Fascin1 Promotes Cell Migration of Mature Dendritic Cells

Abstract

Dendritic cells (DCs) play central roles in innate and adaptive immunity. Upon maturation, DCs assemble numerous veil-like membrane protrusions, disassemble podosomes, and travel from the peripheral tissues to lymph nodes to present antigens to T-cells. These alterations in morphology and motility are closely linked to the primary function of DCs, antigen presentation. However, it is unclear how and what cytoskeletal proteins control maturation-associated alterations, in particular, the change in cell migration. Fascin1, an actin-bundling protein, is specifically and greatly induced upon maturation, suggesting a unique role for fascin1 in mature DCs. To determine the physiological roles of fascin1, we characterized bone marrow-derived, mature DCs from fascin1 knockout mice. We found that fascin1 is critical for cell migration: Fascin1 null DCs exhibit severely decreased membrane protrusive activity. Importantly, fascin1 null DCs have lower chemotactic activity toward CCL19 (a chemokine for mature DCs) in vitro, and in vivo, Langerhans cells show reduced emigration into draining lymph nodes. Morphologically, fascin1 null mature DCs are flatter and fail to disassemble podosomes, a specialized structure for cell-matrix adhesion. Expression of exogenous fascin1 in fascin1 null DCs rescues the defects in membrane protrusive activity, as well as in podosome disassembly. These results indicate that fascin1 positively regulates migration of mature DCs into lymph nodes, likely by increasing dynamics of membrane protrusions, as well as by disassembling podosomes.

INTRODUCTION

Upon maturation, dendritic cells (DCs) change their functions from antigen sampling to antigen presentation, in a process involving massive alterations in their morphology and motility. The actin cytoskeleton plays key roles in multiple aspects of DC function (1–5). For example, antigen presentation by mature DCs depends on the integrity of the actin cytoskeleton (5). Rac1/2, a small G-protein that is responsible for ruffling movements, has been shown to be essential for the interaction of DCs with T-cells (1). WASP (Wiskott-Aldrich Syndrome protein), a molecule that controls Arp2/3-dependent actin polymerization, is required for the formation of the immunological synapse (IS) (2, 6). Although there is abundant evidence that alterations in actin dynamics are important for maturation-associated changes in DCs, it is not clear which actin regulatory proteins induce the profound alterations in morphology and motility observed upon DC maturation. Rac and Cdc42, which control the assembly of membrane ruffling and filopodia, respectively (7), are unlikely to play major roles in these alterations. Rac activity appears to be either unchanged or increased minimally upon maturation (8, 9), while Cdc42 activity is decreased upon maturation (9).

In addition to the profound alterations in morphology and motility, DCs show reduced adhesion to the substrate upon maturation, as evidenced by the loss of podosomes, specialized structures for cell-matrix adhesion (4, 10–12). Podosomes consist of an actin core surrounded by a characteristic ring structure containing adhesion molecules including vinculin, talin, paxilin and integrin (13, 14). DCs transiently lose podosomes at about 20min after activation with LPS, then recover podosomes by 2h. Later, they permanently lose podosomes at about 5–7 hr after activation, concomitant with the generation of characteristic veil-like membrane protrusions (4, 10, 11). While the first and transient loss of podosomes is controlled via pathways involving prostaglandin E2, RhoA and rho-kinase (10, 15), as well as ADAM17 (12), it is not clear what causes the second and permanent loss of podosomes in mature DCs.

Fascin1, an actin-bundling protein (see for review 16), is specifically induced to a great extent upon DC maturation while it is not detectable in immature DCs (17–20). Other blood cells such as primary macrophages and T-cells do not express or induce fascin1. The induction of fascin1 in DCs suggests a specific role for fascin1 in DC maturation. Fascin1 has been suggested to be critical for assembly of filopodia or membrane protrusions. Fascin1 is localized to filopodia (21–25) and fascin1 overexpression induces membrane protrusions and increases cell motility of epithelial cells and colon carcinomas (26–28). Conversely, fascin1 knockdown has been reported to block assembly of filopodia in cultured mammalian cells (27–29). In bone marrowderived DCs, fascin1 depletion by anti-sense treatment resulted in the inhibition of both dendrite formation and T-cell activation (19, 20, 30). However, little is known about what roles fascin1 plays in motility and cytoskeletal reorganization associated with DC maturation.

To determine the physiological function of fascin1 in DC maturation, we generated fascin1 knockout (KO) mice (31). Fascin1 KO mice provide an excellent experimental system, allowing us to analyze fascin function in DCs with little experimental manipulation. In addition, fascin1 KO mice provide fascin1 null phenotypes that are complete and uniform, unlike anti-sense or siRNA approaches that tend to be partial and variable. Our results indicate that fascin1 profoundly alters DC cytoskeletal organization and plays critical roles in migration of mature DCs into lymph nodes for antigen presentation.

MATERIALS AND METHODS

Reagents, antibodies and fascin1-deficient mice

The following antibodies were used: FITC-conjugated hamster anti-mouse CD11c monoclonal (BD Biosciences, San Diego, CA), FITC-conjugated rat anti-mouse CD86 (B7-2) monoclonal, FITC-conjugated rat anti-mouse I-A/I-E (MHC-II) monoclonal (BD Biosciences), mouse anti-vinculin monoclonal (Sigma, St. Louis MO), rabbit anti-α-actinin antibody (25), mouse anti-fascin mouse monoclonal (55k-2) (25), chick antifascin antibody (generated by Aves Labs (Tigard, OR) using recombinant human fascin1 as an antigen). GM-CSF and CCL19 (MIP3β) were purchased from Invitrogen (Camarillo, CA)

Fascin1 KO heterozygous mice were generated by Lexicon Pharmaceuticals, Inc (Woodlands, TX) from an ES cell line (OST124903) (Lexicon's OmniBank® library of gene KO ES cell clones), and backcrossed with C57/BL6 female mice for more than 16 generations (31). Both RT-PCR and Western blot analyses revealed that the loss of fascin1 is not compensated with the expression of the fascin1 paralogues (such as retina fascin2 and testis fascin3) (31). For each experiment with homozygous mice, their wild-type littermates were used as a control. All experimental procedures and protocols for mice are approved by the Animal Care and Facilities Committee at Rutgers. Mice were housed in an AAALAC-accredited animal facility at Rutgers.

Preparation of bone marrow DCs

Preparation of mouse bone marrow DCs was according to the method described in Inaba et al (32) with slight modification. Briefly, single cell suspension was prepared from bone marrow of femurs and tibias, and plated on 65mm dishes in DMEM containing 10% fetal calf serum and 10ng/ml of GM-CSF for 7–10 days. Non-adherent cells were collected and DCs were purified by centrifugation over a 13.7% (w/v) metrizamide discontinuous gradient. More than 85% of cells collected at the interface of the gradient were positive for CD11c. Cells were matured by overnight culture in the presence of 100ng/ml of lipopolysaccharide (LPS, Sigma).

FACS analyses

Mature DCs were fixed with methanol or formalin, and stained with FITC-labeled anti-DC markers including CD86, CD11c, and MHC-II. For double labeling, methanol-fixed cells were blocked with a rat anti-mouse CD16/CD32 antibody (mouse Fc Block, BD Pharmingen), incubated with the mouse anti-fascin1 antibody (clone 55k-2) together with the FITC-labeled CD86 antibody, and then the fascin antibody was labeled with a R-PE-labeled goat anti-mouse IgG. Flow cytometry was performed with a Coulter Cytomics FC500 flow cytometer.

Immunofluorescent microscopy and measurements of thickness, area and circularity

For staining with antibodies against CD11c, CD86, MHC-II, and vinculin, as well as for staining with rhodamine phalloidin (Molecular Probes, Eugene, OR), DCs were fixed with 3.7% formaldehyde, and permeabilized with 0.2% Triton X-100 or 100% acetone. Absolute methanol fixation at −20°C was used for double labeling with the anti-fascin1 mouse monoclonal (clone 55k-2) and the anti-CD86 antibody, and for double staining with anti-fascin1 and anti-α-actinin antibodies. Images were taken as Z-stacks (0.2µm spacing) with a DeltaVision Image Restoration Microscope system (Applied Precision Instrument, LLC Issaquah, WA), deconvolved either with the softWoRx software (Applied Precision Instruments) or the Huygens software (Scientific Volume Imaging, Hilversum, Netherlands). Projected images were generated with SoftWoRx or ImageJ (http://rsb.info.nih.gov/ij/). In some experiments, images were taken on a Nikon TE300 microscope with a 60× objective lens (NA 1.4). Exposure times for imaging and settings for deconvolution were constant for all samples to be compared within any given experiment. For presentation, image contrast and brightness were adjusted with Photoshop (Adobe, San Jose, CA).

For measurements of thickness, area and circularity, wild type and fascin1 KO DCs were labeled with the FITC-labeled CD86 antibody, rhodamine phalloidin and DAPI. Because the expression of CD86 is well correlated with that of fascin1 (see FACS analyses shown in Fig. 1A), CD86^high DCs were chosen to compare differences in thickness, area and circularity between fascin1-expressing wild type and fascin1 null DCs. Orthogonal images created by SoftWoRx were used for measurement of thickness. Areas were measured with xy images of DCs at the ventral focal plane and circularities were measured with Z-projected images. Both areas and circularities were measured using ImageJ software.

Figure 1.

Characterization of wild type and fascin1 null DCs. A, FACS analyses of wild-type (red line) and fascin1-deficient (blue line), mature DCs. Black lines, controls without antibody labeling. a, CD11c; b, MHC-II; c, CD86; d, fascin1. e & f, FACS analyses of wild type (e) and KO (f) DCs double-labeled with CD86 and fascin1 antibodies. B, Western blot analyses. Immature (lM, lanes 1 and 3) and mature (M, lanes 2 & 4) DCs prepared from wild type (lanes 1 & 2) and fascin1 KO (lanes 3 & 4) mice were analyzed with the indicated antibodies. C, Immunofluorescent localization of fascin1 and CD86 in mature wild-type (a–c) and fascin1 KO (d–f) DCs. Mature DCs were double labeled with anti-CD86 (a and d, green) and mouse anti-fascin1 (b and e, red) antibodies. c & f, merged images. Bar, 10µm. D, Localization of fascin1 and F-actin in wild type (a–f) and fascin1 KO (g–l) DCs. Cells were labeled with chick anti-fascin1 (a, d, g & j, green) and rhodamine phalloidin (b, e, h & k, red). Images were taken at the ventral surface (a–c & g–i), as well as at the middle (d–f & j–l) of the same cells. c, f, i, & l, merged images. Arrowheads indicate co-localization of fascin1 and F-actin at the cortex of veil-like protrusions. Bar, 10µm.

Live cell imaging, kymography, microinjection and transfection

For phase-contrast, live cell imaging, DCs were placed at 37 °C in a temperature controlled incubator (MS200D, Narishige) and observed under a Nikon microscope (TE300) with a 40X Plan Fluor phase-contrast (NA 0.60) objective lens. Time-lapse images were taken every 10sec for 20–30min by a CCD camera (CoolSnap-fx, Roper Scientific) with IPLab image analysis software (Scanalytics). Two to three kymographs were generated for each cell with randomly chosen, one-pixel lines using ImageJ (NIH) with the Kymograph plug-in (written by J. Rietdorf and A. Seitz, EMBL). Kymographs were then analyzed using ImageJ to determine rates of membrane protrusions and retractions.

Microinjection of GFP-fascin1 into differentiated THP-1 (human acute monocytic leukemia cell line) cells was performed as follows: Cells were first differentiated into macrophages by the treatment of 200nM of 2-O-Tetradecanoylphorbol-13-acetate (TPA) for overnight as described (33). Microinjection was performed as described previously (26) using GFP-fascin1 at a needle concentration of 9mg/ml. As a control, FITC-labeled BSA was injected. After 1hr incubation, cells were fixed with formaldehyde, permeabilized with acetone, and counterstained with rhodamine-labeled phalloidin or the anti-vinculin antibody to determine effects on podosome assembly. To estimate levels of fascin1 in injected cells, injected cells were stained with the fascin1 antibody, and fluorescent intensities were compared with those of un-injected cells and fascin1-expressing DCs. It was found that injection increased the fascin1 level five to ten times over the level of endogenous fascin1 in THP-1 cells, the level of which is comparable to that found in wild type DCs.

Mature or immature DCs were transfected with a human GFP-fascin1 fusion construct (26) using an Amaxa Nucleofector II according to the manufacture’s instructions.

In vitro assay for chemotaxis

Chemotaxis of mature DCs was assayed in triplicate using a modified Boyden chamber assay. CCL19 (MIP3β), a chemokine for mature DCs, was added in bottom wells at the concentration of 0.6µg/ml. Mature wild type and fascin1 KO DCs (2×10⁵ cells) were placed on top wells of a collagen-coated Boyden chamber with 3µm hole (Corning, Lowell, MA). After 24hr incubation, cells transmigrated into the bottom chamber were counted.

Preparation of epidermal sheets and assay for DC migration into lymph nodes

Epidermal sheets before and after stimulation by FITC painting were prepared essentially as described (34). Briefly, the dorsal surfaces of both ears of wild type and fascin1 KO mice were painted with 25µl of 1% FITC in acetone:dibutylphthalate (1:1). After 24hr, epidermal sheets were isolated from dorsal halves of the ears and stained with anti-MHC-II antibody followed by Cy3-labeled donkey anti-rat antibody. Fifteen to twenty randomly selected fields were photographed with a Nikon TE300 with a 20× objective and the number of Langerhans cells per field (area, 149,000µm²) was counted.

Langerhans cells migrated into draining lymph nodes were prepared essentially as described (35, 36). Briefly, mice were painted with the FITC solution as described above except that both dorsal and ventral sides of ears were painted with 15µl of solution (total 30µl per one ear). After 24hr, draining lymph nodes (auricular) were excised and cell suspension was prepared through cell strainers (Falcon, 70µm). DCs were then enriched on metrizamide discontinuous gradients as described above and cytospun onto coverslips. FITC-bearing DCs were identified and counted by fluorescence and phase-contrast microscopy, and Langerhans cell migration was expressed as numbers of FITC-bearing DCs per lymph node.

Scanning electron microscopy

DCs grown on coverslips were fixed with 2% glutaraldehyde in 0.1 M sodium cacodylate, pH 7.3 for 20min at room temperature, and dehydrated by submerging graded ethanol solutions. After critical point drying and platinum coating, images were taken with an Amray 1930I scanning microscope.

Statistical analysis

Statistical analyses were performed using Student’s t-test (http://www.physics.csbsju.edu/stats/t-test.html).

RESULTS

Fascin1-deficiency does not alter expression of DC maturation markers or other actin-binding proteins

As a first step toward characterization of fascin1 null DCs, FACS analyses were performed to determine whether fascin1 deficiency affects expression of DC markers including CD11c, CD86 and MHC-II when bone marrow-derived DCs were activated by 100ng/ml of lipopolysaccharide (LPS). As shown in Fig. 1A, surface expression of the DC marker CD11c (a), and the maturation markers MHC-II and CD86 (b and c), were similar between wild type and fascin null, mature DCs, indicating that fascin1 null DCs are able to “mature” in terms of the expression of maturation markers. As expected, fascin1 was expressed only in wild type but not fascin1 KO DCs (d, f). As shown in panel e, double labeling with anti-fascin1 and anti-CD86 antibodies revealed a strong correlation between fascin1 and CD86 expression in wild type DCs. Before LPS activation, both wild type and fascin1 null DCs expressed similar levels of CD11c, while neither cell type expressed appreciable levels of CD86 (data not shown). These results are consistent with previous reports that fascin1 depletion by anti-sense treatments did not alter expression of DC maturation markers (19, 20, 30).

Western blot analyses confirmed that fascin1 is induced upon maturation of wild type DCs (lane 2 of Fig. 1B), while it is absent in immature, wild type DCs (lane 1), as well as immature and mature fascin1 KO DCs (lanes 3 & 4, respectively). We examined whether fascin1 deficiency alters expression of other actin-bundling proteins including fimbrin and α-actinin. Fimbrin levels are largely unaltered between wild type and fascin1 KO DCs, regardless of maturation. α-actinin was barely detectable in DCs, and its levels appear unchanged. The levels of vinculin, however, are slightly increased in fascin1 KO DCs.

Fig. 1C shows immunofluorescence localization of fascin1 and CD86 of wild type, as well as fascin1 KO DCs. As expected, fascin1 KO DCs showed no fascin1 staining while CD86 staining intensities were similar between wild type and fascin1 KO DCs, confirming the FACS analyses. As previously reported (17, 25, 37, 38), fascin1 was localized at filopodia-like structures.

Fascin1 is present in the cortex of membrane protrusions

We found that fascin1 was also localized at the cortex of veil-like membranes (a more prominent structure than filopodia in mature DCs) of mature wild-type DCs. Fig. 1D shows wild type and fascin1 KO DCs double-labeled with phalloidin and anti-fascin1 antibody and imaged at two different focal planes (the ventral surface and a middle). In wild type DCs expressing high levels of fascin1, both fascin1 and F-actin were co-localized at the cortex of veil-like protrusions (arrowheads in a–f). In contrast, fascin1 KO DCs were more spread and showed many fewer veil-like protrusions either at the ventral surface or at the middle focal plane. In addition, fascin1 KO DCs frequently showed a cluster of many prominent actin dots at the ventral surface, which were reminiscent of podosomes (arrow in h). In contrast, most wild type DCs did not exhibit such large and clustered dots. As described later, vinculin labeling revealed that these dots observed in fascin1 KO DCs indeed represent podosomes (see Fig. 5).

Figure 5.

Fascin1 expression and actin-bundling activity is critical for podosome disassembly of mature DCs. A, Immunofluorescence of immature (a & b) and mature (c & d) DCs from wild-type (a & c) and fascin1 null (b & d) mice labeled with anti-vinculin (red) and anti-CD11c (green) antibodies. Arrows, podosomes with the characteristic ring structure; arrowheads, focal adhesions. Bar, 15µm. B, Mature fascin1^high DCs (arrow) show no podosome assembly. Bar, 10µm. C, Statistical analyses of podosome assembly. CD11c-positive immature (blue bar) and mature (red) DCs with at least five vinculin-positive podosomes, were judged as DCs with podosomes. Wild-type, mature DCs with very high fascin1 expression (more than 10 times higher than background, pink bar) were also examined for podosome assembly. D, Effects of forced expression of GFP control (a–c), GFP-Wild-type-fascin1 (d–f), GFP-A-fascin1 (g–i) and GFP-D-fascin1 (j–l) on podosome assembly. Fascin1 null DCs were transfected with GFP control (a–c), as well as with wild type and fascin1 mutants, and stained with anti-vinculin antibody (red, b, e, h, k). GFP signal (green, a, d, g, j); merged images (c, f, i, l). Arrowheads in j–l show podosomes. Bar, 10µm. E, Statistical analyses of podosome loss. Podosomes were counted in DCs exogenously expressing control GFP, GFP-W-fascin1, GFP-A-fascin or GFP-D-fascin, and cells were categorized as having less than 4 podosomes or more than 5 podosomes.

Fascin1 is essential for dynamics of membrane protrusions

The presence of fascin1 in the cortex of membrane protrusions (Fig. 1D) prompted us to examine whether the dynamics of membrane protrusions is altered by fascin1-deficiency. To this end, we imaged live DCs plated on glass coverslips using phase-contrast microscopy (supplemental movies 1 and 2 for wild-type DCs, movies 3–5 for fascin1-deficient DCs). Fig. 2 (A–C) illustrates representative still images of wild type (A, corresponds to movie1) and fascin1 null (B & C, corresponds to movies 3 and 4, respectively) DCs, respectively. Fig. 2B represents the majority of fascin1 null DCs whereas Fig. 2C represents a minor fraction (less than 20%) of fascin1 null DCs. These images clearly demonstrate that membrane activity of wild-type DCs is much more vigorous than that in fascin null DCs. Kymograph analyses (Figs 2D–F) confirm that fascin1-null DCs (E & F) displayed greatly diminished dynamics compared to wild-type DCs (D). Box plot analyses of protrusion (G) and retraction (H) rates generated from kymographs (representing 7 different live cell imaging experiments) revealed that the median value of the protrusion rate for wild-type DCs (0.11 µm/sec, n=83) was 2.1 times higher than that of fascin1-deficient DCs (0.053 µm/sec, n=85), and the retraction rate in wild-type DCs (0.10 µm/sec, n=84) was 4 times higher than that of fascin1-deficient DCs (0.025 µm/sec, n=87).

Figure 2.

Fascin1 is critical for membrane dynamics. A–C, Time-lapse, phase-contrast microscopy was performed with wild type (A) and fascin1-null (B,C) DCs that had been plated on glass coverslips. B shows a fascin1-null DC with a spread shape, the dominant phenotype; C shows a cell with a rounded shape, a less common phenotype. Numbers are in sec. Bar, 5 µm. D–F, kymographs generated from the white one-pixel lines at time 0 in A–C, respectively. Dashed lines in D–F indicate the boundary of the membrane protrusions. G & H, box plot analyses of protrusion (G) and retraction (H) rates calculated from kymographs. The bottom and top of each box indicate the first and third quartiles, respectively, and dots indicate outliers. Four asterisks, p<0.0001. I & J, Phase-contrast, time-lapse images of fascin1-null mature DCs exogenously expressing GFP (I) or GFP-fascin1 (J). The last images show fluorescent images at the end (20min) of imaging to confirm GFP-expression. Numbers are in sec. Bar, 5 µm. K & L, kymographs of GFP (K)-, and GFP-fascin1 (L)-expressing DCs generated from the time-lapse images in I and J, respectively. M & N, box plot analyses of protrusion (M) and retraction (N) rates of fascin1-null DCs exogenously expressing GFP or GFP-fascin1.

To determine whether fascin1 is responsible for the dynamic membrane movements, we re-expressed GFP-fascin1 in fascin1 KO DCs, and tested whether fascin1 was able to rescue the poor dynamics of membrane protrusions. Phasecontrast, time-lapse microscopy was performed with fascin1 KO DCs expressing GFP alone (control, see supplemental movie 6) or GFP-fascin1 (movie 7). As shown in both still images (Figs 2I and J) and kymographs (K and L), DCs expressing GFP-fascin1 (J and L) exhibited much greater membrane protrusion dynamics than did DCs expressing control GFP (I and K). Box plot analyses (M and N) revealed a significant difference in both protrusion and retraction rates (p<0.0001): The median protrusion and retraction rates for DCs expressing GFP-fascin1 were 0.18 µm/sec (n=103) and 0.13 µm/sec (n=98), respectively, whereas those of DCs expressing GFP were 0.058 µm/sec (n=74) and 0.043 µm/sec (n=50), respectively. Thus, the protrusion/retraction rates of GFPfascin1-expressing DCs were comparable to those of wild-type DCs, while those for GFP-expressing DCs were similar to those of fascin1-deficient DCs. These results indicate that fascin1 induction upon maturation is responsible for the vigorous dynamic membrane movements of mature DCs.

Fascin1 is important for in vitro chemotaxis of mature DCs toward CCL19

The lower membrane protrusive activity of fascin1-deficient, mature DCs would be predicted to impair their migratory efficiency. To test this idea, we examined, using a modified Boyden chamber (3µm holes with collagen coating), whether fascin1 deficiency affects chemotaxis of mature DCs toward CCL19 (MIP3β). As Fig. 3A shows, fascin1 deficiency reduced chemotaxis by 42% (p=0.0095).

Figure 3.

Fascin1 is critical for chemotaxis and Langerhans cell migration into draining lymph nodes. A, In vitro chemotaxis toward the chemokine CCL19, measured with a modified Boyden chamber. (p=0.0095). B, Immunofluorescence imaging of Langerhans cells in ear epidermal sheets from wild type (a & c) and fascin1 null (b & d) mice without (control, a & b) and with FITC painting (+allergen, c & d). Langerhans cells (indicated by arrowheads) were labeled with the MHC-II antibody. Representative images from four independent experiments. C, Box plot analyses of Langerhans cell distribution without or with allergen treatment. NS, no statistical significance; one asterisk, p<0.05; two asterisks, p<0.01, three asterisks, p<0.001; four asterisks, p<0.0001. D, Box plot analysis of FITC-bearing DCs migrated into draining lymph nodes. P=0.0059.

Fascin1-deficient mice show reduced migration of Langerhans cells

Consistent with the impaired chemotaxis of fascin1 KO DCs toward CCL19 in vitro, we found that Langerhans cells of fascin1 KO mice showed reduced migration into lymph nodes. Twenty-four hours after painting of dorsal sides of both ears with an allergen of FITC, epidermal cell sheets were prepared and stained with a MHC-II antibody. Fig. 3B shows representative immunofluorescence images of wild and KO epidermal sheets. Without allergen, both wild type and KO sheets showed a similar Langerhans cell distribution. After stimulation with the allergen, Langerhans cells from wild type mice clearly exhibited decreased cell density when compared to KO. Quantitative data (Fig. 3C) indicate that the mean density of wild-type Langerhans cells after stimulation was about half of that of fascin1 KO (p<0.001) while the density of Langerhans cells before stimulation was statistically similar between wild type and fascin1 KO mice.

To confirm that the above difference in Langerhans cell density is indeed due to migration of Langerhans cells into lymph nodes, we measured the number of FITC-bearing DCs in draining lymph nodes following stimulation by FITC painting for 24hr. As Fig. 3D shows, there were over twice as many FITC-bearing DCs per lymph node in wild type mice as in fascin1 null mice (p=0.0059). Taken together, these results show that Langerhans cell migration is impaired in fascin1 KO mice, and support our hypothesis that fascin1 plays a critical role in DC migration into lymph nodes by promoting podosome disassembly and increasing dynamics of membrane protrusions.

Fascin1-deficient DCs are more spread and thinner with fewer membrane protrusions

Fascin1-deficient, mature DCs are morphologically very different from their wild-type counterparts. Fig. 4A shows scanning electron microscopy of wild type and fascin1 null mature DCs. Fascin1-deficient DCs were much thinner and more spread with fewer and smaller dorsal ruffles than wild-type DCs.To quantitatively assess these shape changes, cells were labeled with rhodamine phalloidin and anti-CD86 antibody, and analyzed by immunofluorescence microscopy, using serial Z-section imaging (0.2 µm spacing) and 3-D rendering. Because the expression of CD86 correlates with that of fascin1 (Fig. 1A–e), CD86^high DCs were chosen for morphological analysis. Fig. 4B shows representative images of wild type and KO CD86^high DCs stained with phalloidin. Orthogonal images in both xz and yz planes clearly showed that fascin1-deficient DCs were thinner than wild type. Statistical analyses using box plots (Fig. 4C) revealed a significant difference in thickness (p<0.0001). The median thickness of fascin1-deficient DCs (n=126) was 7.4 µm while that of wild type (n=196) was 10.7 µm. The difference in the thickness became even more prominent when DCs were centrifuged at 110×g for 4min; fascin1-deficient DCs were greatly flattened, with a median thickness of 3.6 µm (n=55), while wild-type DCs were more resistant with a median thickness of 6.9 µm (n=45). These results suggest that DC stiffness may be impaired in fascin1-deficient DCs.

Figure 4.

Morphological characterization of wild type and fascin1 null DCs. A, Scanning electron micrographs of wild-type (a) and fascin1-deficient (b) mature DCs. Bar, 10 µm. B, Orthogonal views of wild-type (a) and fascin1-deficient (b), CD86^high DCs. Mature DCs were double stained with phalloidin and anti-CD86 antibody. Only phalloidin staining is shown here. The xy images are on the ventral surface. Both xz and yz images are shown with the top and bottom of cells indicated by dashed lines. Bar, 5 µm. C, D and E, Box blot analyses of thickness (C), area (D) and circularity (E) of wild-type and fascin1-deficient DCs. Thickness was determined without (w/o cfg) or with (w. cfg) cytospin.

To determine how fascin1-deficiency affects cell spreading, we made area measurements of xy images on the ventral surface. As the box plot of Fig. 4D shows, fascin1-deficient DCs are 40% more spread: The median area covered by fascin1-deficient DCs was 290 µm² (n=95) whereas that of wild-type DCs was 205 µm ² (n=121) with a statistical significance (p=0.0032). As wild-type DCs were 30% thicker than fascin1-deficient DCs, these measurements suggest that both types of DCs have roughly equal cell volumes.

The size and shape of protrusions varied widely, making simple measurements of the number and length of protrusions inappropriate for quantitative analyses. We thus measured circularity {(4pi)×(area)/(perimeter)²} of projected images generated from Z-section images, because more protrusions results in higher deviation from circularity (the value for a complete circle is 1). Box plot analyses (Fig. 4E) showed that fascin1-deficiency increased the median values of circularity from 0.39 (n=112) to 0.51 (n=120) with the statistical significance of p<0.0001, confirming that fascin1 KO DCs have fewer protrusions. The finding that fascin1 null DCs show reduced numbers of protrusions is consistent with previous studies demonstrating the role of fascin1 in generating membrane protrusions in other cell types (16, 26, 29, 39, 40).

Fascin1-deficient DCs fail to disassemble podosomes upon maturation

Podosomes are disassembled in mature DCs (4, 10, 11). The images at the ventral surface of fascin1 KO DCs (Figs 1D-h & 4B-b) showed many more podosome-like F-actin dots than in wild-type DCs, suggesting a difference in podosome dynamics. We found that this is the case. Fig. 5A shows immunofluorescence images of vinculin-labeled (red) immature and mature DCs from wild type and fascin KO mice. In each case, DCs were identified by counterstaining with CD11c (green). Immature DCs from both wild type and fascin1 KO DCs assembled podosomes to a similar extent. This result is consistent with the observation that fascin1 expression is minimal in immature DCs. As reported (4, 11), podosomes disappeared in most mature wild-type DCs. In sharp contrast, mature fascin1 KO DCs retained podosomes. We performed quantitative analyses of podosome number in CD11c-positive DCs by setting a criterion that DCs with a cluster of more than 5 podosomes (defined as vinculin-positive ring-like structures) were judged as podosome-positive. Such measurements (C) revealed that, while the percentage of wild-type DCs with podosomes decreased from 65% (n=117) to 22% (n=166) upon maturation, the percentage of podosome-positive, fascin1-deficient DCs was unchanged by maturation (61% for immature, n=158 and 59% for mature DCs, n=111).

We found that the loss of podosomes is highly correlated with the extent of fascin1 expression. As shown in our FACS analyses (Fig. 1A, d & e), mature wild type DCs can be grouped into two populations, one (about 40–50% of mature wild-type DCs) shows two orders of magnitude higher expression of fascin1 than the other. We examined the presence of podosomes in such fascin1^high DCs by double staining with the anti-fascin1 and anti-vinculin antibodies (Fig. 5B). We found that virtually all DCs expressing the higher level of fascin 1 (n=100) had no podosomes (pink bar, Fig. 5C), suggesting an important role for fascin1 in podosome loss. This notion is consistent with the observation that the timing of fascin1 induction (about 7hr after LPS treatment) roughly corresponds to the time when mature DCs lose podosomes. The high correlation between high fascin1 expression and podosome loss may point to a specialized DC subset or maturation state with high fascin1 expression.

Forced expression of fascin1 results in podosome loss

The above correlation has prompted us to test whether very high expression of fascin1 in mature DCs is responsible for podosome loss. To this end, we forced expression of GFP-fascin1 in fascin1-deficient DCs and counterstained them with the anti-vinculin antibody. As a control, GFP alone was transfected in a similar way. As Fig. 5D (d–f) shows, the introduction of GFP-fascin1 resulted in podosome loss in most DCs. In contrast, podosomes remained assembled in fascin1 KO DCs expressing control GFP (a–c). Measurements of podosomes in transfected cells (Fig. 5E) revealed that most (76%) of DCs expressing GFP-fascin1 (n=113) exhibited 4 or fewer podosomes while only 28% of DCs expressing control GFP (n=45) displayed fewer than 4 podosomes. These results suggest that high levels of fascin1 in mature DCs are responsible for podosome loss in mature DCs.

Actin bundling by fascin1 is critical for podosome loss

We next examined whether podosome loss depends on the actin-bundling activity of fascin1. Actin-bundling activity of fascin1 is largely down-regulated by phosphorylation at Ser39 (41, 42), because phosphorylation at Ser39 disrupts one of the two actin binding sites of fascin1. We thus expressed unphosphorylatable (A-fascin1, replacing Ser39 with Ala) and phosphomimetic (D-fascin1, replacing Ser39 with Asp) mutants in fascin1 null DCs. As shown in Fig. 5D, g–l, A-fascin1 was much more effective than D-fascin1 at inducing podosome loss. Indeed, quantitative data (Fig. 5E) revealed that A-fascin1 (n=102) was slightly more effective in inducing podosome loss than wild-type fascin1 (W-fascin1), increasing the fraction of DCs without podosomes from 76±8% to 85±6% (p=0.006). In contrast, D-fascin (n=138) was much less effective than wild-type fascin, resulting in only 42±8% of the DCs displaying fewer than 4 podosomes (p<0.0001). These results indicate that the actin-bundling activity of fascin1 is important for podosome disassembly, and suggest that phosphorylation of fascin1 could contribute to the regulation of podosome assembly.

Fascin1 is associated with the actin structure of podosomes

To explore how high fascin1 expression leads to podosome disassembly, we examined whether fascin1 binds to F-actin within podosomes. Because high expression of fascin1 in wild type DCs makes it difficult to determine possible localization of fascin1 at podosomes in these cells, we searched for a hematopoietic cell line that expresses a low level of fascin1 and, at the same time, has podosome structures. We found that in contrast to primary macrophages, THP-1 cells (human monocytic leukemia cells) express a low level of fascin1, yet assemble podosomes when differentiated into macrophages by addition of phorbol ester (33). Double labeling of THP-1 cells with anti-fascin1 and anti-α-actinin antibodies (Fig. 6A) clearly revealed co-localization of fascin1 (green) and α-actinin (red) at podosomes (arrowheads) with fascin1 being slightly inside the α-actinin ring structure.

Figure 6.

Fascin1 is localized to podosomes in THP-1 cells and microinjection of fascin1 induces podosome disassembly. A, Co-localization of fascin1 and α-actinin at podosomes. THP-1 cells differentiated with TPA were labeled with anti-fascin1 (b & e, green) and anti-α-actinin (red, a & d) antibodies. c & f, merged images. Images in d–f show enlargements of the boxed areas in a–c. Arrowheads, podosomes. B, Disassembly of podosomes by microinjection of fascin1. THP-1 cells were microinjected with FITC-labeled BSA (a) or GFP-fascin1 (c) and counter-stained with phalloidin (b & d). Arrows indicate injected cells while arrowheads indicate podosomes. C, Percentage of cells with podosome arrays of un-injected cells (green); or after microinjection of fascin1 (red) or FITC-BSA (blue). Representative data from three independent experiments is shown. Approximately 20 injected cells for each condition were counted for each set of experiments.

The level of fascin1 in THP-1 cells is approximately 10 times lower than that of fascin1^high mature DCs. We thus asked whether an increase in fascin1 concentration to the level observed with DCs could result in podosome disassembly in THP-1 cells. As Fig. 6B shows, microinjection of GFP-fascin1 induced podosome disassembly (c & d) in most (89%, 50 out of 56 injected cells) THP-1 cells within 1hr (see Fig. 6C for quantitative data). Concomitantly, fascin1-injected cells frequently became rounded and detached from the substrate if incubated for a longer time. In contrast, most (66%, 38 out of 58 injected cells) of control cells injected with FITC-labeled BSA retained podosomes (a & b of Fig. 6B), the percentage of which is statistically similar to that of un-injected cells (72%, see Fig. 6C). These results indicate that whereas fascin1 at a low concentration can bind to actin structure of the podosomes, high fascin1 expression as observed in DCs can cause podosome disassembly.

DISCUSSION

We have demonstrated that fascin1 plays a critical role in the alterations in motility, morphology and adhesion associated with DC maturation. Fascin1 null DCs, when fully matured, are more spread, show fewer and less dynamic membrane protrusions, and retain podosomes. Importantly, fascin1 null DCs show reduced directed migration both in vitro and in vivo.

Fascin1 is critical for dynamic dorsal ruffling

How does fascin1 enhance the dynamics of membrane protrusions? We found that fascin1 is co-localized with F-actin at the cortex of veil-like protrusions (Fig. 1D). Judging from its actin bundling and cross-linking activity, fascin1 is likely to form a meshwork of actin filaments at the cortex, which would give the cell cortical rigidity. Recent studies have shown that the actin cross-linking activity of fascin1 is extremely dynamic (29, 43, 44), and suggested that this dynamic cross-linking is required for vigorous movements of filopodia while at the same time maintaining sufficient rigidity for filopodial protrusions (43). A similar mechanism is likely to work for veil-like protrusions of mature DCs. As the DC cortex protrudes and retracts, fascin1 would be able to quickly reorganize the actin meshwork, providing both the rigidity and the flexibility needed to support dynamic membrane protrusions. Such dynamics are likely to be critical for DC migration through tissues and extracellular matrix to reach the lymph nodes. In keeping with this idea, we found that Langerhans cells from fascin1 KO mice show reduced emigration into draining lymph nodes (Figs 3B–D).

High fascin1 expression is likely to cause the disassembly of podosomes in mature DCs

We found that fascin1 expression is closely correlated with the loss of podosomes in mature DCs. Importantly, forced expression of fascin1 in fascin1 null DCs resulted in podosome disassembly (Fig. 5). Two recent studies, however, have shown that fascin1 appears to favor assembly of podosomes, as well as invadopodia, structures closely related to podosomes (45, 46). In PDGF-treated smooth muscle cells, fascin1 depletion has been reported to suppress podosome assembly (45). Likewise, fascin1 has been shown to stabilize F-actin in invadopodia in melanoma cells (46). We speculate that these apparently contradictory functions of fascin1 may be explained by the difference in fascin1 expression levels between mature DCs and other cell types. Mature DCs express fascin1 one order in magnitude higher than do other cells. In support of this notion, we demonstrated that podosome assembly could be controlled by altering fascin1 levels in THP-1 cells: While endogenous fascin1 is present at a low level in THP-1 cells and localized to podosomes, microinjection of a large amount of fascin1 caused disassembly of podosomes (Figs 6B & C).

An important question is how high levels of fascin1 could cause podosome disassembly in mature DCs. Fascin1 at a low level binds to actin structure of podosomes (Fig. 6A), indicating that fascin1 and other actin binding proteins can simultaneously bind to actin filaments. However, very high levels of fascin1 would saturate actin filaments (fascin1 can bind to actin at a molar ratio of one fascin1 to four actin molecules (47)), which would compete with other proteins for actin binding (48). Such competition would result in dissociation of an actin binding protein(s) that is critical for the organization of podosomes, leading to disassembly of podosomes. Indeed, Park et al., have shown that fascin1 de-branches Arp2/3 complex-mediated branched filaments, transforming the dendritic filament assembly into actin bundles in vitro (49). Because Arp2/3 complex is an essential component of podosomes (50–53), debranching of Arp2/3-mediated dendritic filaments could block de novo synthesis and/or maintenance of podosomes. This idea is consistent with the result that A-fascin1 is much more effective in podosome disassembly than is D-fascin1 because D-fascin1 shows much weaker actin bundling activity (Fig. 5).

Loss of podosomes may be critical for migration of mature DCs

Podosomes appear to profoundly alter migration patterns of DCs, at least in vitro. It has been reported that mature DCs without podosomes display “high-speed migration” with low adhesion to the substrate when compared with immature DCs with podosomes (10). Such “high speed migration” with reduced adhesion would be advantageous for mature DCs to travel to a lymph node as quickly as possible for efficient presentation of antigens to naïve T-cells. On the other hand, immature DCs need to move around the peripheral tissues in order to constantly sample foreign and host antigens. Such movement may require an adhesion structure like podosomes for attachment to the extracellular matrix (10, 13, 14, 54, 55). It is worth noting that other primary hematopoietic cells like macrophages have prominent podosomes, while no fascin1 expression was detected. These cells may need podosomes as adhesion structures so that they can move around the peripheral tissues as a sentinel against external pathogens.

The loss of podosomes might also be critical for the assembly of an immunological synapse. Geyeregger et al. have shown that agonists of Liver X receptors (LXRs) blocked fascin1 expression in human DCs and, at the same time, inhibited the assembly of the immunological synapse (56). Importantly, overexpression of fascin1 in LXR agonist-treated DCs restored immunological synapse assembly (56). We have found that LXR agonists block podosome disassembly in mature, wild type DCs (SY, unpublished result), again supporting our notion that the loss of fascin is highly correlated with sustained podosome assembly. Perhaps, the disassembly of podosomes may facilitate the assembly of an immunological synapse because these two structures share molecular constituents (57).

Conclusion

We have demonstrated that fascin1 plays a critical role in chemotactic migration of DCs. Manipulation of fascin1 expression may thus be effective in enhancing DC-based immune therapy. For example, while tumor antigen-loaded DCs have been used as cancer vaccines, only a tiny fraction (1%) of DCs subcutaneously injected are able to migrate into lymph nodes of cancer patients (58). It might be possible to increase the efficiency of DC migration by selecting DCs with high fascin1 expression and/or by exogenously increasing fascin1 expression.

Abbreviations used in this paper

KO

knockout

DCs

Dendritic cells