The cytoplasmic tail of L-selectin interacts with the adaptor-protein complex AP-1 subunit mu1A via a novel basic binding motif

L-selectin regulates leukocyte adhesion and rolling along the endothelium. Proteins binding to the cytoplasmic tail of L-se-lectin regulate L-selectin functions. We used L-selectin cytoplasmic tail peptide pulldown assays combined with high sensitivity liquid chromatography/mass spectrometry to identify novelL-selectintail-bindingproteins.IncubationoftheL-selec-tin tail with cell extracts from phorbol 12-myristate 13-acetate-stimulated Raw 264.7 macrophages resulted in the binding of (cid:2) 1A of the clathrin-coated vesicle AP-1 complex. Furthermore, full-length GST- (cid:2) 1A and the GST- (cid:2) 1A C-terminal domain, but not the GST- (cid:2) 1A N-terminal domain, bind to L-selectin tail peptide, and the intracellular pool of L-selectin

Leukocyte trafficking is an essential mechanism of immune surveillance. The migration to and from peripheral tissues enables leukocytes to patrol the entire body, providing immune responses as needed. The exit of leukocytes from the circulation and migration into tissues is regulated at the level of cell interactions with the vascular endothelium. At least two families of adhesion receptors participate in this interaction: the integrins and the selectins (1). Selectins mediate the initial phase of leukocyte recognition of endothelium, and this takes the form of leukocyte rolling along the vessel wall (1). There are only three types of selectins: L, E, and P. L-selectin is expressed exclusively in leukocytes and acts as a major contributor of leukocyte tethering and rolling (2). Indeed, mice deficient in L-selectin display defects in lymphocytes homing to lymphoid organs and in neutrophil accumulation in inflamed tissues (3). Unlike E-and P-selectins, L-selectin is constitutively expressed on the membrane and possesses a membrane-proximal cleavage site that is hydrolyzed after leukocyte activation, leading to shedding of its extracellular domain (4). L-selectin shedding plays a key role in the regulation of leukocyte functions. It impairs the homing of activated lymphocytes to lymph nodes (5), augments neutrophil's rolling velocity (4), controls transendothelial migration of monocytes (6), and confers lytic activity to T lymphocytes (7). L-selectin is composed of an N-terminal lectin domain, binding endothelial ligands, followed by an epidermal growth factor-like domain, a short consensus repeat domain, a transmembrane domain, and a short cytoplasmic tail of 17 amino acids (8). The cytoplasmic tail of L-selectin is essential for L-selectin function and localization on the tips of microvilli. Indeed, L-selectin mutants lacking the membrane-distal 11 amino acid residues cannot support leukocyte rolling along endothelial venules and adhesion (9). This defect in L-selectin function cannot be explained by changes in ligand binding and is instead due to a defect in the recruitment of binding partners to L-selectin cytoplasmic tail.
To date, only a limited number of L-selectin tail binding partners are known. These include calmodulin (10), the ERM (ezrin-radixin-moesin) family of proteins (11), ␣-actinin (12), Lck (13), and protein kinase C (PKC) family members (14). Regions of the L-selectin tail involved in the interaction with these proteins have been mapped. For instance, calmodulin binds the unusual highly positively charged membrane-proximal N-terminal motif of dibasic residues, 356 RRLKKGKKS 364 . Moesin binds arginine 357 and lysine 362 of this domain (15). On the other hand, the more acidic C-terminal sequence of 11 amino acids of L-selectin cytoplasmic tail binds the cytoskeletal protein ␣-actinin (12) and the tyrosine kinase Lck (13), whereas the binding sites for PKC family members have not yet been identified. Binding of moesin to the RRLKK motif facilitates the anchor of L-selectin to microvilli, which is critical for leukocyte adhesion and rolling (15). Calmodulin binding is thought to induce a conformational change in the extracellular lectin domain of L-selectin that renders the cleavage site resistant to proteolysis (10). Interaction of ␣-actinin with the C-terminal region of the tail is important for leukocyte tethering and rolling but is not essential for ligand recognition (12). Finally, the recruitment of Lck to the C-terminal tyrosine residue (13) and binding of PKC family members to the L-selectin tail (14) are essential for L-selectin outside-in signaling.
Despite this gained knowledge, several aspects of L-selectin functions remain poorly understood. For example, the mechanism controlling the replenishment of L-selectin on the membrane surface and the clustering of L-selectin on the tips of microvilli are not well known. Identification of novel L-selectin cytoplasmic tail-interacting proteins may shed light on the mechanisms regulating L-selectin functions.
The murine L-selectin tail contains only one serine residue at position 364, which is conserved in human L-selectin. L-selec-tin is phosphorylated in response to T lymphocyte receptor cross-linking or treatment with PMA (14) (an activator of PKC) or engagement of chemokine receptors (16). It was proposed, but not proven, that such phosphorylation may be essential for the dynamic association/dissociation of proteins to the L-selectin tail relevant to the regulation of L-selectin functions. Indeed, moesin (11) and PKC family members (14) bind to the L-selectin tail in lymphocytes stimulated with PMA, but not in resting cells, whereas calmodulin binds L-selectin tail only in resting cells (10).
In this study, we aimed at identifying novel proteins interacting with L-selectin cytoplasmic tail domain by using sophisticated mass-spectrometry-based proteomics. We report the binding of the 1A-adaptin subunit of the AP-1 complex (17)(18)(19) to the L-selectin tail and the regulation of this binding by phosphorylation of L-selectin cytoplasmic tail and 1A. We propose that the AP-1-dependent transport of L-selectin may not be constitutive but regulated by the dynamic interaction between L-selectin tail and 1A in specific cellular compartments.

Identification of novel proteins interacting with the L-selectin cytoplasmic domain
To identify novel proteins binding to L-selectin cytoplasmic tail, we synthesized a desthiobiotinylated peptide comprising the entire 17-amino acid-long cytoplasmic tail of murine L-selectin. As a control, a desthiobiotinylated scrambled peptide of similar size was designed. A linker between the desthiobiotin group and the first amino acid of the two peptides was added (Fig. 1A). This was to ensure that the desthiobiotin group, when coupled to Dynabeads MyOne streptavidin C1 beads, does not A, peptide sequences used to identify proteins interacting with the L-selectin cytoplasmic tail. B, Raw 264.7 macrophages were stimulated for 1 min with PMA (1 g/ml), after which cells were lysed. Lysates were incubated with synthetic desbiotinylated peptides (scrambled control or L-selectin tail peptides) immobilized on Dynabeads Myone streptavidin. Proteins bound to the peptides were eluted, acetone-precipitated, and digested with trypsin and Lys-C. Quantitative mass spectrometric analysis was performed on a LTQ Orbitrap mass spectrometer and analyzed using the MaxQuant Software. Peptides deriving from the bait (Sell for L-selectin) and L-selectin-binding proteins are indicated on the graph. The AP1m1 gene encodes for the 1A subunit of the AP-1 complex.
interfere with the binding of proteins to the N-terminal membrane-proximal residues of L-selectin tail peptides.
To perform peptide pulldown assays, Raw 264.7 macrophages were stimulated with PMA (1 min, 1 g/ml) and lysed in mammalian protein extraction reagent (MPER) 2 buffer. Cleared lysates were incubated with either the L-selectin tail peptide (L-sel peptide) or a scrambled peptide (sc peptide) precoupled to Dynabeads MyOne streptavidin C1 beads. Cell lysis and experimental conditions (incubation time, washing conditions) were optimized to allow identification of novel L-selectin tail-interacting proteins. Our liquid chromatography (LC)/ mass spectrometry (MS) analysis was validated by the identification of L-selectin peptides (Sell in Fig. 1), which derived from the bait used. We identified four novel proteins interacting specifically with L-selectin tail peptide. These include 1A-adaptin (AP1m1 of the adaptor-protein complex 1 AP-1, ID: 11767), Cfh (complement component factor h, ID: 12628), Grn (granulin, ID: 14824), and Lrrc48 (leucine-rich repeat containing 48, ID: 74665) (Fig. 1B). The highest MS signal intensity associated with L-selectin tail peptide was obtained for 1A; therefore, we decided to analyze this interaction in more detail.
1A is one of the four adaptin subunits of the adaptor-protein complex 1 (AP-1). AP-1 controls the sorting of transmembrane proteins into clathrin-coated transport vesicles, mediating transport between the trans-Golgi network (TGN) and endosomes (17,18,20). This way, this vesicular carrier regulates the transport of cargo proteins (including receptors) to the plasma membrane (21).

The C-terminal domain of 1A interacts with L-selectin
To confirm the 1A/L-selectin tail interaction, we investigated the binding of full-length 1A as well as of 1A subdomains, expressed in E. coli as GST-fusion proteins, to L-selectin tail peptide. 1A consists of a N-terminal domain (aa 1-153), positioned close to the membrane, and a C-terminal domain (aa 154 -423), which binds the canonical YXXØ sorting motif (where Ø is a bulky hydrophobic residue) (17,18,22,23). Expression of cDNAs encoding for full-length 1A (GST-FL-1A), the C-(GST-C-1A) or the N-terminal domain (GST-N-1A) of 1A in E. coli resulted in the synthesis of GST-fusion proteins with predicted molecular weights as shown on the polyacrylamide gel stained with Coomassie Blue ( Fig. 2A).
We next incubated the L-sel peptide, or the sc peptide with purified GST-FL-1A. Thereafter, the beads were collected, washed, and resuspended in Laemmli buffer. Proteins were resolved on 10% SDS-PAGE, transferred onto a PVDF membrane, and subjected to Western blotting analysis using an anti-GST Ab (Fig. 2, B and C) or an anti-1A Ab (Fig. 2D). As shown in Fig. 2B, GST-FL-1A binds to the L-sel peptide but not to the sc peptide. GST (added in excess) did not bind to the L-sel peptide, confirming binding specificity. In addition, we found that GST-C-1A, but not GST-N-1A, binds the L-sel peptide (Fig. 2, C and D). Thus, the C-terminal domain of 1A binds L-selectin cytoplasmic tail, as it has been demonstrated for proteins with canonical YXXØ sorting motifs, whereas, as expected, the N-terminal domain, which binds the ␤1 subunit of the AP-1 complex (23), did not bind to the L-selectin tail.

1A is associated with anti-L-selectin receptor immunoprecipitates
We next sought to characterize the interaction between 1A, as part of the AP-1 complex, and L-selectin tail in cells in culture. To this end, Raw 264.7 macrophages were stimulated or not for different time periods with PMA (1 g/ml), before being lysed. Cell lysates were incubated with either an anti-L-selectin Ab or a control isotype-matched IgG2a. 1A (AP-1) association with anti-L-selectin immunoprecipitates was investigated by Western blotting analysis using an anti-1A Ab.
We found little association between L-selectin and 1A (AP-1) at steady state in resting cells. However, we observed a time-dependent increase in the amount of 1A (AP-1) associated with anti-L-selectin immunoprecipitates upon stimulation of the cells with PMA (1 g/ml) (Fig. 3A, top). As a control, we showed that 1A (AP-1) is not associated with IgG2a immunoprecipitates (Fig. 3B, top). We also verified that anti-L-selectin Abs precipitated L-selectin (Fig. 3, A and B, bottom panels), as evidenced by the detection of proteins with molecular masses of 75-120 kDa representing the different glycosylated forms of L-selectin (24). However, due to its low level of expression in Raw 264.7 macrophages, the presence of L-selectin in crude cell lysates could not be detected under these experimental conditions.

L-selectin and AP-1 in vivo colocalization
To test for in vivo L-selectin and AP-1 colocalization by confocal microscopy, we used an established THP-1 cell line expressing wild type, full-length L-selectin green fluorescent protein (GFP) (6). Endogenous AP-1 was labeled with an antibody directed against the ␥1 adaptin AP-1 subunit (20). This antibody recognizes the C-terminal flexible region of ␥1, which is even accessible when AP-1 is part of a membrane coat. Actin staining (red) served as control. Images show two pools of L-selectin (green), one at the plasma membrane and one in intracellular vesicles (Fig. 4, A and B). The intracellular pool is clustered around large membrane domains that have AP-1 bound (blue) and therefore represent TGN membranes.
Mander's colocalization coefficient was measured to prove L-selectin/1A (AP-1) colocalization in intracellular vesicles (overlap coefficient between blue and green signals). We compared the colocalization coefficient measured in an area including intracellular vesicles with the colocalization coefficient measured in an area containing the plasma membrane (Fig. 4B). Because there is no AP-1 at the plasma membrane, the colocalization coefficient measured in this area is not a real colocalization signal and is used as a background reference.
The limited L-selectin/AP-1 colocalization is in line with the transient AP-1/cargo protein interactions during protein sorting and transport via clathrin-coated vesicles (CCV). Thus, TGN-associated L-selectin may serve as a reserve pool. These proteins will be transported to the plasma membrane when this pool needs to be replenished. One function of AP-1 is the sorting of mannose 6-phosphate receptors (MPR), which facilitate the transport of soluble lysosomal enzymes from the TGN to endosomes. AP-1-dependent MPR sorting ensures that 50% of the cellular MPR pool resides in the TGN (awaiting newly syn-thesized lysosomal enzymes), whereas only 10% of the cellular pool is at the plasma membrane due to their highly efficient AP-2-mediated clathrin-dependent endocytosis (20,21). Thus, AP-1-dependent L-selectin sorting might be responsible for the formation of a localized L-selectin reserve pool in the TGN.

Phosphorylation of Ser-364 of L-selectin tail peptide prevents binding of 1A
To investigate whether phosphorylation of serine 364 of the L-selectin tail affects 1A binding, we synthesized an L-selectin tail peptide, in which serine 364 was replaced by a phosphoserine (L-sel-p-peptide). We found that GST-FL-1A, GST-C- After induction with isopropyl ␤-D-thiogalactopyranoside, bacteria were collected and lysed, and lysates were incubated with glutathione-Sepharose beads. After 1 h, the beads were collected and washed, and 2ϫ Laemmli buffer containing 1 mM DTT was added. Samples were run on 10% SDS-PAGE, and proteins were stained with Coomassie Brilliant Blue. The positions of GST-FL-1A, GST-N-1A, GST-C-1A, and GST are indicated on the left by arrows. B, purified GST-FL1A or GST were incubated with a scrambled peptide (sc peptide) or L-selectin cytoplasmic tail peptide (L-sel tail peptide) coupled to streptavidin-coated beads. After 1 h, the collected beads were washed and resuspended in Laemmli buffer. The proteins were separated on 10% SDS-PAGE and transferred onto a PVDF membrane, which was subjected to Western blotting analysis using an anti-GST Ab (top). The positions of GST-FL-1A and GST are indicated on the left by arrows. Loading controls (which correspond to the amount of GST-FL-1A or GST used for the peptide pulldown assay) are shown below. Note that an excess of GST versus GST-FL-1A has been used. C, binding of GST-N-1A or GST-C-1A to a scrambled peptide (sc peptide) or L-selectin tail peptide (L-sel tail peptide) was carried out as described in B. D, binding of GST-FL-1A or GST-C-1A to a scrambled peptide (sc peptide) or L-selectin tail peptide (L-sel tail peptide) was carried out as described in B and C with the exception that Western blotting analysis was carried out with an anti-1A Ab.

Interaction of L-selectin with 1A-adaptin
1A, or murine 1A (AP-1) from cell lysates bound the L-sel peptide but not the L-sel-p-peptide (Fig. 5A). We also designed a L-selectin peptide in which serine 364 was replaced by an aspartic acid residue (L-sel S364D) to imitate a constitutively phosphorylated L-selectin tail. Again, we found that GST-FL-1A, GST-C-1A, or murine 1A (AP-1) from a cell lysate bound the L-sel peptide but not the L-sel S364D peptide (Fig.  5A). To ensure that this effect was due to the introduction of a negative charge at position 364 of the L-selectin tail and not to the replacement of serine 364 by another amino acid, we also conducted experiments with a control peptide in which serine 364 was replaced by an alanine residue (L-sel S364A). We found that GST-FL-1A (Fig. 5B, left) or GST-C-1A (Fig. 5B, right) bound equally well to L-sel and L-sel S364A peptides. Thus, we concluded that phosphorylation of the L-selectin tail on serine 364 abrogates its interaction with 1A (AP-1).

Clusters of positively charged amino acids are essential for 1A interaction with the L-selectin tail
The membrane-proximal domain of the L-selectin tail contains an unusual cluster of three dibasic (positively charged) amino acids, 356 RR 357 , 359 KK 360 , and 362 KK 363 . To investigate whether this RRLKKGKK sequence of the L-selectin tail is involved in the interaction with 1A, we designed peptides in which one arginine or one lysine residue within any of the three dibasic motifs was exchanged by an alanine residue (Fig. 6A). These peptides were incubated with GST-FL-1A or GST-C-1A, and an interaction between the proteins and peptides was analyzed as described above.
We found that replacing a single arginine or a single lysine residue within any of the three dibasic clusters was sufficient to prevent GST-FL-1A (Fig. 6B, left) or GST-C-1A (Fig. 6B, right) from interacting with the L-selectin tail. We also investigated whether the C-terminal, negatively charged aspartic acid residues Asp-369 and Asp-370 of the L-selectin tail have a func-tion in 1A binding. We replaced separately each aspartic acid residue by an asparagine residue (D369N or D370N), or both were substituted with asparagine residues (D369N/D370N) (Fig. 6C). We found that replacement of one or two aspartic acid residues by asparagine augmented the binding of GST-C-1A to these L-selectin mutated tails (Fig. 6D). Thus, the highly positively charged membrane-proximal region of the L-selectin tail is essential for 1A binding, whereas its membrane-distal part, with its doublet of negatively charged amino acids, weakens the interaction with 1A.

Molecular docking of the L-selectin tail to 1A
To identify the 1A surface domain binding L-selectin cytoplasmic tail and to interpret the results of 1A binding to the L-selectin tail with substituted amino acids, we explored docking of the L-selectin tail to the crystal structure of 1A, as determined in the tetrameric AP-1 complex. To do this, we took into consideration the fact that the L-selectin tail binding surface domain of 1A has to be accessible and not sterically blocked by the interactions of 1A with the other subunits of the AP-1  , were prepared for confocal microscopy as described under "Experimental procedures." Cells were labeled for endogenous AP-1 (anti-␥1 subunit, blue) and actin (TRITC-phalloidin, red). A, L-selectin (green) is present at the plasma membrane and within discrete punctae that colocalize with AP-1-positive signal within the TGN. B, the rectangles show the areas used to perform quantification of L-selectin/1A colocalization. A region of interest of similar dimension was used to measure intracellular regions and at the plasma membrane. C, the Volocity software was used to measure the overlap coefficient between blue and green signals corresponding to AP-1 and L-selectin, respectively. Each signal was thresholded to remove background signals before analysis. Threshold parameters were maintained in all of the analysis. Instrument parameter settings remained constant between image acquisitions. The data represent means of colocalization co-efficiency Ϯ S.E. (error bars) of 22 cells prepared over three independent experiments. ****, p Ͻ 0.0001 (unpaired Student's t test). Scale bar, 7.5 m. PM, plasma membrane; IC, intracellular pool.
complex. Fig. 7A shows the favorable docking pose, which explains the binding specificities of 1A to L-selectin tail peptides. The L-selectin tail/1A interaction is predicted to have a strong electrostatic nature generated by two complementary charged surface areas at the binding interface. The positively charged cluster of arginine and lysine residues of the N-terminal region of the L-selectin tail are matched by several negatively charged residues, namely the aspartate and glutamate residues Glu-311, Glu-337, Glu-340, and Asp-417 of 1A. In addition, the positively charged residues of the L-selectin tail form hydrogen bonds with residues Asn-318, Asn-338, and Gln-419 of 1A. The interactions cited above are disrupted when the positively charged residues of the RRLKKGKK motif of the L-selectin tail are substituted to alanine, and this explains the loss of 1A binding. Ser-364 of the L-selectin tail does not form a specific interaction with 1A, in line with the unaltered binding of the S364A tail to 1A, but it is in close proximity to the negatively charged residues Glu-311 and Glu-340 of 1A. Therefore, introducing a negatively charged group at position 364 of the L-selectin tail by the addition of a phosphate group or by substituting serine 364 by an aspartic acid residue will cause electrostatic repulsion, thus preventing binding of 1A to the L-selectin tail.
The aspartic acid residues at positions 369 and 370 of the L-selectin tail form polar interactions with Lys-199 and Arg-410 of 1A, respectively, which would stabilize the binding of 1A. However, Lys-199 and Arg-410 have also polar inter-actions with Glu-261 and Asp-174 of 1A. The aspartate residues at positions 369 and 370 of the L-selectin tail could have some repulsion with Glu-261 and Asp-174 as well as the nearby Glu-381, which would weaken the binding. Asparagine residues at these positions could form interactions not only with the positively charged Lys-199 or Arg-410, but also with the negatively charged Glu-261, Asp-174, or Glu-381 residues of 1A, thereby enhancing the binding of the two proteins. This explains the augmented binding of 1A to the L-selectin tail peptides in which the aspartic acid residues 369 and 370 were replaced by asparagine residues. Our predicted binding mode is in good agreement with our biochemical binding experiments.
Since 1A (AP-1) and 2 (AP-2) are highly homologous and because both bind YXXØ-based sorting motifs and have overlapping cargo specificities, we also modeled the L-selectin tail onto the C-terminal 2 domain to examine whether a binding mode could permit an interaction, although 2 was not identified in our L-selectin tail pulldown experiments (Fig. 1). Docking predicts an interaction energy of 150 -275 kcal/mol for L-selectin/2 binding poses, contrasting with the 400 -500 kcal/mol for L-selectin/1A binding poses. Thus interactions of the L-selectin tail with 2 are much less favorable, confirming the pulldown experiment. It is likely that L-selectin is not endocytosed by the AP-1-homologous plasma membrane AP-2 complex.

Discussion
In this study, we performed L-selectin cytoplasmic tail pulldown assays combined with a high sensitivity LC/MS platform to identify novel L-selectin tail-binding proteins in cell extracts from stimulated Raw 264.7 macrophages. By using LC/MSbased peptide sequencing, we identified the AP-1 adaptor-protein complex subunit 1A as a novel binding protein of the L-selectin tail. We confirmed this interaction by showing that 1) 1A expressed as a GST-fusion protein binds to the L-selectin tail, 2) 1A as part of a fully assembled AP-1 complex is associated with anti-L-selectin immunoprecipitates, 3) L-selec-tin is present as an intracellular pool, which colocalizes with AP-1, and 4) the C-terminal 1A domain binds selectively the non-phosphorylated L-selectin tail.
1A is one of four adaptin subunits of the AP-1 complex. AP-1 selects the cargo proteins for CCV, and it coordinates CCV formation as well as CCV uncoating, a prerequisite for vesicle-organelle membrane fusion. Besides 1A, the AP-1 complex contains the adaptin subunits ␥, ␤1, and 1A (18,19). To incorporate cargo proteins into transport vesicles, 1A binds to specific sequences (called sorting motifs or cargobinding domains) within the cytoplasmic domains of such proteins. 1A is composed of two domains. The N-terminal domain (amino acids 1-153) binds the ␤1 subunit of the AP-1 complex and thus contributes to the assembly of the AP-1 complex. The C-terminal 1A domain (amino acids 153-422) binds sorting motifs of cargo proteins incorporated into CCV (23). In keeping with these structure/function studies of the AP-1 complex, we discovered that GST-C-1A but not GST-N-1A binds to the L-selectin tail.
Two canonical sorting motifs in cytosolic tails of transmembrane proteins have been described for AP-1: the tyrosinebased motif YXXØ, which is bound by the 1A C-terminal domain, and the dileucine-based motifs (e.g. (D/E)XXXL(L/I), which are bound by 1A and/or 1B adaptins (25,26). L-selectin cytoplasmic tail does not contain either of these two sorting motifs. This means that other sequences must be involved in the interaction with 1A.
In addition to these two canonical motifs, non-canonical binding motifs have been identified that interact with AP-1 as well. For example, yeast AP-1 binds a 12-residue domain of the pheromone receptor Ste13p, which contains a tribasic sequence, KRK (27). In addition, AP-1 binds to the highly basic RARHRRNVDR sequence of the polymeric immunoglobulin receptor (28). Based on these examples, the three dibasic clusters of amino acids in L-selectin tail most likely represent a novel, basic, non-canonical binding motif sequence for 1A. To investigate this possibility, we designed L-selectin tail peptides in which one arginine or lysine residue within any of the three dibasic motifs was exchanged by an alanine residue (R357A, K360A, or K363A). We showed that GST-FL-1A or GST-C-1A did not bind R357A, K360A, or K363A L-selectin tail peptides. This experiment demonstrated that each dibasic residue within the N-terminal membrane-proximal region of the L-selectin tail is essential for binding 1A. To understand the role of residues Arg-357, Lys-360, and Lys-363 for 1A interaction, we used molecular docking to predict the predominant binding mode of the L-selectin tail with the 3D structure of 1A. We found that the positively charged cluster of arginine and lysine residues of the N-terminal region of the L-selectin tail 356 RRLKKGKKS 364 are matched by aspartate and glutamate residues Glu-311, Glu-337, Glu-340, and Asp-417 of 1A and form hydrogen bonds with residues Asn-318, Asn-338, and Gln-419 of 1A. Whereas this L-selectin tail region also binds moesin/ezrin, it has been shown that only Arg-357, and not Lys-360 or Lys-363, is required for this interaction (15). Thus, the motifs mediating the binding of moesin/ezrin and 1A overlap, but they involve different amino acid residues. It was reported that association of calmodulin with the L-selectin tail Figure 6. Replacement of arginine or lysine residues within the clusters of dibasic residues prevents binding of 1A. A, amino acid sequence of L-selectin cytoplasmic tail and the different substitutions in the three clusters of dibasic residues is shown. B, purified GST-FL-1A or GST-C-1A were incubated with a scrambled peptide (sc peptide), L-selectin cytoplasmic tail (L-sel tail peptide), or peptides in which one amino acid within the clusters of dibasic residues was replaced by an alanine residue. These include L-sel R357A peptide, L-sel R360A peptide, and L-sel R363A peptide. Thereafter, association of GST-FL-1A (B, left) or GST-C-1A (B, right) with the different peptides was determined as described in the legend to Fig. 2D. C, amino acid sequence of L-selectin cytoplasmic tail with substitutions of aspartic acid residues by asparagine. D, purified GST-C-1A was incubated with sc peptide, L-sel tail peptide, or peptides in which one or two of the aspartic acid residues was replaced by an asparagine residue. These include L-sel D3369N peptide, L-sel D370N peptide, and the double-substituted L-sel DDNN peptide. Thereafter, association of GST-C-1A with the different peptides was determined as described above.
is influenced by the membrane bilayer. Thus, in a phospholipid bilayer (which mimics the inner leaflet of the plasma membrane), the positively charged N-terminal part of the L-selectin tail interacts with anionic phosphatidylserine lipids at the membrane interface through electrostatic interactions. This prevents calmodulin to interact with the N-terminal part of the L-selectin tail (29) and promotes L-selectin shedding (15). Disruption of the interaction between the positively charged N-terminal part of the L-selectin tail with the phospholipid bilayer is a prerequisite for the binding of calmodulin to the L-selectin tail. Such disruption is brought about by the interaction of the FERM domain of moesin with L-selectin to induce a separation of the cationic N-terminal part of the L-selectin tail from the phospholipids (30). We do not know whether the phospholipid bilayer plays a similar role in intact cells for at least two reasons. First, the lipid bilayer is itself a complex mixture of lipids. Its structure in the plane of the bilayer is that of a microscopic mosaic of regions that differ in composition (i.e. domains or clusters) (31). In particular, receptors are found in domains, such as the caveola, that are rich in cholesterol. Second, the cited authors (29, 30) have not taken into account in their model that calcium interacts with phospholipid bilayers. Such interactions induce changes in lipid dynamics, structure, and affinity of phospholipid-binding proteins (32).
AP complexes are targeted to their respective membrane via binding to specific phosphoinositides, AP-1 to phosphatidylinositol 4-phosphate and AP-2 to phosphatidylinositol 4,5-bisphosphate. There are no data in the literature describing inhibition of a sorting motif/AP complex binding by basic residues located in the vicinity of the sorting motif (23,25,26). In fact, phosphatidylinositol 4-phosphate incorporated into liposomes binds AP-1, and the interaction between adaptins and sorting motif-containing proteins is further stabilized by phospholipids during the polymerization of the vesicle coat (33). These results indicate a role of phospholipids and phosphoinositides in the interaction of AP-1 with its binding partners and vesicular assembly.
The 2 adaptin of the plasma membrane AP-2 complex collects proteins for clathrin-mediated endocytosis, and many proteins are transported by AP-1 as well as AP-2 CCV (17, 18). 2 was not isolated in our screen, despite the fact that it is highly homologous to 1A. Therefore, we asked why 2 could

Interaction of L-selectin with 1A-adaptin
not bind L-selectin tail. This is probably explained by the fact that the key amino acids in 1A involved in L-selectin tail binding are not found in 2. Indeed, only the first glutamic acid residue of the L-selectin binding motif in 1A is conserved in 2 (Glu-311 Ͼ Glu-321, Glu-337 Ͼ Ser-347, Glu-340 Ͼ Ala-350, Asp-417 Ͼ Ile-430). Also, the asparagine residues 318 and 338 and glutamine 419 of 1A are not conserved in 2, where they are replaced by leucine and glutamic acid residues. These data strongly indicate that the AP-2 complex, which is highly homologous to the AP-1 complex, does not play a role in the regulation of L-selectin cell surface expression.
We tested for L-selectin and AP-1 colocalization in vivo by confocal microscopy. Besides the plasma membrane pool of L-selectin, we detected an intracellular pool of L-selectin localized next to AP-1 decorated domains of the TGN. This indicated that AP-1-dependent L-selectin sorting might be responsible for the formation of this putative L-selectin reserve pool.
Murine L-selectin tail contains one unique serine residue at position 364, which is conserved in human L-selectin. It was shown that upon stimulation of lymphoblastoid cell lines with PMA or chemoattractants, the L-selectin tail becomes rapidly phosphorylated (16), and this is probably caused by recruitment of PKC family members to the L-selectin tail (14). However, little is known of the relationship between L-selectin tail phosphorylation and its interaction with binding partners in vivo. We found low amounts of 1A associated with anti-L-selectin immunoprecipitates in resting Raw 264.7 macrophages. However, this association increased when the cells were stimulated with PMA. This result could be explained in at least two ways, involving phosphorylation of 1A and/or L-selectin tail. One phosphorylation step would be the enhanced phosphorylation of 1A in response to PMA, because such phosphorylation would augment its interaction with the L-selectin tail. Indeed, the cytoplasmic AP-1 complex is present in a closed conformation, in which the cargo binding domains of 1A and 1A are blocked by ␥1 and ␤1 adaptins. The complex has to undergo a conformational change to an "open" state. This causes the release of the 1A C-terminal domain from the large adaptins, enabling its movement toward the membrane and the cytoplasmic sorting domains of membrane proteins. The open conformation is favored by threonine phosphorylation of 1A in the flexible linker region connecting the N-and C-terminal domains (34 -36). Such a mechanism of activation has also been described for the L-selectin tail-interacting protein moesin. Phosphorylation of moesin by PKC␣ on a threonine residue located in the C-terminal part unfolds the protein, thus allowing it to interact with the L-selectin tail (37).
The second phosphorylation reaction to explain a PKC-stimulated association between 1A and L-selectin tail would be the L-selectin tail phosphorylation. However, 1A does not bind the phospho-L-selectin tail peptide, and this is due to the fact that the phosphate group on serine 364 is in close proximity to the negatively charged residues Glu-311 and Glu-340 of 1A.
In keeping with our results, other investigators showed inhibition of calmodulin binding to L-selectin-p-tail in vitro and in intact THP-1 cells (6).
The AP-1 coat is found on the TGN and on early endosomes (EE), and it mediates constitutive protein transport from the EE to the TGN as well as from the TGN to EE, as exemplified by the mannose 6-phosphate receptors, furin and sortilin (20,21,26). Interestingly, besides its role as cargo binding and constitutive regulator of protein transport, the role of AP-1 in regulated secretory pathways has also been shown (38 -40). Our data showing that phosphorylation of both 1A and the L-selectin tail controls the interaction between these two proteins indicate a regulated, rather than a constitutive, AP-1-dependent secreted pathway for L-selectin. We developed a model of AP-1-dependent L-selectin sorting based on our data and the known functions of AP-1 (Fig. 7B). We propose that stimulation of 1A phosphorylation activates AP-1 and enhances AP-1-dependent protein transport from the TGN to EE. Phosphorylation of the L-selectin tail in activated leukocytes would prevent its binding to 1A (AP-1) in EE, leading to less incorporation of L-selectin into AP-1 CCV for retrograde EE to TGN transport. Consequently, L-selectin would transiently accumulate in EE, from which it can be readily transported to the plasma membrane through exocytosis (21). Such an indirect function of AP-1 in the regulation of plasma membrane protein transport has been demonstrated by us for both mannose 6-phosphate receptors in mouse fibroblasts (20,21). Thus large amounts of L-selectin would be exported to the plasma membrane and could be readily clustered on microvilli relevant for leukocyte rolling and adhesion. We are currently investigating this working hypothesis of stimulation-dependent compartment-specific L-selectin tail and 1A phosphorylation in leukocytes.
In summary, we have made several novel and important observations regarding L-selectin functions. First, we identified 1A as a novel protein interacting with the L-selectin tail. Second, we demonstrated colocalization of an intracellular pool of L-selectin with AP-1. Third, a novel binding motif formed by a triplet of dibasic amino acid motifs in L-selectin (RRXKKXKK) mediates this interaction. Fourth, L-selectin tail phosphorylation inhibits 1A binding. From these data we can conclude that AP-1, but not AP-2, is involved in the transport pathway regulating L-selectin secretion to the plasma membrane upon stimulation.

Chemicals
Protease inhibitor tablets were from Roche Applied Science (Penzberg, Germany), and phosphatase inhibitor mixtures 1 and 2 were purchased from Sigma-Aldrich. Dynabeads Myone streptavidin C1 and MPER buffer were from Invitrogen/ Thermo Fisher Scientific (Karlsruhe, Germany), and Protein A-Sepharose TM CL-4B and streptavidin-Sepharose beads were purchased from GE Healthcare Life Sciences (Munich, Germany). RPMI 1600 medium was purchased from Gibco/ Thermo Fisher Scientific. Fetal bovine serum and octyl-␤-glucoside were from Sigma-Aldrich. TAPI-O was from Peptides International (Louisville, KY).

Peptides
Automated solid phase synthesis of peptides using Fmoc (N-(9-fluorenyl)methoxycarbonyl)/t-Bu chemistry was carried out at the Biochemistry core facility of the Max Planck Institute for Biochemistry (Martinsried, Germany). Characterization and quality control (purity) was done by RP-HPLC. All peptides were desthiobiotinylated, and a spacer (hexanoic acid (Ahx)) between the desthiobiotin group and the first amino acid of the peptides was introduced. The following peptides were used usbin this study: desthiobiotin-GAhxEGRMKQPDKRDS

Cell culture
Raw 264.7 macrophages were grown in 75-cm 3 flasks at 37°C in an atmosphere of 5% CO 2 in RPMI 1600 medium supplemented with 10% FCS, antibiotics, and L-glutamine (20 mM).

Immunoprecipitation and Western blotting
Raw 264.7 macrophages were pretreated with the sheddase inhibitor TAPI-O (10 min, 10 g/ml) and then incubated in the absence or presence of PMA (1 g/ml) for different time periods. Thereafter, cells were collected by centrifugation (200 ϫ g, 2 min) and washed with PBS. The tubes were put on ice for 5 min, after which 1 ml of lysis buffer was added. The lysis buffer consisted of 50 mM Tris-HCl, pH 7.4, 30 mM octyl-␤-glucoside, 1 mM EDTA, 110 mM NaCl, protease inhibitors, and phosphatase inhibitors. Cell lysates were clarified by centrifugation, and the supernatants were subjected to immunoprecipitation. This was performed by exposure to the anti-L-selectin Ab (1 g/ml) or the isotype-matched IgG2a Ab (1 g/ml) for 1 h and then to 40 l of a 50% slurry of protein A-Sepharose for 1 h. The beads were subsequently collected by centrifugation (200 ϫ g, 1 min) and washed three times with lysis buffer. The beads were then resuspended in 2ϫ concentrated Laemmli sample buffer and boiled under reducing conditions (1 mM DTT) for 5 min. The immunoprecipitated proteins were subjected to electrophoresis on 10% SDS-PAGE and transferred to polyscreen PVDF transfer membranes. The membranes were blocked in TBS supplemented with 0.2% Tween 20 and 3% fat-free skimmed milk and then incubated for 1 h with a primary antibody (1:500 dilution of the rabbit anti-1A Ab (19) or a 1 g/ml dilution of the mouse anti-L-selectin Ab) and thereafter washed three times for 5 min in TBS supplemented with 0.2% Tween. The membranes were subsequently incubated for 1 h with peroxidase-conjugated goat anti-rabbit IgGs (1:10 000) or goat antimouse IgGs (1:10 000) in TBS supplemented with 0.2% Tween 20. The blots were washed three times for 5 min with TBS supplemented with 0.2% Tween 20, and antibody binding was visualized by enhanced chemiluminescence (ECL).

Peptide pulldown assays
L-selectin peptides were dissolved in PBS plus 0.01% Tween 20, and then 20 l of streptavidin-Sepharose beads were added for 1 h, followed by the addition of 2% BSA for 30 min. The beads were collected by centrifugation (200 ϫ g, 1 min) and washed once with PBS plus 0.01% Tween 20.
L-selectin tail peptides coupled to streptavidin-Sepharose beads were then incubated in 1 ml of cold buffer composed of 50 mM Tris-HCl, pH 7.4, 0.5% CHAPS, 110 mM NaCl. 50 ng of purified 1A expressed as GST-fusion proteins was added. The Eppendorf tubes were rotated for 1 h in the cold room, after which pellets were collected by centrifugation (200 ϫ g, 1 min). The pellets were washed three times with washing buffer (50 mM Tris-HCl, pH 7.4, 0.5% CHAPS, 300 mM NaCl). The beads were subsequently resuspended in 2ϫ concentrated Laemmli sample buffer and boiled under reducing conditions for 5 min. The proteins were subjected to electrophoresis on 10% SDS-PAGE and transferred to polyscreen PVDF transfer membranes, and Western blotting analysis was carried out as described above using a rabbit anti-1A Ab or an anti-GST Ab.
L-selectin peptides coupled to streptavidin-Sepharose beads were also incubated with lysate extracts of Raw 264.7 macrophages. Briefly, the cells were lysed in buffer composed of 50 mM Tris-HCl, pH 7.4, 0.5% CHAPS, 1 mM EDTA, 110 mM NaCl, and protease and phosphatase inhibitors. Lysates were then added for 1 h to L-selectin tail peptides coupled to beads. Thereafter, the beads were collected by low speed centrifugation (200 ϫ g, 1 min), washed three times with lysis buffer, and finally resuspended in 2ϫ concentrated Laemmli sample buffer and boiled under reducing conditions. 1A bound to the L-selectin tail was detected by Western blotting analysis as described above.

Proteomics analysis
Raw 264.7 macrophages were stimulated for 1 min with PMA (1 g/ml), after which the cells were pelleted and washed with ice-cold PBS. The cells were subsequently lysed with MPER buffer. Clarified lysates were then incubated with Dynabeads Myone streptavidin C1 that had been coupled to either L-selectin tail peptide or scrambled peptide (see above). After a 2-h incubation in the cold room under rotation, the beads were collected with a magnet and washed gently three times with MPER buffer. The beads were then resuspended in 50 mM Tris-HCl, pH 8.0, containing 10 mM biotin to dissociate the biotin analogue desthiobiotin from streptavidin beads. After 1 h in the cold room, supernatants were collected, and proteins were precipitated overnight at Ϫ20°C with 100% acetone. Proteins were pelleted by centrifugation (15,000 ϫ g, 15 min) in a cold centrifuge. Subsequently, the pellet was air-dried under a fume hood. 20 l of 6 M guanidium HCl containing a 10 mM concentration of the reducing agent tris(2-carboxyethyl)phosphine and 40 mM of the alkylation agent 2-chloroacetamine were added to each pellet, and the samples were boiled for 10 min. 180 l of LT-Digestion buffer (25 mM Tris-HCl, pH 8.5, containing 10% acetonitrile) was added, followed by 1 g of each trypsin and Lys-C endopeptidase. The tubes were left overnight at 37°C under shaking. Thereafter, 8 l of 25% TFA (1% final) was added to stop the reaction. The peptides were bound to triple filter surface (catalog no. SDB-RPS 3X), and after washing the filters three times with 0.2% TFA, proteins were eluted with 60 l of buffer containing 80% acetonitrile and 5% ammonium. Peptides were analyzed on an LTQ-Orbitrap analyzer, identified, and quantified using the MAxQUant software (41).

Purification of GST-fusion proteins
Expression of 1A as GST-fusion proteins has been described in detail (42). Briefly, pGEX-5X3 plasmids were transformed into E. coli BL21. 1 mM isopropyl ␤-D-thiogalactopyranoside was added overnight to bacterial cultures grown at 26°C. Bacteria were harvested, put on ice, washed once with ice-cold PBS, and lysed by the addition of a buffer made up of 50 mM Tris-HCl, pH 7.4, 1% Triton X-100, 110 mM NaCl, and protease inhibitors. This was followed by sonication. The lysates were clarified by centrifugation (10,000 ϫ g, 15 min), supernatants were collected, and 10% glycerol was added. These supernatants were incubated in the cold room with glutathione-Sepharose beads pre-equilibrated in PBS buffer. After 1 h, the beads were collected by centrifugation (200 ϫ g, 1 min) and washed three times with ice-cold PBS. The GST-fusion proteins were eluted by adding 200 l of buffer containing 50 mM Tris-HCl, pH 8.0, and 10 mM reduced glutathione. Eluted solutions were passed through a G-25 Sephadex column, which was subjected to centrifugation (500 ϫ g, 1 min), and GSTfusion proteins free of glutathione were collected. Protein concentrations were estimated by spectrometry.

Immunofluorescence confocal microscopy
THP-1 cells stably expressing wild type L-selectin tagged to green fluorescent protein (WT L-selectin-GFP) were generated and cultured as described previously (6). Cultured THP-1 cells were harvested and resuspended in warmed undiluted RPMI at a density of 0.5 ϫ 10 6 cells/ml. Cells were subsequently seeded onto poly-L-lysine-coated coverslips and allowed to settle for 10 min at 37°C. RPMI was aspirated, and coverslip-bound cells were fixed in 3% paraformaldehyde for 20 min at room temperature. Cells were washed three times in PBS and then permeabilized in 0.6% saponin for 10 min at room temperature. After washing fixed cells with PBS, cells were blocked in 5% fetal calf serum containing Fc receptor block (Miltenyi Biotec Ltd.) for 1 h at room temperature before staining with 1:300 (v/v) of 0.1 mg/ml TRITC-phalloidin and anti-␥1 AP-1 mouse monoclonal antibody (250 g/ml; BD Biosciences) diluted at 1:75 (v/v) overnight at 4°C. Cells were washed free from excess phalloidin and antibody, blocked as before, and then stained with secondary goat anti-mouse (Life Technologies) Alexa Fluor 633 (1:400 dilution). Specimens were mounted onto glass slides using mounting medium (Dako). Images were acquired using Leica SP5 confocal microscope, using a ϫ63 objective lens. Scale bar, 7.5 m.

Molecular modeling
Molecular docking of the L-selectin tail to 1A was conducted using the HEX version 8.0 protein-protein docking program (43). The crystal structure of 1A (Protein Data Bank code 4P6Z) and the structural model of 17-residue long L-selectin tail was modeled using the crystal structures of the 8-residue-long L-selectin tail (Protein Data Bank code 2LGF). The remaining 8 amino acid residues of the L-selectin tail were modeled using the loop prediction option of the Prime module within Schrödinger software (44). The L-selectin tail in a wild type as well as mutated forms (S364pS, S364D, D369N, and D370N) were used for docking to facilitate a choice of docking solutions. The 1A/L-selectin binding surface interactions were mapped based on shape and electrostatics complementarities as implemented in HEX version 8.0. Default settings for other docking parameters were used. HEX version 8.0 has generated over 100 solutions for each docking run. Besides the HEX energy function, the obtained docking pose of the L-selectin tail in the wild type and mutated forms was evaluated based on its ability to explain the mutagenesis data. In addition, only docking poses on the surface of 1A, which is not in an interaction with other subunits within, AP-1 were considered.
Author contributions-K. D. designed the project, performed the research, analyzed the data, and wrote the manuscript; I. G. T. performed the modeling of 1a-L-selectin tail interaction, analyzed the data, and wrote part of the manuscript; A. I. performed the colocalization studies; P. S. provided scientific expertise, analyzed the data, and wrote part of the manuscript. All authors read and approved the final manuscript.