Tryptophan- and Dileucine-based Endocytosis Signals in the Neonatal Fc Receptor*

The neonatal Fc receptor, FcRn, transports immunoglobulin G across intestinal cells in suckling rats. FcRn enters these cells by endocytosis and is present on the apical and basolateral surfaces. We investigated the roles of aromatic amino acids and a dileucine motif in the cytoplasmic domain of rat FcRn. We expressed mutant FcRn in which alanine replaced Trp-311, Leu-322, and Leu-323, or Phe-340 in the inner medullary collecting duct cell line IMCD. Individual replacement of the aromatic amino acids or the dileucine motif only partially blocked endocytosis of 125I-Fc, whereas uptake by FcRn containing alanine residues in place of both Trp-311 and the dileucine motif was reduced to the level obtained with the tailless receptor. Leu-314 was required for the function of the tryptophan-based endocytosis signal, and Asp-317 and Asp-318 were required for the dileucine-based signal. Nonvectorial delivery of newly synthesized FcRn to the two cell surfaces was unaffected by loss of the endocytosis signals. However, the steady-state distribution of endocytosis mutants was predominantly apical, unlike wild-type FcRn, which was predominantly basolateral. This shift appeared to arise because the loss of endocytosis signals inhibited apical to basolateral transcytosis of FcRn more than basolateral to apical transcytosis.

The major histocompatibility complex class I-related Fc receptor FcRn 1 (1) carries IgG into a vesicular pathway across epithelia. In this way FcRn protects from degradation much of the IgG that enters epithelia by endocytosis. Such protection is necessary for two physiological processes: the transmission of IgG from mother to offspring, and the protection from catabolism of IgG entering and leaving the blood circulation. This necessity is illustrated by the low serum IgG concentration in the first days after birth (2), the failure to obtain IgG by suckling (2), and the rapid catabolism of IgG (3)(4)(5), in ␤2microglobulin knockout mice, which lack functional FcRn (6). Materno-fetal transmission of IgG appears to be mediated by FcRn in the yolk sacs of mice and rats (7,8) and in the human placental syncytiotrophoblast (9 -11). FcRn expressed at high levels in the small intestines of suckling mice and rats medi-ates the uptake of IgG from milk (for review, see Ref. 12). FcRn expressed at lower levels (3) at widely dispersed sites including capillary endothelium (13) appears responsible for protection from catabolism.
In intestinal epithelial cells from neonatal rats, less than 10% of FcRn is on the cell surface; most is intracellular (14). Both FcRn and IgG enter these cells at coated pits in the plasma membrane and are delivered to tubular and vacuolar structures that appear to be endosomes (14 -16). At the slightly acidic luminal pH of the neonatal duodenum and jejunum (17) IgG can bind FcRn (17)(18)(19)(20) and may therefore enter epithelial cells bound to its receptor. IgG can also cross isolated intestinal segments at neutral pH (21), which is unfavorable for its binding to FcRn (17)(18)(19)(20). Under these conditions IgG may enter epithelial cells in the fluid phase and bind FcRn only after delivery to acidified endosomes (21). Likewise, IgG may enter yolk sac endoderm (7), syncytiotrophoblast (22,23), and capillary endothelium (3)(4)(5) in the fluid phase at neutral pH before binding FcRn in endosomes. It is probable that an endocytic pathway analogous to that in the neonatal intestinal epithelium delivers FcRn to endosomes even in tissues in which IgG does not bind the receptor at the plasma membrane (although there may also be a direct route from the trans Golgi network). Consistent with this, there is some FcRn on the plasma membrane of the yolk sac endoderm (R. Rodewald personal communication cited in Ref. 24), although almost all is intracellular (7).
The signals known to direct receptors and other membrane proteins to undergo endocytosis at coated pits are almost all in the cytoplasmic domains (for review, see Ref. 25). The majority of these signals are tyrosine-based motifs and dileucine-based motifs (for review, see Refs. 25 and 26). A few endocytosis signals contain phenylalanine residues (e.g. in bovine cationdependent mannose 6-phosphate receptor (27)) or tolerate substitution of phenylalanine for tyrosine (e.g. in bovine cationindependent mannose 6-phosphate receptor (28)); none is reported to contain tryptophan, although tyrosine residues in the endocytosis signals of the human low density lipoprotein receptor (29) and human transferrin receptor (30) can be replaced by tryptophan without loss of function. Amino acid sequences overlapping some but not all endocytosis signals are used additionally to target membrane proteins to the basolateral surfaces of polarized cells (for review, see Ref. 31).
The ␣ chain of rat FcRn has a 44-amino acid cytoplasmic domain (1), which includes the dileucine motif Leu-322, Leu-323. There is no tyrosine, but there are two aromatic amino acid residues in this region, Trp-311 and Phe-340. It has recently been reported that in a chimeric protein comprising the extracellular and transmembrane regions of mouse Fc␥RII-B2 fused to the membrane-proximal 24 amino acids of the cytoplasmic domain of rat FcRn, expressed in Madin-Darby canine kidney cells, the dileucine motif is an endocytosis signal, but not a basolateral targeting signal (32). Our goal in this study was to examine the roles in endocytosis of Trp-311 and Phe-340 as well as Leu-322 and Leu-323, in full-length rat FcRn in a rat kidney cell line (33). We also examined the effects of these amino acids on the distribution of FcRn between the apical and basolateral cell surfaces.

EXPERIMENTAL PROCEDURES
Cell Culture-Rat inner medullary collecting duct (IMCD) cells and their transfected derivatives were cultured as described previously (33). Except where noted, cells were plated on 24-mm Transwells (polycarbonate, 3-m pores; Corning Costar, Acton, MA) and grown for 3-6 days before experiments to allow a polarized monolayer with resistance greater than 300 ⍀cm 2 to form (33).
Surface-specific Fc Binding Assay-The Fc fragment of human IgG (Jackson Immunoresearch, West Grove, PA) was labeled with Na 125 I (PerkinElmer Life Sciences) using IODO-GEN (Pierce Chemical Co.). The Fc fragment of human IgG binds rat FcRn as effectively as rat Fc (37). Cells were cooled on ice and washed twice with ice-cold Dulbecco's modified Eagle's medium (DMEM) pH 6.0 at the loading surface (apical or basolateral) and pH 8.0 at the nonloading surface. 125 I-Fc (100 ng/ml, 2 ϫ 10 Ϫ9 M, iodinated to 0.8 mCi/mol) in ice-cold DMEM, 1 mM KI, 1.5% fish gelatin (Sigma), 20 mM HEPES (DMEM-KIGH) pH 6.0 was then added to the loading surface with or without 5 mg/ml unlabeled human IgG (Jackson Immunoresearch, 3.3 ϫ 10 Ϫ5 M). The cells were allowed to bind 125 I-Fc for 6 h at 4°C. Radioactivity in the nonloading compartment was measured in a CliniGamma 1272 gamma counter (LKB Wallac, Piscataway, NJ), and wells in which more than 1% of the applied 125 I-Fc had crossed the monolayer were rejected. Cells were rapidly washed five times with ice-cold DMEM pH 6.0 at the loading surface and pH 8.0 at the nonloading surface. Lastly, the filters were cut from the Transwells, and cell-associated 125 I-Fc was measured. Specific binding to each surface of each cell line was calculated as the difference between the cell-associated radioactivity in the absence and presence of unlabeled IgG, and the mean amounts of 125 I-Fc specifically bound at the apical and basolateral surfaces were represented as percentages of their sum.
Fc Uptake Assay on Nonpolarized Cells-Cells were grown on 12-well plates until nearly confluent but still fibroblast-like in appearance. The cells were cooled on ice for 1 h, washed twice with ice-cold DMEM pH 6.0, and allowed to bind 125 I-Fc (100 ng/ml) in DMEM-KIGH pH 6.0 for 6 h. After the cells were washed four times with ice-cold DMEM pH 6.0, they were incubated at 37°C with prewarmed DMEM-KIGH pH 6.0 for 0, 2, 5, 15, and 30 min. Then the cells were cooled on ice, and the medium was removed and counted (cpm med ). The cells were washed with ice-cold DMEM pH 6.0 and then incubated on ice for 45 min with chymotrypsin and proteinase K, 50 g/ml each in phosphate-buffered saline pH 8.0 to digest 125 I-Fc from the cell surface (cpm surf ). Finally, the cells were washed with ice-cold DMEM pH 8.0 and then dissolved in 0.1 M NaOH and counted (cpm int ). For each time point of each cell line, the percentage of 125 I-Fc internalized was calculated as cpm int ϫ 100/ (cpm med ϩ cpm surf ϩ cpm int ).
Domain-specific Uptake Assay-Because the signal obtained by loading at 4°C was low, uptake of 125 I-Fc from the apical and basolateral surfaces was measured by loading briefly at 37°C. After each surface was washed twice with DMEM pH 8.0, the cells were cooled on ice. Cells were pulsed for 5 min at 37°C with 125 I-Fc (100 ng/ml) in DMEM-KIGH pH 6.0 at either the apical or basolateral (loading) surface, and DMEM pH 6.0 at the opposite (nonloading) surface. Immediately the cells were washed five times with ice-cold DMEM pH 6.0 and then proteins were digested at both cell surfaces as in the uptake assay above. Proteins released from the loading and nonloading surfaces were collected, and 125 I-Fc was counted (cpm loading and cpm nonloading , respectively). Each Transwell filter was cut out, and cell-associated radioactivity was counted (cpm int ). The fraction of 125 I-Fc internalized was calculated as (cpm int ϩ cpm nonloading )/(cpm loading ϩ cpm nonloading ϩ cpm int ) and was expressed as a percentage of the fraction taken up by wild-type FcRn.
Surface Biotinylation-To measure the steady-state distribution of FcRn between the apical and basolateral cell surfaces, the apical, basolateral, or both surfaces were treated with sulfosuccinimidyl 2-(biotinamido) ethyl-1,3-dithiopropionate (NHS-SS-biotin, Pierce), twice for 15 min on ice, essentially as published (38). After the biotinylation was quenched with DMEM, the filters were cut from the Transwells (three for each surface). The cells were lysed and precipitated with rabbit anti-rat FcRn ␣-chain and protein A-trisacryl and then with streptavidin-agarose essentially as described (33). Proteins were eluted from streptavidin-agarose into reducing sample buffer, pooled from each set of three Transwells, and resolved on 8% polyacrylamide denaturing gels. Gels were electroblotted onto polyvinylidene difluoride membranes, and FcRn was detected with antipeptide antibodies as above. The FcRn ␣ chain bands (upper and lower bands in Fig. 6) were quantified using a Gel Doc 1000 work station and Molecular Analyst 2.1.1 software (Bio-Rad).
Pulse-Chase Biosynthetic Labeling-To measure the delivery of newly synthesized FcRn to the apical and basolateral cell surfaces, cells were washed twice with Met -, Cys -DMEM, starved in the same medium for 45 min at 37°C, and then labeled for 15 min with 0.5 mCi/ml 35 S-labeled MetϩCys (PerkinElmer Life Sciences) in the basolateral compartment. After 0, 30, 90, and 210 min of chase in DMEM containing 10% fetal bovine serum (Hyclone Laboratories, Logan, UT), the cells were washed twice with ice-cold phosphate-buffered saline containing 1 mM CaCl 2 and 0.5 mM MgCl 2 , cooled on ice for 1 h, then biotinylated on either the apical or basolateral surface (three Transwells each) twice for 30 min, as above. The cells were lysed and precipitated with anti-FcRn and then with streptavidin-agarose as above. Proteins were eluted, pooled, and subjected to electrophoresis as above. Gels were incubated in 1 M sodium salicylate for 20 min at room temperature and dried. 35 S-Labeled proteins were detected using a Molecular Dynamics Phos-phorImager (Sunnyvale, CA).
Transcytosis Assay-125 I-Fc was allowed to bind the loading surface (apical or basolateral) as in the Fc binding assay, above. The loading and nonloading surfaces were then washed five times each with ice-cold DMEM-KIGH pH 6.0 and pH 8.0, respectively. The nonloading surface medium was counted (cpm 0 ) and replaced with DMEM-KIGH pH 8.0, prewarmed to 37°C. Cells were incubated at 37°C for 0, 2, 7, 22, and 52 min, after which times the medium from the nonloading surface was removed and counted (cpm 2 , cpm 7 , etc.) and replaced with fresh DMEM pH 8.0. At the last time point, the loading surface medium was counted (cpm medium ), and the Transwell membrane was cut out and cell-associated radioactivity counted (cpm cell ). For each well at each time point (t), the cumulative percentage of 125 I-Fc transported was calculated as (cpm 0 ϩ cpm 2 ϩ . . . ϩ cpm t ) ϫ 100/(cpm 0 ϩ cpm 2 ϩ cpm 7 ϩ cpm 22 ϩ cpm 52 ϩ cpm medium ϩ cpm cell ).

Expression of FcRn Mutants in IMCD Cells-To test whether
the dileucine motif and aromatic amino acids in the cytoplasmic domain of rat FcRn were required for endocytosis and basolateral targeting we expressed in IMCD cells FcRn ␣ chains with alanine replacing Trp-311, Leu-322 plus Leu-323, and Phe-340, individually or combination (Fig. 1). Fig. 2 shows a Western blot of extracts of cell lines selected for high and fairly uniform levels of expression of the mutant forms FcRn. Purified anti-peptide antibodies against rat FcRn ␣ chain (33) typically reveal two bands, as here, when FcRn is expressed in either IMCD cells (33) or Rat1 fibroblasts. In Rat1 cells, both bands migrate with the same decreased apparent molecular weight after digestion with peptide N-glycosidase F; the lower band is also shifted by endoglycosidase H treatment, but the upper band is resistant. 2 These data suggest that the lower band is the high mannose form of FcRn usually found in the endoplasmic reticulum, whereas FcRn in the upper band contains complex-type oligosaccharide chains modified in the Golgi. Consistent with this interpretation, the upper band is greatly enriched at the cell surface compared with the lower band (see Fig. 6). We confirmed that all of the IMCD cell lines expressing these and other mutant forms of FcRn could bind 125 I-Fc (below).
Endocytosis of FcRn and Mutants-We first compared the uptake of Fc by subconfluent, nonpolarized IMCD cells expressing wild-type and mutant FcRn ␣ chains. We allowed cells to bind 125 I-Fc at 4°C and measured uptake when they were warmed to 37°C. Cells expressing wild-type FcRn and the tailless truncation 304t were used as positive and negative controls, respectively. were, respectively, 55 and 65% lower than for wild-type FcRn, although both took up near wild-type amounts of Fc over 30 min. In contrast, endocytosis by FcRn containing alanine residues in place of Trp-311, Leu-322, and Leu-323 was decreased by 80% (closed squares), somewhat more than uptake by the tailless receptor (closed circles). Furthermore, the rates of endocytosis by W311A/L322A/L323A and tailless FcRn were both less than 10% of that obtained with the wild-type receptor. Similar results were obtained with Rat1 cells expressing these mutants (data not shown).
These results suggested that in nonpolarized cells Trp-311 and the dileucine motif were components of independent endocytosis signals. We next measured uptake from the apical and basolateral surfaces of polarized IMCD cells to determine whether either endocytosis signal was used preferentially at one or the other surface. We allowed cells to take up 125 I-Fc from the apical or basal medium for 5 min at 37°C. Disruption of either signal alone reduced endocytosis by less than 25% (Fig. 3B); disruption of both signals reduced apical endocytosis by 75% and basolateral endocytosis by 50%.
We explored the sequence requirements of the tryptophanbased endocytosis signal by alanine-scanning and additional substitutions, in the context of a disrupted dileucine motif. Again we used subconfluent, nonpolarized IMCD cells. Replacement of Trp-311 with tyrosine (open circles in Fig. 4A) or phenylalanine (open triangles) allowed endocytosis of Fc at levels similar to or only slightly below those obtained with wild-type FcRn (closed triangles). However, the rate of endocytosis by W311F/L322A/L323A was only 35% of that seen with the wild-type receptor. Mutations in positions 307, 308, 309, 310, 312, 313, and 315 had little or no effect on the extent of Fc uptake (Fig. 4B), although uptake by S315A/L322A/L323A (open diamonds) was 35% slower than by wild-type FcRn. Substitution of isoleucine for Leu-314 was without effect on the amount of Fc taken up (open circles in Fig. 4C) but decreased the rate of endocytosis by 65%. In contrast, substitution with alanine (closed circles) decreased endocytosis to the level and rate of tailless FcRn.
Similarly, we measured endocytosis by FcRn with mutations in the vicinity of the dileucine motif in the context of a disrupted tryptophan-based signal. Mutations in positions 315, 319, 320, and 321 had little or no effect on Fc uptake (Fig. 4D), although replacement of Ser-315A reduced the rate of uptake by 25% (open triangles). In contrast, substitution of alanine residues for Asp-317 and Asp-318 (open diamonds) reduced the rate of uptake by 90% and amount of Fc taken up by 60%, compared with wild-type FcRn.
Apical/Basolateral Distribution of FcRn and Mutants-We used two types of assay to measure the steady-state distribution of FcRn mutants between the apical and basolateral surfaces of polarized IMCD cells, one based on ligand binding and the other on biotinylation. In both assays we used cells grown on Transwell filters for 3 days after they had become confluent. We have shown previously that these conditions allow the cells to become polarized (33).
First we measured the binding of 125 I-Fc from the apical or basal compartment at 4°C. Fig. 5, which summarizes data from 3-10 experiments on each mutant, shows the specific binding to the apical and basolateral surfaces (calculated as the radioligand competed off by unlabeled IgG) as percentages of the total specific binding. In IMCD cells expressing wild-type FcRn we detected ϳ75% of the cell surface FcRn on the basolateral plasma membrane. We found similar percentages of the FcRn mutants W311A, L322A/L323A, F340A, W311A/F340A, and L322A/L323A/F340A at the basolateral surface. Strikingly, the distribution of FcRn with alanine replacing both Trp-311 and the dileucine motif (W311A/L322A/L323A) was reversed, with only 30% on the basolateral membrane. The distribution of the tailless FcRn 304t was intermediate between the wild-type and W311A/L322A/L323A receptors, with ϳ45% on the basolateral surface.
We used surface-specific biotinylation as an independent measure of the steady-state distribution of the W311A/L322A/ L323A FcRn mutant. We found that 65% of the wild-type FcRn was biotinylated at the basolateral surface (Fig. 6). The distribution of FcRn with alanine replacing both Trp-311 and the dileucine motif was again reversed, with less than 5% at the basolateral plasma membrane. We detected 35% of the tailless FcRn at the basolateral surface.
Using the domain-specific binding assay we compared the distributions of FcRn with mutations close to Trp-311 in the context of the L322A/L323A mutation and mutations near the dileucine motif in the context of W311A. L314A/L322A/L323A and W311A/D317A/D318A, which were impaired in endocytosis, were predominantly apical (Fig. 7). Mutants that showed normal endocytosis were found mostly on the basolateral cell surface.
Apical/Basolateral Delivery of FcRn and Mutants-We measured the delivery of newly made FcRn to the apical and basolateral cell surfaces by pulse labeling the receptor, and biotinylating one surface or the other after various chase times. Some labeled FcRn ␣ chain was biotinylated at the zero time point (Fig. 8). This probably represents FcRn that reached the cell surface during the time the cells cooled on ice between the chase and biotinylation. It is unlikely that it represents intracellular FcRn because actin was not biotinylated under these conditions (data not shown). FcRn became available for biotinylation simultaneously on the apical and basolateral cell surfaces. At early time points we detected only the low molecular weight form of FcRn (see above). Generally, between 90 and 210 min of chase the abundance of this material decreased. In parallel, the amount of biotinylated high molecular weight form of FcRn increased. In cells expressing wild-type FcRn, the high molecular weight FcRn was biotinylated predominantly at the basolateral cell surface. In contrast, in cells that expressed W311A/L322A/L323A and W311A/D317A/D318A, high molecular weight FcRn was biotinylated predominantly at the apical surface (Fig. 8). The high molecular weight form of tailless FcRn was biotinylated in similar abundance at either surface.
Transcytosis by FcRn and Mutants-We compared the abilities of the FcRn mutants to transport a cohort of Fc bound at 4°C. In the basolateral to apical direction, W311A/L322A/ L323A, W311A, and L322A/L323A all transported Fc at a level similar to that of wild-type FcRn (Fig. 9B). In the apical to basolateral direction, W311A and L322A/L323A showed slight impairment, transporting, respectively, 90 and 85% as much Fc in 52 min as wild-type FcRn (triangles and circles, respectively, in Fig. 9A). Transcytosis by W311A/L322A/L323A (squares) was reduced to 15% of the wild-type level (diamonds).

Tryptophan-and Dileucine-based Endocytosis Signals in
FcRn-We set out to test the hypothesis that the dileucine motif and the two aromatic amino acid residues in the cytoplasmic domain of rat FcRn are required for rapid endocytosis. Adjacent leucine residues are used as endocytosis signals in several membrane proteins (39 -42). Similarly, many endocytosis signals contain an essential tyrosine (43)(44)(45) or, less commonly, phenylalanine (27).
We compared endocytosis of Fc by wild-type FcRn, tailless FcRn, and FcRn in which alanine had replaced Trp-311, Leu-322 and Leu-323, and Phe-340, singly or in combination. FcRn 304t is truncated after the four amino acids that immediately follow the hydrophobic predicted membrane-spanning region of the ␣ chain (Asn-301, Arg-302, Met-303, and Arg-304). We left these four amino acids in place because they might act as a stop-transfer signal during translocation of the FcRn ␣ chain through the endoplasmic reticulum membrane. The ability of this tailless receptor to undergo endocytosis was greatly reduced but not eliminated. This is consistent with the observation that a chimera comprising the extracellular and transmembrane regions of Fc␥RII-B2 fused to the amino acids Asn-Arg-Met-Arg is internalized much less efficiently than a chimera that includes the complete FcRn tail (32).
The replacement of either aromatic amino acid or of the dileucine motif caused a modest reduction in the level of endocytosis. Of the possible combinations of substitutions, only simultaneous replacement of Trp-311 and the dileucine motif had a greater effect on endocytosis, reducing Fc uptake to a level similar to or less than that seen with the tailless receptor. These data suggest that Trp-311 and the dileucine motif are I-Fc at the apical or basolateral surface in the absence or presence of excess unlabeled IgG. Binding in the presence of competing IgG was considered nonspecific and was subtracted from the total. Specific Fc binding at the apical and basolateral surface was calculated as a percentage of the total specific binding. Columns represent the mean apical and basolateral percentages Ϯ S.E. from 3-10 measurements. WT, wild-type. necessary components of independent endocytosis signals.
Our results differ from those of a recent study in which the dileucine motif alone was critical for efficient endocytosis (32). Three features of that study could explain this discrepancy. First, the Fc␥RII-B2/FcRn chimera lacked the last 20 amino acids of the cytoplasmic domain. It is possible that the truncation inactivates Trp-311 by altering the conformation of the polypeptide in this region. Second, absence of the extracellular and transmembrane domains of FcRn from the chimera might preclude the formation of dimers in which cytoplasmic domains interact (46,47), again altering the contexts of the tryptophan-and dileucine-based signals. Third, the rat FcRn cytoplasmic domain might not interact normally with adaptor complexes in the dog cells in which the chimera was expressed. We avoided these potential complications by studying full-length rat FcRn in a rat cell line.
The tryptophan-based endocytosis signal we identified is, to our knowledge, the first of its kind. In mutagenesis studies, tryptophan can be made to replace tyrosine residues in some endocytosis signals (30) but not in others (48). We found that Trp-311 could be replaced functionally by tyrosine or phenylalanine, although the Phe substitution slowed endocytosis. Tyrosine residues essential for endocytosis mostly occur in the motif YXXØ (49,50), where X is any amino acid and Ø has a large and hydrophobic side chain. Of the amino acids near Trp-311, only Leu-314 was critical to the endocytosis signal. This suggests that in FcRn Trp-311 is part of an unusual endocytosis signal of the YXXØ type.
The endocytosis signal based upon the dileucine motif at positions 322 and 323 was inactivated by the replacement of Asp-317 and Asp-318. One or more negatively charged groups are present in positions Ϫ3 to Ϫ5 with respect to the first residue of most dileucine motifs that function in endocytosis (51,52). Leu-322 and Leu-323 of FcRn, together with Asp-317 and/or Asp-318, are thus parts of a typical dileucine-based endocytosis signal.
The two endocytosis signals in rat FcRn are partly redundant. Several other receptors contain multiple endocytosis signals that are redundant to various extents. The purpose of such redundance is not clear. Because similar signals may direct endocytosis, basolateral targeting and sorting to endosomes from the trans Golgi network, it is possible that signals redundant in one function have distinct roles in others (40).
Nonvectorial Delivery to the Cell Surface-FcRn was delivered to the apical and basolateral surfaces of IMCD cells simultaneously. None of the mutants tested affected this process. The nonvectorial delivery of FcRn to the surface of IMCD cells is consistent with reports of the biosynthetic targeting of the Fc␥RII-B2/rat FcRn chimera (32) and human FcRn (53) expressed in Madin-Darby canine kidney cells.
In our pulse-chase studies in IMCD cells, wild-type FcRn first appeared on the cell surface in a low molecular weight form. We have shown previously that this form is sensitive to endoglycosidase H. 2 Because endoglycosidase H resistance is acquired in the cis or medial Golgi, this implies either that FcRn is delivered to the plasma membrane directly from the endoplasmic reticulum or that most of it passes through the Golgi without becoming fully glycosylated. The low molecular weight form of FcRn was mostly removed from the cell surface during the chase period and was replaced by the high molecular weight form. More of the high molecular weight form of wild-type FcRn was biotinylated at the basolateral surface than the apical surface, whereas the opposite was true for the FIG. 6. Biotinylation of mutant and wild-type FcRn at the apical and basolateral surfaces of polarized IMCD cells. Cells expressing wild-type (WT) FcRn, W311A/L322A/L323A, and the tailless receptor 304t were grown on Transwells. Proteins at the apical or basolateral surfaces were labeled with a membrane-impermeant biotinylating reagent. FcRn was immunoprecipitated from cell lysates with a rabbit polyclonal antiserum, and the biotinylated fraction was reprecipitated with streptavidin-agarose. FcRn eluted from streptavidinagarose was detected on Western blots using affinity-purified antipeptide antibodies against rat FcRn ␣ chain. endocytosis mutants W311A/L322A/L323A and W311A/ D317A/D318A. In all cases the high molecular weight form was preferentially biotinylated at the cell surface where FcRn was more abundant in the steady state (below). These observations suggest that the low molecular weight form of FcRn undergoes endocytosis and is further glycosylated during subsequent trafficking that establishes the steady-state distribution.
The delivery of incompletely glycosylated proteins to the cell surface is unusual but not unique. The glycosylation of FcRn after its delivery to the cell surface indicates that it enters the Golgi stack during recycling or transcytosis. Several other membrane proteins, including transferrin receptor (54) and mannose 6-phosphate receptors (55), recycle through the Golgi. A mutant low density lipoprotein receptor lacking the cytoplasmic domain retains the ability to return to the Golgi (56), as does tailless FcRn in the present study. The anti-transferrin receptor antibody OX26 also enters Golgi cisternae during transcytosis through the blood-brain barrier endothelium (57). Similar trafficking would allow glycosylation of FcRn after its initial delivery to the plasma membrane.
Complex-type glycans on the ␣3 domains form part of the interface between FcRn molecules in dimers seen by x-ray crystallography (20). The immature glycoform of FcRn might therefore be impaired in dimerization, with two significant consequences. First, because in a dimer the cytoplasmic domains are close enough to interact (58), the unpaired tail of the low molecular weight form of FcRn might cause this glycoform to be sorted differently from dimers of the high molecular weight form. Second, because FcRn must dimerize to bind IgG with high affinity (47,59), the low molecular weight form might be impaired in IgG binding. We speculate that these mechanisms allow the immature glycoform to return to the Golgi unoccupied by IgG. This is consistent with the detection of FcRn (14) but not internalized IgG (15) in the Golgi cisternae of intestinal epithelial cells in neonatal rats.
Steady-state Surface Distribution-In the steady state, there was approximately twice as much wild-type FcRn at the basolateral surface as at the apical surface, consistent with previous observations in IMCD cells (33). Studies with BeWo and Madin-Darby canine kidney cell lines show a more apical steady-state distribution (32). Whether this reflects differences in the receptors or the cells remains to be seen. The basolateral surfaces of polarized lines such as Madin-Darby canine kidney (60) are larger than the apical, and the greater abundance of FcRn at the basolateral surface of IMCD cells might result in part from nonvectorial delivery to a smaller apical and larger basolateral surface. For such an initial distribution to be maintained, apical to basolateral transport of FcRn would have to exceed basolateral to apical transport, which is consistent with studies of Fc transport across IMCD cells expressing FcRn (33). All of the mutants we identified as defective in endocytosis, FcRn-304t, W311A/L322A/L323A, L314A/L322A/L323A, and W311A/D317A/D318A, showed a reversed steady-state distribution in which there was more FcRn at the apical surface than the basolateral surface. In all of these mutants that we have FIG. 8. Surface delivery of newly synthesized receptors. Cells were pulsed for 15 min with 35 S-labeled MetϩCys. After the chase times shown, cells were cooled on ice, and proteins were biotinylated at either the apical or basolateral surface. Biotinylated FcRn was detected as in Fig. 6. WT, wild-type.
FIG. 9. Transport of I-Fc across monolayers of IMCD cells that express wild-type or mutant FcRn. Cells on ice were loaded with 125 I-Fc from either the apical (panel A) or basolateral (panel B) compartment, and free ligand was removed. The medium in the loading compartment was at pH 6, and in the nonloading compartment it was at pH 8. The cells were warmed to 37°C to stimulate transcytosis, and medium was collected from the nonloading compartment at the indicated times. After 52 min, the medium was also collected from the loading compartment, and the cells were lysed. Media and lysates were precipitated with trichloroacetic acid and counted for 125 I. The cumulative percentage of each cohort transported to the nonloading compartment was calculated (mean Ϯ S.E., n ϭ 6). tested (FcRn-304t, W311A/L322A/L323A, and W311A/D317A/ D318A), newly synthesized FcRn is delivered to both surfaces at approximately similar rates. Therefore the distribution of the mutant receptors does not arise from altered biosynthetic targeting. Rather, it appears to be attributable to the effects of the mutations on subsequent trafficking. Specifically, W311A/ L322A/L323A was inhibited in apical to basolateral but not basolateral to apical transcytosis. Likewise we have found that FcRn-304t is inhibited in apical to basolateral but not basolateral to apical transcytosis. 2 Thus mutations that eliminate the endocytosis signals in FcRn appear to perturb its steady-state distribution in favor of the apical cell surface by affecting apical to basolateral transcytosis more than basolateral to apical transcytosis.