Characterization of basolateral-targeting signals in the neonatal Fc receptor

It is a general characteristic of polarized cells that many proteins are non-uniformly distributed between the apical and basolateral domains of the plasma membrane (Mostov et al., 2000; Simons and Wandinger-Ness, 1990). This asymmetry underlies the transport across epithelia of ions and small molecules (Muth and Caplan, 2003) and of macromolecules (Tuma and Hubbard, 2003). The neonatal Fc receptor (FcRn) is a macromolecule transporter comprising a membrane-spanning α chain non-covalently associated with beta 2-microglobulin (β2m) (Simister and Mostov, 1989). FcRn is present in many epithelia, including the neonatal intestine for which it is named (Simister and Mostov, 1989), fetal intestine (Shah et al., 2003), adult intestine (Israel et al., 1997), yolk sac (Ahouse et al., 1993; Roberts et al., 1990), placental syncytiotrophoblast (Kristoffersen and Matre, 1996; Leach et al., 1996; Simister et al., 1996), capillary endothelium (Borvak et al., 1998), hepatocytes (Blumberg et al., 1995), renal proximal tubule (Haymann et al., 2000; Kobayashi et al., 2002), airway epithelium (Spiekermann et al., 2002) and mammary epithelium (Cianga et al., 2003). Its functions include both apical-to-basolateral transport of IgG from milk across intestinal epithelial cells (Israel et al., 1995; Roopenian et al., 2003) and basolateral-to-apical secretion of IgG into the gut lumen (Yoshida et al., 2004). In addition to IgG, albumin was recently found to bind FcRn (Chaudhury et al., 2003).

Endogenous FcRn in neonatal rat intestinal epithelial cells is mostly intracellular (Berryman and Rodewald, 1995). FcRn is present on both the apical and basolateral surfaces of these cells (Berryman and Rodewald, 1995), but the relative amounts have not been quantified. When expressed in rat inner medullary collecting duct (IMCD) cells, most of the rat FcRn in the plasma membrane is at the basolateral cell surface (McCarthy et al., 2000; Wu and Simister, 2001). Surface human FcRn α chain co-expressed with human β2m in Madin-Darby canine kidney (MDCK) cells is also predominantly basolateral (Claypool et al., 2004).

Targeting of proteins to the basolateral plasma membrane is directed by sorting signals (Matter and Mellman, 1994). Such signals have been identified in the cytoplasmic domains of many basolaterally sorted proteins. Basolateral-targeting signals are of two kinds, some partially collinear with endocytosis signals (Hunziker and Fumey, 1994; Hunziker et al., 1991; Matter et al., 1994; Prill et al., 1993; Simonsen et al., 1998), and others not (Matter et al., 1994; Miranda et al., 2001; Okamoto et al., 1992). There are tyrosine-based and dileucine-based signals in both categories. While investigating endocytosis signals in rat FcRn, we previously identified three mutants with predominantly apical surface expression (Wu and Simister, 2001). FcRn has two endocytosis signals, one based on Trp311 and Leu314 (Wu and Simister, 2001) and the other on Leu322/Leu323 and Asp317/Asp318 (Stefaner et al., 1999; Wu and Simister, 2001). The apical mutants have non-conservative substitutions in both endocytosis signals – W311A/L322A/L323A, L314A/L322A/L323A and W311A/D317A/D318A – and are severely impaired in endocytosis (Wu and Simister, 2001). Thus, the basolateral sorting signals in FcRn overlap the endocytosis signals.

In the present study, we looked at the effects of additional mutations in and around the endocytosis signals, and found that A309L/L322A/L323A, L314F/L322A/L323A and D317A/D318A impair basolateral sorting without reducing endocytosis. These results further define the unique tryptophan-based basolateral-targeting signal in FcRn, and reveal differences between this signal and the collinear endocytosis signal. Unexpectedly, these results suggest that the acidic cluster Asp317/Asp318 is a part of both the tryptophan-based and dileucine-based basolateral-targeting signals.

DNA encoding the following FcRn mutants was created using the QuikChange method (Stratagene) with the forward primers indicated and their exact complements: D317A/D318A, CGCCAGAGGCTGCACCACTGAGAGAA; E331A/E333A, GAACTTGCCCCCGGCGGCCGCGCCTCAAGGTGTAAATGA. The remaining mutants were created using overlap exchange PCR (Horton et al., 1989) with the following forward primers and their complements; A309L, CGAAGTGGGCTGCCACTCCCATGGCTT; L314F, CCAGCCCCATGGCTTTCTTTCAGTGGT; L322A, GGCGACGCATTGCCTGGTGG; L323A, CTGGCGACCTAGCGCCTGGT.

IMCD cells are derived from rat kidney inner medullary collecting ducts, and were kindly given by J. Schwartz (Alexander and Schwartz, 1991). The lines expressing wild-type rat FcRn and mutant W311A/L322A/L323A have been reported previously (Wu and Simister, 2001). The methods for cell culture have also been described (McCarthy et al., 2000).

IMCD cells were stably transfected with FcRn or FcRn mutants in the neomycin-resistance vector pRc/RSV (Stratagene) using the calcium phosphate method essentially as described before (Gorman et al., 1990). Colonies of cells resistant to 1 mg/ml G418 sulfate (Calbiochem-Novabiochem) were expanded and tested for FcRn expression by western blotting.

Western blots were done as described previously (McCarthy et al., 2000), using rabbit anti-FcRn antiserum (Simister and Mostov, 1989), goat anti-rabbit IgG conjugated to horseradish peroxidase (Bio-Rad), and Renaissance chemiluminescent reagent (NEN-Dupont). Band mobilities were compared with those of pre-stained molecular weight markers (Bio-Rad).

Cells were plated in triplicate on Transwell (Costar) inserts in six-well plates and grown until they formed tight monolayers with electrical resistance of at least 300 Ωcm². The Fc fragment of human IgG (mixed subclass; Jackson Immunoresearch) was labeled with Na¹²⁵I (Perkin Elmer) using Iodogen (Pierce). The monolayers were incubated for 1 hour at 37°C in serum-free Dulbecco's modified Eagle's medium (DMEM), 1 mM KI, 1.5% fish gelatin (Sigma), 20 mM HEPES (DMEM-KIGH), pH 7.4. The cells were then washed twice in ice-cold, serum-free DMEM-KIGH pH 6.0 on the surface to be labeled with ¹²⁵I-Fc and pH 8.0 on the opposite surface, and incubated on ice for 1 hour. The monolayers were then incubated for 6 hours on ice with ¹²⁵I-Fc (2 nM) in 1 ml DMEM-KIGH pH 6 plus or minus unlabeled human IgG (4 μM; Sigma) at either the apical or the basolateral surface. The opposite surface was maintained with ice-cold DMEM-KIGH pH 8. The cells were then washed five times at the binding surface with ice-cold DMEM-KIGH pH 6 or three times at the non-binding surface with ice-cold DMEM-KIGH pH 8.0. The cells were then lysed in 0.1 M NaOH, and counted on a CliniGamma 1272 gamma counter. Specific binding at each surface was calculated as the difference between the cell-associated counts in the absence and presence of cold competitor. The amount of receptor on each surface was represented as the amount of radioactivity detected at that surface as a percentage of the total radioactivity associated with those cells, i.e. % apical=apical CPM ×100/(apical CPM+basolateral CPM).

Cells were grown on six-well plates until they were nearly confluent, but still fibroblast-like in appearance (i.e. not yet polarized). The cells were washed once in serum-free DMEM-KIGH pH 8 and then starved in the same medium for 1 hour. The cells were washed once in pre-warmed DMEM-KIGH pH 6 and then incubated at 37°C for 0, 2, 4, 6 or 8 minutes with ¹²⁵I-Fc (2 nM) in DMEM-KIGH pH 6 with or without unlabeled IgG (4 μM). After incubation, the cells were placed on ice and washed five times quickly with ice-cold DMEM-KIGH pH 6 to stop endocytosis. ¹²⁵I-Fc remaining on the cell surface was then removed by incubating the cells in DMEM-KIGH pH 8 for 45 minutes, followed by rinsing the cell surface once with ice-cold, DMEM-KIGH pH 8. This surface medium, including the rinse, was collected and counted in a CliniGamma 1272 gamma counter (CPM_surface). Finally, the cells were lysed in 0.1 M NaOH and the internal counts were measured in a gamma counter (CPM_internal). Specific binding and endocytosis were calculated by subtracting counts in the presence of competing IgG from counts in the absence of unlabeled IgG. The data were analyzed using an In/Sur plot (Wiley and Cunningham, 1982). The specific CPM_internal/specific CPM_surface ratio was plotted as a function of incubation time. A straight line was fitted to the points using the least squares method, and the slope was calculated. This slope is the endocytic rate constant, and has the units minutes^–1.

Cells were grown in triplicate on Transwell inserts in six-well plates until they formed monolayers. Cells were washed twice with Met^–, Cys^– DMEM (Mediatech), starved in the same medium for 45 minutes at 37°C and then labeled for 25 minutes through the basolateral surface with 1.5 mCi/ml ³⁵S-labeled Met+Cys (Perkin Elmer Life Sciences). After 0, 30, 90 or 210 minutes of chase in DMEM with 10% fetal bovine serum, the cells were washed twice with ice-cold PBS containing 1 mM CaCl₂ and 0.5 mM MgCl₂, (PBS++) pH 7.4, chilled on ice for 1 hour and then biotinylated on either the apical or the basolateral surface. Biotinylation was performed twice on ice for 30 minutes each with 0.5 mg/ml sulfosuccinimidyl 2-(biotinamido) ethyl-1,3-dithiopropionate (NHS-SS-biotin; Pierce) in PBS++ pH 9.0. The biotinylation reaction was then quenched with ice-cold DMEM containing 10% FBS for 15 minutes. The cells were washed twice with ice-cold PBS++, lysed in 500 μl SDS lysis buffer (0.5% SDS, 150 mM NaCl, 20 mM triethanolamine, 5 mM EDTA, 0.02% NaN₃, pH 8.6) and collected in 1.5 ml tubes. Lysis was aided by vortexing for 30 seconds and boiling for 10 minutes. Cellular debris was removed by centrifugation at 14,000 g for 10 minutes at 4°C. Samples were diluted with an equal volume of Triton dilution buffer (2.5% Triton X-100, 100 mM NaCl, 50 mM triethanolamine, 5 mM EDTA, 0.02% NaN₃, pH 8.6). Ten μl of each sample were counted in a Beckman LS 2000 scintillation counter and the samples were normalized based on counts. Normalized samples were pre-cleared with 2 μl normal rabbit serum for 1 hour and 100 μl of a 20% protein-A tris-acrylamide slurry (Pierce) for 30 minutes, both at room temperature. FcRn was precipitated by incubating with rabbit anti-FcRn antiserum (Simister and Mostov, 1989) for 2 hours at 4°C followed by 100 μl of a 20% protein-A tris-acrylamide slurry for 1 hour at 4°C. The precipitated proteins were then washed four times in mixed micelle buffer (1% Triton X-100, 0.2% SDS, 150 mM NaCl, 5% w/v sucrose, 5 mM EDTA, 0.1% NaN₃, 20 mM ethanolamine-HCl, pH 8.6), and twice in final wash buffer (15 mM NaCl, 5 mM EDTA, 0.1% NaN₃, 20 mM ethanolamine-HCl, pH 8.6). Proteins were eluted by boiling in 20 μl of 5% SDS and biotinylated proteins were then precipitated from this solution with 25 μl of a 50% slurry of streptavidin-agarose beads (Invitrogen) in 1 ml of a 1:1 mixture of SDS lysis buffer and Triton dilution buffer. The precipitated proteins were then washed four times in mixed micelle buffer and twice in final wash buffer as above. After the final wash, proteins were eluted by boiling for 10 minutes in 10 μl of Laemmli reducing sample buffer and then resolved by electrophoresis on a 4-20% gel. Gels were soaked in 1 M sodium salicylate for 30 minutes, 3% glycerol for 15 minutes, dried and then exposed to Molecular Dynamics PhosphorImager screen for 2-3 weeks.

To quantify intracellular biotinylation, actin was precipitated from the cell lysates after zero chase with a mouse monoclonal antibody (Sigma), and eluted as above. Biotinylated actin was precipitated from half of the eluate with streptavidin. A fifth of 1% of the total actin and all the biotinylated actin, representing half of the original cell lysate, were run on a gel and exposed to a PhosphorImager screen as described above. The bands were quantified using ImageQuant software (Amersham Bioscience).

We expressed DNA encoding the α chains of the FcRn mutants A309L/L322A/L323A, L314F/L322A/L323A, D317A/D318A and E331A/E333A in IMCD cells (Fig. 1). Western blots showed α-chain doublets typical of FcRn expressed in IMCD cells (McCarthy et al., 2000) (Fig. 2).

Surface distribution of mutant and wild-type FcRn was assayed by measuring the specific binding of ¹²⁵I-Fc at the apical and basolateral surfaces of polarized IMCD cells. Like wild-type FcRn, the E331A/E333A mutant was predominantly (80-90%) on the basolateral surface (Fig. 3A,B). As shown previously (Wu and Simister, 2001), the mutant W311A/L322A/L323A was predominantly apical (Fig. 3A). The other mutants studied resembled W311A/L322A/L323A rather than wild-type FcRn; 80-90% of surface A309L/L322A/L323A, L314F/L322A/L323A and D317A/D318A were apical (Fig. 3A,B).

To measure endocytosis, sub-confluent non-polarized cells were incubated briefly for various times with ¹²⁵I-Fc at 37°C and the endocytosis rate constant was calculated by plotting the ratio of internal/surface ligand as a function of time. The internalization rate constant is the fraction of surface FcRn internalized per minute. In this assay, the rate of endocytosis of wild-type FcRn was 0.16-0.26 minutes^–1 (Fig. 4). The endocytosis rates of A309L/L322A/L323A, L314F/L322A/L323 and D317A/D318A were within this range (Fig. 4A,B,D). E331A/E333A took up Fc at a rate of about 0.1 minutes^–1 (Fig. 4E). The mutant W311A/L322A/L323A, which we previously found defective in endocytosis (Wu and Simister, 2001), did not take up Fc in this assay (Fig. 4C).

Because endocytosis was not impaired in any of the FcRn mutants with reversed surface distribution, we were able to look at their abilities to transport and recycle Fc. IMCD cells were grown as tight polarized layers on Transwells. We allowed the cells to take up ¹²⁵I-Fc from either the apical or basolateral surface at 14°C. Endocytosis continues at this temperature, although transcytosis is blocked (McCarthy et al., 2001). Upon warming to 37°C, the cells were allowed to release ¹²⁵I-Fc at either surface. Typically, at the end of 60 minutes, 10% or less of the counts remained associated with the cells. The remainder was either recycled or transported across the cell layer.

Over the course of 60 minutes, wild-type FcRn transported 58±6% (mean±s.d., n=7) of the total ¹²⁵I-Fc loaded from the apical compartment to the basolateral compartment, whereas 37±6% was recycled back to the apical surface (Fig. 5A-D). Wild-type FcRn transported less efficiently from the basolateral compartment: 22±8% (n=7) of loaded ¹²⁵I-Fc went across the cell layer to the apical compartment, whereas 73±10% was recycled (Fig. 5E-H). The movements of A309L/L322A/L323A, D317A/D318A and E331A/E333A were quite similar to those of wild-type FcRn. A309L/L322A/L323A transported 52±7% (n=5) of loaded ¹²⁵I-Fc in the apical to basolateral direction (Fig. 5A), and 25±7% (n=3) from the basolateral to the apical compartment (Fig. 5E). D317A/D318A transported 50±6% (n=4) of loaded ¹²⁵I-Fc from the apical to the basolateral compartment of the cell layer (Fig. 5C), and 36±8% (n=5) from the basolateral surface to the apical surface (Fig. 5G). E331A/E333A transported 58±12% (n=3) of ¹²⁵I-Fc loaded from the apical to the basolateral compartment (Fig. 5D), and 26±8% (n=7) of ¹²⁵I-Fc loaded from the basolateral compartment to the apical compartment (Fig. 5H).

Like wild-type FcRn, L314F/L322A/L323A transported little across the cell layer from the basolateral compartment to the apical compartment: only 31±8% (n=4) of the ¹²⁵I-Fc loaded (Fig. 5F). However, in the opposite direction, this mutant transported much less than wild-type. Only 31±5% (n=5) of ¹²⁵I-Fc loaded at the apical compartment was transported to the basolateral compartment, whereas 64±7% was recycled (Fig. 5B).

The delivery of newly made FcRn to the cell surface was measured by pulse labeling the receptor with ³⁵S-Met+Cys, and biotinylating either the apical or basolateral surface after various chase times. The upper band of the FcRn doublet seen by SDS-PAGE was previously shown to contain glycoprotein modified in the Golgi, whereas the lower band contains the immature form typically found in the endoplasmic reticulum (McCarthy et al., 2001).

Labeled immature forms of wild-type FcRn and all of the mutants were detected after biotinylation at either cell surface at the zero time point (Fig. 6). Because this glycoform is associated with the endoplasmic reticulum, we checked for biotinylation of internal proteins by assessing biotinylation of actin. Actin was biotinylated at extremely low levels, although ³⁵S-actin was readily immunoprecipitated from lysates (Fig. 6B). From quantitation of the bands in Fig. 6B, the amount of biotinylated labeled actin on the gel was less than a quarter of the total labeled actin (we considered only basolateral biotinylation because we could not accurately quantify actin biotinylated from the apical compartment). Because the biotinylated band was from 250-fold more cell lysate than the total actin band, we estimated that less than 0.1% of intracellular protein was biotinylated.

Mature wild-type FcRn was biotinylated at the apical and basolateral cell surfaces 90 minutes after biosynthetic labeling. At that time, much more labeled FcRn was biotinylated at the basolateral than the apical surface, apically biotinylated FcRn being barely detectable. By 210 minutes chase, mature FcRn biotinylated at the apical surface was no longer detectable, and slightly less was biotinylated at the basolateral surface than after 90 minutes (Fig. 6A). E331A/E333A appeared to be delivered similarly to wild-type FcRn, but was expressed at a lower level. Mature E331A/E333A biotinylated at the basolateral surface was seen after 90 minutes chase, but was undetectable at 210 minutes (Fig. 6A). Mature E331A/E333A biotinylated at the apical surface could not be detected.

A309L/L322A/L323A, L314F/L322A/L323A and D317A/D318A showed a pattern of surface delivery different in several ways from that of wild-type FcRn. The mature forms of all three mutants were biotinylated at the apical surface at 90 minutes in amounts that were readily detected (a little L314F/L322A/L323A biotinylated apically at 30 minutes was seen). The amounts biotinylated apically at 210 minutes were the same or more than after 90 minutes chase. The mature forms of these mutants were also biotinylated at the basolateral cell surface 90 minutes after biosynthetic labeling, as was seen for wild-type FcRn. However, in contrast to wild-type, the amount of labeled mature form of the mutants biotinylated basolaterally at 210 minutes was greater than at 90 minutes.

We examined the effects of mutations on the distribution of rat FcRn between the apical and basolateral surfaces of IMCD cells. Specifically, we studied FcRn molecules altered near Trp311 and Leu322/Leu323, which we had previously found necessary for basolateral targeting (Wu and Simister, 2001), and at a separate acidic cluster. Ala309 is conserved in all known FcRn sequences: human (Story et al., 1994), macaque (GenBank accession number AF485818), rat (Simister and Mostov, 1989), mouse (Ahouse et al., 1993), cow (Kacskovics et al., 2000), sheep (Mayer et al., 2002), pig (Schnulle and Hurley, 2003; Zhao et al., 2003) and possum (Adamski et al., 2000). We had previously shown that its conservative replacement by glycine did not alter endocytosis or distribution (Wu and Simister, 2001). In the present study, we substituted leucine for Ala309. Leu314 is conserved in rat, mouse, human and macaque FcRn but phenylalanine is present in the corresponding position in FcRn from cattle, sheep and pigs. Previously, the L314A mutation disrupted both the endocytosis and basolateral-targeting signals based on Trp311 (Wu and Simister, 2001). Here, we replaced Leu314 of rat FcRn with phenylalanine. Asp317 and Asp318 are conserved or replaced by glutamate residues in all known FcRn α chains. Because of the redundancy of the tryptophan- and dileucine-based endocytosis signals in FcRn, we previously tested D317A/D318A only in the context of the W311A mutation. Now we report the effect of D317A/D318A alone. A second pair of acidic amino acids, Glu331 and Glu333 in rat FcRn, are conserved or replaced by aspartate residues in mouse, human, macaque, bovine and sheep FcRn (Ahouse et al., 1993; Kacskovics et al., 2000; Mayer et al., 2002; Story et al., 1994) (GenBank accession number AF485818). Pig FcRn contains a similar pair but in a different position (Zhao et al., 2003), and possum FcRn has an aspartic acid residue corresponding only to Asp331 (Adamski et al., 2000). We replaced Glu331 and Glu333 with alanine residues.

A309L/L322A/L323A, L314F/L322A/L323A and D317A/D318A were predominantly apical, in contrast to wild-type FcRn. None was impaired in endocytosis, unlike the apical mutants we have previously identified (Wu and Simister, 2001). To understand the reasons for the altered distributions, we looked at biosynthetic delivery and transport between the apical and basolateral cell surfaces.

We studied the delivery of FcRn to the plasma membrane by combining pulse-labeling with biotinylation at the apical or basolateral cell surface. As before (Wu and Simister, 2001), extremely long exposures were needed to detect biotinylated, ³⁵S-labeled mature FcRn, implying that there is very little at the cell surface. Biotinylation of a low-molecular-weight form, previously identified as the high mannose glycoform (McCarthy et al., 2001), was detected. We previously speculated that this represented the delivery of immature FcRn to the plasma membrane, and that its occurrence at the zero time point reflected delivery while the cells chilled before biotinylation (Wu and Simister, 2001). Three factors, taken together, suggest an alternative explanation. First, in the present study, we detected biotinylation of internal actin, albeit with an efficiency less than 0.1%. Second, FcRn α chain was recently shown to accumulate in the endoplasmic reticulum in the absence of β2m (Claypool et al., 2002; Praetor and Hunziker, 2002; Zhu et al., 2002). Third, the level of α chain we express in IMCD cells exceeds that of endogenous β2m (E.E.N. and N.E.S., unpublished). It is therefore possible that a very small fraction of a large pool of excess FcRn α chain is biotinylated in the endoplasmic reticulum. It is unlikely that significant amounts of mature α chain are biotinylated internally because it is much less abundant at later stages of the secretory pathway (Zhu et al., 2002). The observation that there are differences between the biotinylation of some mutants and wild-type FcRn (see below) is also consistent with the biotinylation of mature FcRn occurring primarily at the cell surface, rather than internally.

Mature FcRn was biotinylated at both cell surfaces after the same chase time. This is consistent with previous studies suggesting direct delivery to the apical and basolateral membranes (Praetor et al., 1999; Stefaner et al., 1999; Wu and Simister, 2001). In comparison with wild-type FcRn, substantially more A309L/L322A/L323A, L314F/L322A/L323A and D317A/D318A were biotinylated from the apical compartment, indicating enhanced apical delivery. This delivery would contribute to the predominantly apical surface distribution of these mutants. Nonetheless, biotinylation of mutant FcRn molecules at the basolateral membrane showed that sorting to that surface was not eliminated. This behavior resembles the reduced levels of basolateral delivery that persist after the disruption of basolateral sorting signals in other proteins [e.g. lysosomal acid phosphatase (Prill et al., 1993), and CD147 (Deora et al., 2004)].

We examined post-secretory sorting of the FcRn mutants by measuring the relative efficiencies of transcytosis and recycling of ligand from the two surfaces. Apical to basolateral transcytosis by L314F/L322A/L323A was reduced but basolateral to apical transport was similar to wild-type. This change would contribute to the apical accumulation of this mutant. Transport by the remaining mutants was similar to transport by wild-type FcRn.

Earlier studies showed that FcRn contains two redundant basolateral-targeting sequences sharing elements of its endocytosis signals (Wu and Simister, 2001). Substitution of alanine for either Trp311 or Leu314 inactivated one basolateral-targeting signal and substitution for Asp317 and Asp318 or Leu322 and Leu323 inactivated the other (Wu and Simister, 2001). In the present study, we found that the D317A/D318A mutations by themselves redirect FcRn to the apical plasma membrane. This suggests that this pair of acidic residues is required for both the tryptophan- and dileucine-based basolateral-targeting signals (although it is possible that one aspartate is required for the tryptophan-based signal and the other for the dileucine signal). It is unlikely that this effect of the mutations is due to gross misfolding, because the tryptophan motif retains its function as an endocytosis signal. We are aware of only one other example of overlapping basolateral-targeting signals; in CD1d, the Y+3 residue of a tyrosine-based motif is the first residue of a valine-leucine basolateral-targeting signal (Rodionov et al., 2000) (Fig. 7).

Acidic residues are components of several dileucine-based basolateral-targeting signals [e.g. in the major histocompatibility complex (MHC) class II-associated invariant chain) (Simonsen et al., 1998)]. By contrast, there are dileucine basolateral-targeting signals that lack acidic residues [e.g. in FcγRII-B2) (Matter et al., 1994)]. This might reflect a requirement for acidic residues in some sequence contexts but not others, or the existence of distinct recognition mechanisms. The recognition mechanisms for dileucine signals are imperfectly understood, although recent work suggests that DDXXXLL motifs bind complexes of AP-1 σ and γ subunits and of AP-3 σ and δ subunits (Janvier et al., 2003). The requirement for nearby acidic residues in a YXXΦ-related basolateral-targeting signal is not unique to FcRn either; a signal in lysosomal acid phosphatase is compromised by replacement of an aspartate residue in the Y+5 position (Prill et al., 1993). Acidic amino acids are also found in other types of basolateral-targeting signal. For example, both tyrosine-based signals in LDL receptor require acidic clusters (Matter et al., 1992). A phenylalanine-isoleucine signal in furin likewise needs an acidic cluster (Simmen et al., 1999), and acidic residues enhance basolateral targeting by a monoleucine signal in stem cell factor (Wehrle-Haller and Imhof, 2001).

Two additional mutations disrupted the tryptophan-based basolateral-targeting signal. We first discuss A309L, which replaces the residue in the –2 position with respect to Trp311. Recently, we discovered that the Trp-based endocytosis signal in FcRn binds the μ subunit of AP-2 (Wernick et al., 2005). The binding of μ adaptins to YXXΦ-type motifs is influenced by the residue in position Y-2 (Boll et al., 1996; Ohno et al., 1998; Ohno et al., 1995). Two μ adaptins are implicated in basolateral targeting, μ1B (Folsch et al., 1999) and μ4 (Simmen et al., 2002). The binding specificity of μ1B is not known. Adaptin μ4 does not show a preference for alanine over leucine in the –2 position that would explain the impaired sorting of A309L (Aguilar et al., 2001).

The second mutation, L314F, replaces a residue already known to be important for the tryptophan-based basolateral sorting and endocytosis signals. The effect of phenylalanine in this position on basolateral sorting but not endocytosis shows that the two signals have different constraints. Such differences are seen for other dual-function YXXΦ-type signals, including that in lysosomal acid phosphatase (Prill et al., 1993), presumably because of differences in the fine specificities of the μ adaptins mediating endocytosis and basolateral sorting. In this regard, the effect of the L314F mutation in FcRn is not explained by the specificity of μ4, which prefers phenylalanine to leucine in the +3 position (Aguilar et al., 2001). Nor are residues in position –3 through +2 favored by μ4. This suggests that the tryptophan-based basolateral-targeting signal is recognized by another protein, perhaps μ1B. Phenylalanine occurs naturally in the position corresponding to Leu314 in FcRn from cows, sheep and pigs (Kacskovics et al., 2000; Mayer et al., 2002; Zhao et al., 2003). These species might not use the tryptophan-based sequence for basolateral targeting, relying instead on their conserved DDXXXLL sequences. Alternatively, their basolateral-targeting machinery might differ from that in rat cells and recognize WXXF.

In addition to Asp317 and Asp318, another pair of acidic amino acid residues is highly conserved among FcRn molecules. The replacement of Asp331 and Asp333 with alanine residues caused a moderate reduction in the rate of endocytosis of rat FcRn but did not affect its delivery to the plasma membrane, the balance between transcytosis and recycling, or its steady-state distribution between the apical and basolateral cell surfaces. Because these aspartate residues are not close to either endocytosis signal in the primary structure of the cytoplasmic domain, the means of their effect on endocytosis is not obvious and requires further investigation.

We thank Naomi Wernick for critical reading of the manuscript. This work was supported by NIH grant HD27691.