The human chemokines CCL19 and CCL21 were purchased from R&D systems and PeproTech (Accession # Q99731.1 and Q6ICR7, respectively). Tailless-CCL21 was from ALMAC (sequence SDGGAQDCCL KYSQRKIPAK VVRSYRKQEP SLGCSIPAIL FLPRKRSQAE LCADPKELWV QQLMQHLDKT PSPQKPAQG). Goat anti-human CCL21 was from R&D systems (Cat. no. AF366). The human CCR7 cDNA was cloned from a spleen-derived cDNA library. The promiscuous chimeric G protein Gα_Δ6qi4myr (G_qi4myr) that converts Gα_i-related signaling into a Gα_q readout (23, 24) was kindly provided by Evi Kostenis (University of Bonn, Germany). Antibodies against p44/42 MAPK (ERK1/2) and phosphorylated p44/42 MAPK (ERK1/2, Thr202/Tyr204) were purchased at Cell Signaling. Neuraminidase (NA) was from Sigma.

HEK293 cells were grown at 10% CO₂ and 37°C in Dulbecco’s modified Eagle’s medium 1885 supplemented with 10% fetal bovine serum, 2 mM glutamine, 180 U/ml penicillin, and 45 μg/ml streptomycin. PathHunter U2OS β-Arrestin 2 Parental cell line (DiscoveRx, Birmingham, United Kingdom) were grown at 5% CO₂ and 37°C in MEMα Glutamax medium supplemented with 10% fetal bovine serum, 180 U/ml penicillin, 45 μg/ml streptomycin, and 0.25 μg/ml Hygromycin B (Invitrogen). Transient transfection of HEK293 cells for phosphatidylinositol (PI)-turnover was performed using the calcium phosphate precipitation method as previously described (25, 26). U2OS cells were transfected using FuGENE^® 6 Transfection reagent (Roche, Mannheim, Germany).

HEK293 cells were cotransfected with receptor cDNA and G_qi4myr, which converts the Gα_i signal into a Gα_q signal, making it possible to measure the chemokine receptor activation as PI-turnover (23, 24). One day after transfection, the cells were seeded in 96-well plates (3.5 × 10⁴ cells/well) and incubated with 0.65 μCi of ³H-myo-Inositol in 0.1 ml growth medium for 24 h. Cells were washed twice with Hank’s buffered salt solution (HBSS) supplemented with CaCl₂ and MgCl₂ and incubated for 15 min in 0.1 ml buffer supplemented with 10 mM LiCl prior to ligand addition followed by 90 min incubation. The generated [³H]inositol phosphate was detected directly by addition of SPA-YSI bead solution. Determinations were made in duplicates.

Recruitment of β-arrestin was measured using the PathHunter™ β-arrestin assay (DiscoveRx). CCR7 was fused with the ProLink™ (PK) 1-tag [a small fragment of the enzyme β-galactosidase (β-gal), the enzyme donor] and cloned into a pCMV-vector. Assays were performed in a U2OS cell line stably expressing β-arrestin 2 coupled to the large β-gal fragment (enzyme acceptor). Cells were seeded in 96-well plates, 20,000 cells/well, and transfected the following day with 50 ng DNA using FuGENE^® 6 reagent (0.15 μl/well). Twenty-four hours after transfection, the medium was removed and 100 μl Opti-MEM^® I (Gibco^®) was added. The following day, cells were stimulated with varying concentrations of ligand for 90 min at 37°C. The Detection Reagent Solution^® was added prior to 60 min incubation at room temperature and the β-arrestin recruitment was measured as chemiluminescence using Perkin Elmer EnVision 2104 Multilabel Reader.

The internalization of CCR7 was measured in the U2OS cell line from DiscoveRx. The cells coexpressed enzyme acceptor-tagged β-arrestin 2 and a PK-tag linked to endofin, which is localized to endosomes. The experiment was carried as described for the β-arrestin recruitment assay, yet with 3 h incubation with the ligands.

Monocyte-derived DCs (DCs) were prepared by plastic adhesion from peripheral blood mononuclear cells (PBMCs) kindly provided by Professor Inge Marie Svane, Herlev Hospital, Denmark. The monocytes were differentiated into immature DCs in X-Vivo medium supplemented with 1000 U/ml GM-CSF and 250 U/ml IL-4 for 6 days, followed by maturation for 2 days with 1000 U/ml TNF-α, IL-1β, and IL-6 plus 1 μg/ml PGE₂. Harvest of DCs was performed by medium aspiration followed by cold incubation of remaining cells in PBS with EDTA (5 mM) and subsequent scraping of cells followed by freezing of DCs in aliquots.

Mature DCs were seeded in 96-well plates (8.0 × 10⁴ cells/well). DCs were incubated for 2 h at 37°C, 5% CO₂ before addition of chemokine and incubation at 37°C for 10 min. Following, the cells were lysed and prepared for measurement of ERK1/2 phosphorylation by using the Meso Scale Discovery (MSD)^® MULTI-SPOT Assay system Phospho (Thr202/Tyr204; Thr185/Tyr187)/Total ERK1/2 Assay (Meso Scale Discovery, MD, USA) according to the manufacturer’s instructions. Determinations were made in duplicates and measurements were performed on MSD SECTOR Imager.

Mature DCs were washed twice in serum-free X-vivo medium and incubated in this medium for 2 h at RT to starve them from serum factors. In experiments with NA treatment, the enzyme was added during this incubation at a concentration of 2.5 × 10⁻² U/ml and the cells kept at 5% CO₂ and 37°C. The cells were seeded in 96-well plates (8 × 10⁴ cells/well) and stimulated with CCL19, CCL21, or tailless-CCL21 for the indicated time period. The cells were lysed in ice-cold RIPA buffer (Millipore) with protease (mini complete Roche) and phosphatase inhibitors (mix 2 and 3, Sigma). The lysates were run on SDS-gel and transferred to PVDF for detection of total and phosphorylated ERK1/2.

The Ibidi^® assay was carried out according to the manufacturer. Briefly, the collagen mixture was prepared and DCs were added to a final concentration of 0.5 to 1 × 10⁶ cells/ml. The cell/collagen mixture was loaded into the Ibidi channel, according to the protocol, and left to polymerize for 35 min in a 5% CO₂, humidified incubator at 37°C. The final collagen concentration was 1.67 mg/ml. Finally the source and sink reservoirs were filled with medium containing 2% human serum with or without chemokine, respectively, and the slide with loaded DCs was mounted on a computer controlled stage, holding a temperature controlled (37°C), humidified incubation chamber with 5% CO₂. DC migration was followed for 12 h by time-lapse microscopy using a time interval of 2 min. Cell migration (approximately 20–40 cells) was tracked using a commercial tracking program (Autozell) and subsequently analyzed to get a population-based chemotactic index (CI) value (MATLAB). CI is a measure of net translocation distance to the source relative to total distance traveled.

In order to determine ligand surface localization, DCs were stained and fixed in suspension by incubating in serum-free medium with or without 100 nM CCL21, or tailless-CCL21 for 30 min on ice, followed by washing in PBS and fixation for 15 min in 3.7% formaldehyde in PBS. After fixation the cells were washed and incubated for 1 h with primary anti-human CCL21 antibody. The DCs were washed and incubated with secondary rabbit anti-goat antibody (Alexa-488 coupled) for 1 h. The DCs were washed, in PBS, with an additional final wash in sterile water and resuspended in 70% ethanol. DCs were mounted on microscope slides and sealed under coverslides.

CCL21 and tailless-CCL21 were adsorped on Maxisorp plates in PBS buffer over night at 4°C. The wells were washed three times in PBS and incubated for 60 min in blocking buffer (PBS, 1% BSA) at room temperature. Wells were incubated for 60 min with primary goat anti-CCL21 antibodies in blocking buffer. After PBS washing, the wells were incubated with secondary rabbit anti-goat HRP antibodies for 60 min. After PBS washing, TMB substrate was added. The reaction was stopped with H₂SO₄ 0.2M and absorbance at 450 was read in an Envision plate reader.

IC₅₀, EC₅₀, and K_d/K_i values were determined by non-linear regression and B_max values were calculated using the GraphPad Prism 4.0 software (GraphPad, San Diego, CA, USA).

Statistical significance was calculated using unpaired t test, and data are represented as mean ± SEM. Significance is indicated as follows: NS, not significant; *P < 0.05; **P < 0.01.

To assess the relative importance of the two CCR7 ligands in stimulating various DC responses, we investigated the effect of CCL19 and CCL21 on a range of biological readouts. One very important readout in immune cell biology is the ligand-controlled migration of cells toward physiologically relevant ligands. We measured chemotaxis of human monocyte-derived DCs induced by either CCL19 or CCL21 using Ibidi^® 3D chemotaxis slides. At 10 nM, CCL21 induced a migration response characterized by a significantly lower CI compared with CCL19. However, at 100 nM chemokine concentrations, the cells migrated with equal directionality (CI) toward the two ligands (Figure 1A). The migration patterns are depicted in the spider diagrams (Figure 1B), which show that DC migration is less directional in response to CCL21 compared with CCL19 at low concentrations (10 nM).

ERK is often associated with migration, and therefore we tested ERK activation profiles in mature human DCs stimulated with either CCL19 or CCL21. Surprisingly, CCL21 induced more ERK activity than CCL19 (Figure 2). At 100 nM, CCL21 induced an approximately fourfold increase in pERK1/2 over cells stimulated with buffer, whereas the same amount of CCL19 only induced an approximately twofold increase (Figure 2C). The same tendency was seen at 1 and 10 nM chemokine concentrations (Figures 2A,B).

In order to determine G protein contribution to the MAP kinase activity, ERK activity was measured in the presence of pertussis toxin (PTx), a substance that inhibits Gα_i irreversibly by ADP-ribosylation. As seen in Figure 2D, 10 μg/ml PTx completely abolished CCL19- and CCL21-induced ERK1/2 activation in DCs, indicating that Gα_i is the main mediator of CCR7-induced pERK1/2 activation in DCs.

Due to the established broad spectrum GAG binding of CCL21 through its C-terminal tail (12, 13), a local CCL21 reservoir is expected to form at the DC surface, which may gradually become available for CCR7 activation through various mechanisms, influencing the CCL21 ERK activation profile over time. Thus, we investigated time-dependent MAP kinase activation. At 10 nM chemokine concentrations, there was a tendency that CCL19-induced ERK activation at an earlier time-point than CCL21, which shows a tendency to peak later (Figure 2E). At 100 nM chemokine concentrations, ERK activation was in general more pronounced. ERK activation by CCL21 was prolonged and peaked at 50 min after induction (or even later), whereas that of CCL19 peaked earlier – at 30 min (or possibly before) (Figure 2F). The stronger CCL21-induced ERK activation giving rise to the hypothesis of CCL21 forming a local reservoir.

The GAG-subtype polysialic acid was recently shown to act as coreceptor for CCL21 during CCR7 activation (14), whereas others, e.g., chondroitin sulfate-B (CS-B) have been shown to inhibit CCL21 activity. To determine the impact of polysialic acid on CCL21-induced ERK activation, DCs were pretreated with NA that was also present during the experiment. Indeed, this treatment negatively affected CCL21-induced ERK1/2 activation (Figure 2G), but not CCL19-induced activity (Figure 2H), supporting an important role of polysialic acid in CCL21-induced CCR7 activation and thus ERK signaling in DCs.

Since the release of the C-terminal tail of CCL21 from its polysialic acid interaction negatively affects CCL21-mediated ERK activation (Figure 2G), we wanted to investigate the role of the extended C-terminal tail of CCL21 in chemotaxis and therefore designed a truncated version of CCL21 lacking the C-terminus and investigated its properties in inducing 3D migration and ERK activation of human DCs.

Indeed, tailless-CCL21 was more potent than full-length CCL21 in inducing directed migration of DCs and at a concentration of 10 nM induced chemotaxis with similar CI as CCL19 (Figure 3A), whereas there was no significant difference between the three chemokines at 100 nM (Figure 3A). Tailless-CCL21 also resembled CCL19 with regard to ERK activation and thus in contrast to what was observed for CCL21, neither CCL19 nor tailless-CCL21-induced ERK activation was affected by DC NA treatment (Figure 3B).

To test if CCL21 could be detected on the DC surface to a higher degree than tailless-CCL21, we performed immunostainings against the two chemokines on cells that had been incubated on ice with 100 nM of the respective chemokines. In contrast to tailless-CCL21, CCL21 localized to the DC membrane forming discrete puncta, whereas cells incubated in the absence of ligand were devoid of staining (Figures 4A–E). Membrane localization appeared as an extended rim around the cell that was not sharply defined due to the dendrites that make the membrane villi-like in appearance in non-adherent DCs (Figure 4D). CCL21 stainings in general took form of either a few large puncta or multiple small puncta, with both types of stainings following the rim of the cell (Figure 4E). Importantly, polyclonal antibody against CCL21 recognized tailless-CCL21 to the same degree as it recognized CCL21 (Figure 4F).

Intrigued by the observations that tailless-CCL21 resembled CCL19 more than it resembled CCL21 in both migration and ERK activation, we investigated the potency of CCL19, CCL21, and tailless-CCL21 in inducing signaling pathways “closer” to the membrane (G protein activation, β-arrestin recruitment, and CCR7 internalization) as compared to the more downstream MAP kinase activation and migration.

The ability to induce signaling via Gα_i – the major G protein pathway induced by endogenous chemokine receptors (27) – was investigated in a PI-turnover assay using HEK293 cells transfected with CCR7 and the chimeric protein G_qi4myr that converts a Gα_i-coupled response into a Gα_q readout (24).

We found that the potency of CCL21 was ~10-fold lower compared with that of CCL19 with an estimated EC₅₀ of 50 nM (−log 7.3 ± 0.18, n = 3) for CCL21 compared with 5 nM for CCL19 (−log 8.3 ± 0.27, n = 3), and with no significant difference in efficacy between the two (Figure 5A). Surprisingly, in contrast to what was observed in migration (Figure 3A), tailless-CCL21 did not exhibit increased ability to signal via Gα_i but displayed low activity in this pathway similar to full-length CCL21 (estimated EC₅₀ of 142 nM).

To determine receptor mediated β-arrestin recruitment, CCR7 tagged at the C-terminus with the catalytic N-terminal domain of β-gal, also referred to as the enzyme donor fragment, was transiently transfected in U2OS cells stably expressing β-arrestin 2 fused to an N-terminal deletion mutant of β-gal, the enzyme acceptor fragment. The binding of β-arrestin 2 to CCR7 reconstitutes donor and acceptor parts of β-gal into a functional enzyme and the readout was measured as chemiluminescence. As with Gα_i signaling, CCL19 exhibited a significantly higher potency compared with CCL21 (~17-fold), and CCL21 only reached half of CCL19-induced activity at 100 nM (Figure 5B). EC₅₀ was estimated to 15 nM (−log 7.8 ± 0.1, n = 3) for CCL19 and 263 nM (−log 6.6 ± 0.32, n = 3) for CCL21. Again, tailless-CCL21 was weak; in fact it did not induce β-arrestin 2 recruitment at concentrations up to 100 nM. β-arrestin-mediated CCR7 internalization induced by CCL19 and CCL21 was measured in U2OS cells coexpressing untagged receptor DNA, β-arrestin 2 fused to the β-gal enzyme acceptor fragment, and endosomes tagged with β-gal enzyme donor fragment. Tailless-CCL21 internalization was not investigated due to lack of β-arrestin coupling. Quantitative fusion of β-arrestin and endosomes was detected as chemiluminescence. CCL21 induced internalization; however, only to ~40% of the level induced by CCL19 at 100 nM (Figure 5C). Its potency was also lower (~37-fold) with an EC₅₀ estimated to 152 nM (−log 6.8 ± 0.08, n = 3) compared with 4.1 nM for CCL19 (−log 8.4 ± 0.14, n = 3).

Thus CCL21 is weaker than CCL19 in G protein activation and β-arrestin recruitment, and tailless-CCL21 resembles CCL21 more than CCL19 although it seems to be less potent compared with CCL21.

The fact that tailless-CCL21 displays distinct properties compared with CCL19, and resembles CCL21 more with regard to upstream signaling events, implies that the tail of CCL21 plays no major role in controlling differences in CCL19 and CCL21 receptor engagement. In Figure 6A, a backbone solution structure (top) and a surface view (bottom) is presented of the NMR structure of CCL19 and CCL21, excluding the flexible C-terminal tail of CCL21 (28, 29). Considering overall structural and functional similarities between CCL19 and CCL21 (19–22), we compared the tertiary structures of these two chemokines. As the solution structure reveals, CCL21 maintains the triple-stranded beta-sheet and C-terminal alpha-helix typical to chemokines. The surface presentations underscore obvious differences between CCL19 and CCL21 in GAG but also in receptor docking domains. CCL21 has a much more pronounced clustering of positively charged amino acids in the GAG-binding domain, which probably facilitates the highly increased GAG-binding capability of this chemokine residing in the positively charged C-terminus extending directly from this region (Figure 6A), a feature not matched in CCL19. The primary sequence alignment (Figure 6B) highlights similarities and charges of CCL19 and CCL21. The highly basic GAG-binding tail of CCL21 clearly underscores the charge difference between CCL19 and CCL21 located to the GAG-binding domain.

It is generally believed that receptor activation occurs as a result of the N-terminal chemokine domain docking deep into the main binding pocket; a process thought to be a major contributor to the ligand-specific receptor activation mechanism (30, 31). CCL19 has a slightly more negative and relatively smaller (721 versus 649 Da) receptor docking domain compared with CCL21. Thus differences in size and distribution of electronegativity in the N-terminal chemokine domain probably dictate different receptor docking modes for CCL19 and CCL21.

To investigate possible differences in the docking of CCL19 and CCL21 during CCR7 activation, we introduced mutations in CCR7 in the main binding pocket, in areas previously identified to be involved in ligand binding. The impact of single point mutations for CCL19- and CCL21-induced receptor activation was measured in PI-turnover in HEK293 cells transiently transfected with CCR7 and chimeric protein G_qi4myr (Figure 7). Most chemokine receptors carry a Glu in the top of TM7 (position VII:06/7.39) – a residue that function as an anchor for positively charged residues in chemokines as well as in small molecules (32). We use the generic numbering system suggested by Schwartz, followed by the Ballesteros−Weinstein numbering system (33–35). As the only endogenous chemokine receptor, CCR7 carries a Tyr at this position (at the border between the major and minor binding pocket), and thus this residue could be of selective importance for CCR7 interaction with its ligands. Introduction of Glu (Y312E) resulted in a very high degree of constitutive activity, corresponding to 50% of E_max for CCL19 on wt CCR7, and completely abolished ligand-induced activation (Figure 7A). A subsequent substitution to Ala (Y312A) uncovered that this Tyr is not needed for ligand-induced activation, as both ligands induced CCR7 activity with wt-like potencies [CCL19: EC₅₀ of 5.4 nM (wt) and 7.8 nM (Y312A) and estimated CCL21: EC₅₀ of 156 nM (wt) and 77 nM (Y312A)] (Figure 7A). Like for Y312E, Y312A induced constitutive activity in CCR7 to a basal activity of 30% of E_max for CCL19 on wt CCR7, demonstrating that the Tyr at position VII:06/7.39 restricts basal CCR7 activity – a property that ensures efficient migration toward a chemokine gradient.

Closer to the cell surface, helix 7 in CCR7 contains an acidic residue (Asp309^VII:03/7.36) that could be envisioned to take over the usual role of Glu312^VII:06/7.39 as chemokine recognition motif. Mutation of D309 to A (Figure 7B), revealed that Asp is not needed at this position for neither CCL19 nor CCL21 interaction with CCR7, in fact substitution with Ala increased the potency of both ligands, possibly due to removal of a negative charge that could repel the mildly acidic N-terminus of these ligands.

Mutation of K130^III:02/73.24 to A, previously shown to affect both CCL19 and CCL21 signaling in CHO and COS-7 cells with up to 15-fold decreased potency (37), had no effect in our setup (Figure 7F). We used HEK293 cells and a different receptor construct (untagged CCR7 compared with C-terminally V5-his-tagged CCR7), possibly explaining this difference. Also in our model of CCR7, K130 points away from the binding pocket, and it is therefore not included in our pocket view model (Figures 7G,H).

Mutagenesis of K137^III:09/3.33 to A in the bottom of the major binding pocket, sheltered under ECL2 of TM3 has previously been shown to selectively lead to diminished CCL21-induced G protein signaling through CCR7 with up to 22-fold decrease in potency of CCL21 compared with less than 2.5-fold decrease for CCL19, whereas changing Q227^V:08,5.42 to A in the top of TM5 (same position on basolateral axis in major binding pocket, but further away from minor binding pocket than K137) affected both ligands with up to 20- and 28-fold decrease in potency for CCL19 and CCL21, respectively (37).

We confirm the data on K137A and Q227A (Figures 7C,F) and inferred that if this ligand selectivity reflects direct interaction differences, more residues in this receptor region would be similarly selective, and define a CCL21 recognition domain.

Focusing on the area of the major binding pocket delimited by TM3, -4, and -5, we mutated Glu193^IV:20/4.60 – another putative negatively charged CCR7 anchor point for CCL19 and CCL21 (in the absence of any impact of Tyr312 and Asp 309 in TM7). Intriguingly, mutating E193 to A selectively impairs CCL21 signaling, whereas it has the opposite effect (enforcement of signaling) on CCL19 (Figure 7D). Even the most conservative substitution to Asp (E193D) impaired CCR7 activation by CCL21 with no impairment of CCL19, indicating that the length of the negatively charged side chain matters, and that we manipulated a finely tuned CCL21 interaction domain. How K137 and E193 affect CCL21 signaling is still open to interpretation; however, as they are positioned close to each other (Figure 7H), they likely affect CCL21 as a pair.

Having established that the top of TM3, -4, and -5 seems more important for CCL21 than for CCL19, primarily by focusing on charged residues, we looked into an aromatic residue (Phe133 in position III:05/3.29) in TM3, located between the two identified lysines (K130 and K137) and pointing right into the binding crevice. When F133 was mutated to Ala (F133A), this too resulted in a CCR7 receptor selectively unresponsive to CCL21 (Figure 7E). Even substitution with a histidine that has pseudo-aromatic properties, selectively interfered with CCL21 signaling. This residue, neither as Phe nor His, could be an interaction partner of K137 or E193, and accumulating experimental evidence suggests that this small region constitutes a part of a selective and direct CCL21 recognition domain. Finally, amino acids shown to be important for CCL21 activation, without affecting CCL19 induced signaling, was shown to be of equal importance to tailless CCL21, indicating that these variants of CCL21 probably activate the receptor in the same way (Table 1).