The Hemochromatosis Founder Mutation in HLA-H Disrupts β2-Microglobulin Interaction and Cell Surface Expression*

We recently reported the positional cloning of a candidate gene for hereditary hemochromatosis (HH), calledHLA-H, which is a novel member of the major histocompatibility complex class I family. A mutation in this gene, cysteine 282 → tyrosine (C282Y), was found to be present in 83% of HH patient DNAs, while a second variant, histidine 63 → aspartate (H63D), was enriched in patients heterozygous for C282Y. The functional relevance of either mutation has not been described. Co-immunoprecipitation studies of cell lysates from human embryonic kidney cells transfected with wild-type or mutant HLA-H cDNA demonstrate that wild-type HLA-H binds β2-microglobulin and that the C282Y mutation, but not the H63D mutation, completely abrogates this interaction. Immunofluorescence labeling and subcellular fractionations demonstrate that while the wild-type and H63D HLA-H proteins are expressed on the cell surface, the C282Y mutant protein is localized exclusively intracellularly. This report describes the first functional significance of the C282Y mutation by suggesting that an abnormality in protein trafficking and/or cell-surface expression of HLA-H leads to HH disease.

Hereditary hemochromatosis (HH) 1 is an autosomal recessive disorder of iron metabolism and represents one of the most common inherited disorders in individuals of Northern European descent with an estimated carrier frequency between 1 in 8 and 1 in 10 (1, (2). In patients with HH, excessive iron deposition in a variety of organs leads to multi-organ dysfunction. Recently, we reported a mutation in a novel MHC class I-like gene, called HLA-H (3). Eighty-three percent of HH patient DNAs were found to be homozygous for this mutation, which consists of a single base transition of G to A and results in a change of cysteine 282 3 tyrosine (C282Y). Subsequent reports have confirmed the high frequency of this founder mutation in other HH patients (4 -6), providing further support that HLA-H is the primary HH locus. A second missense mutation, histidine 63 3 aspartate (H63D), was also reported that was enriched in heterozygotes with the C282Y mutation (eight of nine cases) (3). The specific role that either of these mutations in HLA-H play in the etiology of HH disease has not been elucidated.
The HLA-H protein is similar to MHC class I family molecules including HLA-A2, nonclassical class I molecules such as HLA-G, and the human neonatal Fc receptor (FcRn). All four of the invariant cysteine residues that form disulfide bridges in the ␣ 2 and ␣ 3 domains of MHC class I family members are present in the HLA-H protein. One of these conserved cysteine residues is altered in the C282Y mutation. The integrity of the conserved disulfide linkages has been suggested to be critical for proper maintenance of the secondary and tertiary structure of the protein allowing interactions with accessory molecules such as ␤ 2 -microglobulin (7). Importantly, the functional significance of an interaction between ␤ 2 -microglobulin and an unknown class I-like molecule in HH disease was suggested by ␤ 2 -microglobulin-deficient mice; these mice display a progressive hepatic iron overload similar to that observed in human HH (8 -10). Other studies have demonstrated that mutation of cysteine 203 in the ␣ 3 domain of the mouse MHC class I family member H-2L d prevented intracellular transport of the molecule from the endoplasmic reticulum to the plasma membrane (11).
As a step toward understanding the role of HLA-H in HH disease, we examined the effects of the C282Y and H63D mutations on HLA-H cellular processing. In this report we demonstrate that wild-type HLA-H binds to ␤ 2 -microglobulin and that the C282Y mutation completely abrogates this interaction and disrupts intracellular protein trafficking. The data provide support for the hypothesis that the C282Y mutation results in intracellular sequestration of the HLA-H protein, which leads to HH disease.

EXPERIMENTAL PROCEDURES
Cloning of HLA-H Wild-type and Mutant cDNAs-The HLA-H cDNA was fused to the FLAG octapeptide sequence (Eastman Kodak Co.) to utilize specific available antibodies for detection of the HLA-H protein.
To isolate the H63D mutant cDNA we started with first strand cDNA from a patient known to have that mutation and utilized the primer pair from above to amplify the desired clone. For the C282Y mutation we utilized a standard PCR mutagenesis approach: two overlapping fragments with the appropriate G to A base change were produced in a first-round PCR reaction, and the fragments gel-purified and combined with the 5Ј and 3Ј primers used above to yield the appropriately mutated cDNA with the FLAG sequence attached to the 3Ј end.
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Polyclonal Antipeptide Antibodies-Three peptides were synthesized (Multiple Peptide Systems) corresponding to amino acid sequences of the HLA-H protein. Two were to the predicted extracellular region of the molecule: peptide EX1 comprising amino acids 164 -177 of the ␣ 2 domain (CPRAWPTKLEWERHK) and peptide EX2 comprising amino acids 246 -260 of the ␣ 3 domain (CKDKQPMDAKEFEPKD). The third, CT1, was from the putative cytoplasmic tail comprising amino acids 326 -343 (CRQGSRGAMGHYVLAERE). For each, an NH 2 -terminal cysteine residue was incorporated to enable directed coupling to keyhole limpet hemocyanin. Rabbit antisera were produced and the resulting polyclonal antisera named utilizing the peptide nomenclature.
Immunoprecipitations and Western Blotting-Cells were lysed in 25 mM Tris-HCl, pH 7.5, 150 mM NaCl (TBS) plus 0.5% Nonidet P-40 and the HLA-H or ␤ 2 -microglobulin proteins immunoprecipitated from 1 mg of total cell protein with 50 g of FLAG M2 monoclonal antibodies (Eastman Kodak Co.) or 6 g of ␤ 2 -microglobulin monoclonal antibody B1G6 (Immunotech) followed by addition of protein G-Sepharose (Pharmacia Biotech Inc.). After washing three times with TBS containing 0.25% Tween 20, the antibody-antigen complexes were dissociated by heating at 100°C for 4 min in standard Laemmli sample buffer and the material separated on a 4 -20% Tris-glycine gradient polyacrylamide gel (Novex). Gels were electroblotted onto PVDF membranes (Novex), incubated with either 2 g/ml FLAG M2 antibody, 25 g/ml polyclonal ␤ 2 -microglobulin antibody (Boehringer Mannheim), or 2 g/ml CT1 HLA-H antibody and the immune complexes detected by ECL (Amersham Corp.) utilizing horseradish peroxidase-linked sheep anti-mouse antibodies or horseradish peroxidase-linked donkey anti-rabbit antibodies as appropriate.
Immunofluorescence Microscopy-Parental and transfected 293 stable cell lines were seeded on rat-tail collagen (Biomedical Technologies)coated glass coverslips and grown overnight in standard medium. Cells were fixed in 3.5% paraformaldehyde, stained with 1:1 mixture of affinity-purified antibodies EX1 and EX2 at 50 g/ml or FLAG-M2 antibodies at 25 g/ml, and immune complexes detected with fluorescein isothiocyanate-conjugated goat anti-rabbit or rabbit anti-mouse secondary antibodies, respectively (Zymed Laboratories Inc.). To detect intracellular antigen, cells were first fixed and then permeabilized with 0.05% saponin in phosphate-buffered saline for 3 min at room temperature prior to exposure to the primary antibodies. For peptide competition experiments, peptides (EX1, EX2, and M2) were incubated at a 100-fold molar excess with their respective antibodies for 1 h at room temperature prior to application to the cover slips.

RESULTS AND DISCUSSION
We first sought to demonstrate an interaction of the HLA-H protein with ␤ 2 -microglobulin and to examine the effects of the C282Y and H63D mutations on that interaction. Human embryonic kidney cells (293 cells) were transfected with vectors containing the wild-type HLA-H cDNA or the cDNA with either the C282Y or H63D mutation. The FLAG octapeptide sequence was fused onto the carboxyl terminus of each, providing a specific tag for detection of the expressed proteins (13). We established individual stable cell lines expressing the three proteins. Immunoprecipitation of cell lysates with monoclonal antibodies directed to the FLAG sequences (M2 antibodies), to precipitate the HLA-H/FLAG fusion protein, followed by Western blotting with ␤ 2 -microglobulin polyclonal antibodies demonstrated a clear interaction between the HLA-H protein and ␤ 2 -microglobulin (Fig. 1A, left panel, Wild type lane). Significantly, ␤ 2 -microglobulin was not detected in immune complexes from cell lines expressing the HLA-H protein with the C282Y mutation (Fig. 1A, left panel). This failure to detect ␤ 2 -microglobulin was not due to lack of HLA-H protein in the mutant cell lines, since stripping the blots and immunodetection with rabbit polyclonal antibodies directed to the COOHterminal 17 amino acids of HLA-H (CT1 antibodies) demonstrated that the amount of HLA-H protein in the three cell lines was similar (Fig. 1A, right panel). The results with the H63D mutant were similar to the wild-type HLA-H protein; ␤ 2 -microglobulin was co-immunoprecipitated along with that mutant protein (Fig. 1A, left panel, compare H63D and Wild type lanes). It is of interest to note that the wild-type or H63D HLA-H proteins detected in the right panel appeared to migrate as a doublet of 49 and 46 kDa in lighter exposures, whereas the C282Y appeared as only a single band of approximately 46 kDa.
The ␤ 2 -microglobulin/HLA-H interaction results were corroborated by performing the inverse experiment in which cell lysates were initially immunoprecipitated with ␤ 2 -microglobulin antibodies followed by Western blotting with antibodies directed toward the COOH-terminal sequence of HLA-H (CT1 or HLA-H containing the C282Y or H63D mutations. 1 mg of each extract was first immunoprecipitated with 50 g of FLAG-M2 antibodies, resolved by gradient SDS-PAGE, and transferred to PVDF membrane. In the left panel, ␤ 2 -microglobulin was detected by incubation with 25 g/ml of a polyclonal ␤ 2 -microglobulin antibody followed by horseradish peroxidase-linked sheep anti-rabbit antibodies and visualized with Amersham ECL reagent. In the right panel, the blot was stripped, reprobed with 2 g/ml polyclonal antibodies specific for the COOH-terminal 17 amino acids of HLA-H (CT1 antibodies) and detected as above. In both panels the doublet band at approximately 30 kDa is due to the horseradish peroxidase-linked secondary; by elimination of the primary antibodies during the Western blotting process, only these bands are detected (data not shown). B, co-immunoprecipitation of HLA-H with ␤ 2 -microglobulin. Cell extracts were prepared as in A and immunoprecipitated with 6 g of ␤ 2 -microglobulin monoclonal antibodies, resolved by gradient, SDS-PAGE, and transferred to PVDF membrane. In the left panel, HLA-H proteins were detected with 2 g/ml HLA-H CT1 antibodies followed by horseradish peroxidaselinked sheep anti-rabbit antibodies and visualized with ECL reagent. In the right panel, the blot was stripped, reprobed with 25 g/ml ␤ 2microglobulin polyclonal antibodies, and detected as above. The background doublet band detected at approximately 30 kDa is due to the horseradish peroxidase-linked secondary antibodies as in A. antibodies). In this experiment, the ␤ 2 -microglobulin antibodies co-immunoprecipitated HLA-H protein from the wild-type and H63D mutant expressor cell lines, but failed to do so in the C282Y mutant expressor cell line (Fig. 1B, left panel). Stripping the blots and reprobing with ␤ 2 -microglobulin antibodies demonstrated that similar amounts of ␤ 2 -microglobulin protein were immunoprecipitated from each cell line (Fig. 1B, right  panel). These results further confirm an interaction between wild-type HLA-H protein and ␤ 2 -microglobulin and demonstrate that the C282Y, but not the H63D, mutation disrupts this association.
Previous reports have suggested that association of the MHC class I heavy chain with ␤ 2 -microglobulin is critical for cellsurface expression (14,15). Because of the failure of the HLA-H protein containing the C282Y mutation to interact with ␤ 2microglobulin, we next investigated whether this mutation would also affect cell-surface presentation of the HLA-H protein. Parental 293 cell lines and those expressing the wild-type HLA-H protein or the C282Y mutant were examined for cellsurface protein expression by immunostaining with rabbit polyclonal antibodies specific to sequences residing in the predicted external domain of the HLA-H protein (EX1 and EX2 antibodies) followed by detection with immunofluorescence. Parental 293 cells displayed no detectable surface labeling by these antibodies (Fig. 2A), consistent with the undetectable levels of HLA-H protein observed in the Western blotting experiments (Fig. 1A, right panel). Wild-type HLA-H-expressing cells demonstrated a distinct pattern of surface labeling as evidenced by a punctate pattern of labeling that was much more intense at the edges of the cells (Fig. 2B). By contrast, cells expressing the C282Y mutation displayed no surface labeling and were indistinguishable from the parental controls (Fig. 2, compare C and A). The specificity of the antibody labeling was demonstrated by preincubating the EX1 and EX2 antibodies with their respective peptides; in these experiments the punctate surface labeling observed in the wild-type HLA-H expressor cells was completely abolished (data not shown).
We examined the possibility that the C282Y mutant protein was expressed in the transfected cells but remained intracellularly localized. Immunostaining was performed following treatment of the cells with saponin to permeablize them. Staining of these cells for the FLAG-tagged C282Y mutant HLA-H protein demonstrated strong perinuclear fluorescence, which was absent in the parental control cells (Fig. 2, compare D and  F). Permeablized wild-type HLA-H protein expressor cells showed similar intracellular staining with the FLAG-M2 antibody, suggesting that not all of the wild-type protein in these transfected cells reaches the cell surface (E). Experiments utilizing the EX1, EX2, or CT1 antibodies yielded the same results (data not shown). These results clearly demonstrate that the C282Y mutation specifically disrupts cell-surface presentation of the HLA-H protein.
To examine the distribution of wild-type and mutant HLA-H proteins within the cell in more detail, we performed subcellular fractionations on stepwise sucrose gradients to separate the various membrane components. Three separate postnuclear membrane fractions were obtained; the 20/35% interface contained the lightest density membranes (L); dense membranes (D) partitioned at the 40/50% interface, whereas the 35/40% medium-density (M) interface contained a mixture of light and dense membrane-derived components. We initially characterized the efficacy of our subcellular membrane fractionation scheme by assaying these fractions for marker proteins. Antibodies to Na ϩ /K ϩ -ATPase were utilized as markers for plasma membrane, ␤-coatomer protein (␤-COP) for Golgi, and calnexin for ER membrane identification. Samples representing membranes from equal numbers of cells from each interface were analyzed by Western blotting and quantitated on a Molecular Dynamics scanning densitometer. Plasma membranes were found primarily in the light-density interface and to a lesser extent in the medium-density layer: L, 90%; M, 10%; D, 0%. Golgi membranes were distributed nearly equally throughout the three interfaces: L, 30%; M, 40%; D, 30%. ER membranes were found mostly in the dense membrane interface: L, 0%; M, 20%; D, 80%. The fractionations from each of the three cell lines gave equivalent results.
We determined the specific distribution of HLA-H proteins in the sucrose gradient interfaces by Western blotting and probing with HLA-H antibodies. As with the co-immunoprecipitation results (Fig. 1), the immunostaining suggested that the wild-type HLA-H protein migrated as a doublet of the 49 and 46 kDa forms (Fig. 3, left panel). The slower migrating 49-kDa form was found principally in those fractions containing plasma membranes, whereas the lower molecular mass 46-kDa form was distributed in a pattern similar to that of the Golgi marker, ␤-COP. By contrast, the C282Y mutant HLA-H protein consisted only of the faster migrating 46-kDa species. Like the wild-type 46-kDa protein, the mutant 46-kDa protein was distributed in a pattern that most closely resembled that of the Golgi marker protein, suggesting the possibility of incomplete posttranslational processing or modification (Fig. 3, middle  panel). The H63D mutant proteins migrated in a pattern resembling that of the wild-type protein, implying that this mutation had little or no effect on intracellular HLA-H protein trafficking (Fig. 3, right panel). In other studies, no HLA-H protein was detected in the cytosolic fractions of any of the wild-type or mutant cell lines, suggesting that neither the C282Y nor the H63D mutation cause a redistribution of the protein to the cytoplasm (data not shown).
Taken together these results demonstrate that the C282Y mutation prevents the HLA-H molecule from interacting with ␤ 2 -microglobulin and eliminates cell-surface presentation. Cysteine 282 is one of four cysteine residues that are invariant in both classical and nonclassical MHC class I molecules and forms a critical disulfide bridge in the ␣ 3 -immunoglobulin-like domain (7). Thus, the integrity of this structure is critical to the formation of the heterodimer of ␤ 2 -microglobulin and HLA-H and also for proper intracellular processing of the protein.
Class I MHC molecules are noncovalently linked heterodimers between an ␣ heavy chain and ␤ 2 -microglobulin (light chain) (7). The role of the ␤ 2 -microglobulin/heavy chain interaction is to facilitate and stabilize the folding of the heavy chain during biosynthesis through interactions with the ␣ 1 -␣ 2 platform and the ␣ 3 domain (16) (7). Previous work demonstrated that mutating the reciprocal cysteine residue (cysteine 203) in mouse H-2L d protein abolished cell-surface presentation (11). Interestingly, the mutant H-2L d molecule retained the ability to associate with ␤ 2 -microglobulin. Mouse H-2L d belongs to the family of classical antigen-presenting molecules, whereas HLA-H is nonpolymorphic and, therefore, resembles nonclassical molecules such as HLA-G or the human Fc receptor (3). It is conceivable that the integrity of the ␣ 3 domain may be less stringent for ␤ 2 -microglobulin association with classical antigen-presenting proteins (such as H-2L d ) than for nonclassical molecules (such as HLA-H).
The biogenesis of MHC class I molecules is well documented. The heavy chain of a MHC class I molecule is synthesized on membrane-bound polysomes, and N-linked glycosylation occurs co-translationally in the ER (17). Subsequently, the modified heavy chain associates with chaperone proteins and ␤ 2microglobulin in the ER and is then transported through the cis-Golgi network, to the middle and trans-Golgi cisternae where the glycosyl side chain is modified to a more complex form en route to the plasma membrane (18 -20). Class I molecules that fail to assemble properly are recycled between the ER and Golgi, rather than being retained exclusively in the ER (21). In our studies the C282Y mutant of HLA-H is retained on intracellular membranes in a pattern that would be consistent with these earlier observations. The mutant protein migrates similar to the Golgi marker protein ␤-COP in subcellular fractionations, but because of the limited resolution of the stepgradient, we cannot rule out that some protein may also be in the ER. The perinuclear pattern of staining noted in the immunofluorescence studies does not definitively resolve this. More detailed studies will be necessary to ascertain the specific point at which intracellular transport of the C282Y mutant is disrupted.
In contrast to our results with the C282Y mutation, we found no detectable changes in the ␤ 2 -microglobulin interaction or intracellular processing of the H63D mutant form of HLA-H, which is enriched in C282Y heterozygous patients (3). Other studies have demonstrated alterations in intracellular transport of class I molecules by mutations in the peptide-binding groove of HLA-A (22). The H63D mutation is localized in the ␣ 1 domain between the third and fourth ␤ strands of the external peptide-binding region. It is possible that the effect of this mutation is to alter the affinity of the HLA-H protein for an as yet unidentified ligand or to alter the manner that the mutant protein interacts with other proteins in the cell membrane. Alternatively, this mutation may represent a common polymorphism with little or no effect on the biological functioning of the protein. The definitive answer to this question will await further investigation as we elucidate how the HLA-H molecule regulates iron metabolism in the body.