A common autoimmunity predisposing signal peptide variant of the cytotoxic T-lymphocyte antigen 4 results in inefficient glycosylation of the susceptibility allele.

A common T17A polymorphism in the signal peptide of the cytotoxic T-lymphocyte antigen 4 (CTLA-4), a T-cell receptor that negatively regulates immune responses, is associated with risk for autoimmune disease. Because the polymorphism is absent from the mature protein, we hypothesized that its biological effect must involve early stages of protein processing, prior to signal peptide cleavage. Constructs representing the two alleles were compared by in vitro translation, in the presence of endoplasmic reticulum membranes. We studied glycosylation by endoglycosidase H digestion and glycosylation mutant constructs, cleavage of peptide with inhibitors, and membrane integration by ultracentrifugation and proteinase K sensitivity. A major cleaved and glycosylated product was seen for both alleles of the protein but a band representing incomplete glycosylation was markedly more abundant in the predisposing Ala allele (32.7 +/- 1.0 versus 10.6% +/- 1.2 for Thr, p < 10(-9)). In addition, differential intracellular/surface partitioning was studied with co-transfection of the alleles fused to distinct fluorescent proteins in COS-1 cells. By quantitative confocal microscopy we found a higher ratio of cell surface/total CTLAThr(17) versus CTLAAla(17) (p = 0.01). Our findings corroborate observations, in other proteins, that the signal peptide can determine the efficiency of post-translational modifications other than cleavage and suggest inefficient processing of the autoimmunity predisposing Ala allele as the explanation for the genetic effect.

and in the maintenance of peripheral tolerance (1,12,13). The CTLA-4 knockout mouse exhibits a severe lymphoproliferative disorder, autoimmune disease, and early lethality, demonstrating the importance of CTLA-4 in the modulation of T-cell response (14,15). Patients with the Chediak-Higashi syndrome present symptoms not unlike those of the CTLA Ϫ/Ϫ mouse, because of a defect in the CTLA-4 cycling pathway caused by mutations in the lysosomal trafficking regulator gene (LYST) (16). More subtle reductions in expression or function of CTLA-4 may determine susceptibility to common autoimmune diseases.
Indeed, common CTLA-4 polymorphisms have been found to confer susceptibility to type 1 diabetes (17)(18)(19)(20)(21)(22)(23), thyroid disease (22, 24 -33), and several other autoimmune disorders (34 -38). By the transmission disequilibrium test, association with diabetes has been narrowed down to a haplotype encompassing CTLA-4 but not adjacent genes (21,27). The haplotype consists of three CTLA-4 polymorphisms in tight linkage disequilibrium (LD) with each other that includes a C/T Ϫ318 transition in the promoter, a signal peptide amino acid substitution (T17A) (49) and a microsatellite (AT) n repeat in the 3Ј-untranslated region (21,27,39). Because of the tight LD, contribution to diabetes susceptibility cannot be genetically dissected and functional studies are required to define the etiological variant(s). We decided to focus on the nonsynonymous signal peptide polymorphism as the most likely candidate. Two recent reports present evidence that T-lymphocytes from subjects homozygous for the diabetes predisposing G (Ala) allele of the CTLA-4 signal peptide showed enhanced proliferation and cytokine production after in vitro stimulation compared with cells from homozygotes for the protective A (Thr) allele (40,41). Although Muerer et al. (41) presented some nonquantitative evidence of defective CTLA-4 targeting to the cell surface by confocal microscopy, neither study addressed the molecular mechanism for the differential behavior of T-lymphocytes from homozygotes for each genotype.
Because the signal peptide is co-translationally cleaved in the endoplasmic reticulum (ER) and is not a part of the mature protein, we hypothesized that the T17A polymorphism determines differential targeting to the cell surface by altering early intracellular trafficking of CTLA-4. Signal peptides function in directing ribosome-bound nascent polypeptides to the (ER) membrane where they assure the translocation of growing polypeptide chains into the ER lumen. In conformity with most signal sequences, the CTLA-4 signal peptide has three distinct regions (Fig. 1): a predicted hydrophobic sequence of 12 amino acids flanked by two helix-breaking prolines, a small polar COOH terminus (C-) region encompassing the recognition site for signal peptide cleavage, and a long NH 2 terminus (N-) region of 20 amino acids that includes the Thr to Ala substitution at position 17. As shown in Fig. 1A, the Ala allele introduces a hydrophobic amino acid in a highly conserved position, occupied by a serine or threonine in 24 of 25 other species found in a BLAST search we performed (four species are shown for illustration purposes in Fig. 1A). This change somewhat alters hydrophobicity and ␣-helix propensity, two properties known to be important in signal peptide function (Fig. 1B). Because Ala homozygotes are very frequent in the general population, the functional consequences of this substitution must be subtle, as expected of common alleles predisposing to complex disorders. These allelic effects may involve differential binding to the signal recognition particle, anchoring of the signal peptide to the ER membrane, entry of the growing polypeptide into the lumen, signal peptide cleavage, and possibly, association with chaperones and other quality control elements of the ER lumen that might affect folding and glycosylation (42).
The work reported here was aimed at testing the hypothesis that the T17A substitution in the signal peptide of CTLA-4 alters the early ER trafficking and/or processing of CTLA-4 and leads to its differential expression on the cell surface. To test our hypothesis we used a cell-free in vitro translation system suited for examining early ER transport events. To also look for decreased expression at the cell surface, the hypothesized ultimate consequence of defective early processing, we examined allelic differences in intracellular versus cell surface CTLA-4 levels in a dual-transfection system with fusion of each allele to a different fluorescent protein and simultaneous quantification by confocal microscopy. The results suggest defective ER processing of a significant portion of the CTLAAla 17 molecules resulting in an aberrantly glycosylated product and decreased cell surface expression.  (59). The predicted signal peptide cleavage site is indicated by an arrow. B, prediction of factors that influence signal peptide function (60) was done using algorithms available at ProtScale (us.expasy.org/cgi-bin/protscale.pl). Hydrophobicity was calculated by the Kyte and Doolittle method (62) and ␣-helix propensity by Deleage et al. (61). We found that the threonine to alanine change resulted in increased hydrophobicity and in a higher propensity to form ␣-helices in the area directly adjacent to the change.

EXPERIMENTAL PROCEDURES
Construction of DNA Plasmids-Full-length CTLA-4 was amplified by reverse transcriptase-PCR from cDNA of a heterozygous individual for the signal peptide polymorphism at codon 17. For the fluorescent fusion proteins, a forward primer containing the linker NheI (underlined), ATAGCTAGCATGGCTTGCCTTTGGATTTCAG, and an antisense primer containing the linker AgeI (underlined), with two additional bases (lowercase) CACACCGGTgcATTGATGGGAATAAAATA-AGGC, ensured an in-frame, seamless fusion protein. PCR was performed with a high fidelity polymerase/Taq mixture. Adenine 3Ј overhangs were added and the PCR product was cloned into pCR2.1 (Invitrogen, Carlsbad, CA). Clone genotyping was done by PCR and digestion with Fnu4HI (New England Biolabs, Beverly, MA). Fulllength CTLAThr 17 and CTLAAla 17 digested with NheI ϩ AgeI purified from plasmid were fused to restricted NheI ϩ AgeI ECFP and EYFP (Clontech, Palo Alto, CA).
For in vitro transcription, full-length CTLA-4 with the termination codon was amplified from plasmid with the mutagenic antisense primer CACACCGGTtcaATTGATGGGAATAAAATAAGGC (the stop codon is in lowercase and the AgeI linker is underlined). Adenine 3Ј overhangs were added to the PCR product and ligated to pCR2.1 T-vector and clones were selected for orientation downstream of the T7 promoter.
Truncation of amino acids Tyr 201 -Asn 223 from the CTLA-4 COOHterminal results in a mutant constitutively directed to the cell surface with no requirement for the machinery that normally directs this translocation in activated T-cells (43). The corresponding construct was prepared by PCR amplification from cDNA of a heterozygous individual using the antisense primer ATAACCGGTgaCCCCTGTTGTAA-GAGGGCTTC (AgeI underlined) and T/A was cloned into pCR2.1. COOH-truncated CTLAThr 17 and CTLAAla 17 will be denoted herein as CTLAGlyThr 17 and CTLAGlyAla 17 in reference to Gly 200 , the last amino acid in the truncated constructs (numbering of all constructs assumes an uncleaved signal peptide of 37 amino acids). The truncated DNA plasmids were then subcloned in-frame into ECFP and EYFP plasmids, using the NheI ϩ AgeI sites. All resulting DNA constructs were sequenced.
Cell Culture, Transfection, and Antibodies-COS-1 cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 units/ml streptomycin, at 37°C and 5% CO 2 . Transient transfections were carried out using FuGENE 6 (Roche Molecular Biochemicals) in 6-well plates on glass coverslips according to the manufacturer's instructions. Cells were processed 24 -48 h after transfection.
In Vitro Transcription, Translation, and Endoglycosidase H Digestions-Full-length cDNA encoding CTLAAla 17 and CTLAThr 17 (or their glycosylation mutants) in pCR2.1 was linearized by SpeI and in vitro transcribed using T7 RNA polymerase (Promega, Madison, WI). The resulting mRNA was standardized to 1 mg/ml and equal amounts were added to a premixed mixture containing rabbit reticulocyte lysate with [ 35 S]methionine (Ͼ1000 Ci/mmol) (Amersham Biosciences), in the presence or absence of canine pancreatic microsomal membranes according to the manufacturer's instructions (Promega). Translation products were denatured by boiling in 2ϫ SDS loading buffer and 10 -15-l aliquots were run on 15% SDS-PAGE, followed by autoradiography at Ϫ70°C. Endoglycosidase H (EndoH) (New England Biolabs) digestions were performed on in vitro translated protein. Briefly, proteins were treated with 75 mM dithiothreitol and 10% SDS and boiled for 5 min. The resulting supernatant was incubated with EndoH, for the indicated times at 37°C. Products were analyzed by SDS-PAGE and visualized by autoradiography as described above.
Proteinase K Digestions, Ultracentrifugation, and Signal Peptidase Inhibition-To distinguish in vitro translated CTLA-4 molecules integrated in the ER membrane from those in the cytosolic phase, we used proteinase K digestion. Translation products were treated with 100 mM CaCl 2 , and incubated for 1 h with 20 g/ml proteinase K (Sigma) with or without Triton X-100 (1%) at 0°C. The reaction was terminated with phenylmethylsulfonyl fluoride (2 mM final concentration). As an addi-tional measure of cytosolic/ER partition the translation products were separated by ultracentrifugation. Briefly, reactions were diluted 100fold with Na 2 CO 3 (pH 11.5), and centrifuged at 100,000 rpm for 30 min at 4°C. The pellet was resuspended in 1ϫ SDS sample buffer and denatured by boiling for 3-4 min. The supernatant was concentrated by column filtration (Microcon, Bedford, MA) to 1/10 of its original volume. The sample was prepared for SDS-PAGE by denaturation in 2ϫ SDS sample buffer. All products were resolved by SDS-PAGE and visualized by autoradiography. Finally, inhibition of the signal peptidase was achieved by incubation of the translation mixture with 5 mM (final concentration) N-methoxysuccinyl-Ala-Ala-Pro-Val-chloromethyl ketone (Sigma) in 10% Me 2 SO for 1 h at 30°C.
Image Processing and Quantification and Confocal Microscopy-Transfected cells were washed 3 times in phosphate-buffered saline and fixed with ice-cold 50:50 acetone/methanol for 3 min or fixed with freshly prepared 4% paraformaldehyde for 30 min. The fixed cells were mounted on standard microscope slides with conventional mounting media. Samples were analyzed by confocal microscopy (LSM 510, Zeiss Axiophot, Germany). Simultaneous double acquisitions were performed using the 458-and 514-nm laser lines to excite cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP), respectively, using ϫ63 oil immersion Neofluar objectives. The fluorescence was selected with double fluorescence dichroic mirror and a band pass filter of 480 -520 nm for CFP and a long pass filter of 560 nm for YFP. The images were processed and analyzed using NIH Image freeware available from the National Institutes of Health (NIH) website (rsb.info.nih.gov/nih-image/Default.html). An outline of the cell surface was drawn with the cursor and the mean density was calculated for each color acquisition simultaneously. The same was done for the intracellular fluorescence (both procedures are illustrated in Fig. 7C). For densitometry measurements taken from autoradiographs, background density was subtracted from each band and the same area was selected for all measurements within one experiment. 17 and CTLAThr 17 cDNA constructs were in vitro transcribed and translated in the absence of microsomal membranes, [ 35 S]methionine-labeled protein products from both allelic forms migrated with an apparent molecular weight of 26,000. This is consistent with the calculated molecular mass of monomeric, uncleaved, unglycosylated CTLA-4 ( Fig. 2A, lanes 1 and 4). Upon addition of microsomal membranes translation products of 29 kDa were apparent for both alleles, corresponding to the size change expected from high mannose glycosylation in two positions and signal peptide cleavage. In addition, the CTLAAla 17 reaction ( Fig (Fig. 2B) (p Ͻ 1 ϫ 10 Ϫ9 , n ϭ 11 independent experiments). This ratio was independent of the amount of microsomal membranes in the reaction, within a range from 0.8 to 2.8 eq membranes added to the reticulocyte lysate. One observation worth noting was that translation efficiency was substantially increased in the presence of microsomal membranes suggesting that co-translational CTLA-4 processing increases translation efficiency. Having established that a portion of CTLAAla 17 molecules is differentially processed in the ER, we next examined the nature of this allele-specific intermediate.

CTLAAla 17 Is Inefficiently Processed in the ER-When otherwise identical full-length CTLAAla
The 25-kDa CTLA4Ala 17 fraction, close in size with the band seen in the absence of microsomes, could represent an uncleaved and unglycosylated form that has not undergone any processing in the ER. Alternatively, it could represent correctly translocated but incorrectly cleaved and/or glycosylated products.
Signal Peptide Is Cleaved in the Major Products of Both CTLAAla 17 and CTLAThr 17 -According to SignalP, a signal peptidase recognition site prediction algorithm (45) (www. cbs.dtu.dk/services/SignalP/index.html), both alleles had the same cleavage site at Ala 37 -Met 38 (Fig. 1A), resulting in a 37-amino acid cleaved signal peptide. Cleavage by the signal peptidase was experimentally confirmed by using a signal peptidase inhibitor (N-methoxysuccinyl-Ala-Ala-Pro-Val-chloromethyl ketone). For both allelic forms, inhibition of cleavage gave rise to a full-length intermediate of ϳ33 kDa resulting from a gain of 4 kDa by the 29-kDa intermediate, corresponding to a gain of the 37-amino acid signal peptide (Fig. 3). Thus the major bands in the translated products of both CTLAAla 17 and CTLAThr 17 represent cleaved intermediates. The cleavage status of the additional CTLAAla 17 25-kDa band was less clear because of band overlap (Fig. 3), as inhibition was not complete at the maximum effective concentration of inhibitor (higher concentrations inhibited translation). Thus the major bands for both alleles must represent fully glycosylated and fully cleaved product.
CTLA-4 Has Two N-linked Glycosylation Sites-CTLA-4, a cell surface receptor, has two predicted N-linked glycosylation sites located in the extracellular domain, at Asn 113 and Asn 145 , thought to be important for structural integrity (46). Although not all instances of the consensus sequon Asn-X-Thr/Ser (X ϭ any amino acid) are necessarily glycosylated (47,48), the observed size of the major band clearly indicated full glycosylation of most of the molecules for either allele. This was further confirmed with partial digestion of the in vitro translation products with EndoH that, in glycoproteins with several glycosylation sites, produces a ladder of partially digested molecules differing by only one N-linked chain. Two additional major bands with an estimated molecular weights of 25,000 and 22,000 appeared after 1-2 h of digestion (Fig. 4A, lanes 2-5) representing cleavage of, respectively, one or two N-linked glycosylation moieties from the 29-kDa form that was still visible. Further digestion for up to 16 h never produced smaller  2 and 3) abrogating the difference between the two alleles. This confirms that the 25-kDa precursor observed following microsomal processing of the CTLAAla 17 represents a cleaved but improperly glycosylated CTLA-4.
products. The 25-kDa band was indistinguishable in size from the aberrant intermediate seen with CTLAAla 17 . This allowed us to conclude firmly that, in addition to being cleaved, the 29-kDa product was glycosylated at both sites and, tentatively, that the aberrant CTLAAla 17 product represents cleaved, monoglycosylated protein.
This was further confirmed with complete EndoH deglycosylation. As expected, prolonged digestion for 16 h with EndoH resulted in a major band of 22 kDa for both alleles, with no extra band for CTLAAla 17 (Fig. 4B). Thus complete deglycosylation abolished any difference between allelic forms of the protein, allowing us to conclude that the extra band seen after microsomal processing of nascent CTLAAla 17 represents cleaved but aberrantly glycosylated CTLA-4.
Differential Processing of the Two Alleles Depends on Glycosylation-The deglycosylation results regarding the nature of the unprocessed intermediate observed in CTLAAla 17 were further corroborated by site-specific mutagenesis designed to abrogate glycosylation by removing each glycosylation site separately and both together in each allele. Asn 113 and Asn 145 , the predicted N-linked glycosylation sites, were replaced with Asp residues at the same positions (Fig. 5). The resulting constructs were in vitro transcribed and translated, as described above.
All glycosylation mutants translated in the absence of microsomal membranes behaved as the wild-type alleles and a 26-kDa translation product was detected. In the presence of ERcontaining membranes, the double glycosylation mutant migrated at 22 kDa, as expected of an unglycosylated CTLA-4 with a cleaved signal peptide. No extra band corresponding to the differentially processed CTLAAla 17 band was detectable (Fig. 5, lanes 10 and 11), clearly indicating that the allelic difference in processing requires the presence of glycosylation sites. The uniform 25-kDa electrophoretic mobility of the single-site mutants CTLAThr-17(N113D) and CTLAAla-17(N113D), CTLAThr-17(N145D), and CTLAAla-17(N145D) (Fig. 5, lanes 2, 3, and 6, 7) is consistent with loss of the signal peptide (4 kDa) and gain of a single high mannose glycosylation site (2-3 kDa). Again there was no difference between alleles, indicating that differential processing occurs only in the presence of both glycosylation sites. These results also confirm with certainty that both CTLA-4 signal peptide alleles are ER-translocation competent and can be cleaved, because a completely deglycosylated product migrates at 22 kDa, 4-kDa less than the monomeric CTLA-4 product translated in the absence of membranes.
The Partial Glycosylation Intermediate Is Not Integrated in the ER Membrane-We next addressed the question of whether the cleaved but aberrantly glycosylated CTLAAla 17 intermediate, representing almost one-third of all molecules of the auto-immunity predisposing allele in the ER, can be correctly targeted to the cell surface. Misprocessed proteins are retained in the ER bound to chaperones such as calnexin and calreticulin and eventually translocated back to the cytoplasmic phase where they undergo ubiquitin-driven, proteasome-dependent degradation (49). Partition of the aberrantly glycosylated CTLAAla 17 between cytosolic phase and ER membrane was evaluated by two independent assays. Protection from proteinase K digestion and resistance to extraction by alkaline high salt are both indicators of integration into the microsomal membranes. Because CTLA-4 is a type I transmembrane glycoprotein, its 36-amino acid COOH tail is exposed in the cytoplasm and will be digested, resulting in a 4-kDa loss.
Indeed, proteinase K digestion resulted in the expected 4-kDa reduction of the 29-kDa major bands seen with both alleles (Fig. 6A, lanes 2 and 5). A band corresponding to a 4-kDa reduction in the size of the 25-kDa intermediate seen with CTLAAla 17 is also seen (Fig. 6A, lane 5) but this form is considerably less intense relative to the upper band, indicating partial sensitivity to proteinase K digestion. By densitometric quantification, prior to digestion it constitutes 34.5% Ϯ 0.6 of total CTLAAla 17 product, but only 18.5% Ϯ 0.8 after digestion (p ϭ 0.0001, n ϭ 3 independent experiments) (Fig. 6B, in reference to lanes 5 and 6 in Fig. 6A). The corresponding percentages for CTLAThr 17 are 10.8 Ϯ 1.0 and 9.9% Ϯ 1.8, respectively (NS, n ϭ 3) (Fig. 6B, in reference to Fig. 6A, lanes  1 and 2). Thus, the allelic difference in the abundance of the incompletely processed form can be accounted for almost entirely by molecules located outside the ER.
Strikingly similar results were obtained upon treatment of translation reactions with high Na 2 CO 3 , which releases soluble and peripheral membrane proteins while transmembrane proteins remain inserted in the ER lipid bilayer, followed by separation of pellet (P) and supernatant (S) fractions by ultracentrifugation. The major 29-kDa bands were largely recovered in the pellet (Fig. 6C, lanes 5 and 6) for CTLAThr 17 and CTLAAla 17 . This supports the results found in the protease sensitivity assay and confirms that CTLAAla 17 and CTLAThr 17 are both translocation competent and integral ER membrane proteins. It was apparent, however, that significantly less incompletely glycosylated CTLAAla 17 intermediate was recovered in the pellet. By densitometric quantification it was found to be 36.2% Ϯ 1.0 of the total prior to separation by ultracentrifugation and 24.9% Ϯ 0.4 in the pellet (p ϭ 0.0078, n ϭ 3) (Fig. 6D). The corresponding values obtained for the CTLAThr 17 intermediate were 8.9% Ϯ 2.7 prior to ultracentrifugation and 10.0% Ϯ 1.2 recovered in the pellet (NS, n ϭ 3). This remarkable concordance between two independent methods demonstrates that roughly half of the aberrantly processed CTLAAla 17 is in the cytoplasmic phase. Because it has undergone signal peptide cleavage and glycosylation, this fraction does not represent failure of translocation to the ER but rather retrotranslocation back to the cytoplasmic phase for proteasomal degradation.
CTLAAla 17 and CTLAThr 17 Do Not Co-localize in Co-transfected COS-1 Cells-To determine whether the differential processing observed in the cell-free system would translate into cell surface expression differences, we devised a dual-transfection system in COS-1 cells with fusion proteins of CFP and YFP downstream of otherwise identical CTLAAla 17 and CTLAThr 17 . In these experiments reciprocal constructs were always tested to exclude possible effects because of the different properties of the fluorescent proteins. In addition, we directly demonstrated that co-transfections with the same allele tagged with each of the two FPs (e.g. CFP-CTLAThr 17 and YFP-CTLAThr 17 ) showed that the same allele did not behave differently as a result of the different fluorescent protein fusion and was targeted and expressed in the same compartments (Fig. 7A, panels a-c). Similar results were observed for CFP-CTLAAla 17 and YFP-CTLAAla 17 co-transfections (Fig. 7A, panels d-f).
CTLA-4 is known to localize to endosomal compartments in T-cells as well as in non-T cell systems (43,50). To demonstrate that the fusions of CTLA-4 resulted in a properly targeted protein, each allele was transfected independently and the cells were stained with an antibody to the transferrin receptor followed by detection by a Cy-5 conjugate. The transferrin receptor has been previously shown to co-localize with CTLA-4 in post-Golgi compartments as was the case for our fusion proteins (data not shown).
When the two alleles, now labeled with different fluorescent proteins (CFP-CTLAThr 17 and YFP-CTLAAla 17 ) were introduced into COS-1 cells clear differences in targeting were seen, with distinct regions where only one allele could be found (Fig.  7B, panels a-c, see arrows). Similar observations were made for reciprocal transfections of YFP-CTLAThr 17 and CFP-CTLAAla 17 (Fig. 7B, panels d-f). These results are consistent with our in vitro observations of cytoplasmic retrotranslocation of CTLAAla 17 . The expected resulting difference in targeting to the cell surface was then investigated.
Cell Surface Levels of CTLAAla 17 in COS-1 Cells Are Lower Than CTLAThr 17 -Using the same dual transfection system described above, CTLA-4 constructs with signal peptide allelic variants were tested for quantitative differences in expression at the cell surface of COS-1 cells. Surface expression is specific for activated T-lymphocytes, but truncating the last 22 amino acids of the carboxyl tail allows partial cell surface expression in a non-T cell system (43). By co-expressing fluorescent fusion proteins of the COOH-truncated CTLA-4 alleles in one single cell, it was possible to simultaneously measure fluorescence intensity at the cell surface for both alleles and compare it with intracellular fluorescence (Fig. 7C, panels a-f). Ratios of cell surface/intracellular mean fluorescence were quantified in single cells for both CTLA-4 alleles and compared with a paired, two-tailed Student's t test. Mean cell surface/intracellular fluorescence intensity ratios are 9.4 Ϯ 2.2 for YFP-GlyThr 17 and 1.5 Ϯ 0.8 for CFP-GlyAla 17 (p ϭ 5 ϫ 10 Ϫ4 , n ϭ 9 cells, example illustrated in Fig. 7C). The average ratios of cell surface/intracellular fluorescence intensity for YFP-GlyAla 17 and CFP-GlyThr 17 co-transfections were 6.0 Ϯ 2.4 and 10.0 Ϯ 2.8, respectively (p ϭ 0.01) (Fig. 7C). The ratio varied somewhat from cell to cell, accounting for the relatively large mean Ϯ S.E. but, because the comparisons were paired within the same cell, the difference was highly significant statistically. The results presented here were calculated from at least 9 different cells obtained in at least two independent transfections for each allele. Cells were selected for expressing the two colors at  5). We detected a minor product in the CTLAAla 17 digest (lane 5, arrow) migrating at ϳ22 kDa, which represents the incompletely glycosylated precursor minus the COOH tail. Because we recovered considerably less intermediate upon digestion this suggests it is outside the ER membrane. B, quantification of relative densities of the major and minor bands before and after proteinase K digestion. The bar graph shown represents three independent experiments evaluated by a one-tailed Student's t test. C, as an alternative assay of ER membrane integration, translation products were extracted with Na 2 CO 3 (pH 11.5) and ultracentrifuged. ER-integrated proteins were found in the membrane pellet, whereas cytosolic molecules stay in the supernatant. D, quantification of the proportion of incompletely glycosylated intermediate recovered in the pellet, as a fraction of total. Checkered bars represent percentage in the pellet. Statistics as in B, n ϭ 3 experiments. This independently confirms that a substantial portion of the incompletely glycosylated fraction in the Ala allele retrotranslocates to the cytoplasmic phase. roughly equal intensities, and their selection was finalized prior to any knowledge of the quantitative results. By choosing to measure ratios within the same cell for both alleles correction for transfection efficiency differences is inherent as well as correction for any differences in fluorophore efficiency. Again, in this case same-allele co-transfections were performed and similar calculations were done but no statistical significance was found (p Ͼ 0.05). This allowed us to conclude that, for any given level of expression, there is significantly less CTLAG-lyAla 17 expressed at the cell surface of COS-1 cells than CTLAGlyThr 17 .

DISCUSSION
Our data demonstrate that a common amino acid polymorphism in a signal peptide can have measurable consequences on the efficiency of processing the protein and ultimately on its expression level at the cell surface. In the case of CTLA-4, a molecule involved in the inhibitory regulation of the immune response, this mechanism offers an attractive explanation for the higher frequency of diabetes and other autoimmune diseases in individuals homozygous for the Ala allele (19,20,22,23,25,26,33). Signal peptide mutations completely abrogating proper targeting of a protein have been described in Mendelian disorders (51)(52)(53)(54) but, to our knowledge, this is the first demonstration of a subtler effect in the context of susceptibility to a common complex disease. However subtle, such effects are extremely important toward composite molecular prediction and better understanding of the mechanism leading to effective prevention/treatment.
To summarize our findings: by a widely used model of in vitro reconstitution of translation and ER processing, we have shown that both alleles of the signal peptide are capable of translocation to the ER and both are completely and correctly cleaved. The difference appears to lie in the fact that up to one-third of CTLAAla 17 molecules are glycosylated on only one of the two possible sites.
The absence of an effect on translocation or cleavage was not surprising, as the polymorphism does not alter any of the known consensus elements required for translocation or the signal peptidase cleavage site. Somewhat less expected was the effect on glycosylation, as this modification is not ordinarily thought of as depending on the signal peptide. However, evidence is arising that in the time interval between entry of the amino-terminal of the nascent polypeptide into the ER and signal peptidase cleavage, the signal peptide may participate in the process of protein folding and alter its interactions with ER chaperone proteins and modifying enzymes (42,55). Thus, the highly inefficient processing of the human immunodeficiency virus glycoprotein 120 was shown to be due strictly to its signal peptide, as the exact same mature-protein sequence is processed efficiently with a control signal peptide (42). The native human immunodeficiency virus glycoprotein 120 signal sequence is translocation-competent, but affects processing by prolonging glycoprotein 120 association with ER chaperones and results in incorrect folding, which causes ER retention (42).
Moreover, it is now clear that although the oligosaccharyltransferase complex is tethered to the ER membrane, transfer of N-linked high mannose moieties to proteins is not synchronous with translocation to ER, but may follow folding of the protein and is profoundly influenced by it (56). It is known that not all consensus Asn-X-Ser/Thr N-linked glycosylation sites are equally glycosylated or glycosylated at all. A proline in position X (or immediately after Ser/Thr) eliminates glycosylation (47,57), whereas Trp, Asp, Glu, and Leu decrease glycosylation efficiency (57). By introducing glycosylation sequons into different positions of the Saccharomyces carboxypeptidase Y, Holst et al. (56) showed dependence of glycosylation on FIG. 7. CTLAAla 17 and CTLAThr 17 do not co-localize in transfected cells and there is significantly less Ala 17 expressed at the cell surface. We transfected COS-1 cells with CTLA-4 fused to cyan or yellow fluorescent protein (respectively CFP, green, and YFP, red). A, the same allele fused to different color fluorescent proteins was cotransfected into COS-1 cells: YFP-CTLAThr 17 ϩ CFP-CTLAThr 17 (panels a-c) and YFP-CTLAAla 17 ϩ CFP-CTLAAla 17 (panels d-f). We processed the cells 24 -48 h following transfection and analyzed them by confocal microscopy at ϫ63 magnification. All transfections were repeated independently at least 5 times. Complete co-localization of the different colored fusion proteins validates the approach despite inherent differences in the fluorophore intensities. B, panels a and b represent the co-transfection of YFP-CTLAAla 17 ϩ CFP-CTLAThr 17 into COS-1 cells. The two alleles do not co-localize in COS-1 cells (panel c, see arrows, note areas where no red is found and areas where no green is found). Panels d-f are a representative of the reciprocal transfection YFP-CTLAThr 17 ϩ CFP-CTLAAla 17 . C, we performed quantitative fluorescence analysis in COS-1 cells co-transfected with carboxyl-terminal-truncated YFP-CTLAGlyAla 17 ϩ CFP-CTLAGlyThr 17 . Only cells where the cell surface was clearly visible were taken into account, and the reciprocal transfection was likewise analyzed. Measurements were collected by outlining the cell surface and the intracellular fluorescence with a cursor (see panels a-f) and measuring pixel densities. All raw data was processed in NIH Image and the cells shown are a representative of at least two independent transfections and at least 9 different cells.
protein folding and the position of the sequon in the context of the folded protein, results that led them to conclude that glycosylation does not necessarily precede folding, and can be affected by it. In addition, deletion of downstream sequences was shown to abrogate glycosylation of the hepatitis C virus E1 protein (58). Taken together, these observations suggest a mechanism whereby a signal peptide variant may affect the glycosylation efficiency through altered chaperone association and folding.
We performed the single mutant experiments to define which of the two sites remained unglycosylated in the partially processed CTLAAla 17 fraction. Assuming that each mutation did not otherwise change CTLA-4 processing, we had expected that elimination of the specific site that remained unglycosylated in a fraction of CTLAAla 17 would result in a single monoglycosylated band for both alleles, whereas elimination of the site fully glycosylated in both alleles would give a monoglycosylated band in both alleles, plus an additional minor unglycosylated band in CTLAAla 17 . Instead we found a single monoglycosylated band in both alleles, regardless of which site was mutated. The simplest explanation for this is that abolition of the consistently glycosylated site leads to alterations in protein folding and/or interactions with ER chaperones that abolish the glycosylation inefficiency seen with CTLAAla 17 . Alternatively, eliminating one of two glycosylation sites through mutagenesis may relieve competition between them, although a mechanism for such competition is not obvious and no previous paradigm exists.
Our cell surface targeting studies used a truncated CTLA-4 mutant that by-passes the activated lymphocyte-specific regulatory mechanism of translocation from Golgi to the cell surface. We believe that this does not detract from the validity of our conclusions, as passage from Golgi to the cell surface is a post-ER event, occurring after cleavage of the signal peptide and its efficiency should not differ between the Ala 17 and Thr 17 . Given the allelic differences in early processing we found in the in vitro system of isolated ER, the obvious explanation is that less of the Ala 17 makes it to the Golgi, and therefore less is translocated to the cell surface.
Throughout this discussion we have assumed that doubly glycosylated CTLA-4, the major band in both alleles, represents correct ER processing leading to functional expression at the cell surface; it is practically the only form seen with the Thr 17 allele, homozygotes for which are common in the general population and healthy. Extrapolating from our in vitro assay, we propose that Ala homozygotes express one-third less CTLA-4 on the surface of their T-cells than Thr homozygotes, which might tip the balance in favor of immune response and predispose them to autoimmune disease. One might speculate that survival of this allele in evolution despite this disadvantage is because of the counterbalancing effect of better defense against infectious diseases.