A novel conserved targeting motif found in ABCA transporters mediates trafficking to early post-Golgi compartments.

The ATP binding cassette, class A (ABCA) proteins are homologous polytopic transmembrane transporters that function as lipid pumps at distinct subcellular sites in a variety of cells. Located within the N terminus of these transporters, there exists a highly conserved xLxxKN motif of unknown function. To define its role, human ABCA3 was employed as a primary model representing ABCA transporters, while mouse ABCA1 was utilized to support major findings. Transfection studies showed colocalization of both transporters with surfactant protein C (SP-C), a marker peptide for successful protein targeting to lysosomal-like organelles. In contrast, alanine mutation of xLxxKN resulted in endoplasmic reticulum retention. As proof of principle, swapping xLxxKN for the known lysosomal targeting motif of SP-C resulted in post-Golgi targeting of the SP-C chimera. However, these products failed to reach their terminal processing compartments, suggesting that the xLxxKN motif only serves as a Golgi exit signal. We propose a model whereby an N-terminal signal sequence, xLxxKN, directs ABCA transporters to a post-Golgi vesicular sorting station where additional signals may be required for selective delivery of individual transporters to final subcellular destinations.

is essential for directing the proteins to post-Golgi vesicles. Mutation of the LLLWKN motif to alanine produced cellular expression patterns of reticular forms that colocalized with the ER marker calnexin. Moreover, scanning alanine mutagenesis of this motif, as well as mutations of its fl anking upstream and downstream residues, showed the importance of this motif for subcellular traffi cking. As proof of principle, chimeric construct and substitution mutation studies were also performed employing an unrelated protein, surfactant protein C (SP-C), whose targeting motif is well defi ned. Together, the results demonstrated that this motif mediates Golgi export but is not a targeting determinant for fi nal disposition of ABCA transporters. This fi nding offers a novel model for protein sorting whereby a common signal motif directs proteins to transitory vesicles (sorting station) prior to their individual distribution to their respective destinations.

Materials
The pEGFP-C1 and N1 and DsRed-C1 and N1 monomer plasmids were purchased from Clontech, Inc. (Palo Alto, CA). Using tagged constructs were necessary to follow traffi cking of the ABCA transporters as we were unable to fi nd reliable antibodies to generate reproducible data. Tissue culture medium was produced by the Cell Center Facility, University of Pennsylvania. Except where noted, all other reagents were electrophoresis, tissue culture, or analytical grade and were purchased from Sigma Chemical, Inc. (St. Louis, MO) or BioRad, Inc. (Melville, NY).

EGFP and DsRed/proSP-C fusion protein constructs
Plasmids consisting of enhanced green fl uorescent protein (EGFP) and DsRed fused to the N terminus of wild-type human SP-C (EGFP/SP-C  ) were cloned into the EGFPC1 and DsRedC1 expression vectors and have been previously characterized ( 17,23 ). N-terminus-tagged (instead of C-terminus-tagged) SP-C isoforms were chosen because the SP-C proprotein is cleaved in the distal C terminus (at residue ‫ف‬ 145) in a non-cell-specifi c manner at the early stage of post-Golgi traffi cking ( 24,25 ) ( Fig. 1 ), and the entire C terminus of the proprotein is required to initiate proper posttranslational processing and targeting ( 26 ). Using these constructs as templates, EGFP or DsRed mutants were produced where the targeting motif of SP-C (SPPDY) was replaced by the ABCA, LLLWKN motif. Moreover, DsRed ABCA3/SP-C chimera containing the N-terminal domain of ABCA3 (ABCA3  ) and C-terminal domain of proSP-C (SP-C  ) were produced. For these constructs, a single round, two primers (include restriction sites) PCR amplifi cation and subcloning strategy was used as described previously ( 17 ) with the oligonucleotide sets in supplementary Table I.

ABCA1/FLAG constructs
The mouse full-length ABCA1 WT /FLAG in pcDNA3.1 vector (Invitrogen, Eugene, OR) was the generous gift of Dr. Daniel J. Rader (Institute for Translational Medicine and Therapeutics, University of Pennsylvania). The ABCA3 WT /FLAG was used as a template to generate mutant constructs using the primers in supplementary Table I and a one-step PCR amplifi cation and subcloning approach using the QuikChange II XL Site-Directed Mutagenesis Kit according to the manufacturer's instructions (Stratagene, La Jolla, CA). To generate all six consecutive alanine mutations using this kit, two rounds of PCR amplifi cation ABCA7, ABCA9, ABCA10, and ABCA12 , suggest a distinct role for each of these transporters in lipid homeostasis functioning in various subcellular compartments as well (8)(9)(10)(11). To date, 12 members of the human and 16 members of the rodent ABCA transporter families have been identifi ed ( 12 ). The complete coding regions and genomic structures have been determined for many of this subfamily of transporters. Among those transporters that have been sequenced, ABCA1, ABCA3, ABCA4, and ABCA7 exhibit high percentage of amino acid homology ( 11 ) ( Table 1 ).
The understanding of the regulation of integral membrane protein traffi cking is constantly evolving. Central to this process is the presence of cytosolic signals (4-25 amino acids) within the primary sequence and/or a conformationally determined epitope that directs proteins to subcellular organellar compartments and/or plasma membrane (13)(14)(15). In addition to large domain signals with complex structural architecture, small-size motifs containing short spanning amino acid residues have been shown to serve as major determinants in directing the intracellular routes and localization of protein to their site of function or secretion. Synaptic vesicle-targeted proteins ( 15 ), insulin receptor sorting to lysosomes ( 16 ), surfactant protein C routing to the lamellar body of alveolar type II cells (17)(18)(19), and the adherens junction-targeted shrew-1 transmembrane proteins ( 20 ) all represent proteins containing small targeting motifs that play a role in guiding proteins through proper intracellular routes and/or to their fi nal destinations.
Previously, we reported that ABCA3 is preferentially traffi cked to lamellar bodies in lung epithelial cells, and to lysosomes and lysosomal-like organelles in A549 and HEK293 cell lines ( 5,21 ). Since the mechanisms governing proper traffi cking of the ABCA3 protein or any of the other ABCA transporters are largely undefi ned, we sought to identify domain(s) mediating their normal biosynthetic routing. Structural and sequence analysis of ABCA transporters have been reported ( 1 ). These reports, coupled with computer-based modeling, revealed that nearly all ABCA transporters share signifi cant homology in each protein's N-terminal domain, including a highly conserved motif (xLxxKN). Moreover, recently developed computerassisted prediction strategies ( 12 ) revealed that all ABCA transporters, with the exception of ABCA10, have putative short N-terminal domains, which were consistent with previous reports ( 11,12 ) Mack,et al. ( 22 ), with unknown function, where "x" may represent any residues other than charged amino acids or proline. We hypothesized that this conserved sequence could serve as a part of an essential motif for signature traffi cking of ABCA transporters.
To investigate the role of this motif, we utilized the human (hABCA3) and the prototypic mouse (mABCA1) isoforms as models of ABCA transporters. We report that the conserved six-residue-spanning motif of these transporters cle exit, cells were treated with 10 µM monensin concomitant with transfection.
The fi rst 50 amino acids of each transporter are displayed. First transmembrane topologies were determined using the TopRed and HMMTOP prediction web-based server tools. Dark-gray highlights indicate the xLxxKN targeting motifs; light-gray highlights indicate predicted fi rst transmembrane domains; asterisks indicate ABCA transporters used in the present study.  Table I, 4a and 4b). In the fi rst round, an ABCA1 plasmid containing the fi rst three amino acid mutation was created. In the second round, the plasmid created in the fi rst round was used as a template to mutate the additional three successive residues.

ABCA3/EGFP and ABCA3/DsRed constructs
The human full-length ABCA3 WT /EGFP was generated by amplifying three overlapping ‫ف‬ 2 kb segments of cDNA by RT-PCR as described previously ( 21 ). The plasmid was used as a template to generate DsRed fusion and mutant constructs by using either the single-round, two-primer PCR amplifi cation and subcloning strategy or the one-step PCR amplifi cation and subcloning method using the QuikChange kit described above.

DNA constructs
To ensure that no random alterations were introduced, the sequences of the coding regions of all wild-type (WT) and mutant constructs were verifi ed using primers (overlapping primers, when necessary) and by automated DNA sequencing performed by the Core Facility in the Department of Genetics at the University of Pennsylvania.

Statistical analysis
Experimental data were analyzed by one-way ANOVA with the Tukey-Kramer posthoc test using GraphPad InStat software, version 3.0 for Windows (GraphPad Software, San Diego, CA). All values are means ± SE. Signifi cance was accepted at P < 0.05.

ABCA3 and ABCA1 showed distinct subcellular expression patterns
Clones of the full-length human ABCA3 (hABCA3) and mouse ABCA1 (mABCA1) were inserted into EGFP, DsRed, or FLAG expression vectors ( Fig. 1 ) to characterize the traffi cking and subcellular compartmentalization of the transporters. Using alveolar epithelial cell line A549, expression of hABCA3/EGFP displayed nearly total colocalization with CD63-positive vesicles (a marker antigen associated with lamellar bodies and lysosome-like vesicles ( 28 ) ( Fig. 2A , top row). In addition to organelle marker antigens, expression of wild-type surfactant protein C (SP-C) precursor (proSP-C) was used as an independent protein marker for traffi cking of the ABCA transporters. In alveolar type II epithelial cells, SP-C is synthesized as a 21 kDa bitopic integral membrane precursor, which undergoes four proteolytic cleavages as it is traffi cked through the biosynthetic pathway to yield a 3.7 kDa mature peptide ( 25,29 ). In A549 cells, only the fi rst of four cleavages (at or near residue 145 of the COOH propeptide) occurs in early post-Golgi compartments in a non-cell-specifi c manner ( Fig. 1 ). Both hABCA3 and SP-C are traffi cked to the lamellar body, a highly specialized lysosomal-like organelle of alveolar epithelial cells responsible for regulated secretion of surfactant lipids and proteins ( 21,25,30 ). Lysosomal-like organelles containing proSP-C can be used as a surrogate for traffi cking in A549 cells ( Fig. 2A , bottom

Characterization of integral membrane proteins: carbonate extraction
Cell pellets collected by scraping cells from 35 mm dishes and centrifuged at 300 g were resuspended with ice-cold PBS (137 mM NaCl, 10 mM Na 2 HPO 4 , 2.7 mM KCl, 1.8 mM KH2PO4, pH 7.4) containing protease inhibitors (1 mM phenylmethylsulfonyl fl uoride, 1 µg/ml leupeptin, 1 µg/ml pepstatin A, and 1.5 µg/ml aprotinin) and sonicated on ice three times with 20 s bursts at 50 watts. The sonicate was centrifuged at 9,000 g at 4°C for 30 s to remove nuclei. The nuclei-free suspension was centrifuged at 4°C for 1 h at 100,000 g to separate cytosolic (supernatant) from integral/peripheral membrane (pellet) proteins. Following removal of the supernatant, the pellet was resuspended with 100 mM sodium carbonate (pH 11.5), incubated at 0°C for 30 min, and centrifuged at 4°C for 1 h at 100,000 g to separate the integral (pellet) from the peripherally associated membrane proteins. Samples were prepared for immunoblotting by resuspending pellets in lysis buffer (50 mM Tris, 190 mM NaCl, 6 mM EDTA, 2% Triton X-100, pH 7.4) containing protease inhibitors or by precipitating supernatants with 10% (w/v) trichloroacetic acid on ice, pelleting by centrifugation at 4°C for 10 min at 12,000 g , washing twice with acetone, drying with N 2 , and dissolving in lysis buffer.

SDS-PAGE
Cell pellets collected from 35 mm culture dish by scraping and centrifugation at 300 g were solubilized with 40 l of lysis buffer. Following centrifugation at 6,000 g for 30 s to remove nuclei, proteins were separated by electrophoresis on a 12% polyacrylamide gel and transferred to nitrocellulose membrane.

Immunoblotting
Immunoblotting of transferred samples was performed using successive incubations with primary anti-EGFP 1:5000 (Clontech, Mountain View, CA) or anti-DsRed 1:1500 (Clontech) and horseradish peroxidase-conjugated secondary antibodies as previously described ( 23 ). The resulting fl uorescence images visualized by ECL were captured either by exposure to fi lm or by direct acquisition using the Kodak 440 image system.  1. Schematic representation of the ABCA3 and ABCA1 expression constructs used in this study. Twenty-seven different fusion constructs tagged with either an EGFP, a DsRed, or FLAG reporter were used. SP-C isoforms were deliberately tagged at the N terminus to follow the protein to its terminal destination since proSP-C is cleaved at the distal C terminus at or near residue 145 in early post-Golgi compartments in a non-cell-specifi c manner. mABCA1 cDNA clones were systematically altered by sitedirected mutagenesis. Polyalanine mutation of the L 9 LL-WKN 14 motif of both ABCA3 and ABCA1 proteins resulted in the retention of each transporter in compartments positive for the ER marker calnexin ( Fig. 3B , E , right columns). Neither mutant transporter colocalized within SP-C-positive vesicles ( Fig. 3B, E , left columns). Control mutations of adjacent residues located on either side of this motif were generated that included V 3 17 for mABCA1. At least two of the juxtamembrane positively charged residues were left unchanged, since mutations of these types may affect membrane topology, causing misfolding and/or disrupted traffi cking ( 21,(31)(32)(33)(34). Transient expression of these controls produced traffi cking patterns for the ABCA isoforms that were similar to those of their wild-type counterparts, with hABCA3 predominantly ( Fig. 3A, C ) and ABCA1 partially ( Fig. 3D ) colocalizing with SP-C-positive vesicles as was shown in Fig. 2 .

Scanning alanine mutagenesis of the N-terminal motif revealed three distinct patterns of expression
Since hABCA3 and SP-C are each routed to lysosomallike organelles in lung epithelial cells ( 21,25,30 ), the relative signifi cance of individual residues of the LLLWKN motif for proper traffi cking of ABCA transporters was assessed. To accomplish this, scanning alanine mutagenesis of the ABCA3 transporter protein in combination with coexpression of SP-C in A549 cells was performed. A total of 14 additional mutant ABCA3 constructs were generated, consisting of various combinations of alanine substitutions within the motif ( Fig. 4A , panels a-n). Representative fl uorescence colocalization images acquired for each mutant ABCA3 isoform using either SP-C co-expression or calnexin staining are shown in Fig. 4A . When quantitated by cell counting and expressed as percentage of cells co-expressing ABCA3 isoforms with either SP-C or calnexin, three distinct patterns emerge. Group I represents ABCA3 expression that resembled those of the wild-type isoform and showed predominant vesicle colocalization with SP-C, which included alanine mutations of L 9 , N 14 , (L 9 and L 11 ), and (L 9 and N 14 ) ( Fig. 4A , panels a-e). Group III mutant isoforms ( Fig. 4A , panels m, n), which included alanine mutations of the middle four and the latter fi ve residues of the motif, showed total ER retention similar to that observed when the entire motif was mutated to alanine ( Fig. 4B ). Group II mutations ( Fig. 4A , panels f- , however, exhibited mixed patterns of expression localizing within calnexin-positive ER as well as in SP-C-positive vesicles ( Fig. 4B ).
Further analysis of cellular expression patterns of the mutant ABCA3 N-terminal motif isoforms categorized the relative signifi cance of the six residues of the motif. Mutations of residues L 9 ( Fig. 4A , panel a) and/or N 14 ( Fig. 4A , panels d, e) showed little or no effect on the transporter's cellular distribution. In contrast, individual mutations of residues L 10 ( Fig. 4A , panel f; 4B) and K 13 ( Fig. 4A , panel row; 2D ). In contrast to hABCA3, mABCA1 showed only partial and variable colocalization with CD63-positive vesicles ( Fig. 2B , top row), with SP-C-containing vesicles ranging 30-75% of total fl uorescing vesicle counts ( Fig. 2B , bottom row; 2D ), as well as hABCA3-containing vesicles ( Fig. 2C ).

mABCA1 and hABCA3 mutants lacking xLxxKN were retained in proximal compartments
To examine the role of ABCA N-terminal xLxxKN motif, the N-terminal domains of full-length hABCA3 and i; 4B ) appeared to be critical for targeting. Moreover, L 9 , L 11 , and N 14 may play a role for the optimal vesicle targeting. That is, single and double alanine mutations of L 9 ( Fig. 4A , panel a), L 11 ( Fig. 4A , panel b), or L 9 and N 14 together ( Fig. 4A , panel e) did not affect proper traffi cking. However, mutations of L 10 combined with L 9 and L 11 ( Fig.4A , panel k) or combined with K 13 and N 14 ( Fig.4A , panel l) resulted in more ER-retained proteins ( Fig. 4B ). Alanine substitution for tryptophan (W 12 ) was excluded since some ABCA transporters do not contain this residue in their motif, suggesting a lesser importance in targeting. In total, based on the data acquired from Fig. 4B , ranking from highest to lowest in infl uencing the translocation of the transporter to post-Golgi vesicles, the relative contribution in protein targeting of the residues examined was: L 10 = K 13 > L 11 = N 14 = L 9 . The subcellular distribution of the mutant isoforms was not infl uenced by the level of protein expression. As shown in supplementary Fig. II, varying the amount of plasmid delivered by transfection failed to signifi cantly alter the intracellular fl uorescence patterns observed with hABCA3 mutants, including the group II mutants ( Fig. 4 , panels f-j). Altering plasmid amounts from 2.5 to 0.25 µg/35 mm dish diminished the effi ciency of transfection from 50% downward, with 0.5 µg representing a cutoff for detectable protein expression (data not shown). Confocal microscopy of cells transfected with 0.5 µg plasmid DNA/dish consistently produced expression patterns of these mutants similar to those observed at the higher concentration (4 µg/dish). At the lesser amount, the majority of the transfected hABCA3 mutant was still colocalized in calnexin-positive compartments and in punctate vesicles. In addition, a minimal amount of several different LLLWKN mutants was also found to colocalize with the Golgi marker P230 (supplementary Fig. II).

The N-terminal motif of ABCA3 can facilitate traffi cking of an unrelated fusion protein to post-Golgi vesicles
The targeting domain of SP-C has been well characterized. Using deletion and site-directed alanine mutagenesis, the single targeting motif of SP-C was localized to a PPDY motif within the 35 amino acid cytosolic N-terminal domain of the proprotein. Alanine substitution mutation of this motif resulted in ER retention of the propeptide ( 19,35 ).
As proof of principle and to ascertain the role of xLxxKN motif in targeting proteins to post-Golgi vesicles, two plasmid constructs were generated. First, the entire 35 amino acid N-terminal domain of proSP-C was swapped with the 21 amino acid N-terminal domain of ABCA3. Second, the targeting motif of proSP-C (SPPDY) ( 19 ) was replaced with the ABCA3 N-terminal motif (SP-C SPPDY → LLLWKN ) ( Fig.  1 ). The fi rst construct resulted in a chimera consisting of N-terminal ABCA3 (N 1 -K 21 ), the proSP-C noncleavable signal peptide transmembrane domain (L 36 -L 58 ), and the C terminus (L 59 -I 197 ). As demonstrated in Fig. 5A (top row), the DsRed-tagged ABCA3/SP-C chimera is targeted primarily to vesicles containing ABCA3 following an over- protein products remained integrally associated and anchored to the lipid bilayer of the membrane similar to the wild-type SP-C isoform.

DISCUSSION
The results presented in this study demonstrate that a conserved xLxxKN signal located within the short N-terminal cytosolic domain of two functionally unrelated ABCA transporters mediates the export of these multitransmembrane proteins from the Golgi. The data suggest that, although the motif is necessary for Golgi exit, it is not suffi cient for sorting these proteins to their ultimate targets. Using human ABCA3 and mouse ABCA1 as models of ABCA transporters, we showed that alanine substitution for the six-amino-acid-spanning motif can actually abrogate transit of the proteins beyond the ER ( Fig. 3 ). For both transporters, mutations of residues adjacent to this motif did not affect protein traffi cking. Moreover, chimeric constructs, where the targeting motif of SP-C was swapped with the ABCA N-terminal LLLWKN motif, demonstrated its role as the mediator of post-Golgi export. Since this motif is found in nearly all ABCA transporters and each transporter has a specifi c subcellular destination, these results imply that ABCA transporters are segregated in proximal vesicles prior to subsequent sorting to their respective terminal targets.
Unlike targeting motifs that specialize in the delivery of proteins to specifi c subcellular compartments -such as lysosomes, mitochondria, or the plasma membrane -the xLxxKN motif appears to deliver its protein to an intermediary sorting station. The uniqueness of this motif was revealed with the employment of a known targeting motif of SP-C, the PPxY Nedd 4-2 binding motif ( 19,35 ). The biosynthesis of SP-C through the regulated pathway of alveolar epithelial cells is well established ( 41,42 ). While the Nedd 4-2 binding motif of SP-C contains the necessary signal for transport and delivery of the peptide to lamellar bodies in alveolar epithelial cells and to lysosome-like vesicles in the A549 epithelial cell line, the xLxxKN motif facilitated ER export, but it failed to deliver the motif-swapped mutants of SP-C to similar vesicles for processing ( Fig. 5 and Fig. 6 ).
Previously, we had functionally characterized a series of disease-related mutations in hABCA3 using similar methodologies ( 5 ). Most prominent of these was L101P, which, when transfected into similar cell lines, produced profound ER retention. In addition to this traffi cking mutant, another mutant, N568D, was in fact traffi cked to lysosomes but failed to concentrate NBD-phosphatydilcholine in this compartment. While the possibility exists that mutation of the xLxxKN motif infl uences the transporter function of hABCA3, the structural and spatial features of this motif relative to the ABC cassette suggest that disruption of targeting rather than function is more likely. Specifi cally, the large distance and structural gap between these two domains (at least over 500 residues and six transmembrane domains) makes them less likely to infl uence one another unless mutations of the xLxxKN motifs can induce considerable conformational night transient expression. Similarly, the second construct, a DsRed-or EGFP-tagged proSP-C fusion plasmid containing the LLLWKN motif (instead of its PPDY targeting motif) showed partial vesicular colocalization with both ABCA3 ( Fig. 5A , middle row) and CD63 ( Fig. 5A , bottom  row). Together, these results suggest that the N-terminal motif of ABCA3 is capable of targeting an entirely unrelated protein to post-Golgi compartments. Further quantitation revealed that, whereas there was nearly total vesicle colocalization between the wild-type ABCA3 and SP-C isoforms, only partial colocalization was observed between the wild-type ABCA3 and either the chimera or SP-C SPPDY → LLLWKN isoforms ( Fig. 5B ).
To further defi ne the proximal compartment targeting mediated by the ABCA LLLWKN motif, transfected cells expressing wild-type EGFP/SP-C or DsRed/SP-C SPPDY → LLLWKN were treated with monensin to inhibit vesicular export from the trans-Golgi ( 36 ). As Fig. 5C shows, in the absence of monensin, SP-C WT and SP-C SPPDY → LLLWKN were localized in punctate vesicles that were negative for the trans-Golgi marker P230 ( Fig. 5C , left columns). However, in the presence of monensin, both SP-C isoforms were localized in P230-positive compartments ( Fig. 5C , right columns), indicating that these proteins reached a trans-Golgi compartment but were blocked from exiting by monensin.

Post-Golgi-targeted ABCA3 N-terminal motif proteins do not process chimeras
The terminal destination for the ABCA3 and SP-C SPPDY → LLLWKN chimeric constructs was functionally assessed using the proSP-C construct as a control. When A549 cells are used as a surrogate for alveolar type II cells, in addition to their colocalization with CD63, a biochemical signature for traffi cking to post-Golgi distal vesicles is their partial cleavage to a lower molecular weight intermediate ( 23,(37)(38)(39). Immunoblotting of the membrane fraction of transfected cells with anti-GFP or anti-DsRed demonstrated processing of wild-type SP-C as shown by bands ( Fig. 6 , second panel, lanes b, d, e) corresponding to the predicted Mr of the primary translation product of the fusion protein (48,000) and a smaller intermediate cleaved product (arrow) as previously reported ( 23,(37)(38)(39). In contrast, neither DsRed/ABCA3-Chimera nor EGFP/ SP-C SPPDY → LLLWKN was suffi ciently processed with disproportionate expression of their respective primary translation products ( Fig. 6 , third and bottom panels, lanes b, d,  e). The processing profi le was similar to those reported in studies related to the deletion or substitution mutation of the targeting domain of SP-C ( 19,40 ). Together, the data suggest that, following Golgi exit, the LLLWKN motif of ABCA3 is insuffi cient to target the two mutant proteins to distal compartments for processing.
With the exception of cells expressing the vector alone ( Fig. 6 , top panel), no translation product was detected in post 100,000 g centrifuged cytosolic fractions ( Fig. 6 , lane  c). In addition, Na 2 CO 3 treatment of the membrane fractions did not release any of the pellet-associated proteins ( Fig. 6 , lane f), indicating that structurally, both fusion Contrary to the original models for protein sorting proposed two decades ago, many recent studies have revealed that delivery of proteins to subcellular compartments and plasma membrane is not a simple mass-transit process but changes of the transporter regions. However, like L101P, xLxxKN mutations would naturally affect functional outcome if the transporters were held from reaching their destined compartments. provide additional resolution (right panels). Bar, 5 µm. (B) Quantitative cell expression patterns for hABCA3/EGFP mutant proteins. Colocalization of DsRed/SP-C or calnexin marker with each hABCA3/EGFP mutant was determined by manual counting 16-24 h following transfection. At least 150 cells expressing both proteins or protein and marker from three separate experiments were evaluated. * P < 0.001, ** P < 0.05 compared with wild-type hABCA3. uses the "Golgi-localized ␥ -ear-containing ARF binding proteins" (GGA), ubiquitous cytosolic proteins that are essential for the accurate traffi cking of MPRs from the TGN to endosomes ( 49 ). The GGAs recognize an acid clusterdileucine signal in the cytosolic tail of MPRs ( 50 ). Thus, although the binding cytosolic partner(s) for ABCA transporters have yet to be determined, it is likely that cytosolic adaptors for ABCA transporters exist not only to mediate delivery to their fi nal destinations but also to allow their initial transient segregation (via xLxxKN motif) within common vesicles prior to their subsequent traffi cking steps.
The mutagenesis data presented provides some indication that this novel six-residue motif functions as a targeting unit. However, further examination of other N-terminal ABCA domains, the relative contribution of each xLxxKN motif residue or the involvement of the neighboring residues fl anking either side of the motif to post-Golgi vesicular traffi cking, as well as identifi cation of binding partners of the motif all remain to be fully elucidated and may be necessary for a complete understanding of how the targeting motif operates. For example, the data suggest the possibility that the actual size of the ABCA targeting motif could be reduced by at least two residues, making it an LxxK motif, since mutation of the L 9 or N 14 , or both did not signifi cantly affect traffi cking ( Fig. 4 ) and the motifs of ABCA6, ABCA8, ABCA9, and ABCA13 all have alanine at fi rst residue. Based on our data, when the relative contribution of each residue examined was graded, the ranking from highest to lowest in infl uencing the translocation of the transporter to post-Golgi vesicles was: L 10 = K 13 > L 9 = L 11 = N 14 . However, while mutation of L 10 retains more than a third of the protein population in the ER ( Fig. 4 , lane f) and mutations of both L 9 and L 11 or combined L 9 and L 10 or L 10 and L 11 did not signifi cantly affect traffi cking ( Fig. 4 , lane c), a combined mutation of all three residues resulted in a signifi cantly higher ER retention of the transporter ( Fig. 4 , lane k). A similar phenomenon was also observed with K 13 and N 14 residues ( Fig. 4B , lanes i, j, l). Taken together, the data suggest that these seemingly innocuous residues, when they are separately mutated, play an essential role contributing to the targeting effi cacy of the motif.
In conclusion, during the last two decades mutagenesis approaches have led to the discovery of amino acid sequences affecting the sorting of membrane proteins in the biosynthetic pathway. In those studies, various short, linear amino acid motifs have been identifi ed as important signals for targeting to specifi c anterograde/retrograde intracellular destinations or secretion ( 23,43,51 ). Although the teleological advantages for such a mechanism is not clear as yet, the overall sorting system demonstrated in the present study appears to rely upon signals constituting short amino acid sequences that operate in a manner analogous to that of the United States ZIP code ( 49 ). The nine-digit ZIP codes were created primarily for more effi cient routing of mail throughout the country (USPS technical guide, http://ribbs.usps.gov/fi les/CASS/TECH-NICAL_GUIDES/CASSTECH.PDF). The fi rst three digits involves sorting cargo into distinct pathways initiated from the trans-Golgi network (TGN) extending to other sorting stations ( 43 ). For example, the biogenesis of mature granules involves the budding of the immature secretory granules from the TGN, followed by a number of maturation and sorting steps ( 44,45 ). The sorting of these granules implies the segregation of vesicles containing constitutive secretory proteins and lysosomal enzymes from those containing peptides, such as prohormones and hormones. Such steps suggest sorting mechanisms that exist beyond the TGN (46)(47)(48).
Whereas the above model focuses primarily on the vesicles carrying the cargo's luminal proteins and the maturation steps taken to concentrate these proteins to mature vesicles, a second model focuses on the transmembrane proteins themselves. Sorting of transmembrane proteins involves recognition of short peptide signals in their cytoplasmic tails by special cytosolic proteins, which function as adaptors to link cargo transmembrane proteins through deformation and budding. One class of transmembrane protein cargo that uses such a system is the mannose-6phosphate receptor (MPR), which is crucial for normal lysosomal function. This class of transmembrane proteins hint at which large region (state), while digits four and fi ve take into consideration more local factors (town or village) and mail delivery routes. Together, letters and packages bearing the same fi ve digits are initially delivered to a central postal station before subsequent distribution to their respective addresses. The extra four-digit code following the fi ve digits provides further precision for local delivery. Comparably, cells appear to have developed sorting machineries to operate in such a manner to promote traffi cking effi ciency beyond simple TGN sorting. That is, short spanning amino acid residues may function as codes not only to target proteins to their destinations but also to segregate proteins containing similar codes to a specifi c intracellular sorting station, such as small vesicles and endosomes. Additional signals (analogous to the extra fourdigit postal codes) are likely located elsewhere within the protein to mediate fi nal delivery.