Multiple Roles for the C-terminal Tail of the Chemokine Scavenger D6*

D6 is a heptahelical receptor that suppresses inflammation and tumorigenesis by scavenging extracellular pro-inflammatory CC chemokines. Previous studies suggested this is dependent on constitutive trafficking of stable D6 protein to and from the cell surface via recycling endosomes. By internalizing chemokine each time it transits the cell surface, D6 can, over time, remove large quantities of these inflammatory mediators. We have investigated the role of the conserved 58-amino acid C terminus of human D6, which, unlike the rest of the protein, shows no clear homology to other heptahelical receptors. We show that, in human HEK293 cells, a serine cluster in this region controls the constitutive phosphorylation, high stability, and intracellular trafficking itinerary of the receptor and drives green fluorescent protein-tagged β-arrestins to membranes at, and near, the cell surface. Unexpectedly, however, these properties, and the last 44 amino acids of the C terminus, are dispensable for D6 internalization and effective scavenging of the chemokine CCL3. Even in the absence of the last 58 amino acids, D6 still initially internalizes CCL3 but, surprisingly, exposure to ligand inhibits subsequent CCL3 uptake by this mutant. Progressive scavenging is therefore abrogated. We conclude that the heptahelical body of D6 on its own can engage the endocytotic machinery of HEK293 cells but that the C terminus is indispensable for scavenging because it prevents initial chemokine engagement of D6 from inhibiting subsequent chemokine uptake.

Inflammatory stimuli induce rapid, transient production of many chemokines that coordinate leukocyte recruitment by signaling through G-protein-coupled heptahelical receptors (7TMRs) 3

present on circulating leukocytes and other cells.
The human chemokine family, which contains 44 members, is subdivided into four subfamilies (CC, CXC, XC, and CX 3 C) based on variations of a cysteine motif, and each family is typically restricted to a specific group of 7TMRs. Among CC chemokine receptors (CCRs), CCR1 to CCR5 play prominent proinflammatory roles, and 19 of the 26 human CC chemokines can interact with at least one, and often more, of these receptors. This remarkable complexity helps ensure robust responses to a range of potential pathogens (1).
Resolution is a key component of normal protective inflammatory responses (2). This restores tissue homeostasis and prevents the persistent inflammation that lies at the heart of many destructive immunopathologies and promotes tumor formation (3). Resolution is aided by D6, a 7TMR related to CCR1-5, which binds 12 of the chemokine ligands for these receptors (4,5) and is expressed, at least in humans, by lymphatic endothelial cells, trophoblasts, and some leukocyte populations (6,7). D6 null mice show prolonged exaggerated responses to cutaneous inflammatory stimuli and tumor induction (8 -10), enhanced leukocyte infiltration during allergic lung inflammation (11), and increased sensitivity to experimentally induced fetal resorption (7). Conversely, epidermal D6 transgene expression suppresses cutaneous inflammation and tumorigenesis (10). Molecular insight has come principally from studies on D6-transfected cell lines, including human embryonic kidney (HEK) 293 cells that are widely used for 7TMR studies. In these cell lines, D6, unlike typical chemokine receptors (e.g. CCR5), can scavenge large quantities of extracellular chemokines (12,13). Thus, a simple model emerges whereby chemokine scavenging by D6 in vivo suppresses inflammatory leukocyte recruitment by reducing chemokine levels. Consistent with this model, elevated levels of bioavailable D6-binding chemokines are observed in inflamed D6 null mice (7)(8)(9)(10)(11).
Chemokine scavenging is thought to be dependent on rapid, constitutive trafficking of D6 to and from the cell surface. This enables iterative rounds of chemokine internalization through clathrin-coated pits without the need for signaling, with chemokine simply associating with surface D6 molecules destined for internalization (12)(13)(14). Consequently, Ͼ95% of D6 is located in early and recycling endosomal compartments where it shows remarkable stability, presumably by avoiding transit to lysosomes (12). D6 trafficking differs from signaling-competent chemokine receptors (and most other 7TMRs) that typically reside at the cell surface and become rapidly internalized only after activation by ligand. Extensive depletion of extracellular chemokines is limited by receptor desensitization and by receptor down-regulation which, in the case of CXC chemokine receptor 4 and other 7TMRs, is driven by monoubiquitination of intracellular lysines and subsequent passage to lysosomes for degradation (15)(16)(17).
Here we show that in HEK293 cells the Ct of human D6 controls phosphorylation, recycling, and stability of the receptor and drives constitutive association of GFP-tagged ␤-arrestins with membranes only at, and near, the cell surface. Surprisingly, these properties are dispensable for scavenging by D6 in these cells, and in our hands, the subcellular distribution of D6-GFP and its ability to internalize CC chemokine ligand (CCL) 3 are ␤-arrestin-independent in MEFs. Remarkably, even in the complete absence of the Ct, D6 still internalizes CCL3 into HEK293 cells when first exposed to the chemokine. However, progressive scavenging is not possible because exposure to CCL3 prevents subsequent CCL3 uptake. 14 amino acids (aa) from the membrane-proximal domain of the D6 Ct are sufficient to restore progressive scavenging, identifying the putative "eighth helix" as a critical domain in D6 function.
Cell Culture and Transfection-HEK293 cells were maintained and transfected as described previously (12), as were WT and ␤-arrestin null MEFs (provided by R. Lefkowitz, Howard Hughes Medical Institute, Durham, NC) (28,32,33). Stably transfected HEK293 pools were used throughout because transiently expressed D6, or D6-GFP, accumulates in cytoplasmic deposits in these cells causing considerable cell death. MEF transfection also led to some cell death, apparent from the presence of rounded, weakly adherent fibroblasts in the cultures. Confocal imaging of these cells showed they had unusual and inconsistent distributions of D6-GFP and MOR-YFP. Thus, images were only collected from cells that retained obvious fibroblastic morphology, and weakly adherent cells were removed by washing prior to flow cytometric analysis. To generate D6-Ala6, three sequential mutational steps were required using primer pairs 1-3. K324R and K142R were introduced individually into D6-Ala6, and the K324R primers were then used to introduce this mutation into D6-Ala6 K142R, generating D6-Ala6 K142R,K324R.

D6-Ala6
Pair 1 Sense , 5Ј GGC ACT GCC CAG GCC GCT TTA GCT AGC TGT  TCT GAC  Immunofluorescence and Confocal Microscopy-D6 immunofluorescence staining protocols have been described in detail previously (6,12). Cy3-coupled anti-mouse IgG antibodies (Sigma) were used to detect anti-D6 antibodies. Images were captured using a Leica SP-2 confocal microscope configured with Leica confocal software, with a ϫ40 or ϫ63 oil immersion objective and digital zoom. Fluorochromes were excited sequentially with lasers at 488 nm (GFP) or 543 nm (Cy3), and with a UV laser to excite 4,6-diamidino-2-phenylindole (DAPI) when appropriate. Images were superimposed using Leica confocal software and assembled using ThumbsPlus (Cerious Software). In all experiments at least 10 fields of cells were examined and representative images collected, with each image shown being one from a stack of up to five serial z-section images spanning the entire cell.
D6 Detection (Surface and Total)-Surface and total D6 expression was determined by flow cytometry and Western blotting, respectively, using anti-D6 antibodies as described (12). For analysis of MEFs, cells were incubated with mouse Fc block (BD Biosciences) prior to addition of anti-D6 antibodies. In all flow cytometry experiments, control samples were prepared in which the primary antibody (anti-D6) was omitted. The mean fluorescence intensity (MFI) of these samples was subtracted from the test data. Anti-D6 antibodies were detected with phycoerythrin (PE)-coupled anti-mouse IgG (Sigma) and horseradish peroxidase (HRP)-coupled antimouse IgG (Amersham Biosciences) in flow cytometry and Western blot protocols, respectively. Western blots were also probed with mouse anti-actin (Abcam) used at 1:2,000 dilution.
Phosphate Labeling and D6 Immunoprecipitation-ϳ4 ϫ 10 6 cells were incubated in 8 ml of phosphate-free medium for 1 h at 37°C; radiolabeled phosphate (2.5 mCi/ml; Amersham Biosciences) was added, and the cells were incubated at 37°C for a further 3 h. After washing in ice-cold PBS, cells were lysed in 500 l of CellLytic buffer (Sigma), and the lysate was cleared (20,000 ϫ g, 4°C, 10 min). D6 was then immunoprecipitated using anti-D6 antibodies as described (34).
D6 Stability Studies-Receptor stability studies were done as described previously (12). Briefly, a series of cultures each containing an equivalent number of cells (usually 10 6 ) were treated with or without cycloheximide (CHX; Sigma) (20 g/ml). Up to 24 h after CHX addition, cell lysates were prepared and analyzed by Western blotting.
BioCCL3 Scavenging Assay-5 ϫ 10 5 cells were incubated in 1 ml of complete medium containing 50 nM bio-CCL3 at 37°C and 5% CO 2 . 20 l samples of the medium were removed over time, added to 20 l of Laemmli buffer (Sigma), boiled for 5 min, and analyzed by Western blotting. Blots were blocked overnight in 10% milk/PBS; bioCCL3 was detected using HRPstreptavidin (Dako) in PBS, 0.1% Tween, developed using West Pico (Pierce), and exposed to x-ray film.
BioCCL3 Tetramer Uptake-BioCCL3/streptavidin-PE (bio-CCL3/S-PE) tetramers were prepared using 250 ng of bioCCL3 and 3 g of S-PE and used as described (12). Briefly, up to 10 6 cells were resuspended in 50 l of HEK293 medium plus 10 mM HEPES (pH 7.4) containing bio-CCL3/S-PE tetramers or 3 g of S-PE alone, and incubated at 37°C for 1 h with regular gentle agitation. Cells were washed in ice-cold FACS buffer (PBS plus 2% fetal calf serum) and analyzed on a FACScan flow cytometer (BD Biosciences). For each transfected cell type assessed, untransfected or mock-transfected cells, treated with bio-CCL3/S-PE tetramers or S-PE alone, were used as controls. On occasion, adherent cultured MEFs growing in 1 ml of medium were incubated for 1 h at 37°C in situ with bioCCL3/streptavidin-Cy3 (bioCCL3/S-Cy3) tetramers (prepared with 500 ng of bioCCL3 and 6 g of S-Cy3), washed, fixed, and visualized by confocal microscopy.
125 I-CCL3 Binding, Uptake, and Processing-These assays were done as described elsewhere (12,25,31). Briefly, to assess CCL3 binding to receptor, cells were incubated in 6 nM 125 I-CCL3 in the presence or absence of unlabeled competitor CCL3 (at various concentrations) at 4°C for 90 min. After washing with ice-cold PBS, the amount of radiolabel associated with the cells was determined and compared with that bound to cells in the absence of unlabeled CCL3 competitor. For uptake experiments, cells were surface-loaded in 12 nM 125 I-CCL3 at 4°C for 1 h, washed, shifted to 37°C for 10 min, and then washed again in either PBS or acid (0.2 M acetic acid, 0.5 M NaCl), both icecold, for 5 min. The proportion of radiolabel that became acidresistant after a shift to 37°C (indicative of internalization) was determined. To analyze intracellular CCL3 processing, cells were surface-loaded in 12 nM 125 I-CCL3 at 4°C for 1 h, washed, and shifted to 37°C. Up to 150 min later, medium and cells were harvested; the medium was subjected to precipitation in 12.5% trichloroacetic acid at 4°C, and the proportion of radiolabel in the cell pellet, trichloroacetic acid-precipitable (intact 125 I-CCL3 or peptide fragments thereof), and trichloroacetic acid-nonprecipitable (degraded) fractions was determined. In all assays, each time point or condition was performed in triplicate.
Helical Prediction-The entire intracellular C-terminal tail of D6 (from four species) and human CCR1-5 were analyzed for putative helical content using AGADIR software (35).
Statistics-Data were analyzed using GraphPad Prism software applying unpaired t tests.

RESULTS
␤-Arrestins Localize to the Cell Periphery in D6-expressing HEK293 Cells-GFP-tagged ␤-arrestin-1 is reported to co-localize with D6-RFP throughout transfected rat RBL-2H3 cells (14). We investigated the impact of untagged human wtD6 expression on ␤-arrestin distribution in human HEK293 cells. In parental HEK293 cells, GFP-tagged versions of ␤-arrestin-1 or -2 were found throughout the cytosol (Fig. 1A). In contrast, in HEK293 cells expressing wtD6, although some green fluorescence remained uniformly distributed throughout the cytosol, a significant proportion was concentrated at the cell surface and associated with vesicles predominantly at, or just beneath, the plasma membrane ( Fig. 1, B-I). This was clearly visible in optical sections taken through the middle (Fig. 1, B, D, and E), top (Fig. 1C), or bottom ( Fig. 1, F-I) of the cells. GFP-tagged ␤-arrestins showed some co-localization with D6 at these locations, although the receptor was of very low abundance in many of the ␤-arrestinϩ vesicles and the majority of cellular D6 remained ␤-arrestin-free ( Fig. 1, F-I). The peripheral localization of ␤-arrestins in wtD6-expressing HEK293 cells was not detectably altered by incubation with CCL3. Thus, in HEK293 cells, D6 mediates the re-localization of ␤-arrestins exclusively to the cell periphery.
Loss of Constitutive D6 Phosphorylation Prevents ␤-Arrestin Re-localization-Activated 7TMRs associate with ␤-arrestins principally via phosphorylated Ct tails. The human D6 Ct harbors a serine-rich domain that is conserved in D6 proteins from other mammals ( Fig. 2A). We previously reported that wtD6 is constitutively phosphorylated in murine L1.2 cells (34). This is also the case in HEK293 cells; anti-D6 immunoprecipitates from 32 P-labeled HEK293 cells only contained a labeled protein of equivalent molecular weight to D6 when the cells expressed wtD6 (Fig. 2B). As in L1.2 cells (34), there was no quantitative change in phosphorylation after CCL3 exposure (data not shown). We hypothesized that constitutive phosphorylation of the serine-rich motif in the D6 Ct is responsible for re-localizing ␤-arrestins in HEK293 cells. To test this, we mutated all six serines in the serine-rich motif to alanine (D6-Ala6) ( Fig. 2A) and expressed the mutant stably in HEK293 cells. Flow cytometry revealed that, as with wtD6-expressing HEK293 cells, Ͼ95% of cells expressed surface D6-Ala6. However, wtD6 was ϳ2.5-fold more abundant on the surface than D6-Ala6 (Fig.  2C), and a similar difference in expression level was also observed when the total cellular pool was analyzed by Western blotting (Fig. 2D). These data indicated that the surface/cytoplasmic ratio of D6 was not markedly affected by mutation of the six serines to alanines, and consistent with this, D6-Ala6, like wtD6 (12), was abundant in vesicles throughout the cytoplasm (Fig. 2E). Phosphate labeling studies showed that D6-Ala6 had lost the constitutive phosphorylation present on wtD6 (Fig. 2F). This is consistent with the noticeable reduction in apparent molecular weight of D6-Ala6 compared with wtD6 by Western blotting (Fig. 2D). Significantly, D6-Ala6 was completely unable to drive GFP-tagged ␤-arrestins to the cell periphery (Fig. 2G), and this was not altered in the presence of D6 ligand, CCL3.
To investigate this further, two truncated D6 variants were expressed stably in HEK293 cells. These were D6-360, lacking the last 24 aa but retaining the serine-rich domain, and D6-340, in which the last 44 aa are removed to include the serine-rich domain ( Fig. 2A). D6-360 was less well expressed than wtD6 (to levels equivalent to D6-Ala6), and whereas Ͼ95% of cells expressed surface D6-340, there was a considerable reduction in the quantity of total and surface D6-340 expressed (Fig. 2, H  and I). However, as with D6-Ala6, the surface/cytoplasmic ratio was not dramatically affected by truncation, and the truncated proteins were present in vesicles throughout the cytoplasm (Fig. 2J). Consistent with our D6-Ala6 analysis, D6-340 was not detectably phosphorylated (data not shown) and was unable to drive ␤-arrestin-2GFP accumulation at the cell surface, whereas D6-360 behaved like wtD6 (Fig. 2K). Collectively, these data indicate that constitutive phosphorylation of D6, requiring the serine cluster in the Ct, mediates ␤-arrestin re-localization to membranes at the cell periphery in HEK293 cells.
The Serine-rich Domain of the D6 Ct Regulates Receptor Stability and Intracellular Trafficking Decisions-wtD6 is very stable in HEK293 cells, with little change in protein level in cells treated for 24 h with the protein synthesis inhibitor CHX (12). The mutants we had analyzed, particularly D6-340, were less well expressed in HEK293 cells than wtD6 so we examined whether receptor stability had been compromised by mutation (Fig. 3). This revealed that D6-360 had equivalent stability to wtD6, but that D6-340 and D6-Ala6 were considerably less stable and had nearly completely disappeared within 4 h of CHX treatment (Fig. 3A).
Loss of stability might indicate a preference for D6 mutants to traffic to lysosomes after internalization (via early then late endosomes), rather than passing through the early and recycling endosomal compartments like wtD6 (12). To investigate this, we introduced GFP-tagged versions of wt-Rab5 and wt-Rab7 (which mark early and late endosomes, respectively) into cells expressing wtD6 or D6-Ala6, detected by immunofluorescence (Fig. 3B). Consistent with our previous work (12), wt-Rab5-GFP showed extensive co-localization with the intracellular pool of wtD6. D6-Ala6, however, showed markedly less co-localization with wt-Rab5-GFP. Conversely, wt-Rab7-GFP showed minimal co-localization with wtD6 but often co-localized with D6-Ala6ϩ vesicles. Next, we overexpressed constitutively active GFP-tagged forms of Rab5 and Rab7 (Q79L-Rab5 and Q67L-Rab7) that disrupt protein passage through early or late endosomes, respectively. Proteins that normally traffic through these compartments should co-localize extensively with these mutant Rabs. Expression of Q79L-Rab5-GFP led to the accumulation of both wtD6 and D6-Ala6 in enlarged GFPϩ vesicles (Fig. 3C). However, whereas wtD6 rarely co-localized with Q67L-Rab7-GFP, D6-Ala6 showed extensive association with vesicles carrying this mutant Rab (Fig. 3C). D6-340, the other mutant with low stability, behaved like D6-Ala6, whereas D6-360 was like wtD6. Thus, the serine cluster, in addition to mediating ␤-arrestin localization to the cell periphery, is  MARCH 21, 2008 • VOLUME 283 • NUMBER 12 JOURNAL OF BIOLOGICAL CHEMISTRY 7975 required to prevent wtD6 from entering Rab7ϩ late endosomes, ensuring repeated recycling to the cell surface rather than degradation.

Role of the C-terminal Tail of D6
The Reduced Stability of D6-Ala6 Is Dependent on Intracellular Lysine Residues-Chemokine receptors and other 7TMRs can be targeted to lysosomes by ubiquitination of lysine residues (15)(16)(17). There are two intracellular lysines in D6-Ala6 and D6-340 that are conserved across mammalian D6 sequences. These are at aa 324 and, interestingly, aa 142 within the DKYLEIV motif unique to D6. These residues in D6-Ala6 were mutated, individually or together, to arginine to retain charge but lose ubiquitination potential. Remarkably, either mutation alone was sufficient to reverse the low stability of D6-Ala6 (Fig.  3D). Identical mutations in D6-340 were equally effective at restabilizing this molecule (data not shown). Although we have not been able to reproducibly detect ubiquitination of D6-Ala6 or D6-340, our data suggest that preventing D6 phosphorylation allows ubiquitination of lysine residues and subsequent trafficking to lysosomes for degradation.
Chemokine Scavenging Does Not Require Constitutive Phosphorylation or the Last 44 Amino Acids-High receptor stability, recycling, and the recruitment of ␤-arrestins to the cell surface might be anticipated to contribute to effective chemokine scavenging by D6. However, the surface/cytoplasmic ratios of D6-Ala6 and D6-340 were broadly similar to wtD6 and D6-360, indicating that all mutants, like wtD6, were capable of constitutive internalization. Thus, we next compared the ability of wtD6, D6-360, D6-340, and D6-Ala6 to scavenge chemokines in continuous culture over time (Fig.  4). Remarkably, despite their clear biochemical differences, all mutants were as effective as wtD6 at scavenging bioCCL3, with only some slight slowing of scavenging noticeable by D6-340. Thus, in HEK293 cells, residues after aa 340, including the serine-rich domain and the charged extreme end of the mole- cule, are dispensable for D6 internalization and progressive chemokine scavenging.

GFP-tagged D6 Distributes Similarly within Wild-type and ␤-Arrestin Null Fibroblasts and Mediates
CCL3 Uptake-The results above suggested that D6 trafficking is ␤-arrestin-independent in HEK293 cells, consistent with our previous observations using dominant-negative ␤-arrestin constructs (12). To investigate this further, we transiently expressed D6-GFP (D6 with a Ct GFP tag (12)) in WT and ␤-arrestin null MEFs. Confocal microscopy of fixed adherent fibroblasts revealed that D6-GFP was predominantly localized inside these cells, irrespective of their genotype (Fig.  5A). Also, using flow cytometry and anti-D6 antibodies, WT and ␤-arrestin null MEFs expressing equivalent quantities of D6-GFP displayed similar levels of surface receptor (Fig. 5B). Thus, the characteristic distribution of D6-GFP seen in HEK293 (12) is also seen in MEFs and is unaffected by the absence of ␤-arrestins. In addition, adherent WT and ␤-arrestin null MEFs expressing D6-GFP were able to direct internalization of bioCCL3/ S-Cy3 tetramers, visualized by optical sectioning of D6-GFPϩ cells by confocal microscopy (Fig. 5A). S-Cy3 alone was not internalized by D6-GFP-transfected or untransfected MEFs (data not shown).
To quantify BioCCL3 tetramer uptake, WT and null MEFs were transfected with D6-GFP, fed bioCCL3/S-PE tetramers or S-PE alone at 37°C, and analyzed by flow cytometry (Fig. 5C). BioCCL3 tetramers, but not S-PE alone, were readily internalized into both cell types, in a manner directly proportional to the level of expression of D6-GFP. By assessing WT and null MEFs gated for equivalent expression of D6-GFP, it was clear that the absence of ␤-arrestins did not compromise bioCCL3 tetramer uptake (Fig. 5D).
As controls, MEFs transiently expressing the ␤-arrestin-dependent MOR-YFP were imaged with or without exposure to its agonist DAMGO (Fig. 5E). Unlike D6-GFP, MOR-YFP is located principally on the surface of MEFs, independent of cell genotype, but its activation only leads to its intracellular accumulation in WT MEFs.
The Membrane-proximal Domain of the Ct Is Sufficient to Allow the Heptahelical Body of D6 to Mediate Continuous CCL3 Scavenging-D6-340 carries a short Ct that could conceivably coordinate trafficking events required for scavenging. Thus, we generated another D6 mutant, D6-326, which lacks  the Ct from the point at which it diverges from other chemokine receptors, therefore ending with the aa KAF.
In many respects, D6-326 behaved like D6-340 (Fig. 6). Total and surface D6-326 expression was lower than that seen in wtD6 transfectants (but similar to cells expressing D6-340 (Fig.  2, H and I)). The surface/cytoplasmic ratio seemed unaffected by truncation indicating that constitutive trafficking was retained (Fig. 6, A and B). Consistent with this, considerable amounts of D6-326 were inside cells (Fig. 6C). Like D6-340, D6-326 failed to drive ␤-arrestin-2GFP to the cell periphery (Fig. 6D), co-localized to some extent with wt-Rab7-GFP, occasionally with wt-Rab5-GFP, and extensively with Q79L-Rab5-GFP and Q67L-Rab7-GFP, and it was less stable than wtD6 (data not shown). D6-326 had a similar affinity for CCL3 as wtD6 (according to 125 I-CCL3 displacement assays performed at 4°C (Fig. 6E)) and was able to internalize surface-loaded 125 I-CCL3 in short term uptake assays, albeit less effectively than wtD6 but similar to D6-340 (Fig. 6F). Little of this internalized 125 I-CCL3 was released from cells in a trichloroacetic acid-precipitable intact form and was either degraded prior to release or retained inside the cells (Fig. 6G). Surprisingly however, despite all the similarities to D6-340, D6-326 was unable to mediate progressive depletion of bioCCL3 in continuous culture, with only some initial reduction in bioCCL3 levels being observed at 2 h (Fig. 6H).
From these data, we hypothesized that D6-326 was initially competent for CCL3 uptake but that chemokine incubation prevented continuous CCL3 scavenging. To investigate this, we incubated wtD6 or D6-326 expressing cells with CCL3 for 1 h, washed the cells extensively, and then examined their ability to mediate CCL3 tetramer uptake (Fig.  6I). As we observed previously, CCL3 incubation enhances surface expression of wtD6 (12) and its ability to internalize bioCCL3/S-PE tetramers. In contrast, CCL3 incubation completely abrogated the ability of D6-326 to internalize CCL3 tetramers, despite no significant change in surface receptor expression levels. Thus, the membrane-proximal region (aa 326 -340) of the D6 Ct is sufficient to allow continuous chemokine scavenging by the body of the receptor, whereas deletion of the entire D6 Ct creates a molecule whose chemokine scavenging capabilities are limited by incubation with chemokine.

DISCUSSION
D6 scavenges extracellular pro-inflammatory CC chemokines and suppresses inflammation and tumorigenesis. Here we have shown that in human HEK293 cells the Ct of human D6 controls the constitutive phosphorylation, high stability, intracellular trafficking itinerary, and chemokine scavenging properties of the receptor and drives GFP-tagged ␤-arrestins to membranes at, or near, the cell surface. Surprisingly, however, the D6 Ct is only indispensable for scavenging because it prevents chemokine engagement of the heptahelical D6 body from inhibiting subsequent chemokine uptake.
Transfected human wtD6 is constitutively phosphorylated in HEK293 cells, as it is in L1.2 cells (a murine B cell line) (34). In HEK293 cells, this is dependent on a cluster of serine residues in the Ct, which are also required to direct ␤-arrestins to the cell surface. Based on precedent with many other 7TMRs (18,23), the simplest interpretation of these data is that the serines in the Ct cluster are themselves the target of phosphorylation and that this permits direct physical interaction with ␤-arrestin. If so, the localization of ␤-arrestins is likely to directly correlate with the localization of phosphorylated D6, i.e. at or just under the cell surface, indicating the following: (i) wtD6 undergoes transient phosphorylation as it transits the cell surface, and (ii) only a small fraction of the total cellular wtD6 protein is phosphorylated at any one time. It appears as though no specific serine residue from within the cluster is required because mutants in which only the 1st, 2nd, or 3rd two serines, or the first or last four serines, were mutated to alanine behaved like wtD6 in all assays (data not shown).
␤-Arrestin re-localization was not required for D6 internalization in HEK293 cells, but interestingly, all mutated versions of D6 unable to re-localize ␤-arrestins (i.e. D6-Ala6, D6-340, and D6-326) showed markedly reduced stability and were able to traffic to late endosomes. The reduced stability of these mutants was dependent on either of two lysine residues of the intracellular surface of D6, including the one in the unique DKYLE motif in the second intracellular loop. Although we have been unable to reproducibly demonstrate ubiquitination of unphosphorylated mutants, our data are consistent with a model in which D6 phosphorylation and the presence of ␤-arrestins at the cell surface interferes with D6 ubiquitination and subsequent trafficking to the late endosomal compartment en route to lysosomes. D6 may therefore use ␤-arrestins to enhance receptor stability rather than direct internalization.
Constitutive phosphorylation, high stability, and exclusive trafficking through recycling endosomes were all dispensable for effective chemokine scavenging by D6 in HEK293 cells. However, the regulatory potential of these phenomena may have been masked by the very high levels of D6 transcription in our transfected cell lines. The fact that D6-340 was nearly as effective as wtD6 in scavenging assays (despite being 5-6-fold less abundant) indi- cates that the level of receptor expression was not particularly rate-limiting in our wtD6 transfectants. In cells expressing much lower levels of D6, a small change in stability may have a more profound impact on scavenging potential. Moreover, D6 stability will influence the effectiveness of transcriptional regulation of D6. We have shown that inhibiting production of new protein for as long as 24 h has minimal impact on phosphorylated D6 levels, whereas 4 h is sufficient to nearly completely remove unphosphorylated D6. Thus, by controlling D6 phosphorylation in concert with its transcription, it should be possible to effectively manipulate D6 protein levels to modulate scavenging. This possibility needs to be explored in cells expressing endogenous wtD6.
Our observations contrast those of Galliera et al. (14). First, they reported that all RFP-tagged human D6 (D6-RFP) co-localized with GFP-tagged ␤-arrestin-1 in RBL-2H3 cells, perhaps indicating that phosphorylation of human D6 is different in this rat basophilic leukemia cell line. However, it should be noted that in the images presented, GFP-tagged ␤-arrestin-1 did not appear to be specifically associated with D6-RFPϩ vesicles, and green fluorescence from these vesicles was equivalent to that seen in the surrounding cytosol. Second, wtD6 was not detectably phosphorylated in Chinese hamster ovary (CHO)-K1 cells. Cell background might influence the ability of D6 to be phosphorylated as it transits the cell surface, and it will be of interest to examine the stability of wtD6 in CHO-K1 cells and its ability to drive ␤-arrestin re-localization. However, it is notable that the amount of receptor successfully immunoprecipitated was not shown (14), and low level phosphorylation may have been missed. Third, the charged domain carried in the last 28 aa of D6 was required for its internalization into CHO-K1 cells (gauged by antibody uptake) (14), but we have found that it is dispensable for internalization and scavenging in HEK293 cells. Cell type and its species of origin may be responsible for this difference. For example, the charged end of human D6 may be required for successful engagement of the endocytotic machinery of ovarian hamster cells. It may also be capable of driving endocytosis in human HEK293 cells, but other regions can clearly compensate for its absence. The same argument could be made for D6-driven ␤-arrestin re-localiza-tion; this phenomenon may have the potential to contribute to D6 endocytosis in some scenarios, but it can clearly be compensated for in HEK293 cells by membrane-proximal Ct motifs or determinants in the D6 body. This putative redundancy makes some sense; multiple alternative internalization mechanisms could conceivably provide flexibility to ensure robust chemokine scavenging by D6 in a range of different cellular environments.
In our hands, the subcellular distribution of D6-GFP, and its capacity to internalize bioCCL3 tetramers, is independent of ␤-arrestins in MEFs. However, Galliera et al. (14) reported that D6-RFP was only found on the surface of ␤-arrestin null MEFs and that ␤-arrestin-1-GFP could drive its redistribution to intracellular locations, an effect dependent on the last 28 aa of D6. These contradictory observations are difficult to reconcile. However, it may be significant that we imaged cells after adherence to fibronectin-coated coverslips for at least 24 h, whereas Galliera et al. (14) plated cells onto glass 1 h prior to imaging. We have observed rounded MEFs with minimal fibroblastic morphology in our transfected MEF cultures, and these cells show variable patterns of D6-GFP distribution, occasionally with strong surface expression. An alternative explanation is that the use of different fluorescent tags may influence the conformation of the D6 Ct to alter the specific determinants required for internalization. Although these conflicting observations require future resolution, it is nonetheless clear that in human HEK293 cells untagged human D6 can effectively scavenge CCL3 in a manner unaffected by dominant-negative ␤-arrestins (12), and independently of its ability to redistribute ␤-arrestins.
Despite the differences between our study and that of Galliera et al. (14), we agree that WT human D6 has the potential to constitutively drive the re-localization of ␤-arrestins. The possible implications of this for D6 have been discussed above, but it would also seem likely that it will have implications for 7TMRs co-expressed with D6. The subcellular localization of ␤-arrestins is of profound importance to their function as 7TMR regulators and signaling scaffolds. Constitutive localization of these proteins to membranes in D6-expressing cells (be it in the cell periphery (this study) or throughout the cell (14)) would be predicted to influence signaling, trafficking, and stability of ␤-arrestin-regulated 7TMRs co-expressed with D6. If so, it is conceivable that the phenotypes observed in D6 null mice may, at least to some extent, be the result of changes in 7TMR behavior and not exclusively because of loss of chemokine scavenging.
The 14-aa difference between D6-326 and D6-340 is sufficient to allow effective chemokine scavenging into HEK293 cells. D6-326 can internalize some CCL3 after initially encountering the chemokine, but exposure to CCL3 limits subsequent chemokine scavenging without markedly affecting surface FIGURE 7. The membrane-proximal region of the D6 Ct carries an elongated putative 8th helix with a conserved leucine-rich surface. Alignment of the membrane-proximal ends of the D6 Ct from four species and of human CCR1-5. The location of the predicted 8th helix is indicated in boldface in all sequences, and conserved leucines are underlined. Serine residues in the serine-rich region of D6 are italicized and underlined; the predicted end of the seventh transmembrane helix (TM7) is indicated; and the position of truncation in D6-326 and D6-340 is marked with arrows. A helical whorl of human D6 (right) reveals a tri-leucine surface. The letters in the whorl become smaller and the lines thinner toward the C-terminal end. Despite CCR1, -4, and -5 carrying two leucines, these are inappropriately spaced for alignment on a helical surface. The conserved serine (in D6) and glycine (in CCR1-5) at the N-terminal end of the helices are indicated with asterisks.
receptor expression. On the other hand, wtD6-mediated CCL3 uptake and surface wtD6 expression are consistently enhanced by CCL3 exposure (Fig. 6) (12). These data suggest the following: (i) wtD6 and D6-326 are both capable of transmitting CCL3dependent signals that modify scavenging; (ii) the heptahelical body of D6, on its own, is capable of engaging the endocytotic machinery of HEK293 cells, and (iii) the D6 Ct is only indispensable for scavenging because it prevents CCL3 from inhibiting subsequent chemokine uptake. Interestingly, the membrane proximal region of the D6 Ct harbors a putative 8th helix that will be disrupted in D6-326 but intact in D6-340 (Fig. 7). An 8th helix is present in the crystal structure of bovine rhodopsin, anchored by the NPXXY motif at the end of the 7th helix (36,37) and undergoes a marked shift in location after rhodopsin activation (38,39). The recent crystal structures of the human ␤ 2 -adrenergic receptor have an 8th helix underlying the membrane perpendicular to the transmembrane helices which is stabilized by a hydrophobic surface (40 -43). In fact, an 8th helix is thought to be present in most, if not all, rhodopsin-like 7TMRs and play a role in signaling by directly interacting with heterotrimeric G-protein subunits (43)(44)(45)(46)(47)(48)(49), although roles in ligand binding have also been reported (50). The 8th helix of D6 (Fig.  7) is predicted to be the following: (i) longer than that found in CCRs1-5; (ii) have a hydrophobic surface (FFY) that may associate with membrane; (iii) carry a trileucine surface, a possible protein-protein interaction motif; and (iv) be directly preceded by a conserved serine residue (a potential site for phosphorylation) rather than the glycine found in CCR1-5 and other chemokine receptors. We propose that the integrity of the 8th helix is critical for progressive scavenging by D6-340, possibly by interacting with regulatory proteins present in HEK293 cells. Future work is required to explore the role of the 8th helix in the context of full-length D6, and to characterize the "signals" emanating from wtD6 and D6-326 after chemokine engagement that modify their scavenging capabilities. These studies will provide further molecular insight into scavenging by this indispensable anti-inflammatory, tumor suppressor protein.