Glycophosphatidylinositol-anchored Protein Deficiency as a Marker of Mutator Phenotypes in Cancer¹

Mutations in oncogenes and tumor suppressor genes are integral to carcinogenesis. Because the number of mutations required in multistep carcinogenesis in some cancers exceeds that which can be readily explained by basal mutation rates, several investigators have hypothesized that in such cancers a “mutator phenotype” is acquired early in the process (1, 2). By increasing the mutation rate, the mutator phenotype permits the cell to accumulate the requisite number of mutations for both carcinogenesis and progression.

Some Mut⁴phenotypes can result from defects in either DNA replication or DNA repair, and recently such a Mut phenotype resulting from MMR defects has been described in a subset of colon carcinoma patients. These MMR defects have been found both in familial (hereditary nonpolyposis colorectal cancer) and sporadic colon cancers and result in a staggering 100- to 1000-fold increase in the spontaneous mutation rate (3, 4). It has been shown that the Mut phenotype in these cancers can arise as a result of several different MMR enzyme defects. Among those that have been characterized thus far are hMLH1, hPMS1, hPMS2, and hMLH3 (human homologues of bacterial MutL), as well as hMSH2, hMSH6 (GTBP), and hMSH3(human homologues of bacterial MutS). Defects in these MMR genes cause an increase in coding region mutations in addition to microsatellite instability or replication errors.

In mammalian and all other eukaryotic cells, in lieu of customary transmembrane polypeptides, a number of cell surface proteins are linked to the plasma membrane by posttranslationally added GPI-anchoring units (reviewed in Refs. 5 and6). These nonconventional anchoring structures are preassembled in the endoplasmic reticulum and substituted for COOH-terminal signal sequences in the primary translation products of these proteins during their biosynthesis.

The PIGA gene (7) is an X-chromosomal gene(8) that encodes an element required for the transfer of N-acetylglucosamine to phosphatidylinositol, the first step in GPI-anchor assembly (5, 6). Consequently, a single mutational event involving this gene can give rise to loss of GPI-anchored surface protein expression, a circumstance that occurs in hematopoietic stem cells of patients with the acquired hemolytic anemia, PNH (9). Such loss can be readily detected by flow cytometry using available antibodies to any GPI-anchored protein that is expressed by the cell type of interest.

In this investigation, we used established Mut and control colon cell lines to test the feasibility of exploiting loss of GPI-anchored protein expression resulting from PIGA mutation as a new method for identifying Mut phenotypes in cancer. We examined eight microsatellite instability colon cancer lines [RKO,HCT116, and VACO 5 and 6, each defective in hMLH1; LoVo defective in hMSH2 (10); MT1 affected in GT binding protein, as well as HCT116−M2 (cells complemented with chromosome 3 but containing a destroyed hMLH1 gene);and HCT116+Ch2 (chromosome complemented cells lacking a functional hMLH1 gene)]. We included four non-Mutcolon cancers as controls [SW480, SW837, TK6, and HCT116+Ch3(functionally corrected hMLH1 via chromosome 3 complementation; see Table 1)].

HCT116, SW480, RKO, LoVo, SW837, VACO 5, and VACO 6 cells (kindly provided by Dr. James K. V. Willson, Ireland Cancer Center,Cleveland, OH) were cultured at 37°C with 5%CO₂ in MEM containing 10% FBS (Life Technologies, Inc., Gaithersburg, MD). For MT1 and TK6 cells, RPMI 1640 was used in place of MEM. HCT116+Ch2, HCT−M2, and HCT116+Ch3 were generously provided by Dr. Richard Boland (University of California San Diego, La Jolla, CA) and were cultured in MEM 10% FBS containing 400μg/ml neomycin. RKO was graciously provided by Dr. Michael Brattain (University of Texas Health Science Center, San Antonio,TX). Before analyses, adherent lines (Table 1) were detached with versene.

After washing three times in PBS (pH 7.4) containing 1% BSA/0.1%NaN₃/4 mm EDTA, cells(∼10⁸) were incubated on ice for 15 min in 200μl of the same buffer containing 5 μg/ml of biotinylated YTH53.1 rat anti-CD59 mAb (PharMingen, La Jolla, CA) and 5 μg/ml each of mouse IA10 and IIH6 anti-DAF mAb (11). The washed cells then were secondarily incubated in the same fashion with 5 μg/ml each of streptavidin-PE (PharMingen) and FITC-conjugated sheep antimouse IgG (Sigma, St. Louis, MO) and washed again in the same buffer. The stained cells were negatively sorted for DAF and the negatively sorted population reanalyzed for CD59 in a Beckman Coulter Elite flow cytometer. The negative gate for each cell line was determined from studies in which identical aliquots of the cells were incubated with nonrelevant control mAbs and the same fluorochromes.

To generate newly developing mutants for molecular analysis, 100 CD59⁺DAF⁺ cells were sorted initially into wells of 96-well plates, transferred to flasks, and expanded 10⁶- to 10⁷-fold. Single CD59⁻DAF⁻ cells derived from these purged cells were sorted into wells of 96-well plates, and the cells again were expanded. The expanded cells were reanalyzed by flow cytometry to verify that they were homogeneously DAF⁻CD59⁻.

Total RNA (4 μg) from 10⁶ cells was reverse-transcribed at 37°C for 90 min using primer c(5′-AATGATATAGAGGTAGCATAAC-3′) with 200 units of RNase H-free reverse transcriptase (Superscript, Life Technologies, Inc.) in a final volume of 20 μl. PCR amplification of the PIGA coding region was performed using one tenth of the reverse-transcribed product, primers a(5′-GGTTGCTCTAAGAACTGATGTC-3′) and b (5′-TCTTACAATCTAGGCTTCCTC-3′), and 30 cycles of incubations for 1 min at 94°C, 1 min at 55°C, and 2 min at 72°C. Amplification products were phosphorylated, ligated into pGEM3Z vector, and sequenced.

Cells (1 × 10⁸) in 100 μl of complete RPMI were incubated at 37°C for 30 min with 100 μl of a 1:10 dilution of rabbit anti-DAF antiserum (12) in complete RPMI. The sensitized cells then were mixed with 800 μl of a 1:4 dilution of rabbit serum in complete RPMI and incubated at 37°C for 60 min. After centrifugation, the resuspended pellet was spun through Ficoll-Paque (Pharmacia Inc., Uppsala, Sweden), and interface cells were collected. After washing twice with PBS containing 1% BSA and once with FACS buffer, the cells were suspended in 20 μl of IF5(13) anti-CD59 mAb (5 μg/ml) in FACS buffer, incubated for 15 min on ice, washed, secondarily stained with FITC-conjugated sheep antimouse F(ab′)₂, washed a final time, and analyzed on a Becton Dickinson FACScan flow cytometer. In parallel studies, cells were sensitized with 50 μg/ml YTH3.1 anti-CD59 mAb in place of anti-DAF antibody and similarly incubated with 1:4 rabbit serum.

After expansion of 100 sorted CD59⁺DAF⁺ cells 10⁶- to 10⁷-fold, the cells were harvested, counted, and washed with fresh medium. The pelleted cells then were resuspended in 5 ml of medium containing 1 ng/ml activated aerolysin [kindly provided by T. Buckley (Victoria University, Victoria, Canada)], and the mixture was rotated at 37°C for 2 h. The remaining cells were pelleted twice, the pellet was resuspended to 5 ml, and the cell suspension was layered over 5 ml of Ficoll-Paque. After centrifugation at 400 × g for 30 min at 20°C, interphase cells were collected and washed in FACS buffer. After counting, the cells were stained for CD59 with mAb 1F5 and analyzed on a FACScan flow cytometer. The percentage of CD59⁻ cells was quantitated using CellQuest software.

The mutation frequency was calculated by (a) counting the total number of cells that grew after expansion of the 100 sorted tumor cells that had been purged; (b) counting the number of cells surviving after aerolysin treatment; and (c) determining the proportion of surviving cells that were CD59 negative(GPI⁻) by flow cytometry (see Fig. 4). The number of GPI⁻ cells was then divided by the total number of cells present before aerolysin treatment. The denominator, i.e.,total cells, was usually 0.8–1.5 × 10⁸.

Because the mutation frequency in Mut cells is expected to be ∼10⁻⁴/cell and the usual maximum sensitivity of flow cytometric discrimination is 1–2 × 10⁻², we adopted a two-step strategy in which we first sorted for cells that failed to label for one GPI-anchored reporter and then analyzed the negatively sorted cells for the presence of a second GPI-anchored reporter. On the basis of their (a)high expression levels on colonic epithelium and otherwise wide tissue distribution and (b) utility for use in an alternative method of selection (see below), we chose the cell surface complement regulators CD55 (DAF) and CD59 (MIRL) as the primary and secondary GPI-anchored reporters.

In the protocol adopted, negative sorting of DAF⁻ and CD59⁻stained cells was carried out gating on ∼0.5% of the cells with the lowest DAF fluorescence levels. After collection of the sorted cells (usually ∼3000), they were examined for CD59 expression. The results are shown in Fig. 1. As seen from the histograms, discrete peaks of CD59⁻ cells were clearly visible in all cases for the Mut lines and uniformly absent in the control lines.

Because DAF and CD59 are cell surface complement regulators that function to protect self (neoplastic as well as normal) cells from autologous complement attack, we examined the feasibility of enriching for PIGA-mutated cells by complement-mediated lysis. In the first set of studies, two Mut lines and a control line were sensitized with a polyclonal anti-DAF antibody (which blocks DAF function), and the sensitized cells were incubated with rabbit complement (which is not regulated at the C9 step by human CD59). After centrifugation and extensive washing, surviving cells (∼2000) were stained for CD59 and examined by flow cytometry. As shown in Fig. 2, discrete populations of CD59⁻ cells were seen for both Mut lines but not for the control tumor line. Similar results were obtained if the cells were treated first with an anti-CD59 blocking antibody followed by rabbit complement and the survivors analyzed for DAF (not shown).

Among genes encoding the enzymes providing for GPI-anchor processing,only PIGA is X-located. It, therefore, is predicted that GPI-anchor-defective cells arising in Mut lines should derive from mutations of PIGA rather than other (autosomal)genes. To confirm this and ascertain the type of mutation that occurred in each case, we recovered GPI-anchor-defective HCT116 cells from GPI-anchor⁺ cells that first had been purged of mutated cells (see below). We cloned the cells and, after expansion,isolated RNA and subjected PIGA reverse transcription-PCR products to sequence analyses. In all experiments two independent PCRs were done. For a control, PCR of genomic DNA both from the initially sorted CD59⁻DAF⁻ clone and the original line was done to confirm the mutation and establish that it represented a new event. The results are shown in Fig. 3. As can be seen, mutations of the PIGA gene were documented in cells derived from three independent pools (A, B, and C). In each instance more than one mutation was present. In the different HCT116 pools both the same and different mutations were found (see“Discussion”).

Previous studies using the hprt assay system(4) have shown that the mutation frequency of the X-linked hprt gene in the Mut lines studied is 10⁻⁴–10⁻⁵ mutants/cell. To determine whether the mutation frequency of the PIGA gene in the lines is the same or differs, we purged three of the lines (MT1,LoVo, and HCT116) by sorting anti-CD59-stained cells for positivity and deposited 100 CD59⁺ cells per well into the wells of six-well plates. We then expanded the cultures to 10⁸–10⁹ cells and treated them with activated aerolysin, a bacterial toxin that specifically reacts with GPI-anchor structures and induces lysis of GPI-anchored protein-expressing cells (14). As shown in Fig. 4, control studies in which this toxin was added to varying proportions of wild-type (GPI-anchored protein⁺) K562 cells and mutant (GPI-anchored protein⁻) 1A(15) cells (see “Discussion”) verified that the toxin lysed only the wild-type cells. Mut cell lines LoVo, MT1,and HCT116 for which mutation frequency data are available in the literature using the hprt assay were tested. The results of the studies with the Mut cells are given in Table 2. The PIGA gene in the three Mut cell lines exhibited mutation frequencies comparable with those reported for the hprt gene in the same cells (4, 16).

A number of methods have been used for detection of the Mut phenotype in neoplastic cells. These include(a) measurement of the mutation rate of the X-linked hprt gene (17, 18); (b) assessment of ouabain resistance (19), a property conferred by a non-X-linked gene that is acquired in a dominant-negative fashion;and (c) in cells in which one TK allele is first experimentally inactivated, resistance to the drug trifluorothymidine(20). Among these methods, the hprtassay is the most widely used. Major drawbacks, however, of its use and that of the others, are that they are uniformly time consuming and cumbersome to carry out. In addition, the neoplastic cells must be grown for a significant period of time, a requirement difficult to achieve with most primary tumors. For the hprt assay,titration of the 6-thioguanine selectant before plating is required. Moreover, cells must also be grown without selectant to correct for cloning efficiency in mutation rate calculations.

In this study we showed that loss of GPI-anchor expression can be substituted for the above procedures as a sensitive, highly convenient method for identifying Mut phenotypes. Its generality for use and practicality derives from the facts that (a) GPI anchoring is a ubiquitous mechanism used by all cell types (5, 6); (b) many different proteins (>100) are GPI-anchored; and (c) GPI-anchored protein expression is easy to detect. The basis for its applicability is that one of the genes, PIGA, required for GPI anchor assembly, is located on the X chromosome (8) and consequently is functionally inactivated by single-hit kinetics. Such a forward mutation assay using PIGA as the target offers the advantage that it should detect a full range of mutations including large deletions,frameshifts, and base substitutions (21). Perhaps most importantly, a PIGA-based assay does not require growth of the cells and thus could in theory permit direct analysis of tumors for Mut phenotypes without establishing cell lines. In principle, the main complicating factor for this would be the ability to dissociate and prepare homogeneous suspensions of the tumor cells. Work is in progress to accomplish this.

Our experiments confirmed that the presence of subpopulations of GPI-anchored protein-deficient cells distinguishes Mut and non-Mut phenotypes. Sequence analyses verified that PIGA mutation was responsible for the defect in each case. Examination of the mutations found showed that in some cases, they occurred at mononucleotide repeats as is characteristic of MMR defects,but in some instances they did not. Moreover, in certain cases, the same mutation was repetitively identified. Similar results have been found in previous analyses of mutations of the hprt gene in the same lines (17). PCR amplification of genomic DNA from the isolated GPI-anchored protein-negative Mut clones and from the parental lines confirmed that the PIGA mutations were not attributable to PCR errors and did not preexist in the cells.

Our studies demonstrated that an important additional advantage of this newly described methodology is that highly efficient ancillary procedures can be used to enrich for GPI-anchor-defective cells and thereby make their detection easier. These procedures include elimination of nonmutated cells by pretreatment of the original tumor cell populations with antibodies to GPI-anchored proteins, in particular the cell surface complement inhibitors CD55 or CD59 followed by complement or by pretreatment of the cells with the GPI-reactive bacterial toxin, aerolysin (14).

With the use of the later technique, we were able to quantitate the mutation frequency of the PIGA gene, an issue that is important in PNH, where naturally occurring PIGA mutations underlie the disorder (9, 21). Our results show that, at least in the above Mut cells, its mutation frequency is similar to that of the hprt gene, indicating that the PIGA gene is not inherently hypermutable.

In summary, our data taken together offer a novel approach for identifying DNA replication or repair defects in cancer that is not only convenient and consequently useful clinically but also potentially valuable for further work in this mechanistically informative field of research.

We thank Dr. James K. V. Willson of the Ireland Cancer Center for providing most of the colon cancer cell lines, Drs. Hidechika and Noriko Okada (Nagoya City University School of Medicine,Nagoya, Japan) for monoclonal antibody 1F5, and Dr. Tom Buckley for aerolysin. We thank Sara Cechner for manuscript preparation. We gratefully acknowledge Dr. Richard Boland for generously providing the HCT116 and its chromosome-complemented derivatives and Dr. Michael Brattain for RKO.