Requirement of CCL17 for CCR7- and CXCR4-dependent migration of cutaneous dendritic cells

To determine the influence of CCL17 on the development of AD, we treated heterozygous CCL17/EGFP (CCL17E/+) reporter mice and homozygous CCL17-deficient (CCL17E/E) mice with three cycles of tape-stripping and application of an ovalbumine (OVA)-containing patch on the skin of the back (21, 22). Mice treated with tape-stripping but receiving a patch soaked with NaCl alone were used as negative controls. In general, CCL17E/+ mice responded like normal BALB/c mice and were used as controls (Fig. S1). Mechanical stress induced by tape-stripping and application of a NaCl-soaked patch led to only minor thickening of the epidermis. When the model allergen OVA was applied to the patch on the mechanically stressed skin, acanthosis was induced in both CCL17E/+ and CCL17E/E mice to a similar extent (Fig. S2 A and B), indicating that CCL17 has no influence on keratinocyte proliferation. In contrast, thickening of the dermis as a measure of leukocytic inflammation was strikingly diminished in the absence of CCL17 (Fig. S2 A and C). Mast cell and eosinophil infiltration also were significantly lower in the lesional skin of CCL17E/E mice compared with OVA-treated CCL17E/+ controls (Figs. S2 F and G and S3A). To test whether an absence of CCR4 also leads to a similar amelioration of AD symptoms, CCR4-KO mice and C57BL/6 WT mice also were analyzed in the AD model. In contrast to CCL17E/E mice, CCR4KO mice developed allergic skin inflammation to the same extent as WT mice, with strong thickening of both the epidermis and the dermis (Fig. S2 D and E).

As reported previously, CCL17 is not expressed in normal healthy skin, whereas both LCs and dDCs up-regulate CCL17 after maturation and entry into the skin-draining LNs (9). Using the AD mouse model, we tested whether mechanical irritation and OVA treatment were sufficient to induce CCL17 expression within the affected skin. While CCL17/EGFP-expressing cells were present at low frequency in the skin of saline-treated mice, a much higher frequency of dispersed EGFP⁺ cells was detected in the dermis of the OVA-treated mice (Fig. S3B). Because the detection of EGFP⁺ LC in histological sections is inefficient for technical reasons, we also prepared epidermal sheets from dinitrofluorobenzene (DNFB)-sensitized ears to demonstrate that LCs also up-regulate CCL17/EGFP expression in the epidermis after sensitization (Fig. S3C).

Infiltrating CD4⁺ cells were significantly reduced, but not entirely absent, in OVA-treated CCL17E/E mice compared with OVA-treated CCL17E/+ mice (Fig. S4 A and B). In contrast, CCR4-deficient mice demonstrated dermal infiltration with CD4⁺ cells similar to that in WT mice after OVA treatment for AD (Fig. S4C). The number of FcεRI-expressing cells also was significantly reduced in OVA-treated CCL17E/E mice compared with OVA-treated CCL17E/+ mice (Fig. S4 D and E). Taken together, these findings indicate that the allergic reaction was strongly ameliorated in the absence of CCL17, whereas an absence of CCR4 had no significant impact on the development of allergic skin inflammation.

In agreement with previous reports using the AD mouse model (22), we observed a 5-fold increase in total serum IgE level in OVA-treated CCL17E/+ mice compared with saline-treated CCL17E/+ mice (Fig. S5A). In contrast, the total serum IgE level was ∼50% lower in OVA-treated CCL17E/E mice compared with OVA-treated CCL17E/+ controls. More specifically, we also assessed OVA-specific IgE levels and observed a significant reduction in the sera of OVA-treated CCL17E/E mice compared with OVA-treated CCL17E/+ mice (Fig. S5A). The OVA-specific IgG1 and IgG2a titers also were significantly lower in OVA-treated CCL17E/E mice compared with OVA-treated CCL17E/+ mice (Fig. S5A). This indicates that CCL17 plays an essential role in the induction of allergen-specific antibody responses in the AD mouse model.

To gain insight into the cytokine response pattern of lesional skin, we analyzed cytokine expression by quantitative RT-PCR. In contrast to cytokine production in OVA-treated CCL17E/+ mice, levels of Th2-related cytokines (e.g., IL-4, IL-13, IL-10) were significantly reduced in OVA-treated CCL17-deficient mice and were barely distinguishable from those in saline-treated mice (Fig. S5C). After chronic OVA-treatment, the Th1-cytokine IFN-γ also was significantly decreased in CCL17E/E mice compared with CCL17E/+ mice. OVA-treated CCL17E/+ mice showed a strong increase in IL-1β expression in the skin compared with saline-treated CCL17E/+ mice, whereas IL-1β levels were not significantly different in OVA-treated and saline-treated CCL17E/E mice. In line with the severe thickening of the dermis in OVA-treated CCR4KO mice, these mice developed OVA-induced humoral immune responses (Fig. S5B) and local cytokine production (Fig. S5D) similar to what was seen in WT controls. Taken together, we can conclude that Th1- and Th2-cytokine expression is strongly diminished in the lesional skin of CCL17-deficient mice, but not CCR4KO mice, compared with WT mice.

The total cellularity of LNs draining the site of skin inflammation was enhanced by more than 3-fold in OVA-treated CCL17E/+ mice and about 2-fold in CCL17E/E mice compared with the respective NaCl-treated control mice (Fig. S6A). The absolute numbers of CD11c⁺ DCs in these LNs increased by 4- to 5-fold in CCL17E/+ mice, but by only 2-fold in CCL17-deficient mice compared with the respective saline-treated mice (Fig. S6C). We found no major changes in the frequency of CD11c⁺ DCs in the draining LNs (Fig. S6B). To directly quantify the number of skin-derived DCs, we also determined the absolute numbers of LC and dDC descendants in the LNs, based on high expression of MHC class II (MHCII) and differential expression levels of CD24a (23). Both dDCs and LCs (Fig. S6 D and E) were significantly reduced in the draining LNs of CCL17E/E mice compared with CCL17E/+ mice after OVA treatment. In terms of LN size and numbers of skin-derived DCs, again no difference was found between OVA-treated WT and CCR4KO mice (Fig. S6 F–J). Considering that the migration of LCs and dDCs from the inflamed skin to the draining LNs initiates further recruitment of lymphocytes and enlargement of the LNs, we asked whether migration of the DCs themselves was impaired in CCL17-deficient mice. Therefore, we stained sections of the affected skin from all four groups of mice for MHCII and CD11c. Application of a NaCl-soaked patch to the mechanically stressed skin did not induce emigration of LCs in either CCL17E/+ or CCL17E/E mice (Fig. 1 A and B; Fig. S7 A and B). In contrast, after treatment with the OVA-containing patch, the frequency of LCs was reduced by 50% in CCL17E/+ mice (Fig. 1 A and B and Fig. S7 A and B), whereas the number of CD11c⁺ cells in the dermis increased by more than 100% (Fig. S7C). Unexpectedly, the emigration of LCs did not occur in OVA-treated CCL17E/E mice (Fig. 1 A and B and Fig. S7 A and B). In contrast, there was no significant difference in the number of skin-resident dDCs in OVA-treated CCL17E/+ and CCL17E/E mice (Fig. S7C) or in the number of LCs in OVA-treated WT and CCR4KO mice (Fig. 1C).

The observation that CCL17 affects DC migration in the AD model prompted us to also examine emigration of skin-resident DCs in a model of acute contact sensitization. For this purpose, we applied the contact sensitizer DNFB on the ears for 4 h, and then cultured them for 48 h to permit emigration of LCs from the epidermis. Remarkably, LC emigration was significantly attenuated in CCL17E/E mice (Fig. 2A). Under steady-state conditions, ≈1,000 LCs/mm² were found in the epidermis of CCL17E/+ and CCL17E/E mice, and treatment with solvent only did not significantly change this number (Fig. 2 A and B). In contrast, DNFB application induced the emigration of more than 50% of all LCs in CCL17E/+ mice (Fig. 2 A and B) and in WT and CCR4KO mice (Fig. 2C), but not in CCL17E/E mice. In the latter, the number of LCs was not significantly reduced compared with the solvent control and only moderately reduced compared with the steady state (Fig. 2 A and B). These findings are in line with the higher frequency of EGFP⁺ LC in epidermal sheets from DNFB-treated CCL17E/E mice compared with CCL17E/+ mice (Fig. S3C). To further analyze the migration kinetics of skin-derived DCs after allergen exposure, we treated mice with DNFB on the dorsal skin and analyzed the numbers of both LCs and dDCs in the draining LNs after 0, 2 and 4 days. In CCL17E/E mice, the number of LCs was significantly reduced by 4 days after treatment; for dDCs, a significant difference could already be detected on day 2 as well as in the steady state (Fig. 2 D and E). These results are in line with our previous finding of impaired CHS responses in CCL17E/E mice (9).

CCR7 and CXCR4 are known to be key regulators of DC migration from the skin to the skin-draining LNs (5, 6, 24–26). Because migration of mature DCs through peripheral tissues involves direct interactions with the extracellular matrix (3, 27), we first analyzed the migration of bone marrow (BM)-derived DCs in 3D collagen gels using live cell imaging microscopy. In this assay, BM-DCs from CCL17E/E mice showed less random motility and migratory speed (velocity) in the absence of a chemokine gradient compared with CCL17E/+ BM-DCs (Fig. 3 A and B and Movie S1 and Movie S2). When a CCL19 gradient was applied, BM-DCs of CCL17E/+ mice responded with vigorous migration and good directionality (0.52 ± 0.15) (Fig. 3A, Upper Right and Movie S3). In contrast, the directionality of BM-DCs from CCL17E/E mice toward the CCR7 ligand CCL19 was strongly reduced to 0.28 ± 0.18 (Fig. 3A, Lower Right and Movie S4). Likewise, the y-forward migration index, which indicates the directed migration toward the chemokine gradient along the y axis, was reduced by almost 50% (Fig. 3C). Furthermore, the average velocity of CCL17E/E cells was significantly lower than that of CCL17E/+ cells in the presence and absence of a CCL19 gradient (Fig. 3B).

We next analyzed the directional migratory capacity of BM-DCs from either CCL17E/+ or CCL17E/E mice to the CCR7 ligands CCL19 and CCL21 and to the CXCR4 ligand CXCL12 in transwell migration assays. Again, we found that BM-DCs generated from CCL17E/E mice had a strongly reduced ability to transmigrate toward all three chemokines compared with DCs from CCL17E/+ mice (Fig. 3D). Nevertheless, the migratory ability of CCL17E/E DCs was not entirely abrogated. Interestingly, CCL17E/E DCs also demonstrated reduced random migration when incubated without any chemokine and moderately reduced chemokinesis (Fig. S8 A and B). We then asked whether migration of CCL17E/E BM-DCs to CCL19 and CXCL12 could be restored by preincubation with recombinant murine (rec) CCL17. As shown in Fig. 3 E–G, recCCL17 had to be present for more than 2 h and at concentrations of >100 ng/mL to achieve full migratory capacity. In addition, emigration of LCs from the epidermis also could be induced in vivo in CCL17E/E mice by injection of recCCL17 into the ear pinna before challenge with DNFB (Fig. S8 E and F). In contrast, treatment with anti-CCL17 antibodies strongly inhibited CCL17E/+ DC migration (Fig. S8D). Thus, we can conclude that CCL17 is essential for the ability of DCs to respond to CCR7 or CXCR4 ligands with efficient directed chemotaxis in an inflammatory micro-milieu. In contrast, the absence of CCR4 had no influence on the migration of BM-DCs toward CCL19 (Fig. S8G). Although CCL22, which is thought to have similar biological activity as CCL17, is produced at normal amounts in CCL17E/E mice (Fig. S8H), it apparently does not compensate for CCL17-deficiency. But the addition of recCCL22 to CCL17-deficient DCs also restores migration to CCL19 (Fig. S8I), and antibody-mediated blockade of CCL22 slightly inhibits DC migration (Fig. S8J). Furthermore, knockdown of CCL22 by siRNA leads to reduced migration of CCL17-expressing, but not CCL17-deficient, BM-DCs to CCL19 and CXCL12 (Fig. S8 K and L). These findings indicate that CCL22 might have a similar function as CCL17, but nonetheless does not compensate for CCL17 in terms of cutaneous DC migration, perhaps because of local cell type–specific differences in production of CCL17 versus CCL22.

To test whether the impaired responsiveness of CCL17E/E DCs to CCR7 and CXCR4 ligands was caused by reduced expression of CCR7 or CXCR4, we stained BM-DCs for these receptors. Under normal conditions, CCR7 and CXCR4 are expressed by stimulated BM-DCs, and despite the impaired responsiveness of CCL17E/E DC to CCL19, CCL21, and CXCL12, we found no reduction in CCR7 or CXCR4 expression (Fig. 4A). We also stained for CCR4 and found that equivalent amounts of CCR4 were expressed on unstimulated DCs of CCL17E/E and CCL17E/+ mice (Fig. 4B, Left). After LPS stimulation, the expression levels of CCR4 on DCs from CCL17E/E mice were slightly higher than those on DCs from CCL17E/+ mice (Fig. 4B, Right). We also analyzed CCR7 expression on ex vivo isolated LCs of WT, CCL17E/+, CCL17E/E, and CCR4KO mice. Whereas CCR7 expression was not detectable directly after isolation of the cells (Fig. 4 C and D, Left), it was strongly up-regulated after LPS stimulation, with no difference in LCs from WT, CCL17E/+, CCL17E/E, and CCR4KO mice (Fig. 4 C and D, Right). This indicates that CCL17 deficiency does not affect surface expression of CCR7 and CXCR4, but rather alters their functionality. As a consequence of signal transduction through chemokine receptors, cells normally react with the release of intracellular calcium stores. Thus, we compared the ability of CCR7 to induce calcium flux in CCL17E/+ versus CCL17E/E DCs after stimulation with CCL19 or CXCL12 (Fig. 4 E–N). Remarkably, both CCL19- and CXCL12-induced Ca²⁺ flux was strongly reduced in CCL17E/E DCs, whereas Ca²⁺ flux could be induced in all genotypes by treatment with ionomycin (Fig. 4 E–N).

Atopic skin inflammation was induced as described previously (21, 22). In brief, mice were tape-stripped four times using Tegaderm 1624W (3M Medica). First, 100 μg of OVA (grade V; Sigma-Aldrich) in NaCl or NaCl alone was placed on a patch of sterile gauze and attached to the skin. The patches were applied on days 1 and 4 and maintained for a total of 1 week. After 2 weeks of rest, mice were tape-stripped again, and an identical patch was reapplied to the same site. Mice received a total of three 1-week patch exposures.

Mice were treated on both sides of the ear with 10 μL of 0.5% DNFB in acetone/olive oil (5:1). After 4 h, the ears were removed and cultured for 2 days at 37 °C. Epidermal sheets were stained with mouse anti-Langerin (clone 929F3.01; Acris Antibodies). To investigate the in vivo migration of DCs, the dorsal skin was shaved, and 80 μL of 0.5% DNFB was applied. After 2 and 4 days, skin-draining LNs were analyzed by flow cytometry.

BM-DCs were generated as described previously (9). On day 6, nonadherent cells were harvested and activated with LPS overnight. DC migration was measured in 24-well Costar transwell chambers (Corning). After a 3-h incubation at 37 °C, the migrated cells were counted and stained for FACS analysis.

The 3D chemotaxis assays were performed as described previously (27) with some modifications. Collagen-BM-DC mixtures were placed in custom-made chemotaxis chambers and incubated at 37 °C in 5% CO₂ for 90 min. After collagen polymerization, 600 ng/mL of CCL19 (R&D Systems) was added on top of the gel. Time-lapse series were recorded using a fully automated inverted TE Eclipse microscope (Nikon). BM-DC were tracked over 3 h (5 min/frame) using the ImageJ manual tracking plug-in.

Detailed descriptions of all methods used are provided in SI Materials and Methods.