Characterization of the Met326Ile variant of phosphatidylinositol 3-kinase p85α

Abstract

Phosphatidylinositol 3-kinase is a key step in the metabolic actions of insulin. Two amino acid substitutions have been identified in the gene for the regulatory subunit of human p85α, Met-326Ile, and Asn-330Asp, and the former has been associated with alterations in glucose/insulin homeostasis. When the four human p85α proteins were expressed in yeast, a 27% decrease occurred in the level of protein expression of p85α^Ile/Asp (P = 0.03) and a 43% decrease in p85α^Ile/Asn (P = 0.08) as compared with p85α^Met/Asp. Both p85α^Ile/Asp and p85α^Ile/Asn also exhibited increased binding to phospho-insulin receptor substrate-1 by 41% and 83%, respectively (P < 0.001), as compared with p85α^Met/Asp. The expression of p85α^Ile was also slightly decreased and the binding to insulin receptor substrate-1 slightly increased in brown preadipocytes derived from p85α knockout mice. Both p85α^Met and p85α^Ile had similar effects on AKT activity and were able to reconstitute differentiation of the preadipocytes, although the triglyceride concentration in fully differentiated adipocytes and insulin-stimulated 2-deoxyglucose uptake were slightly lower than in adipocytes expressing p85α^Met. Thus, the Met-326Ile variant of p85α is functional for intracellular signaling and adipocyte differentiation but has small alterations in protein expression and activity that could play a role in modifying insulin action.

A major pathway for the metabolic effects of insulin and other growth factors is the phosphatidylinositol 3-kinase (PI3-kinase). This enzyme plays important roles in the stimulation of glucose transport, p70 S6 kinase, glycogen synthesis, and lipolysis (1). The PI3-kinases are heterodimeric enzymes composed of a regulatory subunit (p85) and a catalytic subunit (p110α or p110β). Recent work has demonstrated that the regulatory subunit of PI3-kinase exists in several isoforms derived from the p85α gene, including two short forms termed p50α and p55α/AS53 (Fig. 1), as well as subunits encoded by the p85β and p55^PIK/p55γ genes (1–3). The SH2 domains of the PI3-kinase regulatory subunits bind to phosphotyrosine residues in specific sequence motifs possessing the sequence YMXM or YXXM, in all four insulin receptor substrates (IRS-1, -2, -3, and -4) (4, 5).

Figure 1.

Schematic alignment of the structural features of p85α, AS53 (or p55α), p50α and p55^PIK (or p55γ). AS53 and p50α are 100% homologous to p85α except for the N-terminal 34 and 6 amino acids (AA) which are unique. P55^PIK is the product of a separate gene and is ≈70% identical with p85α at the nucleotide level in the two SH2 domains and the p110-binding region and 44% identical in the N-terminal 34 amino acids (3). The methionine at codon 326 in p85α is conserved among species (human, bovine, rat, and mouse; ref. 12) suggesting a potentially important role of this residue for the function of the protein. A valine is present in human p55^PIK at the corresponding position to Met-326. The numbers below p85α indicate the amino acid positions.

The gene encoding p85α is an obvious candidate gene for the development of diabetes, because of its key role in insulin signaling and studies showing a decrease in IRS-1-associated PI3-kinase in tissues of the type 2 diabetic (6–8). A nucleotide substitution has been detected which predicts a change of methionine to isoleucine at codon 326 (9). This change occurs 6 amino acids from the N-terminal SH2 domain and is present in p85α, p55α/AS53, and p50α (3) (Fig. 1). A study in Danish Caucasian subjects has shown that, although the Met-326Ile variant is frequent (allelic frequency ≈15–16%) and occurs with similar frequencies in diabetic and healthy subjects, homozygous carriers for this variant (≈2% in this population) are characterized by reduction in whole-body glucose effectiveness and decreased rates of clearance of an i.v. glucose load as compared with wild-type and heterozygous carriers (9). The allelic frequency of this polymorphism is similar in both Japanese and Swedes, but no clinical characteristics seem to distinguish the genotypes in these populations (10, 11). On the other hand, a study of Pima Indians has shown that the prevalence of type 2 diabetes is only 49% in women homozygous for the isoleucine allele, whereas the prevalence of diabetes in women who are heterozygous or wild-type carriers is between 71 and 72%, a significant difference (12). Furthermore, the nondiabetic homozygous women had significantly higher acute insulin response and significantly decreased 2-h postload plasma glucose levels when compared with women with a heterozygous or wild-type genotype.

The cDNA libraries that were used for the original cloning of both p85α and p55α/AS53 contained DNA with an asparagine residue present at codon 330 (according to the p85α sequence) (1, 13). In all examined Pima Indian and Danish subjects, an aspartic acid residue is present in p85α at that position and thus seems to be the major allele sequence, at least in these populations (12) (K.A. and T.H., unpublished observation). This substitution in codon 330 has not been associated with diabetes or alterations in glucose homeostasis, although it lies even closer to the SH2 domain than codon 326 and this position could be functionally significant.

The purpose of this study was to express the various forms of the p85α subunit in the yeast two-hybrid system, and mammalian preadipocytes, to determine whether the expression or interaction of p85α with IRS-1 was modified and whether the differentiated adipocytes exhibited alterations in triglyceride concentration, glucose uptake, and AKT/protein kinase B activity because of the variant proteins.

Research Design and Methods

Reagents.

The yeast strain L40 was provided by A. Vojtek (University of Michigan, Ann Arbor) and the pACTII (carrying the Leu⁺ selection marker) by S. Elledge (Baylor University, Houston). The pVJL9 3H yeast expression plasmid (containing pLex and carrying the Trp⁺ marker) was from J. Camonis (INSERM U528, Institut Curie, Paris) (14). The T7 sequencing kit was from Amersham Pharmacia, and the Quick Change Site Directed Mutagenesis kit was from Stratagene. Yeast media were from Bio 101, restriction endonucleases from New England BioLabs. Anti-p85 polyclonal antibody, AKT antibody, and Crosstide were purchased from Upstate Biotechnology. Zeocin was from Invitrogen, and insulin from Boehringer Mannheim.

Plasmid Construction.

The three human cDNAs (p85^{Met-326/Asp-330}, p85^{Met-326/Asn-330}, and p85^{Ile-326/Asn-330}) were prepared by site-directed mutagenesis by using the p85^{Ile-326/Asp-330} as template; the nucleotide substitutions were confirmed by sequencing. The constructs were subcloned in-frame with the DNA-binding domain of lexA into the EcoRI and the SalI sites of the pVJL-HIR vector (15) with the coding sequence of the insulin receptor cytoplasmic domain (amino acids 944-1343) downstream of the repressible promoter MET25 and in-frame with the HA epitope tag and a nuclear localization signal. Repression of the insulin receptor was obtained by adding 5 mM L-methionine to medium and plates. The human IRS-1 plasmid was subcloned into the pACTII vector comprising the GAL4 activation domain (GAD) (15). For studies in brown preadipocytes, the human p85α cDNA constructs (Met/Asp and Ile/Asp) were subcloned into the EcoRI and the SalI sites of the pBabeBleomycin vector.

Transformation of Yeast.

Yeast were grown for ≈16 h at 30°C in YPD medium containing 2% glucose, diluted 10-fold, then grown for another 3 h to a density of 2–6 × 10⁷ cells/ml. The yeast were cotransformed by using the lithium acetate method (15), with the pACTII and pVJL-HIR plasmids expressing the hybrid proteins of interest and grown on dropout-agar plates lacking leucine and tryptophan, with or without 5 mM l-methionine.

Reporter Gene Activity.

After 48 h, the colonies were screened for the blue color characteristic of β-galactosidase expression on a filter by using 5-bromo-4-chloro-3-indolyl-β-galactopyranoside. Quantitation of β-galactosidase activity was performed by using a solution-based assay (15). Results were expressed in Miller's units [one unit = (OD₄₂₀ × 1000)]/[OD₆₀₀ × volume (ml) × time (min)].

Protein Expression in Yeast.

Yeast were collected by centrifugation and lysed in 0.25 M NaOH/1% β-mercaptoethanol on ice. The proteins were precipitated with ice-cold 50% trichloroacetic acid, collected by centrifugation, washed in cold acetone, dried, and resolved on SDS-PAGE. The gels were transferred to Immobilon poly(vinylidene difluoride) membranes, blocked, probed with anti-PI3-kinase antibody, and visualized with 2 μCi of ¹²⁵I-protein A with a Molecular Dynamics PhosphorImager.

Infection of Brown Preadipocytes.

ΦNX packaging cells were transfected by calcium phosphate coprecipitation with 15 μg of cDNA overnight. Media were removed 72–96 h after transfection, and the harvested pBabe virus was frozen until use. Preadipocytes derived from p85α knockout mice (16) were grown to ≈70% confluence in 12-well plates in DMEM-high supplemented with 10% FBS. Infection was performed overnight by removing the media and adding 1 ml of virus and 4 μg of polybrene. The cells were spilt into 15-cm plates, and selection media containing 250 μg/ml Zeocin were added.

Protein Expression, Phosphorylation, and Association with IRS-1 in Preadipocytes.

Cells from confluent 10-cm dishes were lysed for 30 min at 4°C followed by centrifugation. The supernatant (100 μg protein) was boiled in Laemmli sample buffer, separated by SDS-PAGE, and blotted with anti-p85 antibody. For phosphorylation, confluent cells were serum-starved for 16 h, then either lysed directly or stimulated for 5 min with 100 nM insulin. Equal amounts of lysate (≈1 mg) were incubated with anti-p85 antibody (8 μg/ml) for 120 min. Immune complexes were collected with a 50% slurry of protein A-Sepharose, washed twice in lysis buffer, resolved by SDS-PAGE, and blotted with IRS-1 antibody.

Determination of Triglyceride Concentration.

Confluent preadipocytes were induced for 48 h in DMEM-high with 10% FBS supplemented with 20 nM insulin, 1 nM T3, 0.5 mM isobutylmethylxanthine, 0.5 μM dexamethasone, and 0.125 mM indomethacin. The differentiation was stopped 1, 3, and 6 days after induction by freezing the plates in liquid nitrogen. The cells were homogenized in 1 M KOH by rotating for 4 h at room temperature. Triglyceride concentration was measured by using the Trygliceride kit from Sigma.

Glucose Uptake and AKT Kinase Assays.

The assays were performed on differentiated cells as described (17, 18). For glucose uptake, cells were starved for 4 h and treated with insulin for 20 min with 2-deoxy[³H]glucose added during the last 5 min. For AKT kinase activity, cells were starved for 18 h and then stimulated for 10 min with insulin.

Results

Analysis of p85α Variants in the Yeast Two-Hybrid System.

The interaction of IRS-1 with p85α was assayed in yeast as described (14). The interaction was specific, and the vectors themselves did not activate expression of the reporter genes. Thus, when yeast was cotransformed with pACTII/IRS-1 with or without pVJL-HIR/p110 and pVJL-HIR/p85α, respectively, 5-bromo-4-chloro-3-indolyl-β-galactopyranoside showed blue colonies only on plates expressing IRS-1 and the p85α proteins (results not shown).

Because expression level affects the ability of proteins to interact in the yeast two-hybrid system, we first determined whether the amino acid substitutions had any impact on the expression levels of p85α protein by subjecting protein extracts of yeast expressing the normal and variant forms of p85 to SDS-PAGE and immunoblotting with p85 antibody. The levels of expression were similar for p85α with methionine at codon 326 and either aspartic acid or asparagine at codon 330 (p85α^Met/Asp and p85α^Met/Asn, respectively) (Fig. 2A). By contrast, cells expressing the variant with isoleucine at codon 326 and aspartic acid at codon 330 (p85α^Ile/Asp) revealed a 27% decrease in protein expression as compared with the p85α^Met/Asp expression (P = 0.03) (Fig. 2A). Likewise, protein expression of p85α^Ile/Asn was reduced by 43% (P = 0.08) as compared with p85α^Met/Asp (Fig. 2A).

Figure 2.

(A) Expression of human p85α in yeast. The yeast strain L40 was cotransformed with pACTII/IRS-1 and pVJL-HIR/p85α encoding the indicated combination of amino acids at codons 326 and 330. Data represent the average protein levels of p85α from four independent experiments, and error bars indicate the SE (Met/Asp 100%, Ile/Asp 73 ± 9%, Met/Asn 100 ± 12%, Ile/Asn 57 ± 16%). The bar indicated by * is significantly decreased as compared with p85α^Met/Asp (P = 0.03). (Met = M, Ile = I, Asp = D, and Asn = N). (B) Interaction between p85α and IRS-1 in the yeast two-hybrid system. The β-galactosidase assay was performed in the presence (black) [no expression of insulin receptor β-subunit (IRβ)] or absence (white) (expression of IRβ) of L-methionine. The bars represent the β-galactosidase activity corrected for protein expression levels. Data represent 12 independent experiments, each in triplicate, and are expressed as percentage of p85α^Met/Asp (activity during IRβ expression: Met/Asp 100%, Ile/Asp 141 ± 4%, Met/Asn 107 ± 4%, Ile/Asn 183 ± 8%). The values are means ± SE. Bars indicated by * are significantly increased as compared with p85α^Met/Asp (P < 0.001). (C) Expression of human p85α in brown preadipocytes. Brown preadipocytes derived from p85α knockout mice were transfected with human p85α cDNA encoding the indicated combination of amino acids at codons 326 and 330. Data represent the average protein levels of p85α from three experiments performed in duplicate on four independent clones expressing p85^Met/Asp and p85^Ile/Asp, respectively. Error bars indicate the SE (Met/Asp 100%, Ile/Asp 86 ± 10%). (D) Interaction with IRS-1 in brown preadipocytes. Bars represent the interaction with IRS-1 after adjusting for protein expression at basal (black) or after insulin stimulation (white) (Insulin stimulated interaction: Met/Asp 100%, Ile/Asp 114 ± 13%). The experiments were performed three times in duplicate on four independent clones from each cell line (Met/Asp and Ile/Asp). Error bars indicate the SE.

To assess the impact of the amino acid variants on the interaction with IRS-1, we performed a quantitative β-galactosidase assay in the presence and absence of IRβ to stimulate tyrosine phosphorylation of IRS-1. All isoforms of p85α had similar levels of basal interaction with IRS-1 in yeast, and the expression of the active fragment of the IRβ subunit increased the p85α association with IRS-1 by 5- to 6-fold (data not shown). When the β-galactosidase activities were adjusted for the differences in the level of expression, significantly greater interactions occurred between IRS-1 and p85α^Ile/Asp (P < 0.001) or IRS-1 and p85α^Ile/Asn (P < 0.001) (increased by 41 and 83%, respectively) as compared with the interaction of IRS-1 with p85α^Met/Asp (Fig. 2B).

Expression of p85α Variants and Binding to IRS-1 in Brown Preadipocytes.

To assess the expression and function of the p85α variants in a mammalian cell, we used brown preadipocytes derived from p85α knockout mice, because these cells contain no endogenous p85α gene product and could be infected with virus-containing human p85α^Met/Asp and p85α^Ile/Asp. Even though our experiments in yeast cells showed evidence of interaction between 326^Ile and 330^Asn where the biggest changes for both p85α expression and interaction with IRS-1 were observed, we decided to focus our analysis on the p85α330^Asp in the brown adipocytes experiments because this variant seems to represent by far the major allele in humans. In four independent cell lines expressing p85α^Met/Asp and p85α^Ile/Asp all cells containing p85α^Ile had a lower level of expression as compared with cells expressing p85α^Met. Although the effect was reproducible, the average reduction in expression of p85α^Ile as compared with cells expressing p85α^Met in these cells was only 14%, and thus was not statistically significant (Fig. 2C).

To determine whether the variations in p85α affected the binding to IRS-1, we stimulated the preadipocytes with insulin, immunoprecipitated cell extracts with anti-p85 antibody, resolved the proteins by SDS-PAGE, and immunoblotted with IRS-1 antibody. Quantitation of protein bands revealed that the total binding of IRS-1 to both forms of p85α protein in the basal and insulin-stimulated states was similar (data not shown). As in the yeast cells, when corrected for the decrease in expression of p85α^Ile/Asp, the binding between IRS-1 and p85α^Ile/Asp was increased slightly (14%) but not significantly, as compared with binding of IRS-1 to p85α^Met (Fig. 2D).

Effect of p85α Variants on Biological Activities in Adipocytes.

PI3-kinase plays an important role in adipocyte differentiation and glucose uptake. P85α knockout cells showed reduced differentiation as measured by decreased triglyceride accumulation (Fig. 3). Knockout cells expressing p85α^Met/Asp and p85α^Ile/Asp showed reconstitution of differentiation as measured by triglyceride accumulation; however, the level of triglyceride in p85α^Ile/Asp was only 73% of the level in wild-type and p85α^Met/Asp cells (Fig. 3).

Figure 3.

Triglyceride concentration during differentiation. The bars represent triglyceride concentrations measured 1 (gray), 3 (white), and 6 (black) days after induction of the cells. [Triglyceride concentration (μg triglyceride/mg protein) at day 6: WT 163 ± 28, KO 68 ± 16, Met/Asp 163 ± 34, Ile/Asp 119 ± 19).] The differentiation was performed three times and error bars indicate the SE.

Basal and insulin-stimulated 2-deoxyglucose uptake was assessed in differentiated wild-type, p85α knockout, p85α^Met/Asp, and p85α^Ile/Asp cells (Fig. 4). As compared with wild-type cells, the knockout cell line exhibited an 89% decrease in insulin-stimulated glucose uptake (Fig. 4). Adipocytes expressing p85α^Met/Asp and p85α^Ile/Asp were both able to reconstitute glucose uptake to a level of about 50% of the glucose uptake in wild-type cells, with a somewhat better response at low insulin concentrations with the 326^Met cells. Both p85α^Met/Asp and p85α^Ile/Asp were able to reconstitute insulin-stimulated AKT kinase activity to an extent similar to that observed in wild-type cells (results not shown).

Figure 4.

Insulin-stimulated glucose uptake. Differentiated brown adipocytes were serum starved for 16–18 h; insulin was added at the indicated concentrations. Data are expressed as fold glucose uptake from the basal glucose uptake. The graphs present four independent experiments, and error bars indicate the SE.

Discussion

PI3-kinase is central to most of the metabolic actions of insulin (19). The major form of the regulatory subunit of PI3-kinase is p85α. In this study, we have evaluated the expression and function of two sequence polymorphisms that have been observed in the p85α gene of humans (Met-326Ile and Asn-330Asp). We found that in two very different cell types, yeast and brown preadipocytes, p85α^Ile is associated with a lower level of cellular p85α protein. Whether the reduced protein level is caused by instability and increased degradation of the protein or the mRNA or decreased transcription rates will be the subject of future studies. We have shown that mice with heterozygous knockout of the p85α gene and with a 50% decrease in the p85α protein show a significant increase in insulin sensitivity and are protected from development of diabetes induced by other insulin-resistant genes (20). Although the difference in expression level in the fat cells in culture is only modest, it would be of interest to determine whether the difference in expression level in humans is more or less pronounced in other insulin-responsive cell types such as liver and muscle (21, 22).

In addition to the change in level of expression, p85α^Ile has a significantly increased interaction with IRS-1 that seems to compensate for the decrease in expression of the protein, such that no change occurs in the total amount of p85α bound to IRS-1 after tyrosine phosphorylation, which again is similar to the effects observed in the p85α heterozygous knockout mice and is caused by the delicate balance among p85α, p110, and IRS proteins (20). The net effect is that the decreased expression of p85α^Ile is largely cancelled out by an increased binding to IRS-1. The changes in protein expression and IRS-1 binding were less visible in the brown preadipocytes, which may be the result of a better compensation in the mammalian cells than in yeast.

These studies of the effect of the p85α variants with respect to insulin-stimulated IRS-1 binding and AKT activity are in agreement with studies of transiently transfected HEK293 cells in which no alterations of the codon 326 variant on PI3-kinase activity are found (23) and equal inhibition of AKT activity by both Met-326 and Ile-326 (11). However, these results need to be viewed with caution, because the stoichiometry of p85α as compared with the IRS proteins in transient transfection may be significantly different from the stoichiometry in normal or stably transfected cells. Furthermore, the interaction of p85α with other proteins via the proline-rich region, bcr domain, and SH3 domain could alter specific pathways. Recently, it has been shown that the three isoforms of p85α (p85α, p50α and p55α/AS53) modulate the PI3-kinase activity by different mechanisms (18). Although p85α by far is the most prominent isoform in all tissues, including skeletal muscle and liver, examination of the role of this polymorphism in the smaller isoforms of the regulatory subunit could be of value for future studies.

The importance of the presence of p85α during differentiation into adipocytes was also clearly manifested by our experiments. Low concentration of triglyceride occurred in the knockout cell, but by adding either of the human p85α constructs to the cells, we were able to reconstitute differentiation, although the p85α^Met appeared more effective in this regard. Likewise, both variants were able to reconstitute insulin-stimulated glucose uptake to some extent, although the cells expressing the p85α^Met protein tended to have slightly higher glucose uptake at low insulin concentrations.

Taken together, our data suggest that the presence of isoleucine at codon 326 in p85α of PI3-kinase reduces the level of expression of the protein, increases its interaction with IRS-1 and may have slightly reduced efficiency in support of adipocytes differentiation and the insulin-stimulated glucose uptake. Thus, the p85α Met-326Ile variant may have only a minor impact on signaling events, however, when combined with variants in other genes encoding signaling proteins or acquired alterations in protein levels may together have a functional impact on the overall insulin signaling.

Abbreviations

IRS

insulin receptor substrate

PI3-kinase

phosphatidylinositol 3-kinase