A mutated transferrin receptor lacking asparagine-linked glycosylation sites shows reduced functionality and an association with binding immunoglobulin protein.

The function of the transferrin receptor is to transport iron-bound transferrin into the cell. In order to function properly, this dimeric glycoprotein must be expressed on the cell surface and be able to bind transferrin. Site-directed mutagenesis was performed to abolish the three asparagine-linked glycosylation consensus sequences of the human transferrin receptor. The DNA encoding the mutated transferrin receptor was stably transfected into mouse fibroblasts. This form of the human transferrin receptor shows reduced transferrin binding, reduced intersubunit bond formation, and reduced cell surface expression, indicating that the transferrin receptor which lacks asparagine-linked glycosylation is not fully functional. In addition, the mutated form of the receptor is not processed as quickly. It shows an association with an endoplasmic reticular chaperone protein, binding immunoglobulin protein, leading to the hypothesis that the mutated transferrin receptor experiences increased retention in the endoplasmic reticulum.

The function of the transferrin receptor is to transport iron-bound transferrin into the cell. In order to function properly, this dimeric glycoprotein must be expressed on the cell surface and be able to bind transferrin. Site-directed mutagenesis was performed to abolish the three asparagine-linked glycosylation consensus sequences of the human transferrin receptor. The DNA encoding the mutated transferrin receptor was stably transfected into mouse fibroblasts. This form of the human transferrin receptor shows reduced transferrin binding, reduced intersubunit bond formation, and reduced cell surface expression, indicating that the transferrin receptor which lacks asparaginelinked glycosylation is not fully functional. In addition, the mutated form of the receptor is not processed as quickly. It shows an association with an endoplasmic reticular chaperone protein, binding immunoglobulin protein, leading to the hypothesis that the mutated transferrin receptor experiences increased retention in the endoplasmic reticulum.
The human transferrin receptor, a dimer composed of two 94-kDa subunits, is involved in the uptake of iron via the binding and endocytosis of transferrin. The transferrin receptor is synthesized and modified in the endoplasmic reticulum and is further modified as it passes through the Golgi to the cell surface. Serine 24 is phosphorylated (l), cysteine 62 is acylated (2), and asparagines 251, 317, and 727 are glycosylated (3)(4)(5). There is also at least one 0-linked glycosyl group, but its position has not been established (6,7).
It is not clear to what extent many of the modifications affect transferrin receptor function. Proper endocytosis, recycling, and membrane association occur in the absence of phosphorylation and acylation (2,(8)(9)(10). However, some evidence does exist that critical roles are played by the N-linked glycosyl groups (11). N-linked glycosylation has been postulated to have various functions in different proteins, e.g. as a signal for intracellular sorting and cell-cell interactions, promoting resistance to proteases, preventing aggregation of proteins, and maintaining proper conformation and hydrophilicity for protein transport, processing, and/or activity (reviewed in Refs. 12 and 13). If the receptor is to perform its * This work was supported by National Institutes of Health Grant DK-40608. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ T o whom correspondence should be addressed. Tel.:  function, it must be expressed at the cell surface, probably must form dimers, and must be able to bind transferrin. Previous studies in this lab have indicated that in A431 cells (a human epidermoid cell line) treated with tunicamycin, an inhibitor of N-linked glycosylation, the unglycosylated transferrin receptor is not transported to the cell surface, does not bind transferrin, and does not form dimers (11). Because of the pleiotropic effects of tunicamycin, however, we have sought to confirm the putative roles of the N-linked glycosyl groups by more specific means: the oligonucleotide-directed mutagenesis of the consensus sites for N-linked glycosylation. Specific alteration of these consensus sequences to prevent N-linked glycosylation would result in forms of the receptor lacking N-linked glycosyl groups, the behavior of which could then be examined without radically altering normal cell functioning.

MATERIALS AND METHODS
Plasmids pCD-TR1 and pZipNeoSV(X) were gifts from Dr. A. McClelland, Yale University, and Dr. M. Lane, State University of New York Health Science Center, Syracuse, respectively. NIH-3T3 cells were obtained from American Tissue Culture Collection. A431 cells were from Dr. Graham Carpenter, Vanderbilt University. ,'Ins-Labeled methionine/cysteine was from ICN. Geneticin, gentamycin, Dulbecco's modified Eagle's medium and fetal bovine serum were from GIBCO. Sfaphylococcus aureus and tunicamycin were from Calbiochem. Rat anti-binding immunoglobulin protein (BiP)' culture fluid was a gift of Dr. David Bole, Yale University. Molecular weight markers ("C-labeled) were from Bethesda Research Labs (BRL).
Cell Culture and Radiolabeling"A431 and NIH-3T3 cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and gentamycin (50 pg/ml); subconfluent cells in 24-well plates (A431, 3 X 10"-4 X lo5 cells/well) or 6-well plates (NIH-3T3, 2 X l(l-4 X lo" cells/well) were washed twice with sterile phosphate-buffered saline prior to addition of 0.5 (24-well plate) or 1.0 ml (6-well plate) of Dulbecco's modified Eagle's medium minus methionine containing 5% fetal bovine serum. The cells were labeled by adding 10-12.5 pCi per well (24-well plates) or 5-10 pCi per well (6-well plates) of ,"'S-labeled methionine/cysteine for 12-24 h. When cells were treated with tunicamycin, they were incubated in medium containing 1 rg/ml of this inhibitor for 1 h prior to addition of label. In pulse-chase experiments, cells were labeled with 20 pCi of ,'"Slabeled methionine/cysteine in Dulbecco's modified Eagle's medium minus methionine containing 5% fetal bovine serum at 37 "C, 5% COe. After 1 h, the radioactive medium was removed, the cells washed with phosphate-buffered saline and then incubated with complete medium for 0-6 h.
into Ml:lmpl9, and the 1.5-kl) Hindlll-Xhnl fragment WAS sut)cloned into M1:lmplH. for the mutagenesis procedure. Site-directed mutagenesis was performed using the Amersham oligonucleotide-directed mutagenesis svstem. The r o n s e n~u~ sequenre lor asparagine-linked glycosvlation is Asn-X-Ser/l'hr where X is any amino acid except proline. Oligonucleo:ides were designed so that each glvcosylation site had an alteration in the consensus sequence. The oligonucleotides were synthesized hv 13. 'I'he wild type human transferrin rereptor cDNA coding region was also inserted into the same vector and selected for proper orientation; NIH-:IT:X cells were statdv transfected using the Cal'Ol method (14). (ienetirin (0.8 mg/ml)-resistant colonies were cloned and screened for the presence o f human transferrin receptor using a polvrlonal goat antihuman transferrin receptor ant ibodv t)v \Vestern blotting o n MSI Nitroplus 2000 hvhridization memhrane. Detection was by swine antigoat-conjugated alkaline phosphatase (Hio-Had) as per mnnufacturer's instructions.
/mrnunoprrripifnfion-Metaholirall~ laheled cells (2 X 10"-4 X 10'' rells) were washed twice with phosphatrtl-t)r~ffered saline T h e eluates were heated to 95 "C for : i min and suhjected to electrophoresis on SDS-polvarrvlnmide gels (6 or flf';,) overnight at 60-120 V according t o the method o f I,aemmli ( I S ) . T h e gels were agitated for at least :I0 min in 5"; acetic arid. 25' ;' methanol. then for :!O min in Amplifv (Amersham Cnrp). The gels were dried antl autoradiographed at -70 "C using Kodak S-Omat AH film. Autoradiographs were quantitated using the Molecular 1)vnamics :WOA Comprrting 1)ensitometer. Earh experiment was repeated two to five times, antl the autoradiographs were scanned. The averages and standard deviat ions ofthe experiments are presented under "Hesults." T h e polvrlonal antihodv was tested to determine if this antitmdv was capable of quantitatively immunopreripitating the mutated version of the human transferrin receptor (see Fig. I ) . Two methods were used to demonstrate quantitative immunoprecipitation. Different amounts of radioactivity Iat)eled cell extracts were immunoprecipitated using the same amount of antihodv and S . nurcus to deter-  l s t r a n s f e c t r d w i t h w i l d t y p e or m u t a n t u n g l y c o s y l a t e d h u m a n t r a n s f e r r i n receptor. A , untransfected mouse rells (.'17:7) o r mouse cells transfected with the wild type ( W T ) o r unglvcosylated mutant ('/'/tPIA human transferrin receptor were treated (+) or not (-) with 1 pg/ml tunicamvrin for 1 h prior to radioactively Iaheling them overnight in the rontinued presence or ahsence of tunicamvcin. They were soluhilized and immunoprecipitated with human transferrin receptor antiserum. The immunopreripitates were suhjected to electrophoresis in a reducing SDS-8'Y polvarrylamide gel and to autoradiographv. "C molecular weight markers are in order of increasing mohilitv; phosphorvlase, 97,400; hovine serum alhumin, fH,OOO; ovalhumin, 43,000; and carhonic anhydrase, 29.000. H , unlaheled crll extrarts from control cells and cells treated for 24 h with 1 pg/ml tunicamvrin were suhjected to electrophoresis on SI)S-XPO polyacrylamide gels under redwing conditions, transferred to nitrocellulose, immunotletecterl with antiserum to transferrin receptor, and visualized hv chemiluminescence.
as descrihed ahove in the immunoprecipitation section. After pelleting, the supernatant was incuhated at 4 "C with 200 pl oftransferrinagarose in NET-Triton (1:l v/v) on a rotator for 1 h. The transferrinagarose was pelleted as hefore, and the supernatant was then immunoprecipitated with 2 pI of goat, anti-human transferrin receptor and 35 pl of washed S. nurrus (10% solution). The pellets were resuspended in HIPA. spun through 1 ml of IIII'A-I.5p~ sucrose. and eluted in 100-150 pI of 1,aemmli huffer and suhjected to electrophoresis as indicated ahove.
Iflrnliznlion of Trnnsfrrrin Rrcrptor-The medium in wells of metaholically laheled cells was replaced with 1 ml of Hepes-huffered Dulhecco's modilied Eagle's medium, and incuhated on ice with 1.5 pI of goat anti-human transferrin receptor for 1 h. The cells were washed with phosphate-t,uffered saline, soluhilized in NET-Triton, and centrifuged 2 0 min at 14,500 X at 4 "C. After pelleting, the supernatant WAS incuhated nonspecificallv with 7 0 pl of washed S . nurrus (IO";, solution) for 1 h rorking on ire, then pelleted. The final supernatant was immllnoprecipitatecl with 35 pl of washed 5'. nurrus (10% solution) and 1.5 pl of goat anti-human transferrin receptor to ohtain the internal receptors. The first pellet, containing the external receptors, and the final pellet, containing internal receptors, were processed as descrihed previouslv.
All three of these glycosylation sites appear to be used ( 5 , 19). We have used site-directed mutagenesis to alter these sequences so that the cell no longer recognizes them as sites for N-linked glycosylation. The cDNA coding region for this receptor is contained in a 2.5-kb RamHI-XhaI fragment in pCD-TRI; there is a Hind111 site approximately 0.9 kb from the 5' end of the coding region. T h e 5' HamHI-Hind111 portion, containing the codon for the first. N-linked glycosylation site (Asn-251), was inserted into M13mp18, and the Ser-253 codon altered to Ala. The remaining portion of the coding region was inserted into M13mp19 and the asparagine codons at posit,ions 317 and 727 were altered to aspart,ic acid and lysine, respectively. These two fragments were recombined to reconstruct the ent.ire coding region in t.he vector pZipNeoSV(X), which contains the selectable marker for geneticin resistance and is capable of transfecting eukaryotic cells. The wild t?rpe human transferrin receptor was also inserted into this vector to serve as a control.
The unglycosylated human transferrin receptor, also referred to as the triple mutant or TRPL, when transfected into mouse NIH-3T3 cells, was distinguishable from the unmutated, wild type (WT), human transferrin receptor on the basis of mobility on SDS-polyacrylamide gels (Fig. 2). Unlike the human transferrin receptor from human cell lines including A431, K562, and HI, 60 cells, the wild t-ype human transferrin receptor can be resolved into a doublet when the immunoprecipitate is subjected to electrophoresis for an extended period of time (Fig. 2 A ) . T h e mobility of the upper and lower bands correspond to 92 and 87 kDa, respectively. Two major bands at 83 and 78 kDa and one minor band at 97 kDa are immunoprecipitated from 3T3 cells transfected with t.he mutated transferrin receptor. Western analysis of whole cell extracts confirms that the 83-and 97-kDa proteins are transferrin receptor, whereas the 78-kDa protein is not detected by the anti-transferrin receptor serum. The mobility of the 83-kDa protein is consistent with the loss of three carbohydrate side chains in the mutated receptor, and the 97-kDa protein is the mouse transferrin receptor as will be demonstrated in Fig. 3. The fact that much more of the mouse transferrin receptor is precipitated with the mutated human transferrin receptor than is immunoprecipitated in 3T3 cells alone, implies that the two receptors form heterodimers. This result will be shown in Fig. 3.
T h e mobility of the wild type transferrin receptor from cells treated with tunicamycin was compared to the mobility of the mutated transferrin receptor: ( a ) to verify that the mutated transferrin receptor was not glycosylated on asparagine resi- resis o f w i l d t y p e a n d u n g l y c o s y l n t e d m u t a n t h u m a n t r a n s f e rr i n receptor. A , A.t:\I rells lA.l.'l/ ), :!'I3 rells l:1'/?1), untrans!vc.ted or transfected with wild t>pe ( W 7 ' ) or mutant unglycosylatecl I '/'/{/'/A human transferrin receptor, were metahnlically Int)eled nvrrnight and suhjected to immunoprecipitation, elertrophoresis under nonrt.drlcing conditions, and autoradiography. Monomers l r n ) and dimers I d ) nrr indirnted for the unglycosylated transferrin rereptor. I)imers are helieved to he mouse hnmodimer, mouse-human heterodimer. and human homodimer, in order of increasing mohilitv. fj, unlahrled cell extracts were run under nonreducing conditions. transferred t o nitrocellulose, prohed with antiserum t o the transferrin receptor. and visualized hv rhemiluminescence detection. were visualized hv autoradiography. dues and ( b ) to determine if the 78-kDa protein seen associated with the mutated transferrin receptor was also associated with the unglycosylated form of t h e wild type receptor in tunicamycin-treated cells. When cells containing the wild type transferrin receptor are treated with tunicamycin the molecular mass of the receptor decreases to 83 kDa and protein synthesis decreases (Fig. 2). As expected, the molecular mass of the mutant receptor is not altered and remains 83 kDa upon tunicamycin treatment indicating that the receptor is not glycosylated on asparagine residues. Western analysis of whole cell extracts from tunicamycin-treated cells confirms that the 83-kDa protein is t,he transferrin receptor (Fig. 2R). This result indicates that the 83-kDa protein which is metabolically labeled after tunicamycin treatment is the unglycosylated transferrin receptor. In contrast, the 78-kDa protein is not immunodetected by antiserum to the transferrin receptor. This protein is only associated with the unglycosylated form of the receptor (Fig. 2 4 ) . Treatment of cells containing the mutant transferrin receptor with tunicamycin increases the amount of 78-kDa protein in the immunoprecipitates and diminishes the amount of glycosylated mouse transferrin receptor.
Intersubunit Disulfide Bond Formation-The mature transferrin receptor in human cells is a dimer linked by two intersubunit disulfide bonds. Two-dimensional gel electrophoresis and pulse-chase experiments using nonreducing SDSgel electrophoresis were employed to evaluate intersubunit disulfide bond formation in the mutated and the wild t-ype human transferrin receptor transfected into mouse 3T3 cells. Both the wild type and the mutated transferrin receptors are capable of forming intersubunit disulfide bonds but differ with respect to one another in the extent and rates of disulfide bond formation (Figs. 3 and 4). Immunoprecipitates of cells containing the wild type transferrin receptor which are labeled overnight ( o n ) show essentially complete disulfide bond formation using nonreducing SDS-polyacrylamide gel electrophoresis (Fig. 3 A ) . In contrast, immunoprecipitates of cells containing the mutated transferrin receptor show incomplete disulfide bond formation and three bands ranging in molecular mass from approximately 140 to 200 kDa ( Fig. 3.4). T h e 200-kDa band seen in 3T3 cells and the 190-kDa band seen in the immunoprecipitates of 3T3 cells and the cells transfected with the human wild type transferrin receptor are confirmed to be the mouse and human t.ransferrin receptor, respectively, by Western analysis. Likewise the three bands These bands are postulated to he (in descending molecular mobility): the mouse homodimer by comparison with the unt,ransfect,ed 3T3 cells, the mouse human heterodimer, and the human homodimer. These identities were confirmed by excising the hands run under nonreducing condit ions, reducing the proteins in these bands with mercaptoethanol, and suhjecting them to electrophoresis (Fig. :I(.'). Densitomet ric scanning of the nonreduced gels indicates that during a If-h label, 59 +-5% of the mutant remains in a nondisulfidebonded form, with 58 k 1% of the dimers as a mouse-human heterodimer, and with 42 k 106 a s a human homodimer. In cont.rast, over 98% of the wild t-ype receptor in the cells is converted to a covalently dimerized form during a If-h label.
The rate of formation of the two higher molecular weight bands of the mutant transferrin receptor is similar to the rate of formation of the interdisrllfide-linked dimer of the wild t-ype transferrin receptor (Fig. 4, WT, 0-6 h). Disulfide bond formation is essentially complete for the wild type transferrin receptor in 4 h. The rate of formation of the lowest hand in the triplet of mutated transferrin receptor is much slower and is st,ill incomplete at the end of a 6-h chase when compared to the immunoprecipitate from cells laheled overnight (Fig. 4, TRPL, 6 h us. on). These results indicate that in mouse cells. the triple mutant is capable of forming intersuhunit disulfide bonds but does so over a much slower time frame and not nearly as efficiently as the wild t-ype human transferrin receptor transfected into mouse cells.
Cell Surface Expression of the Mutntrd Trnnsfcrrin Rccrptor-To examine cell surface expression, intact cells which were metabolically labeled were exposed to anti-human transferrin receptor antiserum, the cells solubilized, and the surface bound receptors isolated. The remaining transferrin receptors from the supernatant were then immunoprecipitated with anti-human transferrin receptor to determine internal receptor content (Fig. S A ) . In A431 cells, approximatelv I 5 f 1 0 7 of the transferrin receptor appears on the cell surface (11):' In mouse cells transfected with wild t-ype human transferrin recept,or, a similar proportion is expressed on the surface (18 f 10%). Mouse cells transfected with mutant transferrin receptor, on the other hand, express at the most 1.2 f Ir; of the receptor at the cell surface (or 15 times less mutated transferrin receptors on the cell surface than wild type transferrin receptors). This method precludes preadsorptinn oft he cell extracts with S . nureus and results in the visualization of many proteins which nonspecificallv bind to S. nurrus. This complication makes quantitation of the amount of unglycosylated transferrin receptor on the surface of the cells difficult. In order to verify the quantitation of cell-surface transferrin receptor, the same procedure was used on unlaheled cells. The transferrin receptors were visualized hv \Vestern analysis using the anti-transferrin receptor antibodv (Fig.  513). In addition, cell surface iodination of transfectants with Na '.'"I and lactoperoxidase and immunoprecipitation was used to quantitate cell surface mutant and wild type transferrin receptor. This method yields 20 times less of the mutated receptor than wild type on the cell surface (data not shown).
Ahility of thr Mutated Trnnsfcrrin Rrccptor to Hind to Trnnsfcrrin-Ajinrosc-The ability of the transferrin receptor to bind transferrin was measured to indicate i f the mutated receptor has similar quaternary structure to that of the wild t-ype transferrin receptor. Cell extracts were soluhilized and incubated with transferrin-agarose (Fig. 6). After the hinding step, receptors remaining in the supernatant and not capable  T r a n s f e r r i n h i n d i n g o f w i l d type (Wn a n d u n g l yc o s y l a t e d m u t a n t (TRI'L) h u m n n t r a n s f e r r i n r e c e p t o r . ('ells were solut)ilizetl and inc~rl);rted with t rnnsfrrrin-agarose o n a rotator a t 4 "C for 1 h. Alter pelleting the transferrin-agarose hound transferrin receptor, t he supernatant was imrnrrnol)reri~)itnted with washed S . nurvus antl gont ant i-human t ransferrin receptor. Pellets were treated as indicated under "Materials and Methods," suhjected t o electrophoresis on SDS-8";~ polv;lcrvlamidc, t ranslerretl to nit rocel-Inlose, and detected with anti-transferrin receptor antihody hy Western analysis and chemiluminescence. l a n e s marked !fare translerrinagarose hound elrrntes; lanes marked i are immlrnoprccipitates. of binding to the t,ransferrin agarose were immunoprecipitated with anti-human transferrin receptor. Western analysis was used to detect the relative amount of t.ransferrin receptor in each fraction. The mutant receptor shows reduced binding to transferrin-agarose compared to the wild t-ype receptor. Hetween 75 and 90% of the wild t-ype transferrin receptor and only 1 to 5% of the mutated transferrin receptor were isolated with transferrin-agarose.

Rip-Human Transferrin Rewptor Association
Association of thr Ung1.vcosvlatrd Mutatrd Transfrrrin Recrptor with HiP-RiP has been shown to associate with a variety of immature and malfolded proteins in the endo- C t * l l c.xtr;tcts t(.) or anti-human transterrin receptor immrrnol)rrc.il)it:rtc~s t I ) werc' subjectetl to electrophoresis, translerred t o nitrocc~llrrlow. and prot)ed with antihotly t o Hil'. T h e h h f was devrlopd rrsin~ rathit nnti-rat serum, followed by alkaline phosphatnse-r~~~rple~l p m t anti-rathit antihotly and sut)strate. As controls. YIH-3'I':{ cells containing wild type human transferrin receptor ( W 7 ' ) were trr:rtrd with trrnicarnycin for 1 6 h or left untreated. a n d sut)jerted t o the same protocol ;IS t h r cells carrying the mutant receptor. A431 cells were also trr;rtetl with tuniramycin antl subjected t o the same protocol. plasmic reticulum. In tunicamvcin-treated A431 cells, unglycosylated transferrin receptor is associated with Hip, a 78-kDa endoplasmic reticular protein, and other unglycosylated proteins.' We wanted to determine if the mutated human transferrin receptor was associated with HiP in the absence of unglycosylated proteins. Immunoprecipitation of the mutated transferrin receptor with transferrin receptor antiserum reveals the presence of a band with variable intensity at 78 kDa that coprecipitates with mutant transferrin receptor ( Figs. 1 and 2). It coprecipitates with the mutated transferrin recept.or during pulse-chase experiments and does not associate detectably with the wild type human transferrin receptor during its biosynthesis (Fig. 7 ) . To determine if the coprecipitat,ing protein is HiP, cell extracts were immunoprecipitated with anti-human transferrin receptor serum, anti the immunoprecipitates were suhjected to SDS-polyacrylamide electrophoresis. Western immunodetection using a monoclonal antibody to RIP reveals a 78-kDa protein which coprecipitates with the mutated transferrin receptor hut not the wild type transferrin receptor (Fig. 8 ) . If 3T3 cells containing the wild t-ype transferrin receptor or A431 cells are treated with tunicamycin, HIP is detected in these immunoprecipitates also (Fig. 8). I t therefore appears that the unglvcosylated form of transferrin receptor associates with BiP, and the resulting complex is stable enough to be isolated.

DISCUSSION
Little success has been achieved in assigning functions to the posttranslational modifications of the transferrin receptor. Acylation is required for membrane association of some proteins, but Jing and Trowbridge (2) show that membrane association was normal in transferrin receptor in which the cysteine at position 62 was altered to serine by site-directed mutagenesis, making acylation impossible. Phosphorylation has been suggested to play a role in the endocytosis and recycling of transferrin receptor, since phorbol esters increase the phosphorylation state of transferrin receptor in some cell lines and down-regulate transferrin receptor at the cell surface (20). However, other researchers have shown a decrease in phosphorylation of the receptor upon endocytosis in other cell lines (21). Furthermore, elimination of the phosphorylation site (Ser-24) by site-directed mutagenesis did not affect endocytosis and recycling (22).
Glycosylation of asparagine residues appears to have the most profound effect on transferrin receptor function. In A431 cells treated with tunicamycin, transferrin binding to the unglycosylated receptor, surface expression, and dimerization are abolished (11). Tunicamycin is a heterogeneous group of uracil-based compounds which prevents synthesis of oligosaccharide moieties which are transferred from the lipid carrier dolichol phosphate to the asparagine target on the polypeptide chain (12). However, tunicamycin also may inhibit protein synthesis in many cell lines (12) and also could exert an indirect effect by perturbing the synthesis and function of proteins involved in the biosynthesis of the transferrin receptor. Consequently, there must be reservations with regard to the interpretation of results obtained from tunicamycintreated cells. We have therefore used site-directed mutagenesis to eliminate the N-linked glycosylation sites of the human transferrin receptor. Alteration of the consensus sequence Asn-X-Ser/Thr prevents N-linked glycosylation of the initial asparagine residue. The wild type and mutated transferrin receptors were transfected into mouse 3T3 cells as previous investigators demonstrated that this receptor was functional in these cells (10). The present studies indicate that the wild type transferrin receptor is processed slower than in some human cell lines (23) and at a similar rate to that of other human cell lines (5). It does fold correctly and is transported to the cell surface. In contrast to wild type transfected transferrin receptor, the mutated form of the human transferrin receptor shows an electrophoretic mobility and insensitivity to tunicamycin treatment consistent with complete lack of Nlinked glycosylation. While retaining the ability to be immunoprecipitated by a polyclonal antibody specific for human transferrin receptor, this transferrin receptor mutant has a diminished ability to form dimers and to bind human transferrin, and exhibits reduced cell surface expression. Association with BiP, an endoplasmic reticular protein which is believed to associate with improperly folded polypeptides, further suggests that a fraction of this mutated receptor is retained within the cell. Whether the altered properties of the mutant receptor are due simply to lack of the glycosyl groups, or are the result of an altered conformation due to the amino acid substitution is not easy to assess. The similarity of the results we obtained in previous studies using tunicamycin to prevent glycosylation with the site-directed mutagenesis studies strengthens the proposal that N-linked glycosylation plays an important role in the proper folding of the transferrin receptor.
While the mutated receptor shares many of the properties of the unglycosylated transferrin receptor in tunicamycintreated cells, it differs from this form of the receptor in one respect. It appears to have more tertiary structure than the receptor from tunicamycin-treated cells, as it does form heterodimers with the mouse transferrin receptor and to a more limited extent it does form homodimers with itself. In earlier studies using tunicamycin-treated A431 cells, we found the unglycosylated transferrin receptor was a part of a large complex, consisting of BiP and other unglycosylated proteins (11,17).3 This unglycosylated transferrin receptor did not form detectable intersubunit disulfide bonds, did not bind transferrin, and was not transported to the cell surface. Treatment with tunicamycin results in an increase in the amount of BiP immunodetected in these cells. We hypothesize and are currently examining the possibility that the increased amount of BiP and other unglycosylated proteins in these treated cells is contributing to the lack of processing of the receptor. Thus the differences we observe between the unglycosylated transferrin receptor from tunicamycin-treated cells and the unglycosylated mutated receptor could result from the fact that in tunicamycin-treated cells the unglycosylated receptor is part of a large complex with BiP and other unglycosylated proteins.
We do not know if the association of the transferrin receptor with BiP is part of the biosynthetic pathway for the wild type transferrin receptor. An association, even a transient one, between BiP and the normal transferrin receptor has not been detected, although the association may be weak and lost during immunoprecipitation. This latter phenomenon has been observed with other nascent proteins thought to be associated with BiP (24, 25). If BiP only associates with the unglycosylated form of the receptor, then during its biosynthesis it would be released when the transferrin receptor is glycosylated.
We hypothesize that association with BiP results in retention of much of the unglycosylated receptor in the endoplasmic reticulum. Our data indicate that dimerization alone is not sufficient for correct folding of the receptor and transport to the cell surface. Approximately 50% of the mutated transferrin receptors form intersubunit disulfide bonds. Given the distribution of the wild type transferrin receptor in this cell line (18 f 10% outside) we would predict if dimerization were the only requisite for cell surface expression we should see -9% of the mutant receptors on the cell surface. In both the immunolocalization and cell surface "'I-labeling experiments no more than 2.6% (average 1.2%) of the mutated receptor was seen on the cell surface. The intracellular location of the dimers remains to be identified.