein paxillin is phosphorylated by the adhesion kinases and dephosphorylated by the tyrosine phosphatase PTPRO (Zaidel-Bar et al., 2007; Pixley et al., 2001). While the expression of PTPRO increased more than 3-fold in differentiating macrophages, paxillin expression decreased and leupaxin, a paxillin family member, increased commensurately, while vinculin and zyxin decreased markedly (Table 1).
We used structured illumination super-resolution microscopy (SIM) to document maturation changes in the small adhesion structures found in macrophages. BMMs were stimulated with CSF-1 for 10 min to maximize adhesion formation and phosphopaxillin (pY118 paxillin) was used to detect point contacts, which are difficult to discern over background staining when imaged by total paxillin or other adhesion markers (Pixley, 2012; Pixley et al., 2001). Striking changes in adhesion size and distribution were seen as macrophages matured. CSF-1 induced the formation of coarse focal complexes with few point contacts at the leading edge of immature BMMs in contrast to the tiny point contacts and linear focal complexes seen under spreading lamellipodia of mature BMMs (Fig. 2B, left column). Changes in adhesion number, type and distribution are more evident when phosphopaxillin signal intensity is converted to a binary signal (Fig. 2B, right column). Moreover, mature BMMs appeared larger than immature macrophages and this was confirmed by tracing the footprint area of BMMs, which confirmed a more than 2.5-fold increase in size as adherent macrophages mature (Fig. 2C).
Cadherins and catenins may also regulate macrophage adhesion. Similar to the cell-matrix adhesome, the cadherin adhesome undergoes significant expression changes as macrophages mature. Table S8 lists members of the large cadherin superfamily and associated catenins expressed at reliably detectable levels (>5 reads) at one or more stages during macrophage differentiation. The most striking change seen was a >2000-fold upregulation of the atypical cadherin Fat3 (see heatmap in Fig. 1F), and several members of the protocadherin and catenin families show significant upregulation, whereas expression of E-cadherin is strongly downregulated during the development of cell-matrix adhesions (Table S8).
CSF-1-induced macrophage differentiation is associated with extensive changes in the expression of cytoskeletal remodeling genes and an increase in cell motility
Consistent with the requirement for cytoskeletal remodeling in spreading and migration, over 80% of Rho family members showed significant expression changes during macrophage differentiation (Fig. 2D, Table 2). Upregulated molecules included Cdc42, RhoJ and RhoV (Cdc42 subfamily), Rac3 (Rac subfamily), RhoB and RhoC (Rho subfamily), RhoD and Rnd2 (Rnd subfamily) and RhoBTB1 (RhoBTB subfamily). Rac2 and RhoG (Rac subfamily), RhoF and Rnd3 (Rnd subfamily), RhoH, RhoBTB2 and RhoBTB3 (RhoBTB subfamily) and RhoU (Cdc42 subfamily) were downregulated. Of the Rho family GAPs and GEFs expressed in precursor cells or differentiating macrophages, 60% showed significant expression changes when mature macrophages (C3) were compared with non-adherent cells (C1) (Table S9). This increased to 70% when genes with a significant but transient change in expression during development of adhesion, such as Arap3, were also included.
Rho family GTPases activate many downstream effectors that act on the actin cytoskeleton and other cellular targets (Bustelo et al., 2007). We used the open access REACTOME database to identify Rho family actin cytoskeletal effectors (https://reactome.org/). Expression of more than half of the effectors changed significantly during differentiation (C1 to C3), and this increased to almost 70% when transient changes were included (Table S10). Thus, most genes regulating the actin cytoskeleton in macrophages are themselves regulated by the process of differentiation. Finally, intermediate filaments have been shown to regulate cell motility as well as maintain the mechanical integrity of cells (Cheng and Eriksson, 2017). Hence, nestin (Nes), which is the most highly upregulated gene during differentiation (>2600-fold, heatmap in Fig. 1F), may also influence macrophage migration (Tables S2 and S11).
The changes in adhesion number and distribution and the increase in cell size plus the changes in expression of cytoskeletal regulators suggested that migration speed increases as macrophages mature. CSF-1-stimulated chemotaxis was assessed by transwell assay, which showed a 7-fold increase in motility in mature over immature macrophages (Fig. 2E). Thus, widespread changes in the expression of adhesion and cytoskeletal regulatory molecules during macrophage maturation support a profound increase in macrophage migration speed.
CSF-1-induced macrophage maturation is associated with extensive changes in the expression of extracellular protease genes and increased matrix degradation
As interstitial migration requires degradation of matrix proteins, we looked for significant changes in the expression of members of the three ECM-digesting groups expressed in macrophages (Fig. 2F, Table 3). Over 70% of expressed extracellular proteases (23 of 32) demonstrated significant changes in expression. Mmp12, Mmp13 and Mmp27 showed significant upregulation, while several MMPs were strongly downregulated: Mmp8, Mmp9 and the membrane-type MMPs, Mmp14 (MT1-MMP) and Mmp25 (MT6-MMP) (Fig. 2F). The cathepsin family demonstrated even more striking changes. Many cathepsins already highly expressed in non-adherent precursors demonstrated further increases during the development of adhesion and increased motility, notably cathepsins D, L, B, S, A and Z, while cathepsins F and K showed 100-fold and 30-fold increases, respectively, from very low levels in non-adherent cells (Fig. 2F). Finally, expression of the plasminogen activator Plau increased 27-fold (Fig. 2F, Table 3). To examine whether extracellular protease expression changes altered the capacity of maturing macrophages to degrade ECM, we compared the ability of immature adherent (C2) and mature adherent (C3) macrophages to digest Cy3-conjugated gelatin. Degradation was directly associated with actin-rich podosomal rosettes at both stages of maturation although it often extended well beyond rosette outlines in mature macrophages (Fig. 2H, arrows). Maturation produced a 5-fold increase in the rate of rosette formation and some mature macrophages formed podosomal arrays that produced degradation under the entire cell (Fig. 2H, arrowheads). Consequently, the total area digested increased 14-fold in mature macrophages (Fig. 2G). Taken together, these results indicate that CSF-1-induced differentiation produces mature macrophages that adhere and spread better, migrate faster, form more rosettes and degrade matrix more efficiently than immature macrophages.
IL-4-induced activation increases BMM migration and matrix degradation
Determination of the changes associated with CSF-1-induced macrophage differentiation allowed more-precise characterization of the effects of IL-4 activation on macrophage function and gene expression. To examine changes in morphology, mature BMMs were grown for 2 days in CSF-1±20 ng/ml IL-4. Compared to baseline BMMs, IL-4-activated BMMs were elongated and often multipolar (Fig. 3A). Cell tracing of F-actin-stained images demonstrated that IL-4 treatment doubled elongation (Fig. 3B). To quantify the multipolar changes, cells were grouped according to whether they had a single circular or leading-edge lamellipodium, or two, three or more pseudopodia. More than 50% of IL-4-treated macrophages had two or more pseudopodia compared to fewer than 20% of baseline BMMs (Fig. 3C). CSF-1-induced chemotaxis and matrix-degradation assays were then carried out and IL-4 was shown to enhance motility 3.5-fold and degradation 2-fold (Fig. 3D,E).
IL-4-induced activation alters the expression of migratory machinery molecules
To determine if IL-4-induced changes in macrophage function were underpinned by gene expression changes, the transcriptomes of baseline and IL-4-treated BMMs were analyzed. As single-base resolution was not required for this analysis, AmpliSeq was used in the identification of differentially expressed genes (Li et al., 2015). IL-4 treatment produced 750 upregulated and 699 downregulated genes (EdgeR, P≤0.01; Table S11). Consistent with previous studies, well-known markers of IL-4-induced macrophage activation, Arg1 and Retnla, were amongst the most highly upregulated genes, along with Ocstamp, a recently recognized IL-4-induced gene (Fig. 3F, Table 4) (Martinez et al., 2013; Martinez and Gordon, 2014; Yuan et al., 2017). All three showed striking increases in expression from undetectable or low levels, supporting the notion that CSF-1 produces baseline (M0) macrophages in vitro (Murray et al., 2014). Chil3 was not included in the AmpliSeq 20,000 mouse gene expression primer set, but the expression of two recently identified ‘M2’ markers, Egr2 and Tgm2, was also increased (28-fold and 2.3-fold, respectively; Table S12) (Jablonski et al., 2015; Martinez et al., 2013).
As IL-4-activated macrophages migrate and degrade matrix more efficiently than baseline macrophages, we examined whether IL-4 treatment alters the expression of genes regulating the motility machinery. The most striking change in core adhesion molecules was the 100-fold upregulation of integrin αX, with a 2.3-fold increase in expression of its partner β2 integrin (Fig. 3F, Table 4; Tables S12 and S13). Integrins α5, β3, β7 and β8 were also upregulated, whereas expression of integrins α6, α8 and α9 decreased (Table S12). Phosphotyrosine regulators of adhesion were affected, with upregulation of Ptk2 (which encodes FAK), Src and Fyn, downregulation of Hck, Lyn and Ptpro, and leupaxin becoming more abundant than paxillin (Table S12). Associated with the adhesion changes was a 120-fold upregulation of the ECM protein fibronectin (Fig. 3F). Several Rho family GTPases also showed significant changes in expression, with increases in RhoB (3-fold), RhoC (4-fold), RhoH and RhoJ (2-fold each), and RhoV (15-fold), and marked downregulation of Rac3, RhoD and RhoF (Table S12). Expression of the intermediate filament nestin increased further, and several MMPs demonstrated striking increases in expression, particularly Mmp12 (55-fold), Mmp13 (110-fold) and Mmp19 (10-fold), as did cathepsin K (9-fold) (Table S12). Thus, as for CSF-1-induced macrophage differentiation, the increased motility and matrix degradation of IL-4-activated macrophages is supported by widespread changes in the expression of genes regulating adhesion, migration and matrix degradation.
Leupaxin and paxillin contribute to macrophage spreading, adhesion numbers and matrix degradation in a non-redundant manner
Our RNA-Seq data showed a 4.5-fold increase in leupaxin expression, while paxillin decreased by 70% as macrophages developed adhesion (Table 1). SIM was used to compare the relative distributions of paxillin and leupaxin in macrophage adhesions in mature BMMs. Both proteins were found in point contacts and focal complexes, but paxillin stained more strongly than leupaxin in point contacts. Interestingly, leupaxin and paxillin localization did not overlap precisely in focal complexes (Fig. 4A, inset). Adhesion numbers were quantified by binary analysis and showed a 1.4-fold greater number of paxillin-rich adhesions compared with leupaxin (Fig. 4B,C).
These differences led us to examine whether paxillin and leupaxin have similar or distinct functions. Significant biological effects were seen, with a 40% and 60% reduction in paxillin and leupaxin, respectively, using targeted small interfering RNA (siRNA) (Fig. 4D). As demonstrated previously, CSF-1-stimulated spreading produced a 2.1-fold increase in the footprint area of control macrophages (Fig. 4E) (Sampaio et al., 2011). Spreading was attenuated 1.4-fold and 1.6-fold, respectively, when expression of either paxillin or leupaxin was reduced (Fig. 4E). Phosphopaxillin-rich adhesions were visualized by SIM to determine whether the blunted spreading responses were associated with reduced adhesion. Despite reduced total paxillin in paxillin siRNA cells, phosphopaxillin remains the clearest marker of CSF-1-stimulated macrophage point contacts. Consistent with Fig. 2B, control BMMs formed numerous point contacts right to the edge of the lamellipodium (Fig. 4F). In contrast, adhesions in paxillin siRNA cells were disorganized and their density did not increase towards the lamellipodial edge, while in leupaxin siRNA cells, elongated focal complexes anchoring thick F-actin cables formed away from the lamellipodial edge (Fig. 4F and insets). Binary image quantification confirmed a significant reduction in adhesion numbers in paxillin siRNA BMMs with a further decrease in leupaxin siRNA BMMs (Fig. 4G). As the restricted numbers of transduced cells precluded transwell motility analysis, we examined the effect of a reduction in paxillin and leupaxin expression on the ability of macrophages to degrade ECM. In contrast to their effects on the formation of adhesions, paxillin knockdown reduced matrix degradation by 66%, while a reduction in leupaxin expression did not significantly affect matrix-degrading capacity (Fig. 4H).
Thus, paxillin and leupaxin both mediate CSF-1-induced spreading and adhesion but they have distinct, non-redundant functions in the formation of adhesions and in matrix degradation.
DISCUSSION
Here, we have mapped changes in gene expression occurring during differentiation of bone marrow-derived mononuclear precursors into mature BMMs under the stimulus of the primary macrophage growth and differentiation factor, CSF-1. Although cultured macrophages do not fully represent tissue-resident macrophages in vivo, they are useful for the interrogation of cytokine-driven pathways important for macrophage differentiation and activation. To provide insights into progressive CSF-1-induced changes in gene expression, differentiating macrophages were examined in two stages: (1) initial acquisition of adhesion, and (2) subsequent maturation of adhesive, migration and matrix-degrading capabilities in adherent cells. As expected, scatter plots of gene expression demonstrated that immature and mature adherent BMMs were more closely related to each other than they were to non-adherent mononuclear phagocytic precursor cells. This was reflected in the 5.5-fold fewer differentially expressed genes in the C2-to-C3 transition than in the C1-to-C2 transition (Fig. 1). Nevertheless, as the significant downregulation of the monocyte chemokine receptor Ccr2 in the C2-to-C3 transition shows, important expression changes continue as macrophages mature. Neutrophil, B-cell and dendritic cell markers were strongly downregulated, as were the Myb, E2f2–4 and Nfyb,c transcription factors. The transcription factors were also switched off during phorbol ester-induced differentiation of THP-1 monocytic leukemia cells into adherent, mature macrophage-like cells (Suzuki et al., 2009) (Table S3).
Consistent with the development of an adherent phenotype upon exposure to increasing doses of CSF-1, the top upregulated pathways regulated cell adhesion and motility. Biological assays confirmed increased adhesion, motility and matrix-degrading capacity in maturing macrophages. To further validate our findings, we compared our upregulated genes to those characterizing the four macrophage populations found in the mouse by single-cell RNA-Seq (Han et al., 2018). More than 50% of the markers found in all four in vivo populations were upregulated and abundantly expressed in CSF-1-differentiated BMMs. Importantly, our in vitro-differentiated macrophages do not express significant levels of the ‘M2’ markers Arg1, Chil3 or Retnla, indicating that CSF-1-differentiated macrophages are baseline or resting BMMs that can be used to examine the effects of activating cytokines such as IL-4 (Murray et al., 2014).
Since curated GO gene lists likely do not include several of these important motility machinery molecules, a supervised analysis of adhesion, actin regulatory and matrix-degrading molecules was undertaken. Non-adherent, monocyte-like precursor cells and adherent macrophages express all four members of the leukocyte-restricted β2 integrin (Cd18) family: αL(Cd11a)β2, αM(Cd11b)β2, αX(Cd11c)β2 and αD(Cd11d)β2, with abundant expression of αMβ2 integrin (Mac-1) that increases further as macrophages mature (Table 1). In contrast, αL integrin significantly decreases during differentiation, suggesting that Cd11a and Cd11b may play different roles in diapedesis and interstitial migration. Both αMβ2 and αLβ2 integrins have been reported to assist monocyte transmigration (Weerasinghe et al., 1998). β1 integrin is also strongly upregulated during differentiation along with a 25-fold upregulation of one of its partners, α9 integrin. α9β1 integrin is highly homologous to the abundantly expressed α4β1 integrin and both bind the ECM protein osteopontin to regulate macrophage motility (Lund et al., 2013). β1 integrin also mediates adhesion to additional ECM proteins, fibronectin (FN) and laminin (LN) with its highly expressed partners, α5 and α7 integrins.
Expression of FAK, Hck and PTPRO, which regulate adhesion protein phosphorylation, is also increased during differentiation and this is reflected in the marked increase in number and degree of organization of phosphopaxillin-rich adhesions (Fig. 2). Unexpectedly, however, expression of the core adhesion molecule, paxillin, decreased, while that of its selectively expressed family member, leupaxin, increased. The paxillin family of multi-adaptor adhesion proteins consists of the ubiquitously expressed paxillin, the hematopoietically enriched leupaxin and Hic-5, which was not detected in adherent BMMs (Deakin et al., 2012). Although leupaxin was first identified in macrophages (Lipsky et al., 1998), very little is known about its role in these cells. We used super-resolution microscopy to show that, like paxillin, leupaxin is found in macrophage adhesions. However, while paxillin is evenly distributed across focal complexes and point contacts, leupaxin strongly localizes to focal complexes. Moreover, the two adhesion proteins do not precisely colocalize within focal complexes. It is interesting that leupaxin is upregulated during the transition from a focal complex-based adhesion pattern in C2 BMMs to a point contact-based spreading mechanism in C3 BMMs. Consistent with this, lamellipodia of leupaxin knockdown BMMs contain elongated focal complexes with very few point contacts, leading to a striking overall reduction in adhesion numbers (Fig. 4). In contrast, lamellipodia of paxillin siRNA BMMs contain a disorganized mix of point contacts and focal complexes, with a less marked decrease in adhesion numbers than leupaxin siRNA BMMs. Nevertheless, the CSF-1-induced spreading responses of BMMs with reduced paxillin or leupaxin expression are similarly blunted, whereas the matrix-degrading capacity of macrophages was reduced with paxillin but not leupaxin siRNA. It seems that a finely tuned balance of leupaxin and paxillin is required for mature macrophages to assemble their distinctive mix of point contacts and focal complexes under spreading lamellipodia, while paxillin may play a bigger role in matrix degradation. Interestingly, a recent study in RAW264.7 macrophages, which are not CSF-1-dependent, demonstrated that leupaxin recruitment into podosomes regulated paxillin phosphorylation and podosome turnover (Klapproth et al., 2019).
Little is known about the role of cadherins and catenins in macrophage adhesion. Macrophages have been shown to express E-cadherin (Van den Bossche et al., 2012), but we show that its expression drops sharply in adherent BMMs, whereas several protocadherin and catenin molecules increase. The most striking of these is the 2000-fold increased expression of Fat3 (Fig. 1F). Fat-like cadherins have been shown to facilitate migration in many cell types (Horne-Badovinac, 2017), with Fat3 specifically shown to regulate directed migration of amacrine neurons in the developing retina by interacting with Ena/VASP proteins at the leading edge (Krol et al., 2016). Since macrophages and neurons share selective expression of several molecules in the motility machinery and are the only cell types demonstrated to form point contacts (Pixley, 2012), Fat3 may regulate migration similarly in macrophages as in amacrine neurons.
Actin cytoskeletal remodeling is a central element of cell motility and the Rho family of GTPases are core regulators of this process. All 21 are expressed in macrophages or their precursors and significant expression changes during macrophage differentiation were seen in the vast majority (Table 2). In macrophages, classical Rac proteins promote ruffling, lamellipodia formation and adhesion, Cdc42 stimulates filopodia formation and adhesion, and RhoA promotes acto-myosin contractility (Allen et al., 1997). However, it is difficult to determine the overall effect of changes in expression of the various Rho GTPases as the final expression levels of even some of the most highly upregulated genes, such as Rac3 (90-fold), RhoD (100-fold) and RhoJ (22-fold), remain quite low. Nevertheless, changes to ∼70% of Rho family GTPase activators, inhibitors and effectors indicate that regulation of actin cytoskeletal remodeling undergoes profound changes as macrophages become adherent and motile.
Intermediate filaments (IFs) not only maintain the structural integrity of cells but also modulate cell motility. Tissue repair transiently changes IF expression from networks that provide stability to those that permit remodeling (Cheng and Eriksson, 2017). Vimentin, which is abundant in macrophages and non-adherent precursor cells, is upregulated in injured tissues, where it enhances cell migration, possibly through mechanosensing (Cheng and Eriksson, 2017) (Table S11). Nestin, which is upregulated 2600-fold in differentiating macrophages, has previously been shown in microglia and possibly macrophages after brain injury and inflammation (Krishnasamy et al., 2017). It is also highly expressed in neural progenitor cells (Hyder et al., 2014). Indeed, the similar expression pattern of unusual motility molecules such as Fat3, Rac3 and nestin in macrophages and migrating neural progenitors, along with other molecules such as Pyk2, suggests that these two cell types share a similar mechanistic regulation of cell motility.
Depending on the density and organization of the ECM proteins, macrophages can adopt either mesenchymal or ameboid mechanisms of migration (Guiet et al., 2011; Friedl and Wolf, 2010; Linder and Wiesner, 2015). Mesenchymal motility requires the ability to secrete matrix-degrading proteases (Guiet et al., 2011). Basement membranes contain type IV collagen, LN and FN, while the main components of interstitial ECM are fibrous proteins such as type I and III collagens, elastin, FN and tenascin plus the proteo-glycans perlecan and aggrecan (Frantz et al., 2010; Hynes and Naba, 2012). Thus, it is likely that monocytes migrating through the endothelial basement membrane use different ECM proteases to macrophages migrating through interstitial tissue. These differences are reflected in the extensive and highly significant expression changes seen for MMPs, cathepsins and uPa (Fig. 2F, Table 3). The most strongly upregulated MMP was the macrophage elastase Mmp12 (94-fold). Mmp13, which encodes a protein that cleaves interstitial collagens, and Mmp27 were also highly upregulated although their final levels remained low in mature BMMs. Although downregulation of the neutrophil proteases Mmp8, Mmp9 and Mmp25 was expected, downregulation of Mmp14 (MT1-MMP) was surprising, considering that others have shown it to be important in monocyte and macrophage migration (Vérollet et al., 2011). However, the divergent expression pattern of MMPs in human and mouse macrophages prevents inter-species expression profile comparisons (Newby, 2016). The cathepsins demonstrated profound changes in expression of all but one of those expressed in macrophages. Cathepsins A, B, D, L, S and Z, all of which were already expressed at levels between 140 reads per million reads (RPM) and 1500 RPM in non-adherent precursors, underwent additional increases, while cathepsin F and cathepsin K levels increased 100-fold and 30-fold, respectively, albeit from a low baseline (Fig. 2F). In contrast, expression of cathepsins E, G and W was all but switched off, and cathepsin H expression underwent a 4-fold decrease, leaving only cathepsin C unchanged. It is likely that cathepsins, along with MMP12 and uPa, play a central role in the 14-fold increase in the matrix-degrading capacity of mature macrophages. Interestingly, migration by macrophages into gelled collagen is independent of MMPs and relies on cathepsins (Van Goethem et al., 2010), whereas infiltration into tumor spheroids requires MMPs (Guiet et al., 2011). Thus, macrophages modulate expression of both groups of matrix proteases to respond to the various demands of the microenvironment.
The combined alterations in expression of adhesion, motility and matrix-degrading molecules mapped by RNA-Seq underpin development of the highly efficient migratory machinery in macrophages. By characterizing CSF-1-induced changes in mature macrophages, we were then able to clearly distinguish additional changes induced by IL-4. IL-4 is found in the tumor microenvironment, where it induces M2-like TAMs that co-migrate with tumor cells to promote invasion (Wyckoff et al., 2004, 2007). Hence, we focused on the effects of IL-4 on the migration machinery of macrophages. IL-4 altered morphology and increased motility in mature macrophages, which was consistent with previous studies demonstrating elongation (McWhorter et al., 2013) and increased motility (Vogel et al., 2014) in mouse and human macrophages, respectively, and increased microglial migration speed and invasiveness (Lively and Schlichter, 2013; Cougoule et al., 2012). To identify IL-4-induced adhesion, motility and matrix protease gene expression changes, we used AmpliSeq, a reliable and cost-effective method to examine changes in gene expression. However, with primer pairs limited to 20,000, its gene coverage is not as comprehensive as RNA-Seq (Li et al., 2015), as demonstrated by the absence of a widely used M2 marker, Chil3. Nevertheless, our IL-4-upregulated gene set matched that of a previous study of IL-4-activated mouse macrophages (Martinez et al., 2013). Overall, gene expression changes in IL-4-treated macrophages were not as marked as those seen with CSF-1-induced differentiation, but striking changes in the regulators of interstitial motility were seen (Fig. 3F, Table 4). Most importantly perhaps were the striking increases in expression of Mmp12, Mmpp13 and Mmp19, along with cathepsin K, which are likely to be responsible for the increase in matrix-degrading capacity of IL-4-treated BMMs (Table S12). By increasing their migration and matrix degradation, IL-4 is likely to enhance the ability of TAMs to move through the interstitium in vivo and thereby promote invasion of co-migrating tumor cells. TAMs have previously been shown to be a major source of cathepsins B, L and S in the tumor microenvironment (Gocheva et al., 2010). We found that these cathepsins were already abundantly expressed in baseline macrophages with moderate additional IL-4-induced increases demonstrated in cathepsins L and S. Thus, at an RNA level at least, IL-4-treated macrophages express vast quantities of a number of cathepsins along with several MMPs, and, although we do not know the protein levels of these matrix-degrading molecules, IL-4-skewed macrophages are almost certainly very efficient at interstitial migration.
In conclusion, we have comprehensively mapped gene expression changes and measured commensurate increases in adhesion, migration and matrix degradation in macrophages differentiated in the presence of the primary macrophage growth and differentiation factor, CSF-1. Furthermore, we have been able to show definitively that CSF-1-differentiated macrophages are not skewed towards an alternatively activated phenotype but are baseline or ‘M0’ in phenotype. In addition, by extensively characterizing baseline macrophages, we have been able to precisely delineate changes in gene expression and activity of the motility machinery found in IL-4-induced alternatively activated macrophages. Together, these findings are important as macrophages contribute to disease progression in a range of deadly diseases. Furthermore, the identification and selective targeting of core macrophage-specific motility molecules can be used to inhibit the recruitment or activity of disease-enhancing macrophages.
MATERIALS AND METHODS
Macrophage extraction and culture
Bone marrow was flushed from the femurs and tibiae of 8- to 10-week-old male C57BL/6 mice in three independent experiments. All experiments complied with the Australian code for the care and use of animals for scientific purposes and were approved by the University of Western Australia's Animal Ethics Committee. Non-adherent cells were plated in α+ minimum essential medium (MEM; Life Technologies, NY) containing 15% fetal calf serum (FCS; Bovogen, Melbourne, Australia) and a very low dose (0.6 ng/ml) of CSF-1 (kind gift from E. R. Stanley, Albert Einstein College of Medicine, New York, USA) in T75 tissue culture flasks at 37°C and 5% CO2 to maintain mononuclear phagocytic cell survival as previously described (Tushinski et al., 1982). The following protocol was based on a well-established procedure optimized in the Stanley laboratory that initially enriches for non-adherent mononuclear phagocytic cells in a very low concentration of CSF-1 (0.6 ng/ml) before stimulating synchronous differentiation of the precursor cells into adherent but immature macrophages with an increased dose of CSF-1 (12 ng/ml). Once adherent, BMMs require higher doses of CSF-1 (120 ng/ml), which promotes further maturation (Tushinski et al., 1982). Thus, on day 0, bone marrow-derived cells were plated in 0.6 ng/ml CSF-1 and on day 1, non-adherent cells were collected and resuspended in fresh 0.6 ng/ml CSF-1 medium for a further 24 h before lysis and RNA extraction (Fig. 1A, C1). To produce immature adherent BMMs, on day 2, non-adherent cells were then exposed to 12 ng/ml CSF-1 in α+ MEM and 10% FCS for 4 days (Fig. 1A, C2) (Tushinski et al., 1982). To produce fully mature BMMs, day 1 non-adherent precursor cells were exposed to 12 ng/ml CSF-1 for 48 h then 120 ng/ml CSF-1 for a further 7 days (Fig. 1A, C3) (Tushinski et al., 1982). For IL-4-induced activation, fully mature BMMs were kept in CSF-1 and additionally treated with 20 ng/ml IL-4 (Miltenyi, Macquarie Park, Australia) for 48 h. To ensure viability when splitting adherent cells (2 mm EDTA in PBS), proliferating BMMs were expanded in Petri dishes before plating for at least 2 days on either tissue culture plates (Eppendorf, North Ryde, Australia) for RNA extraction, or fibronectin-coated coverslips (Fisher Biotec, Wembley, Australia) or glass-bottomed dishes (MatTek, Ashland, Australia) for biological assays. At least three biological replicates were carried out for each assay.
Transcriptome and differentially expressed gene analysis
For immature and mature adherent BMMs grown in CSF-1, cells were seeded onto 2×10 cm tissue culture dishes and grown to ∼70% confluence (∼0.5–1×106 cells) then lysed and polynucleotides treated with DNase prior to RNA extraction (QIAGEN, Chadstone, Australia). For non-adherent bone marrow-derived precursor cells, a similar number of cells was collected by centrifugation prior to lysis. Eluted RNA from three independent samples for each of the three stages of BMM differentiation and maturation was submitted to the LotteryWest State Biomedical Facility Genomics laboratory at the University of Western Australia for library preparation and sequencing using the Ion Torrent Proton DNA sequencing platform. For each sample, the output fastq files were aligned against the Genome Reference Consortium mouse genome (GRCm38.p6) using the open-source Hisat2 aligner (Kim et al., 2015). The resulting BAM files were uploaded into SeqMonk (version 1.42) with minimum mapping quality set to 60. Aligned sequences passed quality checks outlined in SeqMonk (https://www.bioinformatics.babraham.ac.uk/projects/seqmonk/). PCA analysis and expression scatter plots were generated in SeqMonk using log2 RPM expression. Changes in fold expression were calculated from the log2 RPM data. Lists of differential gene expression were generated using the edgeR platform within SeqMonk using the raw reads as is required in analysis of negative binomial distributions (Schurch et al., 2016). Heatmaps of differentially expressed genes were constructed using the R package pheatmap (Kolde, 2019). For samples treated with and without IL-4, mRNA was again submitted to the LotteryWest State Biomedical Facility Genomics laboratory for conversion to cDNA followed by amplification using the AmpliSeq Mouse Transcriptome V1 PCR primers [Ion AmpliSeq™ Transcriptome Gene Expression Kit mouse (A36553), Life Technologies Australia]. Primers were then partially digested with FuPa and adapters (Ion Xpress™ Barcode Adapters 1-16 Kit, Thermo Fisher Scientific) ligated to amplicons. Libraries were purified, quantitated and six to eight barcoded samples loaded onto chips (Ion PI™, Life Technologies Australia) and sequenced using the Ion S5 System (Thermo Fisher Scientific). Enrichment analysis of the differential gene expression datasets was carried out through GO, using the PANTHER Classification System (Mi et al., 2016) to identify biological pathways up- or downregulated during differentiation and maturation of macrophages from non-adherent precursors.
Morphology assays
For phase-contrast microscopy, cells were grown on tissue culture dishes then imaged with a Nikon Eclipse TS100 inverted microscope with a Nikon CFl Achromat LWD ADL20xF 0.4 NA objective. For immunofluorescence microscopy, 5.0×103 BMMs were seeded onto fibronectin-coated coverslips (Neuvitro, Vancouver, Australia) for 2 days in the appropriate concentration of CSF-1 with or without IL-4, then fixed in 4% paraformaldehyde and permeabilized as described previously (Dwyer et al., 2016). The actin cytoskeleton was stained with Alexa Fluor 568-conjugated phalloidin (Thermo Fisher Scientific) and adhesion structures were stained with an anti-phosphopaxillin pTyr118 antibody (#610051, 1:100, Thermo Fisher Scientific) or a monoclonal anti-paxillin antibody (#610051, 1:200, BD Biosciences) and a polyclonal anti-leupaxin antibody (#STJ111845, 1:100, St John's Laboratory, London, UK) to identify focal complexes and point contacts (Pixley, 2012). Coverslips were mounted in Prolong Diamond with 4′,6-diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific) and imaged on an Olympus IX-81 inverted epifluorescent microscope with an Olympus Plan Apo 60×/1.25 NA oil immersion objective, a Nikon A1 inverted confocal microscope with a Nikon Plan Apo VC 60×/1.49 NA oil immersion objective or a Nikon SIM (Ti2) microscope with a Nikon SR Apo TIRF 100×/1.49 NA oil immersion objective at the Centre for Microscopy, Characterisation and Analysis, University of Western Australia. Images were processed in NIS-Elements (Nikon) and Fiji (Schindelin et al., 2012). To quantify cell footprint area, F-actin stained outlines of at least 50 cells per condition were digitally traced in Fiji. Footprint area (arbitrary units) and elongation (aspect) ratio measurements were calculated and expressed as the mean and s.e.m. for each value. To measure adhesion signal intensity, the mean pY118 paxillin signal was taken from raw data files using Fiji. To quantify adhesion numbers at the leading edge, we adapted the method from Buskermolen et al. (2018) to remove background by thresholding.
Motility assay
For chemotaxis, 1×105 BMMs in 100 µl CSF-1-free medium were seeded into 8 µM pore-size transwell chambers (BD Biosciences) and placed into medium with 120 ng/ml CSF-1 in the lower chamber. BMMs were incubated for 6 h before migrated cells on the lower surface of the membrane were fixed and the membranes carefully excised, mounted and imaged with a Nikon Eclipse TS100 inverted microscope at 20× magnification. Cells from ten fields/membrane were counted, averaged and controlled for cell loading by adding 1×104 cells to insert-free wells.
Matrix degradation assay
BMMs (2×105) were seeded onto Cy3-labelled gelatin-coated 35 mm glass-bottomed dishes in the appropriate dose of CSF-1 plus 20 ng/ml IL-4 for 24 h as described previously (Mouchemore et al., 2013) (Merck, Frenchs Forest, Australia). After fixation and permeabilization, the actin cytoskeleton was stained with Alexa Fluor 488-conjugated phalloidin and nuclei were stained with DAPI. Samples were imaged with five representative fields at 10× magnification by the Olympus IX81 microscope and degraded areas (absence of Cy3 signal) relative to total cell area and total number of cells actively degrading per field were thresholded, quantified and normalized to cell number using ImageJ software.
Transfection of siRNA
Electroporation was performed using the Neon transfection system (Thermo Fisher Scientific) with specific siRNA molecules targeting paxillin (Santa Cruz Biotechnology, sc-36197), leupaxin (Thermo Fisher Scientific, S98816) and a scrambled siRNA (Silencer® Negative Control #2, Thermo Fisher Scientific, AM4613). Electroporation was performed as follows: BMMs were lifted (2 mM EDTA in PBS) and resuspended in buffer R. Two ×105 cells per transfection were mixed with either 1.0 µM paxillin or 0.8 µM leupaxin or scrambled siRNA and collected in a 100 μl Neon electroporation tip. The microporator was set to a pulse voltage of 1000 V, pulse width of 40 ms and pulse number of 2. Cells were plated as required into pre-warmed growth medium onto fibronectin-coated coverslips or matrix-degradation plates, and allowed to recover for 48 h before fixation and staining for immunofluorescent microscopy.
Statistics
Statistical analyses were carried out in GraphPad Prism. Where necessary, the data were transformed to achieve normality. Unpaired Student's t-tests with Welch's correction (Figs 2C,E,G, 3G and 4G) or paired Student's t-tests (Fig. 4C) were used in two-way comparisons. One-way (Fig. 4G,H) or two-way (Fig. 4E) analyses of variance (ANOVAs) were used for multiple comparisons. Statistically significant data are denoted by asterisks as follows: *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.
Acknowledgements
We acknowledge the facilities, and the scientific and technical assistance of the National Imaging Facility at the Centre for Microscopy, Characterisation and Analysis, the University of Western Australia, with particular gratitude for the help of Alysia Buckley with SIM. We thank Dr Richard Stanley for the kind gift of CSF-1, Michaela Rajarathnan and Nick Peters for their technical assistance, and Dr Wally Langdon and Dr Prue Hart for their critical insights and evaluation of the manuscript.
Footnotes
Author contributions
Conceptualization: M.W.M., J.H.S., E.L.G., D.A.J., F.J.P.; Formal analysis: M.W.M., J.H.S., J.M.P., F.J.P.; Investigation: M.W.M., E.L.G., F.J.P.; Writing - original draft: F.J.P.; Writing - review & editing: M.W.M., J.H.S., D.A.J.; Supervision: F.J.P.; Project administration: F.J.P.; Funding acquisition: F.J.P.
Funding
This work was supported by funding from the Cancer Council Western Australia [APP1078830 and APP1122300 to F.J.P.], the University of Western Australia [Australian Postgraduate Award to M.W.M. and E.L.G.] and the Royal Perth Hospital Medical Research Foundation.
Data availability
Raw RNA-Seq data have been deposited in GEO under accession number GSE145437.
References
Competing interests
The authors declare no competing or financial interests.