Genetic engineering of Hoxb8-immortalized hematopoietic progenitors – a potent tool to study macrophage tissue migration

Macrophages are present in all tissues of the organism where they play a central role in the clearance of microorganisms, tissue homeostasis, the mediation of immune and inflammatory responses, and tissue repair. However, tissue infiltration of macrophages also exacerbates pathological processes, such as chronic inflammation, neurodegenerative disorders and cancers (Allavena et al., 2008; Condeelis and Pollard, 2006; Friedl and Weigelin, 2008; Yiangou et al., 2006). Tumor-associated macrophages (TAMs) mainly originate from blood monocytes (Lahmar et al., 2016), and are recruited to the tumor stroma at all stages of cancer progression (Allavena et al., 2008; Condeelis and Pollard, 2006). TAMs can represent more than 50% of the tumor mass, and their number positively correlates with poor prognoses in most cancers (Noy and Pollard, 2014; Ruffell and Coussens, 2015). TAMs are involved in several cancer-promoting events such as angiogenesis, lymphangiogenesis, immunosuppression, resistance to therapy and metastasis formation (Noy and Pollard, 2014; Ruffell and Coussens, 2015). Therefore, the control of human macrophage infiltration in cancers is a current therapeutic challenge (Mantovani et al., 2017; Morrison, 2016; Ostuni et al., 2015). However, many questions regarding the mechanisms of macrophage migration remain unresolved and hinders development of novel therapeutic strategies.

Cell migration in tissues occurs in three dimensions (3D), which deeply differs from 2D migration processes (Doyle et al., 2009; Hooper et al., 2006). Human macrophages are one of the few cell types that can use two migration modes in 3D environments: amoeboid and mesenchymal (Van Goethem et al., 2010). During amoeboid movement, rounded, ellipsoid or moderately elongated cells that form blebs or generate small actin-rich filopodia (Friedl and Wolf, 2003; Lämmermann and Sixt, 2009; Sahai and Marshall, 2003). These cells do not require adhesion to the extracellular matrix (ECM), but use rather a propulsive and pushing migration mode (Fackler and Grosse, 2008; Lämmermann et al., 2009; Paluch et al., 2016). This rapid and non-directional motility involves acto-myosin contractions and depends on the Rho–ROCK pathway. Mesenchymal migration consists of cells that present an elongated and protrusive morphology (Fackler and Grosse, 2008; Friedl and Wolf, 2003; Sahai and Marshall, 2003). The movement is slow and directional, involves cell adhesion to the substratum, and requires proteases to degrade the ECM in order to create paths through dense environments. In both human and mouse macrophages, this migration is dependent on podosomes and is not inhibited but rather stimulated by treatment with ROCK inhibitors (Gui et al., 2018, 2014). It has been shown in vitro that, in contrast to lymphocytes, neutrophils and monocytes (Cougoule et al., 2012; Friedl and Weigelin, 2008; Lämmermann and Germain, 2014), tumor cells or immature human dendritic cells (DCs) can perform these two mechanistically distinct migration modes (Cougoule et al., 2018; Friedl and Wolf, 2003; Sahai and Marshall, 2003). Human and mouse macrophages can also adopt these two migration modes depending on the ECM architecture (Barros-Becker et al., 2017; Čermák et al., 2018; Cougoule et al., 2010, 2012; Gui et al., 2018, 2014; Guiet et al., 2012; Jevnikar et al., 2012; Maridonneau-Parini, 2014; Van Goethem et al., 2011, 2010; Vérollet et al., 2011, 2015). In vivo in mouse tumors and ex vivo in human breast cancer explants, TAMs use the protease-dependent mesenchymal migration mode. At the tumor periphery or in inflamed ear derma in mice, macrophages use the amoeboid motility (Gui et al., 2018). Inhibition of matrix metalloproteases (MMPs) blocks the mesenchymal migration of human macrophages ex vivo and mice macrophages in vivo, which correlates with a decrease in both macrophage infiltration into tumors and tumor growth in vivo (Gui et al., 2018). This provides the rationale for a new strategy in cancer immunotherapy to specifically target TAMs through their motility. The use of the first generation of broad spectrum MMP inhibitors in clinics has been tested, but proved in the past to be toxic (Overall and Kleifeld, 2006). Thus, identification of specific effectors is a key step in the investigation of new potential therapeutic targets affecting specifically the mesenchymal mode of migration, and thus that inhibit a migration mode that is used by TAMs (and possibly by only a few other types of cells).

The comprehensive understanding of the mechanisms used by macrophages to migrate through the mesenchymal mode requires an exhaustive approach that could lead to the identification of a large number of potential targets as recently described (Čermák et al., 2018). All these potential effectors now need to be validated as effective players in macrophage migration both in vitro and in vivo through functional studies. For such screening approaches, many studies use bone marrow-derived macrophages (BMDMs) from wild-type (WT) and knockout (KO) mice. However, this has several drawbacks, such as the use of numerous animals, the limited number of cells and the impossibility of generating stable mutants in primary cells. Macrophage cell lines such as murine Raw 264.7 cells or human U937, HL-60 or THP1 cells are also widely used, but they are distantly related to blood macrophages or BMDMs, particularly because they are cancer cells. We therefore intended to establish immortalized macrophages that counter these drawbacks.

Expansion of murine hematopoietic precursors that were transiently immortalized with a retrovirus-delivered and estrogen-inducible form of the transcription factor Hoxb8 have been described (Wang et al., 2006) and validated for the study of hematopoietic cell biology (Cabron et al., 2018; Chu et al., 2019; Di Ceglie et al., 2017; Gurzeler et al., 2013; Hammerschmidt et al., 2018; Lee et al., 2017; Rosas et al., 2011; Wang et al., 2006; Witschi et al., 2010; Zach et al., 2015). The recent coupling of this long-term hematopoietic progenitor cell line to the CRISPR/Cas9 technology (Hammerschmidt et al., 2018; Roberts et al., 2019) has enabled the creation of new genetically modifiable cell models. During the last few years, expression of Hoxb8 has been used in several studies mainly focused on DC biology (Cabron et al., 2018; Grajkowska et al., 2017; Hammerschmidt et al., 2018; Leithner et al., 2018; Rosas et al., 2011), but only rare studies explored its use to generate surrogate macrophages (Cabron et al., 2018; Roberts et al., 2019; Wang et al., 2006). Only one of these studies directly compared macrophages derived from Hoxb8 progenitors to macrophages derived from bone marrow. This study describes the association of Hoxb8 conditional immortalization and CRISPR/Cas9 to specifically study macrophage biology in infectious settings (Roberts et al., 2019).

We herein combined the unlimited proliferative capacity of estrogen receptor-regulated Hoxb8 expression (ER-Hoxb8)-immortalized hematopoietic progenitor cells with the CRISPR/Cas9 technology to create a potent tool to identify and investigate the role of new effectors of macrophage functions and particularly their motility. We established stable transgenic ER-Hoxb8 progenitor cells and differentiated them into macrophages. They exhibited the main macrophage functions as compared to BMDMs. As a proof of concept, ER-Hoxb8 stable progenitors were used to obtain macrophages depleted for WASP (also known as WAS), an Arp2/3 activator known to be an effector of macrophage motility (Linder et al., 1999). WASP-deficient ER-Hoxb8 cells showed altered mesenchymal migration in vitro and ex vivo as described in BMDMs of KO mice (Park et al., 2014). Therefore, we present here a powerful tool to genetically manipulate macrophages and explore their functionalities in a broad range of applications.

ER-Hoxb8-immortalized hematopoietic progenitor cells were generated as described (Wang et al., 2006). Briefly, hematopoietic stem cells were isolated from the bone marrow of WT C57BL/6 mice and infected with an estrogen receptor (ER)-regulated Hoxb8 retrovirus allowing proliferation of macrophage progenitors in the presence of estrogen (hereafter denoted Hoxb8 progenitor cells). These immortalized progenitors had a high proliferation rate (Fig. S1A) and could be stably maintained in culture for 4 months. After selection of stably transduced progenitors, differentiation into macrophages was induced by estrogen removal and addition of macrophage colony-stimulating factor (M-CSF).

To characterize the capacity of the immortalized Hoxb8 progenitors to differentiate into fully functional macrophages (hereafter called Hoxb8-macrophages), we compared their characteristics to BMDMs in parallel experiments.

First, cell morphology was observed by scanning electron microscopy. As shown in Fig. 1A, while Hoxb8 progenitor cells showed a round morphology and poorly adhered to the substrate, Hoxb8-macrophages were flat and adherent like BMDMs. Spreading of Hoxb8-macrophages and BMDMs on coverslips were comparable (Fig. 1B).

Next, differentiation of Hoxb8 progenitor cells was investigated by analysis of surface expression of macrophage-specific markers by FACS. As shown in Fig. 1C, both Hoxb8-macrophages and BMDMs expressed CD11b, CD115 and F4/80 at comparable levels.

To test the capacity of Hoxb8-macrophages to polarize into pro- or anti-inflammatory macrophages (the M1 and M2 states, respectively), we analyzed their morphology and surface expression of standard markers (Mantovani et al., 2013; McWhorter et al., 2013; Sica and Mantovani, 2012). When incubated overnight with lipopolysaccharide (LPS) and interferon γ (IFNγ) to induce M1 activation, Hoxb8-macrophages spread (Fig. S1B) and expressed higher levels of CD86 and iNOS (also known as NOS2) than M0 (resting) or M2 cells (Fig. 1D). When incubated with IL-4 to induce M2 activation, they acquired an elongated morphology (Fig. S1B) and the expression of Arg1, CD206 (also known as MRC1), Fizz1 (RETNLA) and Ym1 (CHIL3) was enhanced (Fig. 1D).

Next, the microbicidal functions of Hoxb8-macrophages were examined. Reactive oxygen species (ROS) production was measured in response to a phagocytic process triggered by zymosan particles. ROS production was assessed by fluorescence microscopy using a cell-permeant reactive oxygen species (ROS) detection agent (Fig. 2A). As shown in Fig. 2A, while little ROS production was detected in resting Hoxb8-macrophages and BMDMs, it was enhanced in both cell types after stimulation with zymosan. Of note, the signal was enriched around phagosomes (Fig. 2A, arrow in inset), as previously described (DeLeo et al., 1999). Similar results were obtained with a cytochrome c reduction assay, which was used to detect superoxide anion production (Le Cabec and Maridonneau-Parini, 1995) (Fig. 2B). Phagocytosis of FITC–zymosan was next measured by fluorescence microscopy (Fig. 2C,D). Again, the phagocytic capacity of Hoxb8-macrophages was comparable to that of BMDMs, with similar percentages of phagocytic cells and similar number of ingested particles per cell being observed (Fig. 2C), as recently described for engulfment of Mycobacterium tuberculosis and Listeria monocytogenes (Roberts et al., 2019).

Taken together, these results indicate that the immortalized Hoxb8 cells efficiently differentiate into macrophages with comparable polarization, phagocytosis and ROS production capacities to BMDMs.

Since our objective is to use Hoxb8-macrophages as a tool to determine the molecular mechanisms involved in macrophage tissue infiltration, we next analyzed their migration capacities. Hoxb8-macrophages adhered to the substrate and migrated in 2D similarly to BMDMs (Fig. S1C). In the 3D motility assay, Hoxb8-macrophages migrated inside matrices and adopted the expected elongated and protrusive phenotype of mesenchymal migration in Matrigel, and a round morphology characteristic of amoeboid migration in fibrillar collagen similar to what is seen for BMDMs (Fig. 3A; Movies 1–3). Note that the morphology of Hoxb8-macrophages at the surface of Matrigel is more rounded than that of BMDMs (Movie 2, top). Hoxb8-macrophage infiltration in fibrillar collagen was comparable to that of BMDMs with 41.7±3.1% and 47.9±2.3%, respectively (mean±s.d., n=2–4). In Matrigel, it was slightly reduced from 59.3±2.3% (n=4) for BMDMs to 38.2±9.3% (n=3) for Hoxb8-macrophages (Fig. 3B). Migration distances inside matrices were comparable in fibrillar collagen (1110±349 µm for Hoxb8-macrophages versus 757±413 µm for BMDMs; mean±s.d. n=2–4) and slightly reduced in Matrigel (577.5±86.2 µm for Hoxb8-macrophages versus 890±141 µm for BMDMs) (Fig. 3B).

In both mouse and human macrophages, the mesenchymal migration in Matrigel depends on protease activity and is therefore inhibited upon treatment of cells with a protease inhibitor mix (Gui et al., 2018, 2014; Guiet et al., 2011; Jevnikar et al., 2012; Van Goethem et al., 2010). In contrast, amoeboid migration in fibrillar collagen is dependent on the Rho–ROCK pathway and is therefore inhibited by the ROCK inhibitor Y-27632 (Gui et al., 2018, 2014; Guiet et al., 2011; Jevnikar et al., 2012; Van Goethem et al., 2010). To ascertain that Hoxb8-macrophages used the mesenchymal migration in Matrigel and the amoeboid migration in fibrillar collagen, as expected for macrophages, protease inhibitor mix or Y-27632 were used in the migration assay. As shown in Fig. 3B, Hoxb8-macrophages migration in Matrigel and fibrillar collagen were significantly inhibited by the protease inhibitor mix and Y-27632, respectively. The above results show that migration of Hoxb8-macrophages matches characteristics of mesenchymal and amoeboid migration when migrating in 3D in Matrigel and fibrillar collagen respectively, with regard to cell morphology, cell dynamic and drug responsiveness (Gui et al., 2018; Van Goethem et al., 2010).

The capacity of macrophages to perform mesenchymal migration is correlated with their capacity to form podosomes and to degrade the ECM (Cougoule et al., 2018, 2012; Maridonneau-Parini, 2014). Podosomes are submicrometer adhesion cell structures formed at the plasma membrane that protrude into, probe, and degrade the extracellular environment by releasing proteases (Labernadie et al., 2014; Linder and Kopp, 2005; Wiesner et al., 2014). They appear during macrophage differentiation and are not present in monocyte progenitors or in monocytes (Cougoule et al., 2012). They are dynamic cell structures with different spatial organizations (Poincloux et al., 2006; Van Goethem et al., 2011). In resting human macrophages, they are mainly scattered all over the surface of the cell (Poincloux et al., 2006; Van Goethem et al., 2011). In BMDMs, they more often assemble as podosome clusters (Fig. 4A, BMDM, upper row) or rosettes (Fig. 4A, BMDM, lower row), as previously described (Cougoule et al., 2010). The presence of podosomes in Hoxb8-macrophages has previously been observed but their organization has not been compared to BMDMs (Cabron et al., 2018). Here, we show that Hoxb8-macrophages exhibited podosome clusters and rosettes at the same proportion as do BMDMs (Fig. 4A). Consequently, they were able to readily degrade the FITC–gelatin, although less efficiently than BMDMs (Fig. 4B).

Finally, we tested the capacity of Hoxb8-macrophages to infiltrate tumoral tissue ex vivo and in vivo (Gui et al., 2018). For ex vivo experiments, fibrosarcoma were generated by subcutaneous injection of LPB tumor cells (Gui et al., 2018), and then tumors were resected and sliced. Cell Tracker-labeled BMDMs or Hoxb8-macrophages were layered on tumor slices and co-cultured for 3 days. The effects of BB-94, a broad-spectrum MMP inhibitor, and Y-27632 were explored in parallel using tissue slices from the same tumor. As shown in Fig. 5A, Hoxb8-macrophages were able to infiltrate tumor slices very efficiently as compared to BMDMs. Similar to BMDMs, their infiltration into tissue was significantly inhibited by BB-94 but not Y-27632 (Fig. 5B,C) showing that both BMDMs and Hoxb8 macrophages use the mesenchymal mode to infiltrate tumor explants, as previously described for human macrophages (Gui et al., 2018). For in vivo experiments, fibrosarcoma were generated in a surgically implanted dorsal window chamber as previously described (Gui et al., 2018). Hoxb8-macrophages stained with Cell Tracker were co-injected with tumor cells and monitored by intravital microscopy. As shown in Movie 4, Hoxb8-macrophages adopt a mesenchymal phenotype, confirming the results obtained in vitro and ex vivo.

Taken together, these results demonstrate that Hoxb8-macrophages represent a relevant model to study podosome formation, ECM degradation and macrophage migration amongst other functions.

To explore whether the Hoxb8 cell line is a valuable tool to study molecular mechanisms of macrophage tissue migration, we decided to silence the expression of a known effector of macrophage mesenchymal migration, the Wiskott–Aldrich syndrome protein (WASP) (Calle et al., 2008; Linder et al., 1999; Park et al., 2014), using the CRISPR/Cas9 technology. WASP is a hematopoietic cell-specific protein and an actin nucleation-promoting factor that regulates Arp2/3-dependent actin polymerization (Takenawa and Suetsugu, 2007). WASP activity is required for several macrophage functions, including podosome formation, chemotaxis and mesenchymal migration, but not amoeboid migration (Linder et al., 1999; Park et al., 2014).

In contrast to undifferentiated Hoxb8 progenitor cells, Hoxb8-macrophages expressed WASP at a level comparable to that in BMDMs (Fig. S1D). WASP-KO Hoxb8-macrophages (denoted Wasp^−/−) were generated using the lentiCRISPRv2 transfer plasmid containing a guide (g)RNA sequence targeting exon 1 of the Wasp gene. As a control, cells were transduced with ‘empty’ lentiviruses expressing Cas9 but devoid of gRNA (called Wasp^+/+).

As shown in Fig. 6A, WASP was efficiently depleted by 91.6±2.2% (mean±s.d., n=3) in Wasp^−/− Hoxb8-macrophages as compared to Wasp^+/+ or the parental M0 Hoxb8-macrophages. WASP depletion was not compensated for by the upregulation of N-WASP expression (Fig. S1E) as previously described in Wasp-KO mice (Isaac et al., 2010).

We first checked whether knockout of WASP impacted differentiation of Hoxb8 progenitor cells. As shown in Fig. 6B, the expression of CD11b, CD115 and F4/80 was comparable between Wasp^+/+ and Wasp^−/− Hoxb8-macrophages. Cell morphology and spreading of Wasp^+/+ and Wasp^−/− Hoxb8-macrophages were also similar (Fig. 6C). Podosome rosettes appeared similar in both Wasp^+/+ and Wasp^−/− macrophages, but podosome clusters appeared disorganized and less abundant in Wasp^−/− cells (Fig. 7A,B). In addition, the global percentage of cells with podosomes and the proportion of cells with single podosomes and podosome clusters were significantly reduced in Wasp^−/− Hoxb8 macrophages compared to control cells (Fig. 7B). The ability of Wasp^−/− Hoxb8-macrophages to degrade FITC–gelatin was significantly reduced (Fig. 7C,D), as was their capacity to migrate using the mesenchymal mode (Fig. 8A), thus recapitulating results obtained with BMDMs isolated from Wasp knockout mice (Calle et al., 2004; Park et al., 2014).

Finally, we tested the capacity of Wasp^+/+ and Wasp^−/− Hoxb8-macrophages to infiltrate tumors ex vivo. Both the number of tumor-infiltrated Wasp^−/− Hoxb8-macrophages and the distance of infiltration were reduced (Fig. 8B). These results show that genetic invalidation of Wasp affected Hoxb8-macrophage migration in tumor as previously described in vitro in Matrigel and tumor cell spheroids for BMDMs isolated from Wasp-KO mice (Park et al., 2014).

In conclusion, these results demonstrate that CRISPR/Cas9-mediated genome editing in Hoxb8 cell progenitors is highly efficient and represents a potent tool to genetically manipulate macrophage migration effectors.

In the present study, we describe the generation of conditionally immortalized hematopoietic progenitors that can be expanded without limitation, kept in culture, cryopreserved, genetically modified and that fully retain the potential to differentiate into functional macrophages. Using the CRISPR/Cas9 technology, we show that it provides the opportunity to easily manipulate gene expression, a precious tool to study macrophage biology, and particularly macrophage tissue migration. This is complementary to a recent publication using this cell model to study bacterial pathogenesis (Roberts et al., 2019).

Understanding the biology of macrophages is crucial for a large number of therapeutic contexts, such as chronic inflammatory and neurodegenerative diseases, and the majority of solid cancers (Allavena et al., 2008; Condeelis and Pollard, 2006; Friedl and Weigelin, 2008; Yiangou et al., 2006). Macrophage-directed immunotherapy is emerging as a means to combat cancer and inflammatory diseases. Identification of the molecular effectors of macrophage biological functions is thus a requisite to either reduce the damaging effect or boost the protective role of macrophages depending on the disease concerned.

The main caveat in studying macrophages is the lack of tools to manipulate gene expression. siRNA approaches provide only partial inhibition, require specific investigations for each gene to be knocked down and is often inefficient for the removal of stable proteins. CRISPR-Cas9 strategies are more appropriate as this technique allows complete deletion of genes and replacement by a mutant if required. However, it is an experimentally time-consuming approach for cells with a short lifespan that are obtained in a limited number either by drawing blood of humans or preparing bone marrow from mice. Immortalizing progenitors isolated from mouse bone marrow and the subsequent manipulation of gene expression by CRISPR/Cas9 is a powerful alternative with a large number of applications including studies in vitro, ex vivo and in vivo.

Thus HoxB8 conditionally immortalized progenitors cells are a powerful tool to study macrophage functions and genetically manipulate them using the CRISPR/Cas9 technology. Here, we validated the model by analyzing the expression of cell surface differentiation markers, polarization capacity, phagocytosis and production of ROS, migration, podosome formation and ECM degradation capacities, providing complementary properties to HoxB8-macrophages to those described in infectious settings (Roberts et al., 2019).

We report that the phenotypes of BMDMs and Hoxb8 macrophages were similar regarding both differentiation markers and functions including cell adhesion, macrophage polarization, phagocytosis and ROS production. Differences were only observed in Matrigel experiments with a less-efficient in vitro mesenchymal migration for Hoxb8-macrophages, which is consistent with a diminished ability to degrade FITC–gelatin and their rounder morphology on top of Matrigel in comparison to BMDMs. Matrix infiltration is an integrin- and protease-dependent process that also implies re-arrangements of the cytoskeleton. Further work is required to determine the altered mechanism(s) but, interestingly, the capacity of Hoxb8-macrophages to infiltrate tumor tissue ex vivo was slightly increased compared to BMDMs. This discrepancy underlines the differences that can be observed between in vitro experiments performed on acellular ECMs and ex vivo experiments performed on living tissues.

In our hands, Hoxb8-macrophages can be kept in culture for 3–4 months until their migration capacity and macrophage differentiation potential deteriorate. They can also be frozen, which allows one to work with the same source of cells and reduces the number of mice used. This restricted time is in agreement with the 16 weeks delay described by Hammerschmidt et al. (2018) and with the standard time that cell lines can be kept in culture without alteration of their behavior. It provides sufficient time for multiple genetic manipulations but limits the possibility to select sub-clones of the different genetically modified cell lines. Hence, developing advanced culture procedures in the future to optimize maintenance of Hoxb8 immortalized progenitor cells and extend the culture delay would greatly improve the cell system proposed herein.

Within the scope of our study, we propose that Hoxb8 progenitor cells are a source of genetically modifiable macrophages for the identification and investigation of tissue migration effectors. As a proof of concept, we deleted WASP, a known actor of migration that partially disrupts podosomes (Isaac et al., 2010; Linder et al., 1999; Park et al., 2014; Vijayakumar et al., 2015). By combining this cellular tool and our ex vivo tissue migration assay, we describe a powerful experimental model that is adapted to large-scale screening of molecular effectors involved in macrophage migration. This could allow to define the macrophage migration mode within a large collection of healthy or pathological tissues isolated from different mouse models. The use of ER-Hoxb8 monocytes injected in blood circulation to study tissue recruitment of leukocytes from blood has been recently described (Gran et al., 2018). A future objective will be to examine by intravital microscopy the migration of macrophages in tumors (Gui et al., 2018) by labeling genetically modified Hoxb8-macrophages and directly injecting them within tumors generated in dorsal window chambers. Using this approach, the infiltration of blood monocyte/macrophages into tumors will be by-passed, including the step of monocyte trans-endothelial migration.

Besides the exploration of gene function by deletion, CRISPR/Cas9-mediated genome editing of Hoxb8-immortalized hematopoietic precursor cells also provides the opportunity to easily perform multiple gene deletions, gene overexpression or expression of specific mutants or reporter genes (Salsman and Dellaire, 2017). Added to the capacity of Hoxb8 cell progenitors to differentiate into a large scale of hematopoietic cells, such as DCs (Hammerschmidt et al., 2018), granulocytes (Wang et al., 2006), osteoclasts (Zach et al., 2015) or even B and T lymphocytes (Redecke et al., 2013), it represents a potent, rapid and simple model that opens unlimited possibilities for analysis of leukocyte biological functions including tissue migration.

C57BL/6 wild-type mice were bred and housed in the accredited research animal facility of the Institute of Pharmacology and Structural Biology (IPBS) that is fully staffed with trained husbandry, technical and veterinary personnel. All experiments were performed according to animal protocols approved by the Animal Care and Use Committee of the IPBS.

ER-Hoxb8 progenitor cells were generated as previously described (Wang et al., 2006). Briefly, bone-marrow cells were harvested from C57BL/6 mice and hematopoietic progenitor cells were purified by centrifugation on a cushion of Ficoll-Paque. Then, 5×10⁵ cells were pre-stimulated for 2 days with complete RPMI 1640 medium supplemented with 15% fetal calf serum (FCS), 1% penicillin-streptomycin, 1% glutamine, mouse IL-3, mouse IL-6 and mouse SCF (10 ng/ml each) (Peprotech) in a six-well culture plate. Next, 2.5×10⁵ cells were seeded on fibronectin-coated 12-well culture plates in myeloid medium [complete RPMI 1640 medium, mouse GM-CSF (20 ng/ml) and β-estradiol (1 µM)] and transduced with 1 ml ER-Hoxb8 Retrovirus by spinoculation (1000 g, 90 min, 22°C) in the presence of polybrene (24 µg/ml). Polybrene was diluted serially by exchanging half of the medium to obtain a final concentration of 1.4 µg/ml over several days. Antibiotic selection of transduced ER-Hoxb8 cells was performed 3 days post infection with G418 (1 mg/ml) and maintained for at least 2 weeks. Immortalized ER-Hoxb8 cells were maintained and enriched in myeloid medium (supplemented or not with G418) by serial passages of non-adherent cells every 3 to 4 days.

For macrophage differentiation, immortalized ER-Hoxb8 progenitor cells were harvested, washed twice with PBS to remove GM-CSF and β-estradiol and cultured at 1×10⁶ cells in complete RPMI 1640 medium supplemented with mouse M-CSF (20 ng/ml) on glass coverslips in six-well culture plate for 7 days.

gRNA sequence against Wasp gene were designed using the Benchling.com web tool for genome editing, and selected with high ‘on-target’ and low ‘off-target’ activity. A gRNA sequence targeting exon 1 of Wasp gene was obtained (forward, 5′-GCTGAACGGCTGGTCCCCCT-3′; reverse, 5′-AGGGGGACCAGCCGTTCAGC-3′) and was cloned into the LentiCRISPRv2 plasmid (Addgene #49535) using BsmB1 overhangs after hybridization and phosphorylation of the gRNA. After transformation in DH5α-competent bacteria, insertion of the gRNA was confirmed by gel electrophoresis of PCR products and the plasmid was purified using the QiaQuick gel extraction kit (Qiagen) following the manufacturer's instructions. For virus production, 4×10⁶ HEK293T cells (Lenti-X™ 293T Cell Line, Takara #632180) were seeded in 75 cm² flasks in DMEM supplemented with 10% FCS, 1% penicillin-streptomycin and 1% glutamine. The next day, cells were transfected with the lentiCRISPRv2 transfer plasmid (5 µg) and both packaging plasmids [pSPAX2 (3.75 µg) and pMD2G (1.25 µg), Addgene] with the transfection reagent polyethylenimine (PEI). After 24 h, the medium was replaced and, at 48 h post transfection, virus-containing supernatants were harvested, filtered and used for transduction of immortalized ER-Hoxb8 cells. A lentiCRISPRv2 plasmid without any gRNA was used to generate ‘empty’ viruses as a control.

To generate the ER-Hoxb8 Wasp^−/− cells, 10⁶ ER-Hoxb8 progenitor cells were seeded on fibronectin coated six-well culture plates in myeloid medium and transduced with 2 ml of viral suspension (as described above) by spinoculation (1000 g, 90 min, 22°C) in the presence of polybrene (8 µg/ml). Polybrene was serially diluted by renewing half of the medium over several days to obtain a final concentration of 1.4 µg/ml. Antibiotic selection of transduced ER-Hoxb8 cells was performed 3 days post-infection with puromycin (10 µg/ml) and maintained for at least 2 weeks. Immortalized ER-Hoxb8 Wasp^−/− cells were maintained and enriched in myeloid medium with puromycin (2.5 µg/ml) by serial passages of non-adherent cells every 3–4 days into new six-well culture plates. The same protocol was used to generate Wasp^+/+ control ER-Hoxb8 cells, except that ‘empty’ viruses were used.

Bone marrow cells were isolated from femurs and tibias of C57BL/6 wild-type mice, cultured and differentiated as described previously (Cougoule et al., 2010).

Equal amounts of total cell lysates in Laemmli buffer were separated by SDS-PAGE, transferred onto nitrocellulose membranes that were probed with mouse monoclonal anti-vinculin (V9131, Sigma-Aldrich, 1:500), polyclonal rabbit anti-WASP (H-250, sc-8353 Santa Cruz Biotechnology, 1:200) and anti-N-WASP (H-100, sc-20770 Santa Cruz Biotechnology, 1:100) antibodies, and revealed by an enhanced chemiluminescence system (Immobilon™ Western, Millipore).

ER-Hoxb8 cells and BMDMs were cultured on glass coverslips for 16 h, fixed using 0.1 M sodium cacodylate buffer supplemented with 2.5% (v/v) glutaraldehyde and prepared as previously described (Lizarraga et al., 2009) for observation with a JEOL JSM-6700F scanning electron microscope.

Migration assays were performed in 24-transwells (8-µm pores). Empty transwells were used for 2D migration (Cougoule et al., 2012). Transwells were loaded with either 100 µl of Matrigel (Corning) or 120 µl fibrillar collagen [Collagen 2.15 mg/ml (Nutragen), MEM 10×, H₂O, Bicarbonate buffer] and used for 3D migration (Van Goethem et al., 2010). Matrices were allowed to polymerize for 30 min at 37°C, and rehydrated for 3 h with RPMI 1640 supplemented with 1% penicillin-streptomycin and 1% glutamine. The lower chamber was filled with RPMI 1640 containing 10% FCS and 20 ng/ml mouse M-CSF, and the upper chamber with RPMI 1640 containing 1% FCS and 20 ng/ml mouse M-CSF. For the migration inhibition assay, the medium in the upper and lower chambers were supplemented with Y27632 20 µM (VWR) or a cocktail of protease inhibitors [6 µM leupeptine (Sigma), 0.044 TUI/ml aprotinine (Sigma), 2 µM pepstatine A (Sigma-Aldrich), 5 µM GM-6001 (VWR) and 100 µM E64C (Peptides International)]. Vehicle was loaded as a control. Macrophages (6×10⁴) were serum starved for 3 h and seeded in the upper chamber. For 2D migration, after 16 h of migration, cells that did not migrate through the porous membrane were removed and cells that migrated underneath the membrane were fixed, stained with DAPI and counted as previously described (Cougoule et al., 2012). For 3D migration assays, after 48 h of migration, a z-series of images was acquired at the surface of the matrices (Fig. 3, top) and inside the matrices with 30 µm intervals (Fig. 3, inside). Experiments were performed in duplicate. Acquisition and quantification of cell migration was performed using the motorized stage of an inverted video microscope (Leica DMIRB, Leica Microsystems, Deerfield, IL), Metamorph software and the cell counter plugin of the ImageJ software as described previously (Van Goethem et al., 2010). Migration distance was taken as the maximal migration depth crossed by macrophages inside the matrix. The percentage of migration was obtained after counting the number of cells within the matrix and dividing by the total number of counted cells, as described previously (Van Goethem et al., 2010). For live-cell imaging of 3D migration (see Movies 1–3), pictures at the matrix surface and at 300 μm below the surface were recorded every 10 min during 20 h, using the 10× objective of an inverted video microscope (Leica DMIRB, Leica Microsystems, Deerfield, IL, USA) equipped with an incubator chamber to maintain constant temperature and 5% CO₂ levels.

6×10⁴ macrophages were seeded on coverslips for 24 h. Cells were fixed with 3.7% paraformaldehyde (Sigma-Aldrich), permeabilized with 0.1% Triton X-100 (Sigma) and stained with anti-vinculin antibody (clone V9131, dilution 1:100, Sigma-Aldrich) followed by Alexa Fluor 488 anti-mouse immunoglobulin G Fab2 (Cell Signaling Technology, 1:500) and Texas Red-coupled phalloidin (1:200, Molecular Probes Invitrogen). Slides were visualized with a Leica DMLB fluorescence microscope as previously described (Van Goethem et al., 2011).

Coverslips were coated with 0.5 µg/ml poly-L-lysine and 0.2 mg/ml FITC–gelatin (Invitrogen). Macrophages (6×10⁴) were cultured for 24 h on FITC–gelatin, fixed and processed for F-actin staining. Quantification of matrix degradation was performed as previously described (Cougoule et al., 2010).

ER-Hoxb8 macrophages (5×10⁵cell/well in P6-well plate) were incubated in complete RPMI 1640 medium supplemented with LPS (1 µg/ml) and IFNγ (20 ng/ml), or IL-4 (20 ng/ml), for macrophage polarization towards M1 or M2 macrophages, respectively as previously described (Cougoule et al., 2012). After overnight culture, cells were harvested and analyzed by FACS as described previously (Gui et al., 2018).

Macrophages were stained with LIVE/DEAD fixable blue dead cell stain kit (L23105, Invitrogen) for exclusion of dead cells and non-specific binding to Fc receptors was blocked by incubating the cells with TruStain FcX anti-CD16/32 antibodies (93, BioLegend, 1:1000). For extracellular staining, macrophages were incubated for 30 min at 4°C in the dark with the following murine antibodies: anti-Ly6C-FITC (AL-21, BD Pharmingen, 1:100), anti-CD11b-BV785 (M1/70, Biolegend, 1:200), anti-F4/80-AF647 (BM8, Biolegend, 1:200), anti-CD115-APC (CSF-1R)(AFS98, Biolegend, 1:200), anti-CD206-PE-Cy7 (C068C2, Biolegend, 1:200), anti-CD86-PE (GL1, eBiosicience, 1:200) antibodies. For intracellular staining, cells were fixed and permeabilized with fixation (00-8222-49, eBioscience) and permeabilization (00-8333-56, eBiosience) reagents and then incubated for 45 min at room temperature with the following antibodies: anti-iNOS-PE (CXNFT, eBioscience, 1:300), and anti-Arg1 (A1exF5, eBioscience, 1:200). The acquisition was undertaken with a BD LSR Fortessa X-20 flow cytometer under BD FACS Diva software. Data were analyzed with FlowJo software (Tree Star) as described (Gui et al., 2018).

For gene expression analysis, macrophage RNA was extracted and purified according to the manufacturer's instructions (RNeasy kit, Qiagen). Then, RNA were reverse transcribed into cDNA using M-MLV Reverse transcriptase (Invitrogen). Real-time quantitative PCR (RT-qPCR) was performed with specific primers for β-Actin (forward, 5′-GCTGTGCTGTCCCTGTATGCCTCT-3′; reverse, 5′-CCTCTCAGCTGTGGTGGTGAAGC-3′), Ym1, (forward, 5′-TACCCTATGCCTATCAGGGTAATGA-3′, reverse, 5′-CCTTGAGCCACTGAGCCTTC-3′) and Fizz1 (forward: 5′-TGCCCTGCTGGGATGACT-3′, reverse: 5′-AGTTGCAAGTATCTCCACTCTGGA-3′), using TB Green Premix Ex Taq (Takara). Real-time qPCR reactions were carried out using the 7500 Real-Time PCR System and data were analyzed using the 7500 Software version 2.0.6 (Applied Biosystems). RT-qPCR data were normalized to the expression of the β-actin-encoding housekeeping gene, and expressed as ΔCt.

To monitor the intracellular accumulation of ROS by microscopy, the general oxidative stress indicator, CM-H2DCFDA, the carboxylated analog of the cell-permeant agent 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA) (Thermo Fisher Scientific) was used. BMDMs and Hoxb8-macrophage cells (1.5×10⁵ cells/well on 24-well plate) were primed with LPS 100 ng/ml overnight in complete RPMI medium. Cells were washed with phenol-free RPMI and incubated with 200 µg/ml zymosan for 30 min at 37°C. Cells were fixed and slides were visualized with a Leica DM-RB fluorescence microscope.

To quantitatively evaluate the NADPH oxidase activity, the superoxide dismutase (SOD)-inhibitable cytochrome c reduction assay was performed as previously described (Le Cabec and Maridonneau-Parini, 1995). BMDMs and Hoxb8-macrophage cells (5×10⁵ cells/well on 24-well plate) were primed with LPS 100 ng/ml overnight in complete RPMI medium. Supernatant were removed and cells were incubated with cytochrome c (15 mg/ml) with or without SOD (10,000 U/ml) and zymosan (500 µg/ml) in phenol-free RPMI medium for 90 min at 37°C. Reactions were stopped on ice and the reduction rate of oxidized cytochrome c was monitored spectrophotometrically at 550 nm.

5×10⁴ cells were plated on glass coverslips in RPMI medium at 37°C overnight. Next day, the medium was washed away and FITC-labeled zymosan particles were added to cells at the MOI of 2, in serum-free medium. After a 5-h incubation, cells were washed in 5% BSA (Euromedex)-containing PBS, then fixed with 3.7% paraformaldehyde (Sigma), permeabilized with 0.1% Triton X-100 (Sigma), and stained for F-actin with Texas Red-coupled phalloidin (1:200, Molecular Probes) and for nuclei with DAPI. Slides were visualized with a Leica DMLB fluorescence microscope as described previously (Le Cabec et al., 2002).

To generate fibrosarcomas, 10⁶ LPB cells were injected subcutaneously into the mouse flank as described previously (Gui et al., 2018). The LPB cell line is a highly tumorigenic murine clonal derivative of TBL.CI2, a methylcholanthrene-induced C57BL/6 mouse sarcoma cell line. Tumors were allowed to grow until ∼1 cm³, then mice were killed, and tumors were resected. Tumor samples were embedded in 3% agarose prepared in PBS (Life Technologies). Slices measuring 500 µm were obtained with a microtome dedicated to live tissues (Price et al., 1998), with the Krumdieck tissue slicer (TSE Systems) filled with ice-cold PBS (Life Technologies) set to medium blade and arm speeds. Slices were cultured on a 30-mm cell culture insert featuring a hydrophilic PTFE membrane (0.4 µm pore size, Merck Millipore) placed inside six-well plates containing 1.1 ml complete RPMI 1640 medium with or without 10 µmol/l BB-94 (Sigma-Aldrich) or 20 µmol/l Y-27632 (Euromedex). A 5-mm diameter stainless-steel washer was then placed on top of each tissue slice to create a well for macrophages seeding (Gui et al., 2018). The same day, co-cultures were performed by seeding 5×10⁵ murine macrophages (BMDMs or Hoxb8-macrophages or Wasp^+/+ or Wasp^−/−) pretreated or not for 16 h with DMSO, BB-94 or Y-27632 on top of tumor slices and incubated in 37°C, 5% CO₂ environment for 3 days. Murine macrophages were labeled with Cell Tracker (C7025-Green-CMFDA, Invitrogen) according to the manufacturer's instructions prior to seeding on tissue slices. After 16 h of co-culture, the washer was removed. RPMI 1640 supplemented with the indicated inhibitors was replaced daily for 3 days before overnight fixation with formalin at 4°C (Sigma-Aldrich).

Dorsal window chamber surgery and tumor induction for intravital experiments was performed as described previously (Gui et al., 2018). Before attachment of the coverslip onto the window frame, 20 µl of DMEM containing 10⁶ LPB-GFP cells (Life Technologies) and 10⁶ Hoxb8-macrophages labeled with Cell Tracker Red CMPTX (Molecular Probes) were co-injected subcutaneously in the remaining connective tissue of the lower skin layer. After 3 days, cells were imaged. Intravital microscopy was carried out on a customized stage for holding mice using a 7MP upright multiphoton microscope (Zeiss) equipped with a 20×1.0 NA objective (d.f.=1.8 mm) and a Chameleon-Ultra II laser (Coherent). Animals were anesthetized by isoflurane inhalation throughout the imaging session. Animal temperature was maintained at 37°C with an Air-Therm-heated environmental chamber and a heating blanket placed under the mouse. The tumor, localized by the fluorescence of LPB cells expressing GFP, was positioned under the objective and Cell Tracker-labeled Hoxb8-macrophages as well as collagen (through second-harmonic generation) were observed using a 565–610 nm band-pass filter. A 3D image stack of the sections with a z-spacing of 2 µm was acquired every 3 min for 2 hours to assess 3D cell motility through a dynamic z-stack time series.

Tissue slices used in co-cultures with macrophages were embedded in paraffin and processed as previously described (Gui et al., 2018). Sections were stained with anti-fluorescein (ab6556 Abcam) and a peroxidase-coupled secondary antibody. Macrophage infiltration was quantified as previously described (Gui et al., 2018).

Statistical differences were analyzed with the two-tailed paired or unpaired Student's t-test using GraphPad Prism 6.0 (GraphPad Software Inc.). P-values are indicated in each figure. P<0.05 was considered statistically significant.

We gratefully acknowledge David B. Sykes for the generous gift of ER-Hoxb8 tools and Etienne Meunier for the generous gift of CRISPR/Cas9 tools, and both for their advice. We acknowledge Yoann Rombouts for his expertise in macrophage polarization. We thank the Toulouse Reseau Imagerie platform, the Bivic facility and Anexplo facility. We acknowledge Isabelle Fourquaux from the CMEAB (Faculté de médecine de Rangueil, Toulouse, France) for SEM experiments.