The Transmembrane Domains of L-selectin and CD44 Regulate Receptor Cell Surface Positioning and Leukocyte Adhesion under Flow*

During inflammation and immune surveillance, initial contacts (tethering) between free-flowing leukocytes and the endothelium are vitally dependent on the presentation of the adhesion receptor L-selectin on leukocyte microvilli. Determinants that regulate receptor targeting to microvilli are, however, largely elusive. Therefore, we systematically swapped the extracellular (EC), transmembrane (TM), and intracellular (IC) domains of L-selectin and CD44, a hyaluronan receptor expressed on the cell body and excluded from microvilli. Electron microscopy of transfected human myeloid K562 cells showed that the highly conserved TM domains are responsible for surface positioning. The TM segment of L-selectin forced chimeric molecules to microvilli, and the CD44 TM domain evoked expression on the cell body, whereas the IC and EC domains hardly influenced surface localization. Transfectants with microvillus-based chimeras showed a significantly higher adhesion rate under flow but not under static conditions compared with cells with cell body-expressed receptors. Substitution of the IC domain of L-selectin caused diminished tethering but no change in surface distribution, indicating that both microvillus positioning and cytoskeletal anchoring contribute to leukocyte tethering. These findings demonstrate that TM domains of L-selectin and CD44 play a crucial role in cell adhesion under flow by targeting receptors to microvilli or the cell body, respectively.

Neutrophil invasion at sites of inflammation as well as lymphocyte recirculation from blood into secondary lymphoid tissue require that free-flowing leukocytes become activated and subsequently exit the bloodstream. At the molecular level, the cells undergo a closely orchestrated cascade of interactions with the endothelium, each step consisting of a characteristic pattern of receptors, their ligands, cytokines, and chemoattractants (1,2).
Both L-selectin and CD44 are type I integral membrane glycoproteins expressed on most circulating leukocytes. L-selectin knock-out mice show a profound defect of lymphocyte accumulation in secondary lymphatic tissue and of neutrophils at sites of inflammation (11). CD44 is involved in many physiological processes, including lymphocyte activation, adhesion, and proliferation (12). There is also increasing evidence that CD44 and L-selectin contribute to tumor metastasis (13,14).
Under many physiologic conditions, L-selectin initiates the first step of the adhesion cascade (15,16). While surveying the vascular bed, rapidly flowing leukocytes face the challenge to establish contact to static ligands of the endothelial layer at high velocities. Presentation of adhesion receptors on microvilli as the very distal aspects of the cell may be a functional adaptation to facilitate adhesion under high shear stress. Indeed, L-selectin transfectants in murine pre-B cells showed impaired tethering capability under flow when expressed on the cell body using a chimera consisting of the ectodomain of L-selectin linked to CD44 (4). This concept was further confirmed by adhesion assays using the same transfectants in lymphoid tissue in vivo (17). These studies established a close relationship between receptor surface topography and function, thus attaching great importance to the microvillus compartment in the context of leukocyte recruitment.
Although leukocyte surface structures including microvilli, ruffles, filopodia, and lamellipodia are well described (18), a detailed biochemical understanding of their receptor composi-tion is still largely missing. To date, it is well recognized that cytoplasmic interactions can regulate protein sorting (19). L-selectin distribution on microvilli is abolished upon mutation of its intracellular ezrin/radixin/moesin (ERM) binding site (20). The tail-truncated L-selectin mutant L⌬cyto containing only the membrane-proximal ERM binding site of the intracellular (IC) domain still localizes to microvilli (21). As a consequence, cytoskeletal linkage to ERM is commonly believed to be a main determinant of surface localization. However, CD44 links to ERM proteins, too (22), yet shows an inverse surface expression pattern on nonprotrusive sections. Hence, definitive roles for protein domains in trafficking receptors to or away from microvilli have not been established. To investigate the process and implications of receptor segregation on leukocytes, we conducted domain swapping studies between L-selectin (microvilli) and CD44 (cell body) and evaluated the separate contributions of their extracellular (EC), transmembrane (TM), and IC domains on ultrastructural receptor localization and receptor functionality. Surprisingly, we found that the membrane-spanning domain contains the entire recognition pattern for specific receptor surface segregation on microvilli or the cell body, which in turn crucially affected the efficiency of leukocyte adhesion under flow.
Recombinant DNA-Full-length cDNA of human L-selectin and CD44 was used as template for domain swapping employing a splicing by overlap extension PCR protocol (26). In brief, the desired gene region, either only the IC or in combination with the TM domain, were amplified using primers that include an overlap extension site. A second PCR step then allowed linkage of the amplificate to the respective ectodomain. All constructs were subcloned into the eukaryotic vector pIRES2-EGFP (Clontech, Mountain View, CA) containing a cytomegalovirus promoter. Nucleotide sequence authenticity was verified by DNA sequencing.
Cell Culture and Transfection-K562 human myeloid leukemia cells (CCL 243, American Type Culture Collection, Rockville, MD) were transfected by electroporation and cultured according to standard protocols. Cell transfectants were sorted on a FACSAria cell sorter (Becton Dickinson) and resorted periodically. This cell line was chosen because it does not express either endogenous CD44 or L-selectin. Preliminary data regarding microvillus positioning of CD44, L-selectin and the chimera LCL were also obtained using the human pre-B cell line Nalm-6 (data not shown).
Flow Cytometry and Immunoelectron Microscopy-Cells were fluorescently labeled using a standard protocol and analyzed on a FACSCalibur flow cytometer (BD Biosciences). The pre-embedding immunolabeling technique has been described previously (27). Briefly, 2 ϫ 10 6 cells were incubated with the primary antibody at a dilution of 1:10 in phosphate-buffered saline ϩ 1% fetal calf serum at 4°C for 10 min. The secondary antibody conjugated to 10-nm gold particles was added at a dilution of 1:30 for 10 min at 4°C followed by fixation and dehydration. Thin sections were cut on a Reichert Ultracut microtome and stained with 2% uranyl acetate and 2% lead citrate. Images were taken using a Zeiss EM10 microscope. Analysis of the 10-nm colloidal gold distribution was performed by a blinded observer at a magnification of ϫ30,000 -60,000. For each transfectant, 20 -30 cells were analyzed with a total of 700 -1,400 particles. A mock transfectant as negative control did not show any staining.
Static Adhesion assay-Hydrophobic glass slides were coated with 2 mg/ml hyaluronan or 30 g/ml PSGL-1 and blocked with bovine serum albumin. A 10-ml cell suspension (5 ϫ 10 6 cells/ml, Hanks' balanced salt solution medium for L-selectin, RPMI 1640 medium for CD44) was added and incubated at 37°C for 15 min. The number of adhesive cells was assessed with an inverted microscope Zeiss Axio 100. Cells incubated with 10 g/ml blocking mAb for 10 min at 37°C were used as controls.
Adhesion Assays under Flow-Adhesion under flow was investigated using a parallel plate flow chamber from GlycoTech (Gaithersburg, MD) placed on a 35-mm polystyrene dish and connected to a high precision perfusion pump (Harvard Apparatus, Holliston, MA). An inverted microscope Axio 100 (Zeiss, Jena, Germany) was used at ϫ10 magnification equipped with the digital high performance camera SensiCam QE (Cooke Corporation, Romulus, MI). As a specificity control, cells were incubated for 10 min at 37°C with 10 g/ml blocking antibody prior to the experiment.
L-selectin Rolling on PSGL-1-Each dish was coated with 150 g/ml recombinant PSGL-1-Fc for 2 h and then blocked with 2 mg/ml bovine serum albumin for 1 h. Cells were resuspended in Hanks' balanced salt solution containing 1.26 mM Ca 2ϩ (Lonza, Basel, Switzerland) with a final concentration of 1 ϫ 10 6 cells/ ml. Rolling assays were started at 2.3 dynes/cm 2 , and shear stress was gradually decreased to 1.7, 1.3, and 0.7 dynes/cm 2 . Three different fields of view were recorded for a duration of 20 s and extrapolated to 1 min. Velocity determination was performed using a MATLAB (version 7.5.0, R2007B) cell tracking software by John Pickard.
CD44 Adhesion on Hyaluronan-Soluble hyaluronan was plated at 2.0 mg/ml in phosphate-buffered saline, incubated overnight at 37°C, and subsequently blocked with 2 mg/ml bovine serum albumin in phosphate-buffered saline for 1 h. Cells (1 ϫ 10 6 /ml) were washed and resuspended in RPMI 1640 medium. After 7 min of flow, pictures of three different fields of view were taken.

The Intracellular Tail Is Not Sufficient for Defining Receptor
Surface Topology-L-selectin (CD62L) and CD44 show a clearly distinct cell surface topology on microvilli and the cell body, respectively. Because published data mainly emphasize cytoplasmic interactions in determining surface positioning, we first set out to address the question of whether the IC domains of L-selectin and CD44 affect compartmentalized surface distribution.
We generated domain-swapped chimeras between the cytoplasmic domains of human L-selectin and CD44 (Fig. 1, A and  B), referred to by a three-letter code, each letter representing a protein domain starting at the N-(extracellular) and ending at the C terminus (intracellular). The human myeloid cell line K562 (L-selectin-and CD44-negative) was stably transfected, and successful expression was verified by flow cytometry (Fig. 1, C and D) and SDS-PAGE (data not shown). The surface topography of wild type and swapped receptors transfected into myeloid cells was then assessed by immunogold labeling and transmission electron microscopy. No morphologic difference of the microvillus architecture was observed among the transfectants.
Consistent with previous findings (4), 88% of L-selectin (LLL) was associated with microvilli, whereas CD44 (CCC) concentrated on the planar cell body. Surprisingly, the substitution of either cytoplasmic tail did not have significant effects on surface topology. LLC was still positioned predominantly on microvilli, and CCL was found on the cell body at a degree similar to the respective wild types (Fig. 2).
The Transmembrane Domain Governs Receptor Surface Presentation-The unchanged distributions of LLC and CCL compared with LLL and CCC suggest that the motif for the surface localization of L-selectin and CD44 may be encoded in their outer segments. Previous work had established that swapping the EC domains alone had no effect (4). We confirmed this finding because the chimera LCC remained on the planar cell body and CLL was expressed on microvilli in a ratio similar to CCC and LLL, respectively (Fig. 2). These findings in conjunction with the result that swapping of the IC domain did not have significant effects on cell surface topology suggest that the TM domain may be a major determinant of L-selectin and CD44 surface presentation.
To examine this hypothesis, we generated chimeras with a replaced TM but unmodified IC and EC segment (Fig. 1A). Indeed, substitution of the L-selectin TM domain (LCL) provoked a significant loss of the microvillus localization, with 80% of all labeled proteins found on the planar cell body and excluding microvilli (Fig. 2), similar to CD44 typical surface distribution. Likewise, CLC was hardly found on the planar cell body but showed a major shift to the microvillus compartment.
Domain Swapping Does Not Alter Static Adhesion-To address the functional implications of surface localization, we established cell transfectants with tightly matched surface expression levels, which was successful for all constructs except LCL (Fig. 1, C and D). Static binding of wild type L-selectin and chimeras was assessed using immobilized PSGL-1, a known ligand for L-selectin (28). Cells were allowed to settle on coated hydrophobic glass slides, and the number of bound cells was counted. No significant difference in binding activity between the N-terminal ligand-binding lectin domains of L-selectin wild type and chimeras could be detected (Fig. 3A). Similarly, the static adhesion activity of the CD44 transfectants was investigated using its physiologic ligand hyaluronan (29), showing no evidence for differential adhesion (Fig. 3B).
Microvillus Receptor Presentation Enhances Adhesion under Flow-L-selectin supports fast leukocyte rolling on PSGL-1 (30), whereas CD44 mediates very slow rolling on hyaluronan (29). Therefore, we used clones with similar surface expression levels to assess rolling flux and velocity for all L-selectin assays and cell accumulation for CD44 assays, respectively. To mimic leukocyte-endothelium interactions, a parallel plate flow chamber was used, allowing the characterization of the adhesion behavior under defined wall shear stress conditions. L-selectin showed efficient tethering and subsequent rolling at shear stresses between 0.7 and 2.3 dynes/cm 2 . Both mock transfectants and the blocking anti-L-selectin mAb DREG-200 were employed as specificity controls and completely abolished all interaction. Consistent with published data (4), the cell body-based chimera LCC showed a severely reduced rolling flux compared with microvillus-expressed LLL. LLC showed an intermediate phenotype between LLL and LCC at all tested shear rates (Fig. 4, A and B). The rolling velocity of LLL, LCC, and LLC was similar under all conditions (Fig. 4, C and D). Detailed data are provided in Table 1.
Next, we investigated whether microvillus positioning can improve the cell adhesion function of CD44. Expressionmatched pairs of CCC and CLC as well as CLL and CCL, respectively, were compared in a flow chamber assay (Fig. 5). In both cases, microvillus-expressed CD44 receptors (CLC, CLL) supported ϳ3-fold more adhesion than the cell body-based counterparts. Note that both pairs contain the same cytoplasmic tail and thus have the same intracellular anchorage.

DISCUSSION
In this study, the impact of the EC, TM, and IC domain of L-selectin and CD44 on adhesion receptor surface positioning and functionality was evaluated to clarify puzzling data concerning the contribution of the cytoplasmic anchorage of L-selectin and CD44. Using electron microscopy, no significant impact of the IC tail could be detected for both molecules, suggesting that receptor surface pattern is not determined by cytoplasmic interactions. However, mutations of ERM-binding residues within the proximal IC domain of L-selectin are known to abolish expression on microvilli (20), but deleting its distal cytoplasmic tail does not (21), consistent with both L-selectin and CD44 being constitutively ERM-associated (22). Therefore, ERM linkage of compartmentalized cell surface receptors seems to be necessary but not sufficient for defining their sorting to microvilli or the cell body. Further domain swapping revealed the TM domain to be a key determinant of receptor surface compartmentalization irrespective of the EC and IC domains.
In polarized cells, selective targeting and intracellular stabilization at the basolateral and apical cell surface are two putative mechanisms responsible for cell polarization enabling biochemically and functionally distinct plasma membrane domains (31). Epithelial cells are microvilli-bearing on their apical site whereas the basolateral compartment remains rather flat, resembling the protrusive and flat subdomains on leukocytes. Receptors of blood-borne nonpolarized cells may be compartmentalized by a cognate sorting machinery (32). Given our findings, it seems that leukocyte receptor compartmentalization involves two functionally separate processes: selective targeting via the TM domain and perhaps stabilization through ERM linkage present in both CD44 and L-selectin. The  ERM-binding residues and TM domains are highly conserved in mammals, supporting a close functional link (Fig. 6).
As a precondition for functional flow assays, the ability of the constructs to bind ligand was compared in a static environment. Neither receptor surface topology nor domain swapping influenced the adhesive function of L-selectin or CD44 (Fig. 3). Under flow, however, a significant effect on adhesion could be detected at all tested shear levels. At 0.7 and 1.8 dynes/cm 2 , the rolling flux of LLC was approximately 45% of that of LLL. LCC supported about 20% of the rolling flux of LLL at 0.7 dynes/cm 2 and about 12% at higher shear stress (Fig. 4B), suggesting that differential attachment rates favored by microvillus-expressed receptors become more obvious at increasing shear. However, rolling velocity is not a function of either the topographic distribution (17) or of specific cytoplasmic anchorage. This is consistent with the hypothesis that receptor off-rate, determined by the EC domain, determines rolling velocity (33).
Defective tethering may be caused by the missing cytoplasmic ␣-actinin binding site that was shown to play a central role in leukocyte tethering under flow (21,34). Indeed, the substitution of the L-selectin IC domain (LLC) primarily evoked a dramatic loss in initiating adhesion under flow ( Fig. 4) but did not change surface distribution (Fig. 2). Moreover, receptor presentation on microvilli is sufficient for increased adhesion under physiologic shear conditions as shown by LLC (on microvilli) compared with LCC (on the cell body; Fig. 4, A and B). These results are further supported by reciprocal experiments using CD44 chimeras. Both microvillus-expressed constructs containing the TM domain of L-selectin (CLC, CLL) showed superior adhesion compared with their cell body-expressed counterparts under flow (Fig. 5) but not under static conditions (Fig. 3B). Thus, the strong enhancement of CD44 adhesion under flow through the introduction of the L-selectin TM segment is likely to be a functional consequence of the receptor shift from the cell body to microvilli. Together, these findings emphasize that the TM domain is responsible for specific receptor surface positioning with direct consequences on receptor functionality. The CD44 and L-selectin TM domains are both 95-100% conserved among mammalian species (Fig. 6). Our data are the first to demonstrate that these TM helices indeed take   center stage in physiologic receptor function by defining surface localization. E-selectin, another member of the selectin family, is displayed ubiquitously on the surface of pre-B cell transfectants, including flat and protrusive sections (4). Despite the close relationship to L-selectin in the EC domain, its TM domain is barely conserved in mammals. Thus, it is plausible to propose that noncompartmentalization of leukocyte surface receptors constitutes a "default pathway," whereas TM signals lead to an active segregation to the cell body or microvilli.
Research about the organization of the plasma membrane has focused on cholesterol-enriched microdomains (lipid rafts) as a main constituent of distinct subdomains. The tyrosine kinase p56 Lck was shown to play a central role in trafficking the CD4 receptor to microvilli in the lymphoid cell line CEM (35). Notably, lymphocyte L-selectin requires functional p56 Lck for the activation of the Ras pathway (36). Moreover, a PSGL-1 mutant lacking the IC tail and thus missing a putative ERM binding domain still concentrates on microvilli of primary leukocytes (6). Because this mutant was still able to associate with lipid rafts and promote normal rolling on P-selectin, the authors hypothesized that cytoskeletal anchorage of PSGL-1 was achieved through raft molecules that, in turn, link to actin. Similarly, the deletion of the C-terminal tail of prominin, a pentaspan TM glycoprotein associated with lipid rafts, does not perturb its accumulation on microvilli (37). Collectively, these findings support the concept of lipid raft integrity as a precondition for microvillus receptor presentation. However, other studies seem to contradict: cell body CD44 in murine fibroblasts (38) but not microvillus L-selectin in human primary cells (39) was shown to be enriched in rafts using the detergent Triton X-100. This incongruence could be due to cell typespecific discrepancies, the existence of multiple mechanisms for protein sorting, or multiple types of rafts populating different surface subdomains.
The complexity of microvillus receptor expression is illustrated by the failure of transfected K562 cells to display human wild type PSGL-1 on microvilli (40) unlike primary cells (7,6). This points to possible accessory molecules necessary for PSGL-1 but not L-selectin positioning. Similarly, the CD44 receptor localizes to the planar cell body of L1-2 pre-B cells (4), but not of a melanoma cell line (22). Furthermore, domain swapping experiments of ␣4and ␤2-integrins (on microvilli and on the cell body, respectively) suggest that EC interactions may be responsible for microvillus positioning (41). Taken together, despite a very similar phenotype, the underlying mechanisms for the association of different surface receptors with microvilli seem to be diverse. Cytoskeletal anchorage, EC and TM interactions, as well as lipid raft integrity appear to contribute to a microvillus expression pattern. However, the surface expression patterns of CD44 and L-selectin are determined by the TM domain.
Membrane-spanning domains not only function as membrane anchors but also play a critical role in receptor functionality (42,43). Through lateral membrane interactions or dynamic conformational changes, TM domains are known to regulate the formation of heterogeneous protein complexes and protein folding, yet there is to date no evidence about any particular impact on receptor localization. The biological importance is highlighted by several disease-associated mutations within single-spanning TM domains clinically leading to severe pathologies such as achondroplasia, acute myeloid leukemia, or lupus (43). Because the extent of microvilli on cancer cells is correlated with the hematogenous dissemination potential of the tumor (44), our findings suggest that TM domains of adhesion receptors may also play a role in metastasis. Insights about the physiology of TM interactions can therefore guide studies into new therapeutic targets in inflammation and cancer metastasis.