Disruption of Ligand Binding to the Insulin-like Growth Factor II/Mannose 6-Phosphate Receptor by Cancer-associated Missense Mutations*

The insulin-like growth factor II/mannose 6-phosphate receptor (IGF2R) carries out multiple regulatory and transport functions, and disruption of IGF2R function has been implicated as a mechanism to increase cell proliferation. Several missense IGF2R mutations have been identified in human cancers, including the following amino acid substitutions occurring in the extracytoplasmic domain of the receptor: Cys-1262 → Ser, Gln-1445 → His, Gly-1449 → Val, Gly-1464 → Glu, and Ile-1572 → Thr. To determine what effects these mutations have on IGF2R function, mutant and wild-type FLAG epitope-tagged IGF2R constructs lacking the transmembrane and cytoplasmic domains were characterized for binding of insulin-like growth factor (IGF)-II and a mannose 6-phosphate-bearing pseudoglycoprotein termed PMP-BSA (where PMP is pentamannose phosphate and BSA is bovine serum albumin). The Ile-1572 → Thr mutation eliminated IGF-II binding while not affecting PMP-BSA binding. Gly-1449 → Val and Cys-1262 → Ser each showed 30–60% decreases in the number of sites available to bind both 125I-IGF-II and125I-PMP-BSA. In addition, the Gln-1445 → His mutant underwent a time-dependent loss of IGF-II binding, but not PMP-BSA binding, that was not observed for wild type. In all, four of the five cancer-associated mutants analyzed demonstrated altered ligand binding, providing further evidence that loss of IGF2R function is characteristic of certain cancers.

The insulin-like growth factor II/mannose 6-phosphate receptor (IGF2R) 1 has evolved in mammals to carry out multiple functions. Distinct regions in its extracytoplasmic domain in-teract with two classes of ligands, namely the mitogenic growth factor, insulin-like growth factor II (IGF-II), and proteins that bear a mannose 6-phosphate (Man-6-P) marker as a result of post-translational modification in the Golgi (1). The IGF2R is a 300-kDa type I transmembrane receptor that is comprised of a 40-residue NH 2 -terminal signal sequence, followed by 15 homologous repeats made up of 124 -192 amino acid residues, a 23-residue transmembrane domain, and a 167-residue cytoplasmic domain (2,3). Functional mapping studies of the extracytoplasmic domain have revealed the location of distinct binding sites for both Man-6-P (4 -6) and IGF-II (7)(8)(9)(10)(11).
In addition to being localized to unique regions of the receptor, the two binding functions are thought to serve different physiologic roles. Along with the cation-dependent mannose 6-phosphate receptor (CD-MPR), the IGF2R carries out lysosomal enzyme targeting through its Man-6-P binding activity (for review see Refs. 1, 12, and 13). This same Man-6-P binding function of the IGF2R has also been shown to be necessary for the activation of transforming growth factor-␤ (14), which bears Man-6-P residues in its secreted, prohormone form (15). After activation, transforming growth factor-␤ exerts effects on cellular proliferation by interacting with its own serine/threonine kinase receptors, usually resulting in growth inhibition (16 -19). In contrast, IGF-II binding to the IGF2R at the cell surface is thought to result in internalization and degradation of the ligand, thereby down-regulating the level of this mitogenic factor (20,21).
The multiple functions of the IGF2R suggest a role for this receptor as a growth inhibitor. In addition, several observations support the hypothesis that Man-6-P/IGF2R acts as a tumor suppressor gene. Microsatellite instability has been observed at the Man-6-P/IGF2R locus in tumors of the gastrointestinal tract (22,23) and endometrium (22). Increased secretion of cathepsin D and other Man-6-P-bearing proteins has been observed in association with cancer of both the prostate and breast (24 -26). Furthermore, loss of heterozygosity (LOH) at the Man-6-P/IGF2R locus has been correlated with poorly differentiated states in early breast carcinomas (27). IGF2R has also been found at decreased levels in hepatocellular carcinomas (28,29), which may be explained by LOH at the Man-6-P/IGF2R locus in these tumor types (30 -32).
Whereas the observation of LOH in tumor samples suggests that loss of receptor function may be involved in the progression to a transformed phenotype in these cancers, another hallmark of a tumor suppressor is the presence of loss-offunction mutations in copies of the gene remaining in the tumor cells. The screening of tumors that exhibit LOH has led to the discovery of several mutations (31)(32)(33), which could serve to strengthen the hypothesis that the IGF2R is a tumor suppressor if these mutations somehow alter normal receptor function. Many of the identified mutations are frameshift or nonsense mutations that would prevent the translation of the complete, mature IGF2R (33,34). However, several missense mutations, many of which are located in the extracytoplasmic domain of the receptor, have also been identified (33,34).
To address the question whether the five cancer-associated missense mutations affect the function of the IGF2R, receptor constructs bearing these mutations in the extracytoplasmic domain were assayed for their ability to interact with both IGF-II and a Man-6-P-bearing protein. The Cys-1262 3 Ser, Gly-1449 3 Val, Gly-1464 3 Glu, and Ile-1572 3 Thr mutations that were observed in hepatocellular carcinoma and the breast cancer-associated Gln-1445 3 His mutation were expressed as soluble receptor constructs. Ligand binding analysis revealed that four of the mutants, all but Gly-1464 3 Glu, altered either Man-6-P binding, IGF-II binding, or both. These alterations in ligand binding to the IGF2R mutants suggest a mechanism for loss of receptor function that is consistent with the Man-6-P/IGF2R as a tumor suppressor. Preparation of Epitope-tagged Soluble Receptor Constructs-A cDNA encoding the human IGF2R (2) was cloned into pCMV5 using the 5Ј SalI site and the 3Ј XbaI site. The 5,157-nt fragment from nt 162 to 5319 of the receptor cDNA was removed by digesting with EagI followed by re-ligation. This smaller insert allowed for the addition of the 24nucleotide FLAG epitope followed by two stop codons using amplification with Vent TM polymerase and two primers. The 5Ј primer contained an XhoI restriction site preceding the sequence corresponding to nt 94 -113 of the receptor cDNA. The 3Ј primer represented sequence complementary to nt 7602-7620 at the carboxyl terminus of repeat 15 in the receptor cDNA followed by 24 nt encoding the FLAG epitope, DYKDDDDK, two stop codons, and an XbaI site. The product was digested with XbaI and XhoI and subcloned into pBKCMV (Invitrogen). This plasmid was then digested with HindIII and XbaI so that the insert could be moved back into pCMV5. Finally, wild-type or mutant EagI fragments were subcloned into the construct, reconstituting a complete FLAG epitope-tagged receptor construct termed 15F. Sitedirected mutagenesis was carried out using the QuikChange TM mutagenesis kit (Stratagene). Pfu-directed thermal cycling was conducted with a fragment of the human IGF2R cDNA encompassing two PflMI sites (nt 3847-6315) in pCRII (Invitrogen). Complementary primer pairs correspond to the following: (a) nt 3921-3952 substituting T to A at nt 3931 creating the Cys-1262 3 Ser (C1262S) mutation and a silent G to A mutation at nt 3939 to create a HindIII site; (b) nt 4470 -4494 containing a G to T mutation at nt 4482 creating the Gln-1445 3 His (Q1445H) mutation; (c) nt 4478 -4522 substituting G to T at nt 4493 creating the Gly-1449 3 Val (G1449V) mutation and containing a silent CCTG to TTTA mutation at nt 4503-4506 incorporating a DraI site; (d) nt 4530 -4557 substituting G to C at nt 4493 creating the Gly-1464 3 Glu (G1464E) mutation; (e) nt 4840 -4874 substituting T to C at nt 4862 creating the Ile-1572 3 Thr (I1572T) mutation. The resultant product was digested with PflMI and subcloned into a shuttle vector containing the IGF2R cDNA. The presence of each mutation was determined by sequence analysis. Finally, the mutant constructs were digested with EagI, and the 5.2-kilobase fragment was subcloned into the pCMV5 vector containing the FLAG-tagged (15F) human receptor construct.

Materials-Recombinant
Expression of the Wild-type and Mutant Receptor Constructs in 293T Cells-Transient expression of the constructs was carried out in 293T human embryonic kidney cells cultured in Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum plus 5 g/ml gentamycin at 37°C in 5% CO 2 . The transfections were carried out by a modification of the calcium phosphate method described previously (36). The major changes to the published protocol were that the cells were grown in the presence of 5 g/ml gentamycin, and the chloroquine shock was not applied. Conditioned medium was prepared by replacing the transfection media with serum-free Dulbecco's modified Eagle's medium on day 3 and was collected on day 6 of the transfection. Freeze/thaw cell lysates were prepared on day 6 by suspending the cells in 0.5 ml of 150 mM NaCl, 10 mM HEPES, pH 7.4, and freezing and thawing the cells 4 times at Ϫ80°C, followed by centrifugation and collection of the supernatant. Finally, Triton X-100 cell extracts were prepared on day 5 or day 6 following transfection with a solution containing 1% Triton X-100, 1 mM MgCl 2 , 10 mM HEPES, pH 7.4, as described earlier (10). Protein concentrations of the crude cell lysates and extracts were determined using the bicinchoninic acid assay (Sigma).
Immunoblot Analysis of Cell Lysates for Quantification of Receptor Construct Expression-Immunoblot analysis was conducted using the M2 anti-FLAG antibody (VWR Scientific, Chicago, IL) on 0.2 mg of cell lysate protein. The cell lysate aliquots were electrophoresed on 6% reducing SDS-PAGE gels and transferred to BA85 nitrocellulose paper. The blots were blocked with 3% nonfat milk in 15 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Tween 20 and probed with the M2 anti-FLAG antibody (1:1000 dilution) followed by a secondary rabbit anti-mouse IgG (Dako). The resultant antibody complex was developed with 125 Iprotein A (NEN Life Science Products) and detected with autoradiography followed by PhosphorImager analysis (Molecular Dynamics) to quantify relative expression of the receptor constructs.
Immunoadsorption of the 15F Receptor Constructs from Cell Lysates-To separate the 15F constructs from other cellular proteins, especially the endogenous 293T IGF2R, the epitope-tagged receptors were routinely immunoadsorbed to M2 resin. Lysate protein (0.2 mg) was incubated with 12 l of packed M2 resin in HEPES-buffered saline (HBS) with 1% bovine serum albumin (BSA) and 5 mM Man-6-P at 3°C for 16 h. The use of Man-6-P was necessary to prevent co-immunoprecipitation of lysosomal enzymes and other Man-6-P-bearing glycoproteins present in the cell extracts. The affinity resin was collected by centrifugation at 14,000 ϫ g for approximately 12 s. The resultant resin pellets were washed twice with 0.75 ml of HBS containing 0.05% Triton X-100 (HBST). This purified resin-bound form of the receptor was then used for further experimentation.
125 I-IGF-II Binding Analysis-The ability of the 15F constructs to bind IGF-II was measured by incubating equal amounts of immunoadsorbed receptor constructs with 2 nM 125 I-IGF-II and 100 nM unlabeled IGF-I for 3-4 h at 3°C in 25 mM HEPES, pH 7.4, 150 mM NaCl, 0.05% Triton X-100, and 0.5% BSA. The addition of IGF-I to the binding reaction prevented interference from IGF-binding proteins that exist in the cell lysate preparations. The resin was washed twice with HBST to remove unbound ligand. Finally, the amount of bound 125 I-IGF-II was determined using a gamma counter. For affinity labeling, cross-linking was done by adding 0.25 mM disuccinimidyl suberate (DSS) to the binding reaction at the end of the 3-4-h binding reaction. After 30 min on ice, cross-linking was terminated with the addition of 0.8 ml of 0.1 M Tris-HCl, pH 7.4, followed by incubation at 3°C for 15 min. The covalent ligand-receptor complex was resolved on a 6% reducing SDS-PAGE gel followed by autoradiography. Competitive binding analyses of receptor constructs to IGF-II were conducted by incubating equal amounts of immunoadsorbed 15F construct in the presence of 2 nM 125 I-IGF-II with increasing amounts of unlabeled IGF-II from 0 to 500 nM. Again, 100 nM unlabeled IGF-I was included in the binding reaction. At the end of a 3-4-h incubation at 3°C, bound ligand was determined by centrifuging the resin, washing, and counting in a gamma counter as described for the IGF-II binding reactions. The data were fit to a model for one-site competitive binding using GraphPad Prism TM software.
Analysis of Mannose 6-Phosphate Binding-Pentamannose phosphate (PMP) was hydrolyzed and purified from a yeast cell wall phosphomannan following the procedure of Murray and Neville (37). The product of the hydrolysis was conjugated to BSA following the procedure of Braulke et al. (38). Briefly, 15 mg/ml BSA was incubated in the presence of 0.2 M PMP and 160 mM NaCNBH 3 at 37°C for 4 -5 days. The resultant product was purified on a 30-ml G-50 Sephadex column in phosphate-buffered saline. The flow-through fractions were collected, pooled, and stored at Ϫ20°C. Aliquots of the protein (25 g) were iodinated to a specific activity of 30 Ci/g by incubation in 0.5 M phosphate buffer, pH 7.4, with 2 mCi of Na 125 I using pre-coated IODO-GEN tubes (Pierce) for 25 min. The product was separated from free iodine on a G-50 column. The iodinated PMP-BSA was collected from the flow-through fractions and stored at Ϫ20°C until use. Binding analysis with this ligand was conducted in much the same way as for IGF-II. Typically, aliquots of immunoadsorbed receptor constructs were incubated at 3°C for 3-4 h in HBS plus 1% BSA in the presence of 1 nM 125 I-PMP-BSA, with or without 5 mM Man-6-P. The resin pellets were then washed with HBST and counted in a gamma counter to determine the amount of binding. Western ligand blotting was performed on cell lysates with 125 I-PMP-BSA following a modified procedure published earlier for 125 I-IGF-II detection of IGF-binding proteins (IGFBPs) (39). Cell lysates (0.2 mg) were electrophoresed on 6% SDS-PAGE and electroblotted to BA85 nitrocellulose. The proteins were renatured, and the blots were blocked with 1% BSA. Affinity for PMP-BSA was detected by probing the blots with 1.5 ϫ 10 6 cpm 125 I-PMP-BSA in 8 ml of blocking solution for 16 -24 h at 3°C. The blots were then washed and exposed to x-ray film. Competitive binding analysis of receptor constructs using PMP-BSA was conducted as described for IGF-II. Briefly, equal amounts of immunoadsorbed receptor construct were incubated with 1 nM 125 I-PMP-BSA in the presence of increasing amounts of unlabeled ligand. The amount of bound ligand was determined, and regressions for single-site competitive binding were calculated.
IGF-II and PMP Affinity Depletion Analysis-Equal amounts of wildtype, C1262S, and G1449V receptor constructs (approximately 5 mg of transfected 293T cell lysates) were immunoadsorbed to 0.5 ml of anti-FLAG M2 resin (Sigma) in the presence of 5 mM Man-6-P for 4 h at 3°C. The resin was then collected and washed 4 times with 1 ml of HBST. The receptor constructs were then eluted with 0.5 ml of 1 mg/ml FLAG peptide in HBS according to the manufacturer's procedure. Aliquots of the purified constructs (200 l) were then serially incubated 2-3 times with a 4:1 ratio (volume of purified construct:resin) of packed IGF-II-Sepharose (11) or PMP-Sepharose (8). In addition, the wild-type construct was incubated with blank resin as a negative control. The amount of unbound receptor at the end of each 3-h incubation was determined by centrifugation and anti-FLAG immunoblot of the supernatant. The percent remaining was calculated by PhosphorImager analysis.

Expression of Wild-type (WT) and Mutant IGF2R Constructs-
To address the question of how the cancer-associated missense mutations, occurring in the extracytoplasmic domain of the IGF2R (Fig. 1), affect the ligand binding functions of the receptor, the mutations were incorporated individually into truncated receptor constructs that contain all 15 repeats of the extracytoplasmic domain of the IGF2R followed by an 8-residue FLAG epitope tag. These constructs and the empty pCMV5 vector, as a control, were transiently expressed in 293T human embryonic kidney cells. Conditioned medium, freeze/thaw lysates, and Triton X-100 cell extracts were analyzed for the presence of the constructs. Surprisingly, none of the WT 15F constructs were secreted into the media but were found at high levels in both the freeze/thaw and Triton X-100 cell extracts (data not shown). Because they contained the highest levels of transfected construct, the Triton X-100 cell extracts were analyzed for relative expression levels by an M2 anti-FLAG immunoblot (Fig. 2) and were used as a source of the constructs in the remaining experiments. The WT 15F IGF2R construct and all of the mutant cDNA constructs were capable of making proteins. The levels of expression for each construct were quantified by PhosphorImager analysis and were found to be nearly equivalent, depending on the transfection. For example, although in Fig. 2 the I1572T mutant appears to be expressed to about half the level as the other constructs, this difference was not apparent in lysates from two other transfections (data not shown).
IGF-II Binding Analysis-Based on the PhosphorImager data, equal amounts of receptor constructs were immunoadsorbed to M2 anti-FLAG resin so that detailed analysis of IGF-II binding could be made by affinity cross-linking. Initially, the immunoadsorbed receptors were incubated in the presence of 125 I-IGF-II, cross-linked with 0.25 mM DSS, and resolved on a 6% SDS-PAGE gel followed by autoradiography (Fig. 3A). Quantification of the amounts of displaceable 125 I-IGF-II present in the 250-kDa cross-linked bands was carried out by PhosphorImager analysis. Both the Q1445H and G1464E mutant receptors showed the same amount of affinity labeling as the WT 15F, whereas the I1572T mutant demonstrated a complete loss of IGF-II affinity labeling under these conditions. The C1262S mutation caused an approximately 90 -95% reduction in the intensity of the receptor/ligand band, and the G1449V mutation diminished the intensity of the cross-linked band by approximately 60%.
To determine if the changes in affinity labeling among the C1262S, G1449V, and I1572T mutant receptors were due to decreased ligand binding, a more direct assay of IGF-II binding was used. The immunoadsorbed receptor constructs were incubated with 2 nM 125 I-IGF-II for 3-4 h at 3°C. The amount of bound ligand was determined by washing the resin pellet and

FIG. 2. Expression of WT and mutant IGF2R constructs in 293T cells.
The control vector (CMV5) and each of the IGF2R constructs were transfected into 293T cells. Cell lysates (0.2 mg of protein) prepared on days 5 or 6 after transfection were resolved by SDS-PAGE, immunoblotted to nitrocellulose, and probed with the M2 anti-FLAG antibody followed by development with 125 I-protein A. A representative autoradiogram is shown.
FIG. 1. Schematic illustration of the IGF2R showing the cancer-associated missense mutations. The overall structure of the human IGF2R is shown, with emphasis on the extracytoplasmic repeats thought to be involved in ligand binding and the position of the cancerassociated missense mutations. The extracytoplasmic repeats are represented by rectangles and are numbered from the amino terminus according to Lobel et al. (3). The arginine residues thought to coordinate interactions with Man-6-P (6) are also indicated. counting in a gamma counter. A representative binding analysis is shown in Fig. 3B. Some variability in IGF-II binding ability was noted among constructs expressed from different transfections. To obtain an average measurement of binding, the C1262S and G1449V receptor constructs were transfected in three independent experiments, and duplicate binding reactions were conducted on each set so that a mean could be calculated. In accordance with the affinity labeling analysis, the Q1445H and G1464E mutations both showed no difference in IGF-II binding when compared with WT. In addition, the I1572T mutant demonstrated a knockout of IGF-II binding, whereas G1449V caused a 36.5 Ϯ 8.2% reduction in binding. The C1262S mutation, which caused a nearly complete oblit-eration of affinity labeling, demonstrated only a 51.5 Ϯ 5.9% abrogation of IGF-II binding, suggesting a change in the IGF-II cross-linking efficiency for this mutated receptor construct.
To determine if these alterations in IGF-II binding caused by C1262S and G1449V were due to changes in either the affinity or the number of available binding sites (B max ), competitive binding analysis was carried out (Fig. 4A). Equal amounts of WT and mutant 15F constructs were immunoadsorbed to M2 resin and incubated with 2 nM 125 I-IGF-II in the presence of increasing concentrations of unlabeled IGF-II. The amount of radiolabeled IGF-II bound at equilibrium was determined for each concentration of unlabeled IGF-II, and IC 50 values were calculated. The WT receptor had an IC 50 of 5.3 nM, whereas C1262S and G1449V showed no significant difference from WT with IC 50 values of 3.4 and 4.7 nM, respectively.
The competitive binding data for IGF-II were also analyzed by Scatchard plot analysis (Fig. 4B). The WT 15F receptor construct demonstrated a K d of 1-5 nM, which is consistent with the IGF-II affinity previously measured for the rat receptor in our laboratory (10). The C1262S and G1449V mutations did not alter the affinity of the receptor construct for IGF-II. However, the C1262S and G1449V mutant constructs demonstrated decreases in B max of 43 and 61% relative to WT, respectively. These data show that a change in the available number of binding sites, not a uniform alteration of affinity, is responsible for the decreased ability of these mutant constructs to bind IGF-II.
Analysis of Man-6-P Binding Function-Two approaches were employed to probe the Man-6-P binding function of the receptor constructs. First, a Western ligand blotting procedure FIG. 3. Analysis of IGF-II affinity cross-linking and binding to the IGF2R constructs. A, IGF-II affinity cross-linking was carried out on cell lysates from 293T cells transfected with vector (CMV5) or with the WT or mutant IGF2R constructs. Equal amounts of each receptor construct, as determined by PhosphorImager analysis of anti-FLAG immunoblots, were immunoadsorbed using anti-FLAG resin and incubated with 2 nM 125 I-IGF-II in the presence (ϩ) or absence (Ϫ) of 1 M unlabeled IGF-II for 3 h at 3°C. Bound ligand was cross-linked to receptor by adding 0.25 mM DSS. The radioactive ligand-receptor complexes were resolved by SDS-PAGE followed by autoradiography. B, direct IGF-II binding measurements were made to complement the affinity labeling experiments. Equal amounts of each receptor construct immunoadsorbed to anti-FLAG resin, or resin exposed to 0.2 mg of pCMV5 mock-transfected 293T cell lysates, were incubated in the presence of 2 nM 125 I-IGF-II. Bound radioligand was determined by centrifuging the resin pellets, washing, and counting in a gamma counter. Radioactivity retained in the presence of 1 M IGF-II was subtracted from each binding reaction to determine specific binding, which has been expressed as a percentage of wild-type binding. Depicted is a representative binding analysis from one out of three sets of transfected cell lysates. Data are means Ϯ range (n ϭ 2).

FIG. 4. Analysis of IGF-II binding isotherms for the WT, C1262S, and G1449V IGF2R constructs.
A, equal amounts of immunoadsorbed WT (q), C1262S (⅜), and G1449V (f) receptor constructs were incubated with 2 nM 125 I-IGF-II in the presence of increasing concentrations of unlabeled IGF-II. Bound radioligand at each concentration of unlabeled IGF-II was determined by centrifuging the resin pellets, washing, and counting in a gamma counter. The data have been plotted as a percentage of WT binding for each concentration of unlabeled IGF-II. The data were fit to a single binding site model using GraphPad Prism TM software, which is represented as a line. B, Scatchard plots were prepared by replotting the data in A as Bound/Free versus Bound. Linear regression analysis showed that in each case, the major change in binding is due to a decrease in B max .
was used with 125 I-PMP-BSA as the probe (Fig. 5A). For samples transfected with 15F, the ligand blot demonstrated two bands corresponding to the higher molecular weight, endogenous 300-kDa IGF2R, and the smaller, 250-kDa 15F transfected construct (arrow, Fig. 5A). The WT, Q1445H, G1464E, and I1572T receptors all bound PMP-BSA in this assay, whereas C1262S demonstrated no detectable PMP-BSA binding. Interestingly, the C1262S mutation also completely inhibited binding to 125 I-IGF-II in the Western ligand blot procedure (data not shown). The G1449V mutant showed variable ability to interact with PMP-BSA in this assay. Independent experiments were conducted using three sets of lysates from cells transiently transfected with G1449V; two of them demonstrated no ability to bind PMP-BSA, and one set showed some binding, although less than wild type. To determine whether these changes in band intensity on the ligand blot were due to decreased Man-6-P binding, a direct binding assay was employed for measurement of the amount of 125 I-PMP-BSA bound to equal amounts of immunoadsorbed receptor constructs (Fig.  5B). The C1262S mutant showed a 61.6 Ϯ 3.6% reduction, and G1449V caused a 26.7 Ϯ 2% decrease in PMP-BSA binding. The I1572T mutant, although demonstrating a complete knockout of IGF-II binding (Fig. 3), was equivalent to WT in its ability to interact with PMP-BSA (Fig. 5B).
Competitive binding analysis was carried out with PMP-BSA in the same way as with IGF-II, comparing the C1262S and G1449V mutants to WT. PMP-BSA bound to the 15F construct with a K d of 1-2 nM, which corresponds to the affinity of a divalent Man-6-P-bearing saccharide for the bovine IGF2R as reported earlier (40). Both the C1262S and G1449V mutants showed no difference in affinity for PMP-BSA when compared with WT. As with the effects on IGF-II binding, the decrease in PMP-BSA binding by these mutants was due to a decrease in B max (Fig. 6).
Characterization of Loss of IGF-II Binding to Q1445H Receptor Constructs during Storage-Whereas the Q1445H mutant IGF2R construct was identical to the WT control insofar as its ability to bind IGF-II or PMP-BSA when first prepared from cells, subsequent direct binding reactions on stored cell lysates revealed a loss of IGF-II binding ability that was not observed for the WT receptor construct (Fig. 7A) or any of the other mutant proteins (data not shown). This loss was specific to the IGF-II binding function, as PMP-BSA binding was unaltered upon extended storage (Fig. 7B). To characterize this phenomenon further, IGF-II binding was assayed on Q1445H lysates frozen at Ϫ80°C for different times (Fig. 7C). These results indicated that the Q1445H mutant underwent a sharp transition in its ability to bind IGF-II after about 10 days at Ϫ80°C. Competitive binding and Scatchard plot analyses, using Q1445H receptor constructs that demonstrated about a 50% reduction in IGF-II binding, revealed that the loss of the IGF-II binding function was due to a change in B max (Fig. 7D). This loss of binding function during storage at Ϫ80°C was specific to the IGF-II binding function of only the Q1445H mutant, as the other mutant constructs demonstrated no difference from the WT over a 2-month period (data not shown).
IGF-II and PMP Affinity Depletion Experiments-One possible explanation for the decrease in B max for binding IGF-II or PMP-BSA observed for some of the mutant IGF2R constructs is loss of high affinity ligand binding of a subpopulation of the receptors. To test this prediction, the existence of subpopulations of C1262S and G1449V mutant constructs differing in ligand affinity was investigated by chromatography on immobilized IGF-II or PMP. The constructs were first purified on anti-FLAG resin in the presence of 5 mM Man-6-P to remove endogenous phosphomannosylated ligands. They were then subjected to two to three serial exposures to either PMP-Sepharose (Fig. 8, A and B) or IGF-II-Sepharose (Fig. 8, C and D). The amount of unbound receptor construct remaining in solution after each exposure was determined by anti-FLAG immunoblot analysis of the supernatant. Surprisingly, the C1262S and G1449V mutants were able to bind the immobilized ligands to the same extent as WT upon successive exposures to each resin. The C1262S mutant bound to both PMP-and IGF-II-Sepharose resins to the same extent as the WT construct; about 90 -95% bound after only one exposure to either resin. The G1449V mutant, however, demonstrated a slower association with both

FIG. 5. Analysis of PMP-BSA binding by the IGF2R constructs.
A, Western ligand blotting was conducted to reveal affinity for Man-6-P-bearing ligands. Cell lysates (0.2 mg of protein) were transblotted to nitrocellulose following SDS-PAGE. The endogenous receptor and transfected receptor constructs were renatured and probed with 1.5 ϫ 10 6 cpm of 125 I-PMP-BSA. The autoradiogram shows both the higher molecular weight endogenous IGF2Rs and the lower molecular weight 15F IGF2R constructs (arrow). B, PMP-BSA binding was measured by incubating equal amounts of receptor constructs immunoadsorbed to anti-FLAG resin or resin exposed to 0.2 mg of pCMV5-transfected 293T cell lysates with 1 nM 125 I-PMP-BSA in the presence or absence of 5 mM Man-6-P for 3 h at 3°C. Bound ligand was determined in a gamma counter after washing the resin pellets. The data are plotted as a percentage of WT binding (mean Ϯ range, n ϭ 2). The experiment shown is representative of three replicates. resins in comparison to the WT and C1262S constructs, with 52% of the receptor still present after one exposure to PMP-Sepharose and about 26% remaining after one exposure to IGF-II-Sepharose. Whereas the G1449V mutant was less able to bind the immobilized ligands after one round of the depletion when compared with both the WT and C1262S constructs, it demonstrated almost complete binding after 2 exposures to PMP-Sepharose or 3 exposures to IGF-II-Sepharose (Fig. 8, B  and D). No evidence for binding-incompetent subpopulations of either mutant construct could be detected in this assay. DISCUSSION The observation of several missense mutations occurring in the extracytoplasmic domain of the IGF2R in specific tumors raises the question of whether and how these mutations might alter the function of this protein. Sequence analysis of the positions corresponding to the C1262S, Q1445H, G1449V, G1464E, and I1572T mutations in the human, rat, mouse, and bovine receptors revealed that all but G1464E are conserved without exception, suggesting that they are important in the overall structure and function of their respective repeats. In order to determine how these five Man-6-P/IGF2R missense mutations affect the ligand binding functions of the receptor, the mutants have been tested as IGF2R cDNA constructs comprised of the signal sequence and the 15 extracytoplasmic repeats, followed by an 8-residue FLAG epitope tag at the carboxyl terminus. This 15F construct allows isolation and separation of the mutants from the endogenous receptor background found in most cell types, by virtue of its unique epitope tag.
When expressed transiently in 293T cells, the 15F construct was not secreted into the culture medium but was found at high levels in cell lysates prepared by either Triton X-100 extraction or freeze/thaw procedures. This finding was unexpected, as a soluble IGF2R construct bearing a deletion of only the transmembrane domain was found to be secreted in transgenic mice (41). The incorporation of the FLAG epitope may play a role in the retention of the 15F construct, but other truncated FLAGtagged IGF2R constructs were found to be secreted, and replacement of the FLAG tag with a c-Myc epitope did not lead to secretion of the extracytoplasmic domain (data not shown). The observation that the 15F construct could be released into a detergent-free solution by freeze/thaw indicates that it is water-soluble. Retention of the 15F construct within the cell may also indicate the presence of an independent intracellular retention signal in the extracytoplasmic domain. Such a signal has been proposed to exist by Dintzis et al. (42), who found that the IGF2R extracytoplasmic domain was required for a predominantly intracellular localization of epidermal growth factor receptor/IGF2R chimeras. Further experiments are under way to determine the significance and molecular basis of 15F retention within the transfected 293T cells.
Triton X-100 cell lysates were used for the remainder of the current study, because they contained the highest levels of transfected construct. Immunoblotting of these lysates revealed that mRNAs derived from all of the mutant cDNAs were apparently capable of being translated into proteins that were homogeneous by gel electrophoresis, with molecular masses consistent with the predicted 250 kDa. Deletion of the IGF2R transmembrane and cytoplasmic domains does not appear to affect the ligand binding functions of the receptor, as WT 15F binds IGF-II and a Man-6-P-bearing ligand with affinity constants that are consistent with observed values for the intact IGF2R of 3-5 and 1-2 nM, respectively.
To test whether the mutations affect ligand binding to the IGF2R, the mutant receptor constructs were analyzed for the ability to interact with IGF-II and PMP-BSA, a Man-6-P-bearing ligand. It has been shown previously that the two Man-6-P-binding sites localize to repeats 1-3 and 7-9 of the extracytoplasmic domain (5,6). In particular, Arg-435 and Arg-1334 in repeats 3 and 9 of the bovine receptor are thought to be involved in coordinating interactions with the phosphate groups of Man-6-P-bearing proteins (6,43). Previous studies have shown repeats 11 and 13 to be important for high affinity IGF-II binding (8 -11). With binding functions mapping to such an extensive portion of the receptor, mutations occurring in the extracytoplasmic domain could disrupt one or both of its binding activities.
When assayed for their ability to bind IGF-II by affinity cross-linking or by direct binding, three of the cancer-associated mutations demonstrated a measurable reduction in ligand-receptor association. C1262S, G1449V, and I1572T all cause decreased binding of IGF-II, suggesting a disruption of FIG. 7. Analysis of ligand binding by Q1445H receptor constructs following storage at ؊80°C. IGF-II (A) and PMP-BSA (B) binding analyses of lysates that were fresh or stored at Ϫ80°C for 8 weeks were conducted as described under "Experimental Procedures." The results are reported as a percentage of WT binding at the times indicated. C, decay of IGF-II binding function was measured over time for both WT (q) and Q1445H (Ⅺ) receptor constructs. Fresh lysates were collected from 293T cells transfected with either the WT or Q1445H 15F cDNA. These samples were stored at Ϫ80°C and analyzed at the indicated times for their ability to bind IGF-II. Results are reported as a percentage of the binding activity detected at time 0. D, IGF-II binding isotherms for 4-week-old WT (q) and Q1445H (Ⅺ) receptors were obtained as described in the text, and the data were represented as Scatchard plots.
the IGF-II binding domain. These mutations are located in repeats 9, 10, and 11, respectively. The hepatocellular carcinoma-associated mutant, I1572T, substituting a polar residue, Thr, for a bulky, hydrophobic residue, Ile, in the heart of the IGF-II binding domain, causes a complete knockout of IGF-II binding. This mutation had previously been shown to knockout IGF-II binding in a truncated receptor construct comprising the first half of repeat 1 fused to repeat 11 (10). It is less obvious how C1262S and G1449V, which reside outside the IGF-II binding domain, are capable of affecting IGF-II binding. Several possibilities could account for this effect. Destabilization of a single repeat may foster interdomain effects. Also, disruption of the conserved disulfide bonding pattern of the repeats may have global effects, which may account for the observed ligand binding disruption caused by the C1262S mutation. This cysteinyl residue is thought to be involved in the conserved disulfide bonding pattern of the 9th repeat (3). The disulfide bond pattern appears to be disrupted to a degree that when electrophoresed on a non-reducing SDS-PAGE gel, the mobility of the C1262S mutant is altered relative to WT (data not shown).
In addition to disrupting IGF-II binding, C1262S causes a decrease in IGF-II cross-linking efficiency when compared with WT 15F. One possible interpretation of these data is that C1262S causes the displacement of a lysyl residue away from radiolabeled IGF-II in the bound state. The homobifunctional cross-linking agent employed in these studies, DSS, utilizes lysyl side chains for its cross-linking chemistry. With a spacer arm length of 11.4 Å, small alterations of the protein backbone could be responsible for decreased cross-linking efficiency. The G1449V mutation results in the substitution of a bulky, hydrophobic residue, Val, for a small, neutral residue, Gly, that would be predicted to have a high degree of conformational freedom (44). This mutation, also located in repeat 10, outside the major IGF-II-or Man-6-P-binding regions of the receptor, is capable of decreasing the binding interactions of the 15F construct. However, in contrast to the C1262S mutation, the G1449V mutation does not seem to affect the efficiency of IGF-II/IGF2R cross-linking.
The C1262S and G1449V mutations also affect interactions with the Man-6-P-bearing ligand, PMP-BSA. C1262S failed to show any detectable binding to PMP-BSA in the Western ligand blot, whereas G1449V showed variable ability to bind PMP-BSA in this assay. The ligand blotting protocol involves denaturing the proteins with subsequent renaturation after non-reducing SDS-PAGE. Because of this property of the assay, not only are inherent binding interactions detected, but the ability of these mutant receptors to renature on the nitrocellulose membrane also contributes to the measured end point. It was necessary to determine if the mutant receptors were capable of binding PMP-BSA in an assay that did not involve a cycle of denaturation-renaturation. Direct binding analysis confirmed that both C1262S and G1449V have reduced PMP-BSA binding when compared with WT. These mutations did not completely eliminate PMP-BSA binding, as the ligand blot would suggest, but only reduced binding, as reminiscent of the IGF-II binding data. Thus, failure to detect PMP-BSA binding to these mutant IGF2Rs in the ligand blot suggests the possibility of a potential defect in in vitro refolding caused by these mutations.
Competitive binding analysis with both IGF-II and PMP-BSA revealed that the C1262S and G1449V mutations reduce binding by decreasing the number of binding sites (B max ) in the population of receptors, while having no apparent effect on the relative affinity toward IGF-II or PMP-BSA. In addition, affinity depletion with either IGF-II or PMP-BSA covalently attached to Sepharose beads demonstrated that nearly all the C1262S and G1449V receptor molecules are eventually capable of interacting with ligand, even though the G1449V mutant showed a decrease in the rate of depletion with both immobi- lized ligands. Several possible explanations may account for this surprising finding. These data could be explained if the mutations induced a conformation of the receptor that is incapable of binding ligand which is in equilibrium with a conformation that can bind. A difference in the rate of conversion between such conformations could account for G1449V causing a decrease in the rate of association with the immobilized ligands in the context of the experimental time frame. Alternatively, the observed decrease in the rate of ligand association for the G1449V mutant may be compensated by a reduced rate of dissociation, resulting in little or no effect on the equilibrium dissociation constant. Such changes in the binding characteristics of this mutant could have dramatic effects on the ability of tumor cells to bind, internalize, and degrade IGF-II through the IGF2R pathway.
The existence of IGF2R dimers could also explain the discrepancy between the change in B max and the ability of the C1262S and G1449V mutants to interact with immobilized ligand. Previous reports have suggested that the IGF2R exists as a monomer in solution (45), but cross-linking studies of IGF2R molecules in cell membranes imply that the receptor exists in multimeric forms (46). While this manuscript was being prepared for publication, York et al. (47) reported that bivalent ligands are capable of increasing the rate of IGF-II internalization, which was attributed to their ability to crosslink IGF2R molecules, suggesting that multimeric forms of the IGF2R form in the presence of bivalent Man-6-P ligands. If the soluble IGF2R constructs exist as multimers, heterodimers formed between receptors that can bind ligand and those that cannot would interfere with separation of receptor species by affinity chromatography. Finally, it should also be noted that the presence of a bipartite Man-6-P binding domain may complicate the interpretation of the Man-6-P binding analysis. The two Man-6-P-binding sites may be functionally distinct, as recent studies have suggested (43). Thus, an alternative explanation for our data may be that the decreased PMP-BSA binding observed for C1262S and G1449V is due to a loss of function of only a single site.
The Q1445H mutation may provide a tool for further investigation of the possible mechanisms for B max effects on the IGF-II binding domain. This mutant construct is capable of binding and cross-linking IGF-II to the same extent as WT when freshly extracted from cells. However, over time in storage at Ϫ80°C, the IGF-II binding function is selectively lost. Thus, there seems to be some instability of repeats 11-13 comprising the IGF-II binding domain. Other analyses, such as rates of denaturation at 37 or 47°C and the pH dependence of ligand dissociation, showed little change relative to WT (data not shown). Regardless of the cause, competitive binding analysis of the mutant Q1445H construct that showed a 50% loss of IGF-II binding was due to a decreased number of detectable binding sites and not an affinity change. Because the Q1445H mutation does not affect the PMP-BSA binding profile of the receptor construct, it is likely that the B max change observed for this mutation is due to a localized disruption or distortion of the IGF-II binding region. Finally, it is unclear whether the instability of the IGF-II binding domain is manifested in a functional change in the Q1445H mutant receptor in vivo. Fulllength versions of the Q1445H mutant IGF2R in 293T cells have recently been expressed, and preliminary studies have revealed similar instability upon storage of plasma membrane preparations bearing the Q1445H mutant receptor. 2 Experiments to measure the thermal stability and half-life of the mutant receptors in living cells are planned.
It is interesting to note the location of each of the mutations examined in this study relative to the conserved cysteinyl residues within each repeat. Most of the repeats contain 8 conserved cysteinyl residues, and based on the patterns of conservation in the bovine receptor, these have been predicted to form disulfide bonds in the pattern: 1 ϩ 2, 3 ϩ 4, 5 ϩ 7, and 6 ϩ 8 (3). The CD-MPR is a 46-kDa membrane protein that shares sequence homology to each repeating unit of the IGF2R (3,48). Recent analysis of the crystal structure of the CD-MPR revealed that the extracytoplasmic domain is made up of ␤-strands that comprise two ␤-sheets (49). By assuming that the repeating units in the extracytoplasmic domain of IGF2R are similar, then each of the extracytoplasmic repeats is comprised of these two ␤-sheet half-domains connected by a coil between the 4th and 5th cysteinyl residues. The I1572T mutation occurs near this putative linking region in the 4th ␤-strand. It is possible that this mutation alters the folded conformation of the repeat so that the final compact structure is not formed. Substituting a polar residue in place of the Ile at position 1572 could interfere with the van der Waals interactions between hydrophobic residues that are proposed to hold the two ␤-sheets together. Based on the homology to the CD-MPR, the substitution of Val for Gly at position 1449, which decreases binding to both IGF-II and PMP-BSA, may limit the conformational flexibility of the peptide backbone in a turn between the 5th and 6th ␤-strands. The Q1445H mutation also occurs near this turn within the 5th ␤-strand. Interestingly, the G1464E mutation, which shows no measurable change in its ligand binding characteristics, occurs just after the 5th conserved cysteinyl residue. The disulfide bond that occurs at this position may stabilize the overall structure of the 10th repeat containing the G1464E mutation.
If the observed mutations are capable of disrupting the conformation of the repeat in which they occur, the question still remains as to how disruption of a single repeat can obliterate functions that occur in other parts of the IGF2R. It has been postulated that each of the extracytoplasmic repeats is capable of folding independently of the others based on the analogy to the CD-MPR. This hypothesis has been supported by the fact that truncated receptor constructs are capable of binding IGF-II and Man-6-P ligands, suggesting that the two binding functions act independently of each other (6,7,10,11). In addition, reciprocal inhibition of Man-6-P and IGF-II binding has been observed for the IGF2R suggesting some interaction between the binding sites for these two classes of ligands (50 -52). The observation that the C1262S and G1449V mutations, which reside outside of the minimal IGF-II and Man-6-P binding domains, reduce ligand interactions implies that the domains do not act independently of each other.
In summary, we have demonstrated that four of five mutations found in association with LOH in human cancers have altered ligand-binding properties. Both the C1262S and G1449V mutations decreased the IGF-II and Man-6-P binding functions of the IGF2R. The I1572T mutation caused a complete loss of detectable binding to IGF-II, while leaving the Man-6-P binding function intact. Unlike these mutations, the Q1445H mutation caused a loss of IGF-II binding capability but only on extended storage at low temperature. These decreases in ligand interaction occurred as a result of changes in the number of binding sites without changing the apparent affinity of ligand-receptor complex formation. Overall, the observation that these cancer-associated mutations affect the normal function of the IGF2R supports the hypothesis that this receptor is involved in the progression of tumorigenesis. providing 293T cells. We also appreciate helpful discussion with and suggestions from Beverly S. Schaffer, Dr. Jung H. Park, Dr. Robert E. Lewis, Dr. C. Kirk Phares, and Dr. Myron L. Toews. We thank Margaret H. Niedenthal of Lilly Research Laboratories for providing the IGFs, and Drs. William S. Sly and David W. Russell for providing the human IGF2R cDNA and pCMV5, respectively. DNA sequencing costs were subsidized by NCI Core Grant CA36727 from the National Institutes of Health and the Nebraska Research Initiative.