© 2002 Macmillan Magazines Ltd

NATURE CELL BIOLOGY VOL 4 FEBRUARY 2002 http://cellbio.nature.com154

brief communications

AP-4 binds basolateral signals andparticipates in basolateral sortingin epithelial MDCK cells

Thomas Simmen*†#, Stefan Höning*‡, Ann Icking§, Ritva Tikkanen§ and Walter Hunziker†¶**†Institute of Biochemistry, University of Lausanne, CH-1066 Epalinges, Switzerland

‡Biochemistry II, University of Göttingen, 37073 Göttingen, Germany§Institute for Cell Biology, University of Bonn, 53122 Bonn, Germany

¶Institute of Molecular and Cell Biology, 30 Medical Drive, Singapore 117609**e-mail: hunziker@imcb.nus.edu.sg

*These authors contributed equally to the work.#Current address: Vollum Institute, OHSU, Portland, OR 97201, USA

Published online: 21 January 2002, DOI: 10.1038/ncb745

Adaptors are heterotetrameric complexes that mediatethe incorporation of cargo into transport vesicles byinteracting with sorting signals present in the cytosolicdomain of transmembrane proteins. Four adaptors, AP-1(ββ1, γγ, µµ1A or µµ1B, σσ1), AP-2 (ββ2, αα, µµ2, σσ2), AP-3 (ββ3, δδ,µµ3, σσ3) or AP-4 (ββ4, εε, µµ4, σσ4), have been character-ized1,2. AP-1 and AP-3 mediate sorting events at the levelof the TGN and/or endosomes, whereas AP-2 functions inendocytic clathrin coated vesicle formation; no function isknown so far for AP-4. Here, we show that AP-4 can binddifferent types of cytosolic signals known to mediatebasolateral transport in epithelial cells. Furthermore, inMDCK cells with depleted µµ4 protein levels, several baso-lateral proteins are mis-sorted to the apical surface,showing that AP-4 participates in basolateral sorting inepithelial cells.

Epithelial cells establish apical and basolateral plasma mem-brane domains with distinct protein compositions3,4. Sortingof proteins into transport vesicles destined for the apical or

basolateral surface occurs both at the level of the trans-Golgi net-work (TGN) and, after internalization, in endosomes. Basolateralsorting of integral membrane proteins is mediated by short cytoso-lic amino acid motifs, many of which are similar to and/or collinearwith endocytosis signals, and often depend on the same crucialtyrosine or dihydrophobic residues5. Sorting signals interact direct-ly with the subunits of one or more of the four known adaptorcomplexes. Owing to the structural similarities between manybasolateral sorting and endocytosis signals, it has been suggestedthat basolateral sorting might involve adaptors6–8. Indeed, theepithelial cell specific isoform of AP-1, AP-1B, has been shown tomediate basolateral sorting of some proteins in LLC-PK1 cells9.Here, we show that a second adaptor, AP-4, participates in basolat-eral sorting in epithelial MDCK cells.

To characterize AP-4, we stably expressed a µ4 cDNA constructcarrying an N-terminal Myc-epitope tag (Myc–µ4) in MDCK cells.A protein of the size expected for µ4 (~50 kDa) was specificallydetected in transfected cells on a blot probed with anti-Myc anti-bodies (Fig. 1a). Fractionation of cell homogenates into membraneand cytosol fractions showed that approximately half of the totalMyc–µ4 could associate with membranes (Fig. 1a). If cytosol wasfractionated by gel filtration (Fig. 1b), Myc–µ4 was eluted in highmolecular weight fractions, consistent with its incorporation into a

protein complex. Western blot analysis of the gel filtration fractionsrevealed the presence of AP-1 (anti-µ1) and AP-2 (anti-α2, data notshown) in fractions 21–26, whereas AP-3 was recovered in fractions21–29. Myc–µ4 eluted in fractions 27–33, with fractions 30–33being devoid of other APs (Fig. 1c). To confirm the assembly ofMyc–µ4 into an AP-4 complex, we analysed whether other AP-4subunits precipitated together if tagged µ4 was precipitated withanti-Myc antibodies from metabolically labelled cells (Fig. 1d).Labelled proteins corresponding in molecular mass to µ4 (~50kDa), β4 (~80 kDa), ε (~140 kDa) and σ4 (~15 kDa) were precip-itated together from Myc–µ4 producing, but not from control,cells. Immunoblotting of the precipitates with antibodies against ε,β4 and µ4 (antibodies to σ4 were not available) confirmed theidentity of the precipitated proteins as AP-4 subunits. Also consis-tent with the incorporation of Myc–µ4 into AP-4 adaptors, confo-cal light microscopy showed extensive co-localization of epitopetagged Myc–µ4 with endogenous β4 (Fig. 1e).

To analyse the subcellular distribution of AP-4, MDCK cellswere stained with antibodies to µ4, β4 or Myc. Staining with anti-µ4 (Fig. 2a; green) and anti-γ-adaptin antibodies (Fig. 2a; red)showed partial co-localization of AP-4 and AP-1 in the perinuclearregion (Fig. 2a; yellow, arrows). Co-localization was restricted tosmall areas that were often flanked by domains with labelling exclu-sively for AP-1 or AP-4. No co-localization was observed on moreperipheral vesicles. In MDCK cells that produce human furin10

(Fig. 2b; green) and Myc–µ4 (red), both proteins were partiallylocalized to the perinuclear region (yellow, arrows). Staining of theendogenous 46K cation-dependent mannose-6-phosphate receptor(MPR46), (Fig. 2c; green) revealed a more extensive colocalizationwith µ4 (red), but little overlap was observed on more peripheralvesicles. The partial perinuclear colocalization of AP-4 with AP-1,furin and MPR46 is indicative of its localization to subdomains ofthe TGN. Such an interpretation would be consistent with theestablished TGN localization of AP-4 in other cell types11,12 andwith the labelling of the trans region of the Golgi complex ofMDCK cells by immunogold electron microscopy (EM) usingeither anti-µ4 or anti-β4 antibodies (data not shown).

Immunogold-EM with anti-µ4 or anti-β4 antibodies alsodetected AP-4 on vesicles with endosomal morphology, which oftencontained internalized bovine serum albumin (BSA)–gold (Fig. 2e;arrows), showing that some of the peripheral AP-4-positive vesicleswere of endosomal origin. Although labelling efficiencies for AP-4were low, ~20% of the total AP-4 labelling was on membranes con-

© 2002 Macmillan Magazines Ltd

brief communications

NATURE CELL BIOLOGY VOL 4 FEBRUARY 2002 http://cellbio.nature.com158

antibody binding13 were used to monitor the delivery of newly syn-thesized Tac F 780–787 to the apical and basolateral domains. Adecrease in the basolateral delivery of Tac F 780–787 with a corre-sponding increase in apical insertion was observed in µ4-depletedcells compared with control or µ4-sense-expressing cells (Fig. 4c),consistent with a role of AP-4 in biosynthetic transport to the baso-lateral domain. The extent of apical mis-sorting of Tac F 787–780in µ4-depleted cells was similar to that of furin tail chimaera withan inactivated FI motif13. By contrast, the distribution of the apicalmarker gp135 was unchanged in µ4-depleted cells (Fig. 4g).

Similar results were obtained for the endogenous LDLR andMPR46, showing that apical mis-sorting of Tac F 780–787 in µ4-depleted cells was not due to the overproduction of a foreign pro-tein. As expected from the basolateral sorting of the LDLR inMDCK cells6,14, labelled LDL was only internalized from the baso-lateral compartment by control and µ4-sense transfected cells (Fig.4e and data not shown). In µ4-depleted cells, however, significantuptake occurred from the apical side (Fig. 4e), indicating apicalmis-sorting of the LDLR in these cells. Owing to the lack of suitableantibodies to the canine LDLR, a characterization of the biosyn-thetic delivery of newly synthesized receptors was not possible.Also, the MPR46, which is found adjacent to the lateral membranein control cells10, was clustered beneath the apical surface in µ4-depleted cells (Fig. 4f).

Because the inactivation of the basolateral sorting signal oftenresults in apical transport of the corresponding protein, the apicalexpression of Tac F 780–787, LDLR and MPR46 in µ4-depletedcells is consistent with a requirement for AP-4 for efficient basolat-eral sorting. The presence of residual µ4-protein or multiple baso-lateral sorting mechanisms (see below) might explain why basolat-eral sorting was only partially affected in µ4-depleted cells.Interestingly, in LLC-PK1 cells, which are defective in AP-1B-medi-ated basolateral sorting9, the MPR46 was still basolaterally distrib-uted (Fig. 4f), indicating AP-1B-independent sorting. Although wecannot exclude the possibility that the depletion of AP-4 results ina backup of proteins in a pathway distinct from basolateral sorting,thereby leading to apical mis-sorting owing to saturation of thebasolateral sorting machinery, this seems unlikely because basolat-eral sorting of the LDLR is not easily saturated14. The endogenouslysosomal membrane protein lamp-2 could not be detected oneither the apical or the basolateral surface of µ4-depeleted cells(data not shown), indicating that at least lysosomal transport wasnot impaired.

Surprisingly, basolateral expression of the TfR was only margin-ally affected in µ4-depleted cells (Fig. 4b), despite the interaction ofits basolateral sorting signal with AP-4. Residual µ4 protein mightbe sufficient for the efficient basolateral sorting of the TfR,although this is unlikely given its lower apparent affinity for AP-4compared with the other receptors. Alternatively, the interaction ofthe TfR tail with AP-4 in vitro might be unrelated to basolateralsorting in vivo, a phenomenon observed for other adaptors.Intriguingly, basolateral sorting of the TfR differs in several aspectsfrom that of other proteins. Newly synthesized TfR appears first inendosomes before reaching the basolateral surface of MDCK cells19.Furthermore, although the basolateral sorting signals of the LDLRand the polyimmunoglobulin receptor can be decoded at the levelof the TGN and endosomes20,21, this might not be the case for thesorting determinant of the TfR15.

Because not all proteins seem to depend equally on AP-4 forbasolateral sorting, the role of AP-4 might be restricted to a partic-ular subset of cargo molecules, to one of several possible routes tothe basolateral surface (e.g. directly from the TGN, indirectly viaendosomes) or to a particular sorting compartment (e.g. TGN,endosomes). This would then imply that there are other basolater-al sorting mechanisms that function in parallel or in concert withAP-4. Indeed, AP-1B, an epithelium-specific variant of AP-1 (ref.22), can mediate basolateral sorting9. LLC-PK1 cells lack µ1-B and‘mislocalize’ several proteins (including the LDLR and TfR) to the

apical domain, and exogenous expression of µ1B restores their ‘cor-rect’ basolateral localization. However, AP-1B apparently mediatesbasolateral sorting of only a subset of proteins in LLC-PK1 cells.Furthermore, µ1B is not expressed in hepatocytes and hippocam-pal neurons22, both polarized cell types that rely on classical baso-lateral signals to deliver proteins to the sinusoidal or somatoden-dritic surfaces, respectively23–25. In fact, even unpolarized cells thatdo not express µ1B, such as fibroblasts, use different biosyntheticroutes to the cell surface in a process reminiscent of polarized sort-ing in epithelial cells26,27. Thus, different cell types can use alterna-tive polarized sorting mechanisms, perhaps reflecting different rel-ative expression levels of AP-1B and AP-4. The two adaptors mighttherefore act in concert to generate and maintain polarity in epithe-lial cells, with the ubiquitous AP-4 perhaps playing a more promi-nent role in establishing protein asymmetry in non-epithelial cellsor in epithelial cells that lack AP-1B.

MethodsAntibodies and cell linesAntibodies to Tac (H93), the c-Myc epitope (9E10) and furin have been described13. G. Ojakian

(Brooklyn, NY), G. Thomas (Portland, OR) and J. Gruenberg (Geneva, Switzerland) provided antibod-

ies to gp135, furin and lysobisphosphatidic acid, respectively. P. Schu (Göttingen, Germany) provided a

serum against µ1, M. S. Robinson (Cambridge, UK) provided sera against AP-2 (anti-µ2) and AP-4

(anti-β4 and anti-ε). AP-2 and AP-1 were detected with antibodies to the α (Transduction

Laboratories) and γ (Sigma Chemical) subunits, respectively, and AP-3 was detected with an antibody

to the human δ-adaptin hinge region (amino acids 752–839) or a monoclonal antibody to µ3

(Transduction Laboratories). A µ4 serum was produced by immunizing rabbits with a mixture of two

peptides (human µ4 amino acids 12–22 and 431–446) coupled to KLH (Calbiochem). For immunoflu-

orescence and immunogold EM experiments, the µ4 antiserum was affinity purified on the peptides.

MDCK II cells and clones stably expressing Tac–furin constructs have been characterized13.

µ4 cDNA constructs and transfection of MDCK cellsThe cDNA for µ4 was amplified by PCR from a human kidney cDNA library (Clontech) using a

5′ primer designed to introduce an N-terminal or internal (between amino acids x and y) Myc-epitope

tag. The PCR products were sequenced and inserted in the sense or antisense orientation into the

mammalian expression vector pCEP4 (Invitrogen). MDCK II cells or clones expressing Tac–furin con-

structs13 were stably transfected with the µ4 sense (N-terminal tag) or antisense (internal tag) plasmid

as described13. Clones were selected in 0.5 mg ml–1 hygromycin and screened by immunofluorescence

and western blot. Polarized MDCK cell monolayers were established on Transwell units (0.4 µm pore

size; Costar). Expression from the transfected cDNAs was induced 12 h before the experiment by the

addition of 10 mM sodium butyrate20. Establishment of a tight cell monolayer was verified by measur-

ing the transepithelial electrical resistance using an EVOM ohmmeter (World Precision Instruments).

Cells expressing the µ4-antisense cDNA showed a slightly reduced transepithelial electrical resistance

compared with control or µ4-sense-transfected clones.

Gel filtration and cell fractionationCells from one 15 cm plate were washed and scraped into 2 ml buffer A (20 mM HEPES pH 7.3,

150 mM NaCl, 10 mM KCl, 2 mM MgCl2, 2 mM EDTA, 1 mM DTT) supplemented with protease

inhibitors. The cells were collected by centrifugation, homogenized and a 100,000 g supernatant (cytosol

fraction) was prepared. The cytosol was passed over a 0.8×30 cm TSK size-exclusion column (Toso-

Haas) connected to a perfusion chromatography workstation (Applied Biosystems) at a flow rate of 0.8

ml min–1. Fractions (0.5 ml) were collected and analysed for the presence of adaptors by western blot.

Membrane and cytosol fractions were obtained by high-speed centrifugation (100,000g, 60 min).

Immunofluorescence microscopy and immunogold EMDifferent proteins were localized by confocal microscopy as previously described13. Immunogold EM

was performed according to the Tokuyasu method28. To label the endocytic compartment, cells were

incubated for 90 min at 37 °C with BSA–gold (5 nm) and fixed with 2 % paraformaldehyde and 0.2 %

glutaraldehyde. Ultrathin cryosections were cut from gelatin-embedded frozen samples and labelled

with primary antibodies followed by detection with protein A conjugated to 15 nm gold.

Quantification was performed on randomly selected sections showing good morphology by scoring

the gold particles along a fixed track. All gold particles seen were counted and scored for their subcel-

lular location.

SPR analysisPeptides corresponding to wild-type or mutated signals (Fig. 3) were purified using reverse phase

chromatography, analysed by mass spectrometry and immobilized on a sensor surface according to the

manufacturer’s instructions (BIAcore). A streptavidin-coated surface was used to capture biotinylated

peptides and a CM5 surface was used for direct immobilization of peptides via their N-terminal thiol

group. Measurements were carried out using a BIAcore 3000 biosensor as described29. The use of AP-4-

enriched cytosol fractions for the binding experiments precluded the determination of adaptor con-

centrations and therefore the calculation of rate constants. Relative affinities were estimated by calcu-

lating equilibrium rate constants using an arbitrary value for the concentration of AP-4, setting the

rate constant for the binding of AP-4 to the LDLR peptide to 100 and relating the rate constants for

the interaction of AP-4 with the other peptides to this value.

© 2002 Macmillan Magazines Ltd

brief communications

NATURE CELL BIOLOGY VOL 4 FEBRUARY 2002 http://cellbio.nature.com 159

Antibody and ligand internalization and binding assays.The domain specific binding and internalization of antibodies has been described13. For LDL uptake

experiments, cells cultured in medium containing delipidated serum were incubated with DiL(3H-

Indolium,2-[3-(1,3-dihydro-3,3-dimethyl-1-octadecyl-2H-indol-2-ylidene)-1-propenyl]-3,3-dimethyl-

1-octadecyl-, perchlorate)-labelled LDL (2 µg ml–1; Molecular Probes) added to the apical or basolater-

al compartment for 60 min at 37 °C. The distribution of Tac F 780–787 or TfR was quantified in bind-

ing experiments using radioiodinated anti-Tac antibodies13 or canine transferrin30 (TfR), respectively.

Domain specific insertion of newly synthesized proteins.The insertion of newly synthesized Tac furin tail chimaera into the apical or basolateral plasma mem-

brane domain was essentially determined as described31. Briefly, cells on Transwell units were starved

for 30 min in cysteine- and methionine-free medium, and newly synthesized proteins were pulse-

labelled with 35S-Cys/Met (Easytag Express protein labelling mix; NEN; 1 mCi ml–1) for 30 min. H93

antibody (2 µg ml–1) was then added during a 30 min chase to either the apical or the basolateral com-

partment. Filters were then washed on ice, excised and the cells lysed in 500 µl RIPA buffer (10 mM

Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 10 % glycerin, 1 % sodium deoxycholate, 1 % SDS, 1 %

Triton X-100) supplemented with protease inhibitors. Receptor–antibody complexes were precipitated

with protein-G–Sepharose and the supernatant was re-precipitated with fresh antibody to determine

the total amount of receptor. Samples were analysed by SDS-PAGE (10 % polyacrylamide), autoradi-

ographed and quantified by densitometry as described31.

RECEIVED 18 OCTOBER 1999; REVISED 22 AUGUST 2001; ACCEPTED 12 NOVEMBER 2001;PUBLISHED 21 JANUARY 2002.

1. Bonifacino, J. S. & Dell’Angelica, E. C. J. Cell Biol. 145, 923–926 (1999).

2. Robinson, M. S. & Bonifacino, J. S. Curr. Opin. Cell Biol. 13, 444–453 (2001).

3. Keller, P. & Simons, K. J. Cell Sci. 110, 3001–3009 (1997).

4. Yeaman, C., Grindstaff, K. K. & Nelson, W. J. Physiol. Rev. 79, 73–98 (1999).

5. Mellman, I. Annu. Rev. Cell Dev. Biol. 12, 575–625 (1996).

6. Hunziker, W., Harter, C., Matter, K. & Mellman, I. Cell 66, 907–920 (1991).

7. Hunziker, W. & Mellman, I. Semin. Cell Biol. 2, 397–410 (1991).

8. Brewer, C. B. & Roth, M. G. J. Cell Biol. 114, 413–421 (1991).

9. Fölsch, H., Ohno, H., Bonifacino, J. S. & Mellman, I. Cell 99, 189–198 (1999).

10. Bresciani, R., Denzer, K., Pohlmann, R. & Vonfigura, K. Biochem. J. 327, 811–818 (1997).

11. Dell’Angelica, E. C., Mullins, C. & Bonifacino, J. S. J. Biol. Chem. 274, 7278–7285 (1999).

12. Hirst, J., Bright, N. A., Rous, B. & Robinson, M. S. Mol. Biol. Cell 10, 2787–2802 (1999).

13. Simmen, T., Nobile, M., Bonifacino, J. S. & Hunziker, W. Mol. Cell. Biol. 19, 3136–3144 (1999).

14. Matter, K., Hunziker, W. & Mellman, I. Cell 71, 741–753 (1992).

15. Odorizzi, G. & Trowbridge, I. S. J. Cell Biol. 137, 1255–1264 (1997).

16. Jing, S., Spencer, T., Miller, K., Hopkins, C. & Trowbridge, I. S. J. Cell Biol. 110, 283–294 (1990).

17. Stephens, D. J. & Banting, G. Biochem. J. 335, 567–572 (1998).

18. Aguilar, R. C. et al. J. Biol. Chem. 276, 13145–13152 (2001).

19. Futter, C. E., Connolly, C. N., Cutler, D. F. & Hopkins, C. R. J. Biol. Chem. 270, 10999–11003

(1995).

20. Matter, K., Whitney, J. A., Yamamoto, E. M. & Mellman, I. Cell 74, 1053–1064 (1993).

21. Aroeti, B. & Mostov, K. E. EMBO J. 13, 2297–2304 (1994).

22. Ohno, H. et al. FEBS Lett. 449, 215–220 (1999).

23. Yokode, M. et al. J. Cell Biol. 117, 39–46 (1992).

24. Bradke, F. & Dotti, C. G. Curr. Opin. Neurobiol. 10, 574–581 (2000).

25. Winckler, B. & Mellman, I. Neuron 23, 637–640 (1999).

26. Müsch, A., Xu, H. X., Shields, D. & Rodriguez-Boulan, E. J. Cell Biol. 133, 543–558 (1996).

27. Yoshimori, T., Keller, P., Roth, M. G. & Simons, K. J. Cell Biol. 133, 247–256 (1996).

28. Tokuyasu, K. T. J. Microsc. 143, 139–149 (1986).

29. Höning, S., Griffith, J., Geuze, H. J. & Hunziker, W. EMBO J. 15, 5230–5239 (1996).

30. Podbilewicz, B. & Mellman, I. EMBO J. 9, 3477–3487 (1990).

31. Hunziker, W. & Fumey, C. EMBO J. 13, 2963–2969 (1994).

ACKNOWLEDGEMENTS

We thank M. Robinson, P. Schu, J. Gruenberg, G. Thomas and G. Ojakian for antibodies, as well as K.

von Figura and R. Sitia for stimulating discussions and their generous support. We apologize to our

colleagues whose work could not be cited owing to space limitations. This work was supported by

grants from the Swiss and German Science Foundations to W.H. (31-31978.91 and 31-50581.97) and

S.H. (SFB523) respectively, and by the Biomedical Research Council of Singapore to W.H., as well as

fellowships from the Telethon Foundation to T.S. (380bs) and the Leenaards Foundation to W.H. W.H.

is an adjunct staff member at the Department of Physiology, National University of Singapore.

Correspondence and requests for materials should be addressed to W.H.