Plasmacytoid Dendritic Cell Recruitment by Immobilized CXCR3 Ligands¹

Plasmacytoid dendritic cells (pDCs)³ are professional APCs with the inherent capacity to produce large amounts of natural IFNs (nIFNs). pDCs are activated by viral pathogens which, in contrast, induce “paralysis” or apoptosis of conventional myeloid DCs (1, 2, 3, 4, 5). For example, HSV promotes survival, evokes T cell stimulatory potential, and induces high level nIFN production of pDCs (6, 7, 8). Consequently, pDCs can confer immediate innate protection against viral replication (7). Unlike secondary lymphoid tissues, the noninflamed organs are devoid of pDCs. Thus, pDCs must home to virus-infected tissues to execute the first line innate defense against viral infections and, later, to serve as APCs for the induction of cellular antiviral immunity. However, up to now, the molecular mechanisms of pDC homing into the sites of viral replication remain enigmatic.

Functionally different DC precursors have distinctive patterns of chemokine receptor expression which determine their migratory responses. Epitheliotropic DC precursors respond to the CCR6 ligand CCL20 (9, 10), precursors of the dermal/interstitial-type DC lineage respond to CCR1/CCR5 and CCR2 ligands (9, 10, 11), and pDC precursors/immature pDCs express CXCR4 and high levels of CXCR3, a receptor also present on Th1 and CTLs (6, 12, 13, 14). Accordingly, it has been speculated that L-selectin⁺ pDCs use CXCR3 to access T cell-rich areas of inflamed lymph nodes (6). CXCR3 may also be the receptor guiding pDCs into virus-infected tissue, as its three ligands CXCL9, CXCL10, and CXCL11 are induced by IFN and are released rapidly after virus exposure (15, 16, 17, 18). However, CXCR3 on pDCs apparently does not mediate chemotaxis in vitro (12, 14, 19). This does not reflect an indigenous inability of pDCs to chemotax because CXCL12 can induce migration of these cells. As CXCL12 is a chemokine with constitutive expression in many tissues (9, 20, 21), it is unlikely that it is alone responsible for pDC migration into sites of viral infection. Accordingly, it has been shown that CXCR4-mediated migration of pDCs is enhanced by CXCR3 ligand (CXCR3L) priming (14). This suggests that CXCR3 on pDCs functions solely as an amplifier of CXCR4-mediated chemotaxis. Thus, it is believed that pDCs require the combined presence and action of CXCR3 and of CXCR4 ligands to migrate into infected sites. These conclusions are based on data obtained in Boyden-type chamber assays that measure response to soluble chemokine gradients. It remained unknown whether CXCR3 alone could induce pDC migration in physiologically more relevant systems such as adhesion to and migration across endothelial cells (ECs). In this study, we demonstrate a novel mechanism of CXCR3-dependent pDC migration that does not involve CXCR4-mediated chemotaxis but requires immobilization and presentation of CXCR3Ls.

Purified nonlabeled mAbs were anti-CD3, anti-CD11b, anti-CD16, anti-CD19, anti-CD34, anti-CD41, anti-CD56, anti-CD235a (Immunotech, Marseille, France); anti-CCR3, anti-CXCL9, anti-CXCL10, anti-CXCL11, anti-CXCL12 (R&D Systems, McKinley Place, MN); and anti-CD8 and anti-CD45RA (both from BD Biosciences, Mountain View, CA). HECA-452 was kindly provided by Dr. L. Picker (University of Texas Southwestern Medical Center, Dallas, TX). PE-conjugated mAbs were anti-CD123, anti-CXCR1, anti-CXCR2, anti-CXCR4, anti-β₇-integrin, anti-CD104 (BD Pharmingen, San Diego, CA), anti-CCR7, anti-CD4, anti-L-selectin (BD Biosciences), anti-CCR2, anti-CCR6, and anti-CXCR5 (R&D Systems). FITC-conjugated mAbs were anti-CCR5, anti-CXCR3 (R&D Systems), and anti-CD45RA (BD Biosciences). Biotinylated anti-CCR1 and anti-CD123 were obtained from R&D Systems and BD Pharmingen, respectively. PerCP-conjugated anti-HLA-DR was obtained from BD Biosciences. HRP-conjugated phosphotyrosine-specific (PY-20) and focal adhesion kinase (FAK) phosphorylation site (Y397)-specific mAbs were obtained from BD Transduction Laboratories (San Jose, CA). Second step Abs included Alexa 488- and tetramethyl isothiocyanate (TRITC)-labeled goat anti-rabbit IgG (Molecular Probes, Leiden, The Netherlands) and TRITC- and Cy5-labeled goat F(ab′)₂ anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA). Lymphatic vessel-specific anti-podoplanin Abs were generated by repeated immunizations of rabbits with the extracellular domain of human podoplanin expressed in Escherichia coli.

Recombinant human chemokines used included CCL2, CCL3, CCL5, CCL7, CCL8, CCL13, CCL17, CCL19, CCL20, CCL22, CXCL8, CXCL11, CX3CL1 (R&D Systems), CCL22, CCL25, CXCL9, CXCL10, and CXCL12 (Strathmann Biotec, Hamburg, Germany). Human rTNFα and rIL-3 were obtained from Strathmann Biotec. IFN-α2b (intron-A) was obtained from Schering-Plough (Kenilworth, NJ). Heparan sulfate (HS) and chondroitin sulfate (CS) were obtained from Sigma-Aldrich (St. Louis, MO) and heparitinase II and chondroitinase ABC from Seikagaku (Tokyo, Japan). Wild-type HSV-2 and influenza virus strain A/Puerto Rico/8/34/H1N1 were kindly provided by Drs. H. Hofmann (Department of Clinical Virology, Medical University of Vienna, Vienna, Austria) and T. Muster (Department of Dermatology, Medical University of Vienna, Vienna, Austria), respectively. Pertussis toxin (PTX) was from Calbiochem (Darmstadt, Germany).

pDCs were purified from the peripheral blood of healthy donors essentially as described (22). Briefly, PBMCs were isolated by Ficoll-Hypaque (Pharmacia, La Jolla, CA) density gradient centrifugation and sheep RBC-binding T cells were removed. The resulting cell fraction was then depleted of residual T, B, NK, and hemopoietic stem cells, monocytes, basophils, platelets, and erythrocytes by anti-CD3/CD11b/CD16/CD19/CD34/CD41/CD56/CD235a (2 μg/ml each) immunolabeling and anti-mouse IgG immunomagnetic depletion (MACS; Miltenyi Biotec, Auburn, CA). Sixty to 85% of the remaining cell fraction qualified as pDCs by immunophenotype, i.e., CD123⁺CD45RA⁺HLA-DR⁺CD11c⁻, as revealed by FACS analysis (FACScan; BD Biosciences). T cells were obtained from the sheep RBC-bound PBMC fraction by hypotonic lysis. Polymorphonuclear neutrophil (PMN) cells were isolated by centrifugation of heparinized blood over a continuous isomolar 70% Percoll gradient (Amersham Biosciences, Arlington Heights, IL) as described previously (23). Human umbilical vein ECs and dermal microvascular ECs (DMECs) were isolated and propagated as described elsewhere (24, 25).

Total RNA was isolated from pelleted DMECs and flow-sorted (FACStar^PLUS; BD Biosciences) CD123⁺HLA-DR⁺CD11c⁻ pDCs (purity >99%) using TriPure reagent according to the manufacturer’s instructions (Roche Diagnostics, Basel, Switzerland). DNase 1 digestion and enzyme inactivation were performed in RNeasy microcolums (Qiagen, Valencia, CA) as recommended. DNA-free DMEC and pDC RNAs were eluted in 20 μl of RNase-free water and cDNA synthesis was conducted in MμlTI Ultra PCR tubes (Sorenson Bioscience, Salt Lake City, UT), each containing 50 mM Tris-HCl, pH 8.3, 75 mM KCl, 3 mM MgCl₂, 10 mM DTT, 2 mM of each dNTP, 200 pmol random hexamer (Roche Diagnostics), 20 U of recombinant RNase inhibitor (RNaseOUT; Invitrogen Life Technologies, Carlsbad, CA), 9.5 μl of RNA and 200 U of SuperScript III reverse transcriptase (+RT; Invitrogen Life Technologies) or 1 μl of 87% glycerol as mock RT control (−RT) in a total volume of 20 μl. After incubation for 1 h at 48°C in a thermocycler, 2.5-μl aliquots of each reaction mixture were subjected to PCR using two different primer combinations for specific amplification of CXCR3 splice variants A and B (26). For amplification of a 281-bp fragment of variant A, primer 5′ cxcr3 Aj (5′-ccatggtccttgag/gtgagtgacc-3′) including translation start and spanning the exon1/exon2 junction of the A variant was combined with primer 3′ cxcr3 AB (5′-gagcaggaaggtgtcggtgctgc-3′) recognizing both variants. For amplification of a variant B-specific 449-bp fragment, primer 5′ cxcr3 B (5′-gctgagcggatggagttgaggaag-3′) corresponding to exon 2 of variant B was used in combination with the consensus primer 3′ cxcr3 AB. Fifty-microliter reaction mixtures were prepared in MμlTI Ultra PCR tubes, each containing 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 1.5 mM MgCl₂, 1.5 mM of each dNTP, 100 pmol primer 3′ cxcr3 AB, 100 pmol primer 5′ cxcr3 Aj, or 5′ cxcr3 B, 2.5-μl template, and two drops of mineral oil as top layer. After a hot start at 95°C for 5 min and addition of 1.5 U of Taq polymerase/tube, a PCR program consisting of seven touch-down cycles (denaturation at 92°C for 30 s/primer annealing at 72°C-63°C for 45 s (−1.5°C/cycle)/primer elongation at 72°C for 45 s) followed by 38 standard cycles (92°C for 30 s/63°C for 45 s/72°C for 45 s) was conducted in a MiniCycler MJ Research (Biozym, Oldendorf, Germany). Resulting amplification products were visualized on a 1.5% Tris-acetate agarose gel by ethidium bromide staining.

pDCs were seeded in 96-well flat-bottom microtiter plates (Costar, Cambridge, MA) in RPMI 1640 supplemented with 10% FCS, 2.5 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (all from Invitrogen Life Technologies, Rockville, MD). pDCs were cultured in the absence or presence of IL-3 and TNF-α (100 U/ml each), HSV-2 (1 × 10⁶ PFU/ml), influenza virus (strain A/PR-8, 5 hemagglutination units/ml), or IFN-α (100 U/ml) for the indicated time periods. CXCR3-expressing T cells were generated by culture of purified T cells in the presence of IL-2 (100 U/ml) for 14 days as described previously (27).

Chemotaxis assays were performed as described previously (9). Briefly, pDCs either freshly isolated or stimulated as indicated were washed and resuspended in migration buffer (HBSS; Invitrogen Life Technologies), 1 mM CaCl₂, 0.5 mM MgCl₂, 0.1% BSA (Sigma-Aldrich)) at a density of 2–3 × 10⁶ cells/ml. Chemokine solution or buffer alone was added to individual wells of 24-well plates (Costar) on ice before Costar transwell devices (5-μm pore size) were inserted into the wells. Suspended cells were layered on top of the membrane and allowed to migrate for 3 h at 37°C. Transmigrated cells were recovered from the fluid phase of the lower well, stained with anti-CD123-PE, anti-CD45RA-FITC, and anti-HLA-DR PerCP simultaneously, and pDCs were enumerated by FACS analysis as described (9).

In a different set of experiments, we assessed migration in response to chemokines immobilized onto transwell filters. Chemokines (1–100 ng/ml) were allowed to bind to the upper, to the lower, or simultaneously to the upper and lower side of HS or CS (both 100 μg/ml)-coated or noncoated filters for 20 min at room temperature. After washing the filter insert, cells were added into the upper well and allowed to migrate for 3 h at 37°C. Then, membranes were cut out, stringently washed with ice-cold PBS, and adherent cells were fixed in ethanol/acetone (1:1) at −20°C for 15 min and stained with anti-CD123-PE (5 μg/ml). Migrated pDCs, i.e., CD123⁺ cells attached to the lower surface of the transwell membranes, were visualized by laser scanning microscopy ((LSM) LSM 510; Zeiss, Oberkochen, Germany) and enumerated automatically by a previously described algorithm and NIH image software (23). Migration indices were obtained as a ratio of pDCs migrated in the presence of chemokine and pDCs migrated in response to buffer. In addition, pDCs were counted on the upper side of the transwell filters to estimate their adhesion. Adherent pDCs were fixed with 4% paraformaldehyde, permeabilized with 0.1% saponin in PBS/1% BSA, and stained with anti-phospho(p)FAK (1 μg/ml) followed by goat anti-mouse IgG-Alexa488 (2 μg/ml). After quenching in normal mouse serum, cells were incubated with TRITC-coupled phalloidin (5 μg/ml; Sigma-Aldrich) and anti-CD123-biotin (10 μg/ml) followed by streptavidin-Cy5 (1 μg/ml; Molecular Probes). Control stainings were performed with appropriate isotype-matched Abs (2 μg/ml; Sigma-Aldrich).

TEM assays were performed as described (23). Briefly, ECs were seeded on the top of a collagen gel (Vitrogen 100; Invitrogen Life Technologies) and cultured to confluence in IMDM (Invitrogen Life Technologies) supplemented with 20% FCS, EC growth supplement (Promo Cell, Heidelberg, Germany), 2.5 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. After repeated washings with RPMI 1640, EC monolayers were incubated with the indicated chemokine for 10 min at 37°C followed by multiple rounds of washes. pDCs, T cells, or PMN cells were loaded with Cell Tracker Green (Molecular Probes), resuspended in RPMI 1640/5% BSA (10⁶ cells/ml), and seeded on top of the EC monolayer. Cells were then allowed to adhere to and migrate through the EC monolayer for 3 h at 37°C. In chemokine prepulsing experiments, pDCs were incubated with the indicated chemokine (1–100 ng/ml) for 20 min on ice, washed three times, and seeded onto the EC monolayer. Where indicated, pDCs were pretreated with 0.2 μg/ml PTX for 2 h before seeding them onto the EC monolayer. Migration was stopped by fixing the collagen gels with 4% paraformaldehyde and pDCs were visualized by anti-CD123 (5 μg/ml) and anti-mouse IgG-TRITC (2 μg/ml) immunostaining. Adherent and transmigrated cells were identified as Cell Tracker Green⁺ cells (and CD123⁺ cells in the case of pDCs) by confocal LSM and were enumerated using NIH image software as described (23).

Phosphotyrosine blotting was performed as described previously (28) with modifications. Cell pellets were lysed in 1% Brij97 (Sigma-Aldrich) lysis buffer (10 mM Tris-HCl, pH 7.8, 150 mM NaCl, 20 mM NaF (Sigma-Aldrich), 1 mM Na₃VO₄ (Calbiochem)) supplemented with a complete protease inhibitor mixture (Roche Diagnostics). After 30 min at 4°C, lysates were centrifuged to remove nuclei at 15,000 × g for 2 min. Supernatants were mixed with an equal volume of 2× reducing sample buffer and boiled for 3 min. Isolated proteins were resolved by gradient (5–12%) SDS-PAGE and blotted onto polyvinylidene difluoride membranes. Membranes were blocked using a buffer containing 1% BSA, 20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Tween 20, 0.1 mM Na₃VO₄ and incubated with HRP-conjugated PY-20 mAb (0.125 μg/ml). Phosphotyrosine-modified proteins were visualized using ECL Western blotting detection reagents (Amersham Biosciences).

ECs were detached by trypsin/EDTA (Invitrogen Life Technologies), washed, and incubated with the indicated chemokines (1–1000 ng/ml) for 20 min on ice. After several washes in ice-cold Ca²⁺/Mg²⁺-containing PBS, cells were exposed to rabbit anti-CXCL10 or anti-CXCL12 Abs (1 μg/ml each) followed by Alexa 488-labeled anti-rabbit IgG (2 μg/ml). Where indicated, ECs were treated with 25 mU/ml heparitinase II or chondroitinase ABC for 1 h at 37°C, washed three times, and then subjected to the chemokine binding assay. Cell-bound Alexa 488 immunofluorescence was analyzed on a FACScan.

All human tissue specimens were obtained upon informed consent. Pieces of normal human skin were taken during elective plastic surgery. Four-millimeter punch biopsies of lesional tissue were obtained from patients affected by HSV-2 or varicella zoster virus. Tonsils were removed in the process of elective tonsillectomy. All samples were snap-frozen in liquid nitrogen-chilled isopentane. Frozen sections were mounted onto glass slides and fixed with acetone at room temperature. For staining, the slides were hydrated and incubated with mouse anti-CXCL9, anti-CXCL10, or anti-CXCL11 mAbs. The bound primary mAbs were detected by sequential incubations with alkaline phosphatase-conjugated rabbit anti-mouse and an alkaline phosphatase-anti-alkaline phosphatase staining kit (DAKO, Carpinteria, CA) according to the manufacturer’s instructions. Isotype-matched control mAbs (DAKO) were used at equimolar concentrations. Each immunostaining protocol was performed on tissue samples from at least three different donors.

Five-micrometer cryosections were mounted onto glass slides, air-dried, and then fixed with acetone for 20 min. After drying, slides were hydrated with Ca²⁺/Mg²⁺-deficient PBS and exposed to anti-CD45RA or anti-CD8 (1:50 in PBS/1% BSA) for 45 min at room temperature followed by washings and incubation with Cy5-labeled goat F(ab′)₂ anti-mouse IgG (5 μg/ml) for 45 min at room temperature. After blocking with 1% mouse serum in PBS/BSA, the sections were exposed sequentially to biotinylated anti-CD123 (10 μg/ml) and streptavidin-Oregon Green (1 μg/ml; Molecular Probes). After several washings, the sections were stained with PE-labeled anti-CD4 (1/50 diluted in PBS/BSA), mounted with Fluoprep (BioMerieux, Marcy l’Etoile, France), and examined by confocal LSM. To detect lymphatic vessels, sections were incubated with rabbit anti-podoplanin antiserum or preimmune serum (final dilution: 1/2000) followed by TRITC-labeled goat F(ab′)₂ anti-rabbit IgG. In these experiments, pDCs were simultaneously visualized by anti-CD123 and anti-CD45RA immunostaining. To characterize chemokine receptor expression in tissue cells, double labeling experiments were performed with FITC- or PE-labeled anti-chemokine receptor mAbs and biotinylated anti-CD123 followed by TRITC- or Oregon Green-labeled streptavidin.

pDCs, identified as CD123⁺CD45RA⁺ cells in lineage Ag-depleted PBMCs, express CXCR3 uniformly and at high levels (Table I). CXCR4 is expressed by pDCs but its expression varies between pDCs from different donors (range: 7–86% positive pDCs; Table I). CCRs 1 through 7 and CXCR1, CXCR2, and CXCR5 are not or are only weakly expressed by blood pDCs. The majority of blood pDCs display substantial amounts of L-selectin which they lose during the process of purification (Table I). In addition, pDCs express high levels of the P-selectin ligand (PSGL)-1 and its fucosylated variant, the cutaneous lymphocyte Ag (CLA; Table I). Furthermore, nearly all pDCs express α₄β₇ integrin, the receptor for the mucosal addressin cell adhesion molecule. α^E, α₆, and β₄ integrins were not detected on pDCs in significant levels (Table I). Thus, pDCs consistently express one preponderant chemokine receptor, CXCR3, and display the relevant receptors for homing into the skin, the intestine, and secondary lymphoid tissues. We next analyzed isolated pDCs for their ability to migrate to a variety of chemokines in a transwell insert chemotaxis assay. pDCs did not transmigrate through the filters in response to CXCR3Ls irrespective of the chemokine concentration used (Fig. 1) but responded to CXCL12 (Fig. 1,A). Conversely, IL-2-activated CXCR3⁺ T cells were vigorously attracted by soluble CXCL10 gradients (Fig. 1,B). pDCs were not significantly attracted by any of the other chemokines tested (Fig. 1,A). pDCs matured in vitro in the presence of IL-3 and TNF-α migrated efficaciously to the CCR7 ligand CCL19 but not to CXCR3Ls (Fig. 1 B, see below) or to any other chemokines tested (data not shown). These results closely mirror previously published results obtained in Boyden-type chamber assays (12, 13, 14).

We next asked whether the apparent nonresponsiveness of pDCs in Boyden-type chamber assays is due to the expression of a migration-incompetent form of CXCR3. Indeed, recent data indicate that two splice variants of CXCR3, CXCR3A and CXCR3B, exist (26). While CXCR3A mediates chemotaxis, CXCR3B may be involved in transmitting angiostatic signals to ECs. As revealed by a sensitive and splice variant-specific RT-PCR assay, pDCs express CXCR3A but fail to display CXCR3B-encoding RNA (Fig. 2). In contrast, human microvascular ECs are positive for CXCR3B (Fig. 2). This indicates that pDCs are, in principle, equipped with the potential to migrate in response to CXCR3 ligands.

It has been suggested that immobilized chemokines induce leukocyte adhesion to ECs and consequent TEM in vivo (29). Haptotaxis, i.e., migration along a positive gradient of immobilized chemoattractant, was shown to be the mechanism of neutrophil and monocyte migration (30). To investigate the possible dependency of pDC migration on chemokine immobilization, we precoated 5-μm pore-size transwell filters on the apical, on the basal, or on both the apical and the basal side with CXCL10. The coating conditions used allow for the immobilization of significant levels of chemokine on the selected side of the filter (A. Rot, unpublished observations). Purified pDCs were placed into the upper chamber of the transwells and the cells migrated were enumerated on the basal side of the filter and in the lower chamber. pDCs transmigrated when CXCL10 was immobilized onto the upper side of the filter but no response was observed when the chemokine was immobilized on the lower side or on both sides of the membrane (Fig. 3,A). In contrast, CXCL10 immobilized on the basal filter side promoted weak T cell migration (Fig. 3,A). No T cell migration was induced when CXCL10 was immobilized on the apical filter side (Fig. 3 A). In none of the experimental setups above were pDCs or T cells detected in the fluid phase or adherent to the bottom of the lower chamber (data not shown). Thus, pDCs can sense a negative step gradient of immobilized CXCL10 and migrate in response by a mechanism that entails features of haptotaxis (i.e., dependency on chemokine immobilization) and chemorepulsion (i.e., movement away from highest chemokine concentration).

HS saccharides that decorate various proteins are believed to bind and present chemokines in vivo, in particular, on the apical EC surface (29, 30, 31, 32). To deconstruct such a chemokine presentation setup, the top sides of filters were coated with HS or CS before pulsing them with CXCL10 and measuring pDC migration in response. As shown in Fig. 3,B, HS but not CS enhanced CXCL10-mediated pDC migration. HS-bound CXCL10 more efficaciously induced pDC migration than polycarbonate-bound CXCL10 (Fig. 3,B). Binding studies with titrated amounts of radiolabeled CXCL10 showed that, upon exposure to suboptimal chemokine concentrations, the HS-coated filters bound up to 5-fold more chemokine than nonmodified filters (data not shown). For control purposes, HS, CS or mixtures of CXCL10 and HS or CS were placed in soluble form into the lower chamber of the transwells. None of these experimental conditions resulted in migratory responses comparable to those induced by HS-immobilized CXCL10 (Fig. 3,B). To see whether pDC migration in response to HS-bound CXCL10 is restricted to apically bound chemokine, HS coating was implemented in an experimental setup analogous to Fig. 3,A. As shown in Fig. 3 C, pDCs transmigrated when HS/CXCL10 was immobilized onto the upper side of the filter but no response was observed when HS/CXCL10 was immobilized on the lower side or on both sides of the membrane. Thus, CXCL10 immobilized and presented apically in a physiologically relevant fashion is sufficient to induce pDC migration.

The following assays were performed to study the outcome of pDC stimulation by immobilized CXCL10. As shown in Fig. 4,A, pDCs strongly adhered to HS/CXCL10-bound, but not to HS-coated, membranes. In contrast, exposure of pDCs to soluble CXCL10 did not induce adhesion to the substratum (data not shown). Also, only immobilized but not soluble CXCL10 induced intracellular signaling as evidenced by enhanced tyrosine phosphorylation of several cytosolic pDC proteins with a molecular mass of ∼120–140 kDa (Fig. 4,B, left panel). In contrast to pDCs, T cells failed to phosphorylate cellular proteins when exposed to HS-immobilized chemokines (Fig. 4 B, right panel).

pDCs that migrated in response to HS-immobilized CXCL10 formed lamellipodia and long “dendrites” that impinge through the pores of the membrane (Fig. 4,C, left panels). These morphologic alterations were not seen in the few cells that adhered and migrated spontaneously in the absence of chemokine (Fig. 4,C, right panels). Confocal LSM of pDCs during the process of transmigration revealed clustering of phosphorylated FAK (pFAK) as well as of IL-3Rα/CD123 next to the edges of the filter pores (Fig. 4 D). As expected, pan-FAK immunostaining also revealed a redistribution of pDC-expressed total FAK toward the edges of the pores and pFAK clustering occurred only when filters were coated with CXCL10/HS but not when the filters were coated with HS only (data not shown). This suggests that pDCs sense the immobilized negative step gradient of CXCL10 in or around the pores and respond by adhesion and migration.

Next we investigated whether CXCL10 and CXCL12, the two prototypic chemokines that stimulate pDCs, can bind to ECs in vitro. ECs were pulsed with CXCL10 and CXCL12 and chemokines bound were detected by specific Abs and FACS. While nontreated ECs were devoid of CXCL10 and CXCL12, chemokine pulsing of ECs resulted in the appearance of significant surface immunoreactivity (Fig. 5,A). No significant CXCL10 and CXCL12 binding was observed to heparitinase-treated ECs (Fig. 5,A). Control chondroitinase treatment did not significantly affect chemokine binding to ECs (data not shown). Moreover, CXCL10 and CXCL12 cross-competed for EC binding (Fig. 5,A) indicating that ECs display a limited number of moieties capable of HS-dependent chemokine binding. In binding studies with titrated concentrations of CXCL10 we observed that CXCL10 binding to ECs was saturable and optimal at rather low chemokine concentrations of around 100 ng/ml (Fig. 5 B). A similar dose-binding relation was seen with CXCL12 (data not shown).

To explore whether EC-associated CXCR3Ls can be detected in vivo, we studied HSV-infected skin by immunohistochemistry. All three CXCR3Ls were found to be present in the lesions and were detected in the epidermis most prominently (Fig. 5,C). CXCL10 and CXCL11 were further found associated with dermal microvascular ECs (insets of Fig. 5,C; lower left panel and upper right panel, respectively). Staining of normal human skin with CXCR3L-specific mAbs and isotype control staining of lesional skin revealed no immunoreactivity (Fig. 5 C, data not shown).

To further explore the biologic relevance of chemokine presentation-dependent pDC migration, we asked whether EC-bound chemokines can also provide the essential stimulus for pDC transmigration through EC monolayers. Resting and TNF-α-activated EC monolayers grown on collagen gels were pulsed with CXCL10. pDCs were placed apically onto ECs and adhesion and TEM were assessed by multilevel confocal LSM. In accordance with the results obtained in the transwell assays, ECs supported significant CXCL10-dependent adhesion (Fig. 6,A) and TEM (Fig. 6,B) of pDCs. EC-bound CXCL12 also promoted adhesion (Fig. 6,A) and TEM (Fig. 6,B) of pDCs, though mostly to a lesser extent than CXCL10. Resting ECs supported pDC adhesion and migration to CXCL10 more potently than activated ECs (Fig. 6, A and B). In the same experiment, selectively activated ECs but not resting ECs promoted the transmigration of PMN (Fig. 6,C). In agreement with the results of the transwell experiments, activated T cells showed only inefficient migration through EC monolayers presenting CXCL10 apically (Fig. 6 D).

So far we have shown that soluble CXCR3Ls have no promigratory effect on pDCs. We asked whether soluble CXCR3Ls can affect the TEM induced by immobilized chemokines. This is a relevant question because under pathologic conditions CXCR3Ls can be present in the circulation (33, 34). pDCs were pretreated with CXCL9, CXCL10, or CXCL11 before placing them over EC monolayers presenting immobilized CXCL10. Pretreatment with each of the CXCR3Ls abolished the adhesion of pDCs to ECs and their TEM (Fig. 7,A). Inhibition of pDC migration was, however, not ligand specific: pretreatment of pDCs with either CXCR3Ls or CXCR4Ls resulted in the inhibition of both their adhesion and migratory responses to either of these ligands immobilized on the EC surface (Fig. 7,B). It further appears that the signal transduction required for CXCL10-mediated pDC adhesion and TEM is different. Adhesion is insensitive to PTX (Fig. 7,C), while transmigration is abolished by G_i-protein blockade (Fig. 7 D).

Next we investigated whether pDCs infiltrate virus-induced skin eruptions. pDCs can be reliably identified by their CD4⁺CD45RA⁺CD123^bright immunophenotype (35, 36, 37). We found pDCs localized in perivascular clusters in HSV (Fig. 8, A and B; pDCs appear as whitish and yellow cells, respectively) and in varicella zoster virus-induced skin eruptions (data not shown). Clusters of pDCs were surrounded by a mixed infiltrate consisting of CD4⁺ and CD8⁺ T cells (Fig. 8, A and B). Normal skin did not contain pDCs (not shown). Control staining of tonsils revealed pDCs evenly dispersed in the T cell areas (Fig. 8,C). In situ analysis of chemokine receptor expression revealed homogeneous CXCR3 expression in pDCs as well as in the majority of the cells infiltrating HSV-induced lesions (Fig. 8 D; pDCs are yellow). Other chemokine receptors (e.g., CCR5, CCR7, CXCR4) were either not detectable or were identified on a minor part of the cellular infiltrate only (data not shown).

To investigate the effects of HSV exposure on the migratory capacity of pDCs, we cultured these cells in the presence of HSV and determined chemokine receptor expression and function. Within 6 h of HSV exposure, pDCs started to lose CXCR3 and to gain CCR7 expression (Fig. 8,E). After 12 h, the loss of CXCR3 was complete and the expression of CCR7 was at its peak (Fig. 8, E and G). Consistent with the rapid and massive gain of CCR7 expression, pDCs started to migrate to soluble gradients of CCL19 within 6 h of virus exposure (Fig. 8,F). Exposure of pDCs to exogenous IFN-α and to influenza virus down-regulated CXCR3 and induced CCR7 expression and responsiveness to a similar magnitude as seen with HSV (Fig. 8,H, data not shown). Thus, the direct contact with different viruses and/or virus-induced nIFNs stimulates the loss of CXCR3 and up-regulation of CCR7-dependent migration. In in situ analyses of HSV-induced lesions, pDCs were occasionally found in close proximity to podoplanin-expressing lymphatic vessels (Fig. 8 I). This indicates that HSV-induced CCR7 expression allows pDCs to exit the tissue and to migrate toward regional lymph nodes via CCR7 ligand-expressing lymphatic vessels (25).

A successful combat of viral infection requires at least two components: the secretion of virus replication-inhibiting factors such as nIFNs and the induction of virus-specific T cell immunity. pDCs are ideally suited to accomplish both these tasks. To understand their functionality in vivo it is important to learn whether and how pDCs enter virus-infected tissues. This study provides original evidence that CXCR3L immobilization on ECs may be required and sufficient to target pDCs into foci of viral replication.

It has been known that CXCR3 is a predominant chemokine receptor expressed on circulating pDCs (Refs. 6 , 12 , 14 ; this study). In functional terms, however, CXCR3 is inactive in inducing chemotaxis of pDCs (Refs. 12 , 14 , 19 ; this study). Recent work emphasizes that CXCR3 functions only as a threshold regulator of CXCR4-driven chemotaxis (14). In this study, however, we show that the CXCR3-CXCR3L interaction is fully competent to induce adhesion and migration of pDCs. These events are evoked only when CXCR3Ls are presented to pDCs in solid phase, a physiologically relevant condition. Glycosaminoglycans, HS, in particular, are crucial for productive presentation of chemokines in vivo (29, 31, 32). Our study shows that CXCR3Ls which are presented by HS on artificial surfaces or the plasma membrane of ECs, but again not the soluble chemokines, can deliver the promigratory signals to pDCs.

Intriguing is the observation that migration in response to immobilized CXCR3Ls is a cell type-restricted phenomenon. CXCR3⁺ T cells, in contrast to pDCs, failed to migrate to CXCR3Ls immobilized on artificial membranes and apical EC surfaces but migrated vigorously to soluble chemokine gradients. This implies that CXCR3 acquires the specific functionality to respond to immobilized chemokines only in the context of the pDC membrane microenvironment. Mechanistically, transmigration of pDCs takes place only to apically immobilized chemokine, signifying the involvement of a negative step gradient. Positive haptotactic CXCR3L gradients as well as chemokines immobilized in a nongradient fashion failed to induce the transmigration of pDCs.

Thus, the postulated mechanism of pDC migration includes two mechanistic facets: the migration to immobilized chemokine (haptotaxis) and the movement along a negative chemokine gradient (chemorepulsion). These two migratory mechanisms have been described for other chemokines and other leukocyte types (30, 38, 39). However, until now, in sharp contrast to the pDC response to CXCR3Ls, the chemokine-induced haptotaxis and chemorepulsion were shown to take place separately from each other and only as alternatives to also viable chemotaxis. Remarkably, in the case of CXCR3L-induced pDC migration, “haptorepulsion” is the only possible in vitro mechanism of the observed response. Analogous chemokine immobilization-dependent migration has been described in T cells in response to CXCL12 which, however, also requires lateral shear stress (40). Lateral shear stress is not required for TEM of pDCs. Conceivably, haptorepulsion, as shown for pDCs as a paradigmatic example in this study, may be a novel relevant migratory mechanism used by other cell types in response to different chemokines.

CXCL12 is clearly capable of being productively presented by ECs to pDCs in a similar fashion as CXCL10. Thus, our data suggest that CXCR4, besides being a receptor mediating chemotaxis, may be able to execute haptorepulsion in pDCs. Our data also show that CXCR3 and CXCR4 on pDCs can function independently from each other. This argument derives from experiments showing that CXCL10 bound to artificial membranes suffices to deliver the migratory signal, that primary ECs that support pDC migration via immobilized CXCL10 lack CXCL12 immunoreactivity, and that blocking anti-CXCR4 mAbs do not interfere with CXCL10-mediated migration across ECs (data not shown). Our data, however, do not exclude the possibility that the two chemokine receptors use a common signal transduction machinery to execute pDC haptorepulsion. In support of the latter stands the observation of heterologous desensitization of CXCR3- and CXCR4-dependent pDC responses by the respective ligands. However, in the scenario of a viral infection, the possible importance of the CXCR4/CXCL12-dependent migration will be diminished due to the appearance of virally induced CXCR3Ls that compete with CXCL12 for EC binding (Fig. 5, A and C). The involvement of CXCR3 in pDC migration into sites of viral replication is further substantiated by the finding that pDCs express the “classical” form of CXCR3 and not its migration-incompetent B variant (26) as revealed by RT-PCR, cDNA cloning, and sequence analysis.

Our contention that CXCR3L-binding and presenting moieties exist on ECs derives from the observations that CXCR3Ls are present on ECs in cutaneous sites of viral replication and that CXCR3Ls bind primary ECs in a dose-dependent and saturable fashion. At the first glance, candidate EC receptors include CXCR3B that binds CXCR3Ls with lower affinity than CXCR3A (26). The observations, however, that CXCR3L binding to ECs is HS-dependent and that low concentrations of CXCL12 can inhibit EC CXCR10 binding speak in favor of a previously not recognized CXCR10L-binding structure on ECs.

pDCs more efficiently adhered to and migrated across resting than activated EC monolayers. Shedding of the HS-based chemokine binding sites during EC activation could be responsible for this phenomenon (41) but the actual mechanism remains to be determined. More importantly, our data strongly suggest that activation-induced up-regulation of EC adhesion molecules is not required for TEM of pDCs. Granulocytes, in contrast to pDCs, migrated across activated ECs selectively. We speculate that the abundant expression of L-selectin, selectin ligands including CLA and PSGL-1 as well as ligands for intestinal addressins (Refs. 6 , 8 , 14 ; this study) may allow pDCs to roll on resting or on minimally activated ECs and, thus, to constantly screen peripheral sites and lymph nodes for chemokine signatures of viral infection. In this respect, the CXCR3Ls may be of particular significance as they are among the gene products that are extremely rapidly and robustly up-regulated after viral encounter (42, 43). In aggregate, these observations allow for speculation that pDC traffic into virally infected sites is maximal at early time points and may decrease with the onset of overt EC activation and EC activation-dependent recruitment of effector cells of the innate and adaptive immune system.

Soluble chemokines exert a negative signal to pDCs that is dominant over the productive proadhesive and migratory signal induced by immobilized chemokines. This suggests a novel therapeutic possibility to prevent pDC homing to peripheral tissues in autoimmune diseases such as lupus erythematosus and functionally related pathologies such as graft rejection (44, 45, 46). To this end, CXC chemokines genetically engineered to preserve their CXCR3-desensitizing capacity but to abolish their EC/HS-binding ability may be used to prevent pDC-mediated tissue damage.

Upon HSV exposure, pDCs almost instantaneously down-regulate CXCR3 and up-regulate CCR7 and respond to its ligands chemotactically. Our previous results demonstrated that CCR7Ls are secreted basolaterally by lymphatic ECs (25) and, thus, may be in the position to guide the virus-exposed pDCs into the lymphatic channels. In summary, the two distinctive mechanisms of pDC trafficking, the CXCR3L immobilization-dependent, haptorepulsive migration and the chemotactic gradient-induced departure define consecutive stages of maturation-related pDC functionality.

We thank B. Reininger and E. Schwarzinger for excellent technical assistance.