INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Targeting of transmembrane proteins to different compartments of the endocytic and late secretory pathways depends largely on sorting signals contained within their cytosolic domains (Letourneur and Klausner, 1992; Mellman, 1996; Kirchhausen et al., 1997). These signals are thought to interact with specific recognition molecules, which are components of membrane-bound transport intermediates (Schmid, 1997; Hirst and Robinson, 1998; Bonifacino and Dell'Angelica, 1999; Kirchhausen, 1999; Marsh and McMahon, 1999). Most internalized proteins contain sorting signals such as the tyrosine-based motif, Yxx, where x is any amino acid and  is a bulky hydrophobic side chain (Lobel et al., 1989; Collawn et al., 1990; Boll et al., 1996) or the di-leucine-based motif, LL (Letourneur and Klausner, 1992). For the tyrosine-based motif, the interaction is mediated by the medium chains of adaptor protein (AP) complexes (Ohno et al., 1995). Cocrystallization of the signal binding domain of µ2 with two different Yxx peptides provided a structural explanation for this binding specificity (Owen and Evans, 1998).

The Nef protein of primate lentiviruses is the only nontransmembrane protein known to traffic via a di-leucine-based motif (Kirchhausen, 1999). It is required for the internalization of CD4 from the cell surface to endosomes and lysosomes (Craig et al., 1998; Greenberg et al., 1998). Several proteins have been proposed to direct the sorting of Nef. They include the -chain of AP-1 complexes (Bresnahan et al., 1998; Greenberg et al., 1998), the -subunit of the COP-I coatomer (Piguet et al., 1999; Janvier et al., 2001), and the regulatory subunit H (V1H) of the vacuolar ATPase (V-ATPase), previously named NBP1 for Nef binding protein 1 (Lu et al., 1998; Mandic et al., 2001). Although the precise sites of these interactions varied, all protein complexes were reported to bind to the C-terminal flexible loop in Nef, which contains the di-leucine-based motif at its center (Grzesiek et al., 1996; Lee et al., 1996; Geyer et al., 2001). Nef also downregulates major histocompatibility complex (MHC) class I molecules from the cell surface (Schwartz et al., 1996), and targeting of this complex to the trans-Golgi network is mediated by the cellular protein PACS-1 (Piguet et al., 2000).

The V-ATPase is a multisubunit enzyme, consisting of two distinct functional domains, V₀ and V₁ (Stevens and Forgac, 1997). The 260-kDa V₀ domain, which is composed of five different subunits, is an integral complex that is responsible for proton translocation across the membrane. The V₁ domain is a 570-kDa peripheral complex that is responsible for the hydrolysis of ATP. V₁ is composed of eight different subunits of molecular masses of 70-14 kDa (subunits A-H). Although the principal arrangement of the V-ATPase is similar to the F-ATPase, the molecular architecture of the V-type ATPase appears to be more complex. The number of known subunits of the V-ATPase exceeds that of the F-ATPase. Moreover, their low sequence homology and different composition suggests some fundamental differences between the two classes of enzymes. The latest subunit of the V-ATPase to be identified is the regulatory subunit H of the head region V₁ (V1H), which is essential for the catalysis but not for the assembly of the enzyme (Ho et al., 1993). There is no known counterpart to V1H in the F-type ATPases. V-type ATPases are thought to facilitate the acidification of intracellular compartments in eukaryotic cells and therefore play an important role in receptor-mediated endocytosis, intracellular trafficking processes, and protein degradation (Mellman et al., 1986; Nishi and Forgac, 2002).

AP-2 is a heterotetramer protein complex composed of two large subunits ( and 2, ~110 kDa), one medium subunit (µ2, ~50 kDa), and one small subunit (2, ~17 kDa; reviewed in Bonifacino and Dell'Angelica, 1999; Kirchhausen, 1999; Marsh and McMahon, 1999; Pearse et al., 2000; Boehm and Bonifacino, 2001). Although both sorting motifs, Yxx and LL, interact with AP-2 complexes (Kirchhausen et al., 1997) and use distinct saturable components (Marks et al., 1996), the binding site on AP-2 for the di-leucine-based motif is not yet clearly established (Marsh and McMahon, 1999). Recently, using protein cross-linking experiments and limited tryptic proteolyses, the target molecule for the di-leucine-based sorting signal has been narrowed to the N-terminal trunk (~65 kDa) of the -chain of AP-1 complexes (Rapoport et al., 1998).

In this article, we show that the C-terminal fragment of 350 amino acid length in V1H shares significant similarity to a fragment in the N-terminal trunk portion of -adaptins. Expression of these fragments blocks the internalization of transmembrane proteins that depend on di-leucine-based sorting motifs in cells. Our results suggest the determination of a di-leucine-binding domain in -adaptins and the identification of a homologues domain with similar properties for the regulatory subunit H of the vacuolar ATPase.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Plasmid Constructions

Plasmids encoding full-length V1H were described previously (Lu et al., 1998). Various fragments of V1H (AC: AF298777), 1 (M77245), and 2 (M34175) genes were generated by PCR-mediated amplification with primer containing NcoI and EcoRI restriction sites and cloned into the yeast expression vector containing the Gal4 activation domain (pACT2, Clontech, Palo Alto, CA). Fragments 1F (171-518), 2F (171-518), and 2L (133-518) used for in vitro translation and expression in mammalian cells were subcloned into the expression vectors T7 or Myc-tagged pEF-BOS, respectively. Plasmids of Nef for yeast two-hybrid experiments were cloned into the Gal4 DNA binding domain vector pAS2-1 using NcoI and EcoRI sites. Mutation of amino acids LL₁₆₈ and ED₁₇₈ or partial deletion of the flexible loop (residues 158-178) were introduced into the nef gene of HIV-1_SF2 (K02007) by PCR-mediated mutagenesis, resulting in Nef-LLAA, Nef-EDAA, and Nef-fl.loop plasmids, respectively. Nucleotide sequences of novel constructs were confirmed by DNA sequencing. Plasmids encoding GST-Nef fusion proteins of the HIV-1_SF2 nef gene were constructed with the pGEX-2TK vector (Pharmacia, Piscataway, NJ) using BamHI and EcoRI restriction sites.

Plasmids encoding 1 and 2 were generous gifts from Margaret Robinson, Tac-DKQTLL and pTTMb (Tac-YxxL) from Juan Bonifacino, and IL2R-LL and IL2R-AA from Warner Greene. The expression plasmid IL2R-LL encodes the outer and transmembrane region of the IL-2 receptor linked to the di-leucine-based motif from Nef as cytoplasmatic tail (Bresnahan et al., 1998). Tac-DKQTLL (Letourneur and Klausner, 1992) and pTTMb (Marks et al., 1996) contain the di-leucine-based motifs from CD3 and the tyrosine-based motif from HLA-DM, respectively.

Yeast Two-hybrid Binding Assays

Yeast transformation, X-Gal filter assays and liquid assays with CPRG and ONPG reagents were performed with the Matchmaker GAL4 two-hybrid system using the Y187 cell line according to the manufacturers instructions (Clontech). -Galactosidase assays in liquid culture (~10 colonies/culture) were done from three independent Y187 transformations following the Clontech protocol. The N-terminally lengthened 2 fragment (133-518; 2L) was cloned to increase transformation efficiency with Nef, which resulted in an average of eight colonies per plate.

In Vitro Binding Assays

GST-Nef fusion proteins were expressed in Escherichia coli and purified on glutathione-Sepharose beads as described (Fackler et al., 2000). Equal amounts of GST and GST-Nef fusion proteins immobilized on the beads were incubated with fragments of [³⁵S-Met]-labeled in vitro translated V1H, 1, or 2 (Promega, Madison, WI). Binding reactions were performed at 4°C for 3 h in 10 mM CHAPS buffer and 50 mM NaCl at pH 7.4. The beads were then carefully washed two times with CHAPS buffer and two times with PBS andTriton X-100 0.1%, and bound proteins were separated on 10% SDS-PAGE and analyzed by autoradiography.

Protein Sequence Alignments

Protein sequences used for alignments were as follows: V1H: AAG22809; 1: AAC98702; 2: AAA35583; 3A: AAB61638; 3B: AAB71894; 4: AAD20448 (all human) and -COP: CAA40505 (rat). Alignments were performed using the program ALIGN from the GeneStream software and the program packages Multalin and Clustal W (vers. 1.7) for multiple sequence alignments. Sequence labeling was performed with the MacBoxShade software.

Analysis of the Subcellular Localization of Endocytosis Reporters

To monitor the steady state subcellular distribution of IL2R reporter constructs, NIH3T3 cells were grown on coverslips and cotransfected with 0.75 µg of expression vectors for the LL-binding fragments or an empty plasmid vector control and 0.25 µg of the IL2R reporters using Lipofectamine (Life Technologies, Rockville, MD). For the titration experiments, a total amount of 1 µg of DNA was transfected with an empty plasmid vector control compensating for the reduced amounts of LL-binding fragment expression vector. The cells were fixed with PBS 3.7% formaldehyde (5 min, RT) 36 h posttransfection, permeabilized with PBS 0.3% Triton X-100 (5 min, RT), blocked in PBS 3% BSA (30 min, RT), and stained with PE-conjugated anti-CD25 (Becton Dickinson, Mountain View, CA) as well as FITC-conjugated anti-Myc (Santa Cruz Biotechnology, Santa Cruz, CA) antibodies. The cells were analyzed by confocal microscopy using a laser scanning confocal unit (LSM510, Carl Zeiss, Thornwood, NY) attached to an Axiovert microscope (Carl Zeiss) with a 63 × 1.2 NA plan C-Appochromat objective. Images were saved as tiff files and processed using Adobe Photoshop (San Jose, CA). Individual sections of at least 100 transfected cells were analyzed per individual transfection. When cells coexpressed IL2R reporter and a LL-binding fragment, only cells with detectable expression of the LL-binding fragments were considered. For statistical analysis, the percentages of cells lacking IL2R reporters in intracellular organelles were determined from at least 100 positive cells each from three individual transfections.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Molecular Characterization of V1H

Cleavage sites for protein fragmentation of human V1H were determined based on the degree of sequence conservation using seven sequences of V1H from five different organisms (Figure 1A). Additionally, we used information of secondary structure prediction methods (PHD, Columbia University, and Predator, EMBL), which indicated mostly helical structures (~65%) for the different protein sequences. Overall, the C-terminal half of V1H, starting with a hydrophobic and highly uncharged stabilization region at position 235, exhibits a significantly higher degree of conservation (avg. identity 65.1%) than its N-terminal half (avg. identity 51.7%). The analysis of sequence insertion and deletion sites as well as consideration of putative proteolytic cleavage sites led to the determination of four cleavage sites in the protein of 483 residues (Figure 1).

View larger version (36K): [in this window] [in a new window]   Figure 1.   Sequence characterization of the regulatory subunit H (V1H) of vacuolar ATPases from different organisms. (A) Top: Hydrophobicity plot of human V1H (1-483) based on parameter from Kyte and Doolittle (1982). Data were smoothed with a binominal filter of sixth degree. The location of charged residues is indicated by blue (Arg, Lys, His) and red (Asp, Glu) bars. Middle: Schematic sequence display. Insertion gaps in human V1H relative to sequences of other organisms are indicated by dashed lines. Cleavage sites selected for protein fragmentation are indicated by arrows. Bottom: Sequence conservation of V1H proteins based on seven sequences from five different organisms (human, bovine, C. elegans, S. pombe, S. cerevisiae). The degree of sequence identity is displayed as shaded area (average sequence identity is 58.5%), the sequence similarity is indicated by a solid line. (B) Secondary structure of V1H as derived from the structure-based sequence alignment to the yeast homologue VMA13p (Sagermann et al., 2001). The position of the eight armadillo repeats (ARM) is indicated.

The crystal structure of the yeast homologue of subunit H, VMA13p, was published recently (Sagermann et al., 2001). Although the yeast homologue VMA13p contains only 21.7% sequence identity to human V1H, a structure-based alignment enabled us to transform the secondary structure elements of VMA13p to V1H (Figure 1B). The all-helical structure contains eight so-called HEAT or armadillo (ARM) repeats that are also found in Importins (Sagermann et al., 2001). A single ARM repeat consists of ~42 residues that fold into three -helices with an almost triangular cross section. Multiple ARM repeats pack regularly side by side, forming elongated molecules with a superhelical twist that results in an internal concave surface formed by the third helix of each repeat (Groves and Barford, 1999). Although the 10 N-terminal residues of VMA13p are folded back and occupy the first three ARM repeats, the last two repeats (7 and 8) are bent apart from the main trunk by the insertion of two helices and form an independent unit. Indeed, the analysis of our fragments revealed that the cleavage sites at positions 133, 363, and 402 matched very well with the proposed secondary and tertiary structure of V1H. For example, the cleavage site at position 133 is at the beginning of the third helix in the second ARM repeat, which at position 363 is right in between the two helices that separate the sixth and seventh ARM repeat and which at position 402 is in the seventh ARM repeat (Figure 1B). Only the cleavage site at position 228 is set within the second helix of the fourth ARM repeat, which may result in an unstable protein fragment.

Mapping of the Interaction Sites between V1H and Nef

We mapped the binding domains between V1H and Nef using the yeast two-hybrid assay, which has been used previously to study the interaction between medium chains and tyrosine-based motifs (Ohno et al., 1995; Aguilar et al., 1997). Full-length V1H and nine fragments (constructs A to J) were examined for their binding to the wild-type Nef protein from HIV-1, strain SF2. All these fragments were stable and expressed to comparable levels. Results of five independent binding experiments using the X-Gal filter assay and the CPRG liquid assay revealed that a C-terminal fragment of 351 amino acids (V1H-G) bound best to Nef (Figure 2A). In more detail, two separate regions of V1H were identified (aa 133-363 and 402-483) that both interacted independently with Nef, suggesting the formation of a groove that might face Nef from two sides. The binding results observed here could reflect the repetitive character of the ARM repeat structure in V1H and the autoinhibition of the first three repeats by its N-terminus. Indeed, all fragments that were cleaved at the N-terminus bound better to Nef, which holds true also for fragments A (1-483) and G (133-483) and fragments B (1-363) and H (133-363). This suggests that the accessibility of the concave ARM repeat structure is required for the interaction with Nef. A direct comparison of fragments C (1-232) and F (228-483) and I (133-232) and J (228-363) suggested further that ARM repeats 4-6 are more essential for the binding to Nef than the first repeats, although the cleavage sites 228 and 232 within the fourth repeat is not set optimally. Fragments D (402-483) and E (362-483) finally contain the last two ARM repeats of V1H and bound to Nef independently of the N-terminal 6 repeats.

View larger version (22K): [in this window] [in a new window]   Figure 2.   Mapping of the binding site between V1H and HIV-1 Nef. (A) Full-length V1H and nine fragments were cloned in the plasmid vector with the GAL4 activation domain (constructs A to J) and tested for their binding to the wild-type SF2-Nef as the bait. -Galactosidase assays in liquid culture (~10 colonies/culture) with CPRG reagents were done from three independent transformations using Y187 yeast cells, following the manufacturers instructions (Clontech). (B) Binding of different Nef alleles and Nef mutants to V1H indicates that the integrity of the flexible loop structure is required for the interaction. Yeast transformation in the Y187 cell line, X-Gal filter assays and liquid assays were performed with the Matchmaker GAL4 two-hybrid following the Clontech protocol.

Subsequently, we mapped the site of interaction on Nef by site-directed mutagenesis. Mutation of the di-acidic cluster ED₁₇₈ to alanines as well as the mutation of the central di-leucine-based motif at position 168 in the flexible loop of Nef significantly diminished the binding to V1H. Deletion of the entire flexible loop (Nef158-178), which resulted in a stable protein (Grzesiek et al., 1996), completely abrogated this interaction (Figure 2B). A Nef allele from a different strain of HIV-1, NL4-3, displayed similar binding properties as SF2-Nef. We conclude that the entire flexible loop structure in Nef is required for its binding to V1H. Our mapping results suggest that these 33 residues encompassing flexible loop in Nef binds to the cleft that is formed between the two units of ARM repeats 1-6 and 7-8 in V1H.

Homology of V1H to -Adaptins

The initial identification of the -subunit of the COP-I coatomer, -COP, defined this protein by its similarity to -adaptins, which stretches over its N-terminal 460 amino acids and suggests a common main-chain fold (Duden et al., 1991; Serafini et al., 1991). In that publication, a consensus sequence WI(L/I)GEY around position 450 was discovered (Duden et al., 1991). This observation reminded us of a similar motif in V1H. Indeed, the alignment of sequences of V1H and 2 revealed a striking similarity between these two proteins (Figure 3). We varied multiple fragment lengths of both proteins using mutual alignments, which indicated that the N-terminal 133 residues are less conserved (13.5% identity) than the C-terminal residues in this domain. Most strikingly, the highest degree of similarity between V1H and 2 was achieved by aligning the V1H fragment that binds best to Nef (133-483) with a region in 2 from positions 171-518. Both segments exhibited an amino acid identity of 20.3% and an overall similarity of 52.3%, which is even higher than the homology of the respective segments in -COP and 2 (19.9% identity). Multiple sequence alignments of the respective fragments in V1H, the -chains of 1 and 2, and -COP identified additional conserved sites besides the GEY motif throughout these sequences (Figure 3). Importantly, the sequence similarity is spread homogeneously over the entire fragment size of 350 residues and displays the typical pattern of helical secondary structure contents, see e.g., the recurrent 3- to 4-residues conservation from position 440-480 in V1H. A sequence identity of 20% or more is believed to result in a similar overall main-chain fold of two proteins (Elofsson and Sonnhammer, 1999; Thornton et al., 1999).

View larger version (98K): [in this window] [in a new window]   Figure 3.   Sequence similarity of V1H with -adaptins. Multiple sequence alignment of -adaptins (1, 2), V1H, and -COP. Protein fragments aligned are 2 (170-518), 1 (170-518), V1H (133-483), and -COP (170-533). Identical and conserved residues are represented by white letters in black or gray columns, respectively. Similar residues are shaded in gray columns. Protein sequences used for alignments are as follows: V1H, AAG22809; 1, AAC98702; 2, AAA35583; -COP, CAA40505. Alignments were performed using the program ALIGN from the GeneStream software and the program packages Multalin and Clustal W (ver. 1.7) for multiple sequence alignments. Sequence labeling was performed with the MacBoxShade software.

Analysis of the sequence homology of the respective region in other -adaptins led to the suggestion of a preserved structural domain in all four known adaptor complexes. Although the overall identity of the full-length proteins is significantly lower, 2 (170-518) shares 92.3% sequence identity to 1 (170-518), 27.7% to 3A (197-569), 28.8% to 3B (192-574), and 28.5% to 4 (169-510). Importantly, the homologous region identified here is N-terminal to the flexible hinge region containing the clathrin box signals of 1, 2, 3A, and 3B (ter Haar et al., 2000). In support of our finding, recent computational analyses on HEAT and ARM repeat structures suggested -adaptins and -COP to be HEAT repeat containing proteins (Traub, 1997; Andrade et al., 2001).

The Homologous Fragments of -Adaptins Bind di-Leucine-based Sorting Signals

To examine the functional similarity between these fragments, we transferred the mapping of our binding results between V1H and Nef to the -chains of AP complexes. The fragments 1(171-518), 2(171-518), and 2(133-518), referred to as 1F, 2F and 2L, respectively, were tested for their binding to Nef using the yeast two-hybrid liquid assay with the Y187 cell line. Indeed, all fragments of V1H, 1, and 2 bound to Nef in a specific manner and within the margins of error, the recognition of Nef by V1H-G, 1F, 2F, and the longer 2L fragments appeared similar (Figure 4A). Again, the deletion mutant Nef158-178 that lacks the flexible loop abolished completely the binding to all interacting fragments. Moreover, for -adaptins, the LL₁₆₈AA mutation in Nef alone abrogated this binding, which suggested the identification of a di-leucine binding domain. Binding to V1H was less dependent on an intact LL-motif, an effect similar to that of the EDAA mutation. Mutation of the di-acidic cluster in Nef to alanine (ED₁₇₈AA) significantly diminished the binding to all homologous fragments. This effect may indicate the requirement of the acidic residues for the formation of the flexible loop in Nef and its exposure for recognition as discussed later.

View larger version (24K): [in this window] [in a new window]   Figure 4.   Binding of the homologous fragments to the di-leucine sorting signal of Nef. (A) Binding between Nef and V1H, 1, and 2 using the yeast two hybrid liquid assay. -Galactosidase assays in liquid culture (~10 colonies/culture) with CPRG and ONPG reagents were done from three independent transformations using Y187 cells, following the Clontech protocol. (B) Binding between mutant V1H(GEY-AAA) or 2(GEY-AAA) fragments and Nef protein using the yeast two hybrid liquid assay as in A. (C) In vitro GST pull-down experiments indicate di-leucine motif dependent binding of GST-Nef fusion protein and [35S-Met]-labeled in vitro translated V1H, 1, and 2 fragments. Binding reactions were performed at 4°C for 3 h in 10 mM CHAPS buffer and 50 mM NaCl at pH 7.4 and subjected to autoradiography.

Mutation of the GEY consensus sequence in the LL-binding domains of V1H and 2 reduced the binding to Nef and no interactions could be detected to the mutant Nef(LL₁₆₈AA) protein (Figure 4B). This result suggests that the GEY sequence motif participates in the binding interface to the flexible loop in Nef but is not exclusively required for the interaction. This observation is in line with the previous data that the N-terminal and C-terminal ARM repeats (3-6 and 7-8; fragments H and E in Figure 2A) both interact independently with Nef, whereas the sum of both segments (repeats 3-8, fragment G) binds best. Analysis of the GEY motif on the structure of VMA13p reveals that the three residues are located on the third helix of the seventh ARM repeat. On this surface the first two residues, glycine and glutamic acid, are exposed for possible interactions, whereas the tyrosine is found in the core of the helical ARM repeat fold.

The interaction determined between Nef and the homologous di-leucine-binding fragments were confirmed in binding assays between GST-Nef and [³⁵S]-methionine-labeled in vitro translated V1H, 1, and 2 fragments using the high detergent buffer CHAPS (Figure 4C). In this system too, a specific but very weak interaction between wild-type Nef (SF2-Nef) and the homologous fragments of V1H and -adaptins was detected (Figure 4C, lanes 2, 6, and 10). In contrast, binding of mutant Nef(LL₁₆₈AA) was strongly diminished or could not be detected (lanes 3, 7, and 11). We conclude that the homologous fragments of -adaptins and V1H bind the di-leucine-based sorting motif of Nef in a specific manner.

Expression of LL-binding Domains Affects LL-mediated Internalization

Next, we asked whether the identified LL-binding fragments were functional in cells. To this end, we determined if the expression of our fragments could interfere with the internalization of the extracellular and transmembrane portions of the IL2 receptor linked to the flexible loop of the HIV-1 Nef protein and therefore containing a cytoplasmic di-leucine-based motif (IL2R-LL; Bresnahan et al., 1998) and determined its subcellular localization. NIH3T3 cells were cotransfected with plasmids encoding for the IL2R-LL fusion protein and the Myc-tagged di-leucine binding fragments, respectively. After permeabilization, cells were stained and only IL2R/Myc double-positive cells were analyzed subsequently by confocal microscopy (Figure 5). As expected, the IL2R-LL chimera was detected in a punctate cytoplasmatic and perinuclear pattern when expressed together with an empty control plasmid vector (Figure 5A), suggesting that it was internalized efficiently from the plasma membrane into cytoplasmic organelles. In contrast, the mutant IL2R-AA chimera, bearing a substitution of alanines for di-leucines, was found predominantly at the plasma membrane and was excluded from intracellular organelles (Figure 5B). Coexpression of the LL-binding fragments of V1H, 1, and 2 (Figures 5, C, E, and G) markedly shifted its subcellular localization toward that of the internalization defective IL2R-AA, resulting in a intensive staining of the plasma membrane and the exclusion from intracellular organelles. As monitored by the detection of the expressed fragments (Figures 5, D, F, and H), all LL-binding fragments displayed a homogenous distribution throughout the cytoplasm, and the plasma membrane of transfected cells and were expressed to comparable levels. Expression of these fragments did not affect the localization of the internalization incompetent IL2R-AA reporter. These findings suggest that the fragments bound efficiently to the LL-motif of the IL2R-LL chimera and prevented its internalization.

View larger version (31K): [in this window] [in a new window]   Figure 5.   Expression of the homologous fragments affects localization of di-leucine sorting signals. Immunofluorescence of NIH3T3 cells transfected with IL2R reporter constructs in the absence (A and B) or presence (C-H) of the indicated fragments. The subcellular localization of IL2R was visualized with the PE-conjugated anti-CD25 antibody (A-C, E, and G). Cells expressing LL-binding fragments were identified with the FITC-conjugated anti-myc antibody (D, F, and H). Presented are representative individual sections of transfected cells. (A) Internalization competent IL2R with a functional cytoplasmatic LL-based motif. (B) Internalization incompetent IL2R with a mutated cytoplasmatic LL-based motif. (C-H) Internalization competent IL2R in the presence of the LL-binding domain of V1H (C and D), 1 (E and F), and 2 (G and H), respectively.

Quantification and Specificity of the LL-binding Fragments

To quantify the effects of the LL-binding fragments on the subcellular localization of the IL2R-LL reporter, the percentage of transfected cells that lack staining of intracellular organelles (as in Figure 5B as compared with Figure 5A) was determined. We found that the fragments from V1H, 1, and 2 were similarly active in this assay and blocked the internalization of IL2R-LL to levels obtained with the internalization deficient IL2R-LLAA reporter (Figure 6A). In contrast, expression of a V1H fragment, V1H-C, that did not bind to di-leucine motifs (see Figure 2) had no detectable effect on the surface expression of the IL-2RLL reporter. Mutation of the consensus site GEY to alanines in V1H and 2 had only minor effects on the block of internalization exhibited by these fragments. This observation supports the previous suggestion that the binding interface of the LL-binding fragments to the di-leucine-based sorting motif significantly exceeds the binding site of one single ARM repeat. Together, these results demonstrate that the detected molecular interaction of LL-binding domains with di-leucine motifs is functional in cells.

View larger version (18K): [in this window] [in a new window]   Figure 6.   Expression of the homologous fragments affects di-leucine mediated internalization. Data presented are averages from at least three independent experiments with the SEs of the mean indicated by error bars. More than one hundred cells in each transfection were scored for the lack of intracellular accumulation of the IL2R reporters according to Figure 5, and their percentage among all transfected cells is presented. (A) Quantification for various fragments of the experiments shown in Figure 5. (B) Dose dependence of the effect of the di-leucine-binding domain of 2 on the localization of the IL2R-LL reporter. Constant amounts of IL2R-LL expressing plasmid were cotransfected with increasing amounts of a plasmid expressing the 2F fragment, and the effect on IL2R internalization was scored as in A. (C) Specificity of the block of internalization for di-leucine-based signals. The effect of the 2F fragment was monitored on Tac reporter constructs that are either internalization incompetent (Tac) or are internalized via LL- and YxxL motifs, respectively.

The effects of the di-leucine-binding domains coexpressed with IL2R-LL in NIH3T3 cells were strictly dose dependent, with an efficient block of internalization already at a 1:1 M ratio of plasmid reporters and effectors used in our transfections (Figure 6B). Finally, we were analyzing the specificity of this interaction for di-leucine- and tyrosine-based sorting motifs. Coexpression of the 2F fragment with a transmembrane chimera containing a functional YxxL sorting motif (Tac-YxxL; Marks et al., 1996) did not block its internalization, whereas the internalization of the Tac-LL chimera, bearing the di-leucine-based motif of CD3, was efficiently blocked (Figure 6C). These results reflect the specificity of YxxL motifs for the medium-chain of adaptor protein complexes (Ohno et al., 1995; Owen and Evans, 1998) and support the specific targeting of LL-based sorting motifs to -adaptins.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Our data suggest the determination of a shared domain in -adaptins and the regulatory subunit H of the vacuolar ATPase with significant structural and functional similarity. Furthermore, in -adaptins this domain exhibits specificity for di-leucine-based sorting motifs and is involved in endocytic trafficking. This finding supports previous results of limited tryptic proteolysis of AP-1, which suggested that the interaction site of the LL motif in CD3 resides in the N-terminal ~65-kDa trunk portion of 1 (Rapoport et al., 1998).

In this report we show that the di-leucine-based internalization motif of Nef binds to the ARM repeat structure of V1H (133-483), which exhibits sequence homology to -adaptins (Figures 2-4). Expression of the homologous fragments from V1H, 1, and 2 in cells blocks the internalization of transmembrane proteins, which depend on di-leucine-based sorting motifs (Figures 5 and 6). Both -adaptin fragments 1F and 2F from adaptor protein complexes AP-1 and AP-2 display a homogenous distribution throughout the cytoplasm and the plasma membrane of transfected cells (Figure 5). This observation suggests that the specificity of AP-complexes for different subcellular localizations results from subunits that exhibit a higher degree of heterogeneity, namely the large - and -subunits (Boehm and Bonifacino, 2001), whereas the highly homologous fragments 1F and 2F identified here (92.3% sequence identity) form a structurally and functionally similar domain. Thus, expression of the various LL-binding domains should lead to their association with di-leucine motifs at virtually all subcellular locations without discriminating between distinct transport complexes. Because the LL-binding domains miss the hinge region containing the clathrin box signal as well as the successive -appendage domain, the recruitment of transport competent complexes is inhibited upon binding to a di-leucine motif. Interestingly, this dominant negative effect is reminiscent to that of the VHS-GAT construct of the GGA1 protein used by Puertolano et al. (2001).

Very recently, the acidic-cluster-di-leucine motif of the cytosolic tails of sortilin and the mannose 6-phosphate receptor was found to bind the VHS domain of GGA proteins (Nielsen et al., 2001; Puertollano et al., 2001; Zhu et al., 2001). The monomeric GGAs are a multidomain protein family implicated in protein trafficking between the Golgi and endosomes. Previous structural analysis shows that the small 18-kDa VHS domain of the Hrs protein consists of three HEAT or ARM repeats (Mao et al., 2000), a protein fold that has been recently found also in the structure of the regulatory subunit H of the V-ATPase (Sagermann et al., 2001). On the basis of the sequence similarity found, we conclude that also -adaptins and -COP are HEAT or ARM repeat-containing proteins. These observations suggest that HEAT or ARM repeats form the structural scaffold for the recognition of di-leucine-based sorting motifs. The detailed specificity for the sequence motifs recognized, however, has to be determined individually for each protein family. In accordance with our data, additional specificity for the binding to LL-motifs in -adaptins may come from the µ-chain of the adaptor protein complexes as suggested before (Hofmann et al., 1999). Because the N-terminal trunk of -adaptins is supposed to bind the µ-chain in the AP assembly (Hirst and Robinson, 1998; Kirchhausen, 1999), the interface of these two molecules could contribute to a combinatorial surface for di-leucine-based motif recognition.

A major difficulty in the analysis of endocytic trafficking compartments is with the low-affinity recognition of the various motifs in vitro (Marsh and McMahon, 1999; Pearse et al., 2000). For the LL-motif, this observation is additionally paired with a low signature specificity because leucines are the most abundant residues, and two successive hydrophobic residues occur often statistically. Formation of a multiple helix bundle, as is the repetitive HEAT or ARM repeat fold, is often less sensitive to N- or C-terminal truncation than a -pleated sheet, as is, e.g., the YxxL binding domain in µ2. In µ2, the first -sheet of the 280-residue YxxL-binding domain associates with the second last -sheet (Owen and Evans, 1998), and truncations at both ends result in the immediate loss of binding recognition (Aguilar et al., 1997). For the N-terminal trunk portion of -adaptins with its proposed helical structure instead, we suggest that a protein fragment that does not reflect precisely the required LL-binding domain but exhibits flexible linker segments may block its own target site. Therefore, the transfer of the mapping results from V1H based on sequence similarities to -chains may have been key to determine a fragment in -adaptins that binds to di-leucine-based sorting motifs.

On a speculative level we suggest that the clathrin box signal LLNLD (Shih et al., 1995) or other LL sites in the flexible hinge region of -adaptins compete in a dynamic exchange process with di-leucine-sorting motifs for the binding to its target site. This intramolecular interaction would be disrupted by the recognition of a bona fide LL-sorting signal, which leads to continuous exposure of the clathrin box signal and subsequently induces the assembly of AP-clathrin coats. Interestingly, in 1 and 2 adaptins 5 LL and 3 LI sites are present within the 194 residues between the LL-binding domain identified here and the -appendage domain (Figure 7B). This autoinhibition could induce a functional switch by a conformational change that indicates cargo uptake and initiates clathrin coat formation. This model would also explain why empty clathrin cages are never observed in vivo (Kirchhausen, 1999) and correlate with previous observations by electron microscopy that show various dispositions of the appendage domain relative to the trunk portion (Heuser and Keen, 1988). Interestingly, autoinhibition by an internal nuclear localization signal was discovered for Importin  and found to explain the regulatory switch between the cytoplasmic, high-affinity form, and the nuclear, low-affinity form for NLS binding of the Importin (Kobe, 1999).

View larger version (37K): [in this window] [in a new window]   Figure 7.   Model representation of the di-leucine sorting motif in Nef and proposed domain organization of its target sites. (A) Sequence of the flexible loop (residues 152-184) of HIV-1 Nef (SF2) and model structure. Eight adjacent aspartic and glutamic acid residues form a strong negative charged cluster at the N- and C-termini of the flexible loop while the di-leucine-based motif ExxxLL is best exposed at its center. The degree of similarity conservation based on 186 analyzed sequences (Geyer and Peterlin, 2001) is printed aside the sequence. The formation of the negative charged cluster is indicated in the electrostatic surface display (right). The figure was generated with GRASP (Nicholls et al., 1991) using an electrostatic potential display of 16 kBT (red) to +16 kBT (blue). (B) Proposed modular organization for the di-leucine binding domain in 2 and its homologous domains in V1H and -COP. The flexible hinge region containing the clathrin box signal and the -appendage domain are marked. Protein domain sequences share 20.3% identity between V1H and 2 and 19.9% identity between 2 and -COP. The consensus site GEY in all three proteins (position 415 in V1H) is indicated.

Under normal circumstances, CD4 molecules are internalized from the plasma membrane via a di-leucine motif in their cytoplasmic tail. How can Nef enhance CD4 endocytosis from the plasma membrane using a similar di-leucine-based motif for internalization? Most di-leucine-based sorting motifs contain an upstream acidic residue (D/ExxxLL) or additional phosphorylation sites (Wilde and Brodsky, 1996). A minimal spacing from the plasma membrane as well as these acidic residues N-terminal to the LL-motif were found to be critical for internalization (Geisler et al., 1998). Although the cytosolic part of CD4 does not contain these acidic residues but is rather positively charged (pI 11.7), unless phosphorylated on its serine residues (Pitcher et al., 1999), the 33 amino acids encompassing flexible loop of Nef contains 10 acidic residues (pI 4.0) mostly located at its two ends (Figure 7A). These residues face each other and form a highly conserved negative cluster upstream to the LL-motif (Geyer and Peterlin, 2001), which can be described as a EEx₈LLx₈DD internalization signal. In the membrane-bound Nef protein these acidic residues may lead to exposure of the LL-motif to the cytosol and satisfy the preference of negative charges for the recognition of the di-leucine-binding domain and therefore enhance internalization of the CD4-Nef complex compared with CD4 alone. Thus, the interaction of Nef with CD4 would transform the phosphorylation depended di-leucine signal from CD4 into a constitutively active di-leucine signal from Nef.

The similarities found for the fragments in V1H, -adaptins, and -COP suggest a common modular organization of the three different proteins (Figure 7B). They could contribute to recently described shared domain organization of adaptor protein complexes and coatomer assemblies (Eugster et al., 2000; Boehm and Bonifacino, 2001). Also, our results suggest a related function for the regulatory subunit H of the vacuolar ATPase. Because interactions between the V₁ and V₀ sector of the V-ATPase are dynamic and regulated by extracellular conditions (Kane, 2000), V1H could act as a specialized trafficking molecule. Future studies will unravel whether the entire V-ATPase is required for the functions of V1H in intracellular sorting and how these processes are regulated. With the identification of the domain organization in the N-terminal trunk of -adaptins, precise functional and structural studies are now possible. The dominant negative effects of the LL-binding domains should become useful for functional studies on the trafficking of proteins that contain di-leucine-based sorting motifs. Moreover, the stability of the identified domain appears promising for its subsequent crystallization and structural characterization.