A Novel Protein Complex Distinct from Mismatch Repair Binds Thioguanylated DNA

Materials and Methods

Cell Culture and Preparation of Cell and Nuclear Extracts.

The human T-lineage leukemia cell lines CCRF-CEM and Molt4 and the human B-lineage WIL2NS (B lymphoblast) cells were obtained from ATCC (Manassas, VA); the human T-lineage leukemia P12, pre-B-cell leukemia 697, and B-lineage Nalm6 cell lines were obtained from the DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). Cell number and viability were determined in duplicate in a Burker-Turk chamber using trypan-blue exclusion. All experiments were started with an initial concentration of 0.25 × 10⁶cells/ml before addition of MP, which was dissolved in medium and added as a single dose to achieve a final concentration of 10 μM drug in the culture medium (determined by absorbance at 320 nm). Cytotoxicity was expressed as a ratio of viable cells incubated with MP (i.e., cell count × percentage of viable cells) to the number of viable cells (cell count × percentage of viable cells) in the control (without MP). Nuclear extracts were prepared according toDignam et al. (1983).

Synthesis of Duplex DNA.

The synthesis of modified oligodeoxyribonucleotides containing 6-deoxythioguanosine (dG^S) was accomplished using standard phosphoramidite chemistry with S6-DNP-dG-CE phosphoramidite (Glen Research, Sterling, VA) and isolated by 12% PAAG, as described previously (Krynetskaia et al., 1999). The presence of dG^s in oligodeoxyribonucleotide was confirmed by spectral analysis (λ_max 260 and 340 nm) and enzymatic hydrolysis to nucleosides followed by high-performance liquid chromatography separation of the reaction mixture as described (Krynetskaia et al., 1999). For annealing, oligonucleotides were heated to 85°C and cooled slowly to room temperature, and duplexes were purified by nondenaturating gel-electrophoresis (12% PAAG). Oligonucleotide sequences are given in Table1.

Primer Elongation by Deoxynucleoside Triphosphates (dNTPs) Opposite dG^S

The Klenow fragment (exo⁻) Escherichia coli DNA polymerase I and the dNTPs (ultrapure grade) were purchased from Promega (Madison, WI). A 13-mer primer was³²P-labeled and annealed with 10% excess of the 42-mer template, and polymerase reaction was performed as described (Spratt and Levy, 1997). Radioactivity on the gel was determined by PhosphorImager and the ImageQuaNT Software system (Molecular Dynamics, Sunnyvale, CA).

EMSA Assay and Nucleoprotein Complex Isolation.

DNA-protein binding assays were performed as described earlier, in 25 to 30 μl of buffer A (10× incubation buffer; Stratagene, La Jolla, CA) or buffer B (10× 200 mM Tris-Cl, pH 8.0, 50 mM NaCl, 3 mM MgCl₂, 1 mM dithiothreitol, 10% glycerol) (Griffin et al., 1994). DNA-protein complexes were isolated by 5% (w/v) denaturing polyacrylamide gel (PAAG) at 4°C. Fractions A, B, C were extracted by electroelution and analyzed in 7.5% SDS-polyacrylamide gels. N-terminal sequencing was performed in the Hartwell Center for Biotechnology and Bioinformatics at St. Jude Children's Research Hospital, after electrotransfer to the polyvinylidene difluoride membrane.

Immunological Methods.

A monoclonal antibody against hMSH2 (Ab-2; Calbiochem, San Diego, CA), hMLH1 (Ab-2; Calbiochem), and polyclonal antibodies against hMSH6/GTBP (N-20; Santa Cruz, Santa Cruz, CA) were used for Western blotting at dilutions of 1:5000, 1:1000, and 1:1000, respectively. Monoclonal antibody for GAPDH (Chemicon, Temecula, CA) and PCNA (Calbiochem, San Diego, CA) was used at a dilution of 1:500 for Western blotting and 1:50 for immunofluorescence. Quantitative comparisons of total cellular and nuclear GAPDH among cell lines were normalized to the amount of total protein loaded in each lane.

Confocal Microscopy.

To assess GAPDH distribution between nucleus and cytosol and the translocation of GAPDH from nucleus to cytoplasm at various times after incubation with MP, immunofluorescence analysis was performed on ALL cells using a monoclonal antibody to GAPDH and FITC (green)-coupled second antibody. Cells were also stained with propidium iodide to identify the nucleus. Cells were examined in a Leica confocal laser scanning microscope (model TCS NT SP; Leica, Wetzlar, Germany) to derive optical sections of cells to unambiguously localize the nuclear and cytoplasmic labels. Cells were centrifuged onto glass slides using the Cytospin-1 cytocentrifuge (Shandon Southern Instruments, Inc., Sewickley, PA). Slides bearing cells were rinsed in PBS, fixed in 3% paraformaldehyde and permeabilized with acetone as per our published procedures (Murti et al., 1993). The cells were incubated with monoclonal GAPDH antibody, followed by FITC-coupled secondary antibody. Slides processed per the above were also stained with propidium iodide (after incubating the cells with RNase) to label the nuclei. Differences in median intensity in the nucleus at 0 and 48 h were assessed by the Exact Wilcoxon rank-sum test.

Results

Primer Elongation by dNTP on Thioguanylated DNA Template.

Kinetic parameters of the DNA polymerization reaction across the thioguanylated DNA template are summarized in Table2. These data demonstrate that dCTP is preferentially incorporated into DNA opposite dG^S; dTTP is also incorporated opposite dG^S but at a substantially lower efficiency. The incorporation of dATP and dGTP occurs at an even lower rate and affinity, opposite either dG or dG^S. Therefore, modified duplexes containing either G^S·C or G^S·T were used in subsequent experiments (Table1).

Thiopurine Sensitivity and MMR Status of Human Lymphoblastoid Cells.

The relation between MP cytotoxicity, DNA modification and mismatch repair protein expression was assessed in human T- and B-lineage leukemia cells (CEM, Molt4, P12, WIL2NS, 697, Nalm6), after treatment with 10 μM MP. Nuclear extracts from CEM, Molt4, P12, 697, WIL2NS cell lines demonstrated G·T-mismatch binding by the MSH2/MSH6 (MutSα) complex (Fig. 1A). In contrast, this G·T-binding activity was absent in Nalm6 cells. Western blot analysis of nuclear extracts revealed the absence of both components of MutSα, namely hMSH2 and hMSH6, in Nalm6 cells (Fig. 1B) and the absence of hMLH1 in CEM (not shown), in contrast to the other cell lines that had these MMR proteins.

As we have reported previously in detail (Krynetskaia et al., 1999), after 24 h exposure to mercaptopurine, there was essentially no cytotoxicity. Only after 48 h did the viability of cells decline, with even greater cytotoxicity after 72 h except for the most resistant cells. CEM, Molt4, and Nalm6 cells were the most sensitive to MP (<50% viability after 72 h exposure to MP), whereas P12 and 697 cells were more resistant (75–90% viability after 72 h). Using terminal deoxyribonucleotidyl transferase-mediated dUTP nick-end labeling assay, it was shown that Nalm6 cells were highly sensitive, with up to 30% of cells exhibiting apoptotic changes after 24 h incubation, in contrast to P12 cells, which did not show evidence of apoptotic changes after 24 h exposure to 10 μM MP. Parallel experiments assessing BrdU incorporation established that MP did not inhibit DNA synthesis (data not shown). With the exception of the most resistant cells (697 and P12), the nuclei are too fragile after 72 h exposure for isolation and characterization; therefore, post-treatment nuclei were harvested at 24 and 48 h.

MutSα Binds Duplex DNA with dG^S·T Mismatch.

Incubation of either the G·T- or G^S·T containing DNA duplexes (3GT/3G^ST or 4GT4/G^ST) with nuclear extracts from MutSα-proficient cells (Molt4, CEM, 697, Wil2NS, P12) resulted in formation of a nucleotide-protein complex with MutSα (G·T binding shown in Fig. 1A; G^S·T in Fig.2A), that was abrogated by ATP (fraction B in Fig. 2A, lanes 1 and 2). In contrast, no such protein-DNA complex was formed in the MSH2/MSH6 deficient Nalm6 cells (Fig. 1A lanes 1 and 2; Fig. 2A, lanes 3 and 4). However, we observed formation of a second DNA-protein complex in each of these cell lines when nuclear extracts were incubated with DNA duplexes containing either a G^S·T- or a G^S·C, and this complex was not abrogated by ATP (Fig. 2A, fraction A). Similar DNA-protein complexes were formed with G·T- and G·C-containing DNA (duplexes 3GT/3GC or 4GT/4GC), but with lower affinity than its binding to dG^S-containing duplexes (Fig.3C). Biochemical isolation of the G^S-DNA-protein complexes gave fractions A, B, and C as depicted in Fig. 2A. Individual bands isolated by EMSA were further analyzed by SDS electrophoresis and Western blot analysis using hMSH2- and hMSH6-specific antibodies. The MSH2 and MSH6 proteins constituting the MutSα heterodimer were documented by Western blot analysis in fraction B but not in A or C, as exemplified with P12 nuclear extract in Fig. 2B.

Identification of GAPDH in dG^S-DNA-Protein Complex

The DNA-protein complex formed between dG^S- containing oligonucleotide duplexes and nuclear extracts in the presence of ATP was eluted (fraction A) and characterized by SDS electrophoresis. Multiple protein bands ranging from 30 to 200 kDa were detected by silver staining, with four to five major components (Fig. 3A). N-terminal sequencing of the 37-kDa protein band identified GAPDH as one of the components of the dGS-binding complex. GAPDH was documented in fraction A, but not in C by Western blot analysis, as exemplified with Nalm6 nuclear extract in Fig. 2C. As depicted in Fig. 3, B (lane 2) and C (lanes 2 and 4), anti-GAPDH monoclonal antibody added to the reaction mixture effectively dissociated this complex. Control experiments with anti-PCNA monoclonal antibodies revealed no effect on this protein complex (Fig. 3B, lane 3).

Nuclear Localization of GAPDH in ALL Cells.

Western blot analysis revealed similar levels of GAPDH in total cell lysates of these six human lymphoblastic leukemia cell lines (Fig.4A). In contrast, the intranuclear level of GAPDH differed markedly in these cells (Fig. 4B), with the lowest level of intranuclear GAPDH in P12 cells and the highest in Nalm6 cells. Furthermore, immunostaining of cells with anti-GAPDH antibody and confocal microscopy revealed differences in the distribution of GAPDH between nucleus and cytosol in these cell lines (Fig. 4C). Sequence analysis of RT-PCR amplified RNA isolated from Nalm6, WIL2NS, P12, and 697 cells did not reveal any mutations in the GAPDH open reading frame among these cell lines, indicating other mechanisms for the heterogeneity in nuclear GAPDH levels.

As depicted in Fig. 5C, there was a statistically significant relationship between the amount of GAPDH in nuclear extracts of human ALL cells before treatment and the extent of cytotoxicity after 72 h of incubation with 10 μM MP. Cells with higher GAPDH were more sensitive to MP and are localized in the lower right part of the chart, whereas more resistant cells are grouped in the upper left part, reflecting decreased intranuclear content of GAPDH (p = 0.001; Pearson coefficient).

Interestingly, MP treatment resulted in intranuclear accumulation of GAPDH, as exemplified by Western blot analysis of nuclear extracts from Molt4 and P12 cell lines (Fig. 6A). Visual confirmation of this was obtained by direct evaluation of intranuclear GAPDH level by confocal microscopy. Figure 6B depicts intracellular redistribution of GAPDH in P12 cells at different time points after MP treatment, evaluated by immunostaining and confocal microscopy. Confocal microscopy revealed that the treatment of human leukemia cells with 10 μM MP resulted in translocation of GAPDH from the cytosol to the nucleus, evident by 24 h and increasing further at 48 h of MP exposure, as shown for P12 cells (Fig. 6). Before MP exposure essentially all of the GAPDH was cytoplasmic in these cells, as indicated by the green color at the periphery of cells (Fig. 6B, panel 1) and the paucity of yellow color in the nuclei (Fig. 6B, panel 3). At 24 h after incubation with MP (Fig. 6B, panels 4–6), modest translocation of GAPDH into the nucleus was evidenced by the presence of green label in some nuclei (panel 4) and yellow in the merged image (panel 6). After 48 h of exposure to MP (Fig. 6B, panels 7–9), a significant fraction of GAPDH translocated into the nucleus, as revealed by the central distribution of green label (panel 7) and yellow in the merged image (panel 9). The mean fluorescence intensity in nuclei of these cells was 26.9 ± 5.8 at zero hour, 53.7 ± 20.4 at 24 h and 79.2 ± 28.0 at 48 h (p = 0.03). It is noteworthy that even after nuclear translocation of GAPDH, the relatively MP-resistant P12 cells had substantially lower nuclear GAPDH levels compared with the MP sensitive cells, [e.g., Molt4 (Fig. 6A)].

Discussion

Although thioguanine incorporation into DNA is considered asine qui non for thiopurine cytotoxicity, we have shown that the level of dG^S incorporation is not directly related to the extent of cytotoxicity in human lymphoblastic leukemia cells (Krynetskaia et al., 1999). We therefore postulated that heterogeneity in cellular recognition and response to DNA modified by dG^S incorporation plays a critical role in thiopurine sensitivity and resistance. Previous findings from our lab and others indicate that the alterations in DNA structure caused by dG^S incorporation into DNA likely contribute to the pharmacological effects of thiopurines (Ling et al., 1992a;Krynetskaia et al., 1999; Tendian and Parker, 2000). Therefore, we hypothesized that specific cellular proteins detect subtle changes in DNA structure, thereby triggering a chain of biochemical events that culminates in cytotoxicity in sensitive cells. The present work provides new insights into the proteins that participate in these cellular events, after thiopurine treatment.

It was previously shown that dG^STP is utilized efficiently in DNA-dependent DNA polymerase reactions, as well as RNA-dependent DNA polymerase reactions catalyzed by viral, bacterial, and mammalian DNA polymerases (Ling et al., 1991; Ling et al., 1992b;Rappaport, 1993). The present experiments showed that the Klenow fragment of E. coli DNA polymerase I efficiently elongated a 13-mer DNA primer using a thioguanylated template. The presence of dG^S in the template resulted in the formation of G^S·T-pairs, but formation of G^S·C-pair was preferred 300-fold (Table 2), in contrast to published data for S-methyl deoxythioguanosine template (Spratt and Levy, 1997), yet consistent with published reports on polymerization using a dG^S-template (Ling et al., 1992; Rappaport, 1993).

Recently, mismatch repair proteins were reported to be major mediators of the cytotoxic effects of thiopurines (Karran and Bignami, 1996;Swann et al., 1996; Armstrong and Galloway, 1997). Our experiments confirmed that the MutSα complex is able to bind G^S·T-containing DNA in MutSα-expressing cells (Molt4, P12, 697, Wil2NS, CEM) as previously reported (Griffin et al., 1994), but not in MutSα-deficient Nalm6 cells. The involvement of MutSα in G^S·T binding was also demonstrated by abrogation of the DNA-protein complex by ATP (Drummond et al., 1995). Consistent with this observation, there was markedly lower binding of MutSα with DNA containing a G^S·C mismatch (Fig. 2A, lanes 5 and 6, fraction B). Final proof for MSH2/MSH6 involvement in G^S·T recognition was obtained after isolation of fraction B and Western blot analysis of proteins interacting with G^S·T-containing DNA (Fig. 2B).

In contrast to previous reports that MMR-deficient colorectal carcinoma, endometrial carcinoma, and Burkitt's lymphoma cell lines are resistant to thioguanine (Branch et al., 1993; Hawn et al., 1995;Glaab et al., 1998), we have shown that two of the most MP-sensitive human lymphoblastic leukemia cell lines, CEM and Nalm6, lack one or more components of MMR. Western blot analysis demonstrated absence of both components of MutSα, namely hMSH2 and hMSH6, in Nalm6 cells (Fig. 1), and CEM cells were deficient in hMLH1 (E. Y. K., unpublished observations). This latter protein (MLH1) is not directly involved in binding to DNA mismatches (Kolodner and Marsischky, 1999).

These findings indicate that MMR protein expression is not the only determinant of MP sensitivity in human leukemia cells, and that MMR-deficient leukemia cells can be highly sensitive to thiopurines. EMSA experiments with nuclear extracts from human leukemia cells revealed protein complexes distinct from the MMR complex that interacted with duplex DNA containing either G^S·T- or G^S·C-pairs, leading to the identification of a new protein complex involved in recognition of thioguanylated DNA (Fig. 2A, fraction A, lanes 3, 4, 7, 8). This complex is distinct from MutSα (fraction B) because it is present in MMR-deficient as well as MMR-proficient cells, and its interaction with DNA is not sensitive to ATP (Fig. 2A, lanes 3,4,7,8). This protein complex interacted predominantly with G^S·T- and G^S·C-containing duplexes (Fig. 2A), rather than duplexes with G·T or G·C-pairs (Fig. 3C). Another important difference between this complex (fraction A) and MutSα (fraction B) is that the new complex is formed in the presence of both G^S·T and G^S·C duplexes with similar affinity, whereas MutSα predominantly binds the G^S·T -containing duplex DNA (Fig. 2A, lanes 1, 5 and 3,7). Western blot analysis established the presence of MSH2 and MSH6 in fraction B, and established their absence in the new protein complex in fraction A (Fig. 2B).

Subsequent biochemical isolation of this new protein complex (fraction A, from Nalm6 nuclear extract) and N-terminal protein sequencing identified GAPDH as one of its components. The presence of GAPDH in the complex was further confirmed by Western blot analysis after electroelution of fraction A (Fig. 2A line 4) and SDS electrophoresis (Fig. 2C).

EMSA experiments with anti-GAPDH mAb caused dissociation of this complex (Fig. 3, B and C); it is not unusual for an antibody to block or disrupt complex formation rather than to cause a supershift. Because GAPDH readily dissociates from the complex, we speculate that GAPDH is localized on the periphery of the core protein-DNA complex. SDS gel electrophoresis of fraction A revealed at least four other components of the dG^S-binding complex (Fig. 3A), in line with other reports demonstrating the presence of GAPDH in nuclear multiprotein complexes. For example, the presence of GAPDH in nuclear bodies containing promyelocytic leukemia protein was recently reported (Carlile et al., 2000), as well as physical association of GAPDH with DNA-polymerase α and primase (Grosse et al., 1986; Seal and Sirover, 1986). Enzymes participating in structural functions unrelated to their principal enzymatic activity have been recognized, as exemplified by several plant and mammalian proteins (Hendriks et al., 1988; Mulders et al., 1988; Cooper et al., 1993; Morimoto et al., 1999). Identification of other components of this complex should provide additional insights into the biological role of this protein complex, work that is ongoing in our lab.

GAPDH is a well known example of a multifunctional enzyme, with DNA repair as one of its functions (Sirover, 1999). Previous reports demonstrated intranuclear translocation of GAPDH and its participation in apoptotic pathways after cytotoxic treatment with cytosine arabinoside (Sawa et al., 1997), although the upstream events and points of intersection with the known apoptotic pathways remain to be elucidated. Consistent with this report, we observed translocation of GAPDH into the nucleus after MP treatment; the phenomenon was noted in both resistant (P12) and sensitive (Molt4) cells. The current work establishes GAPDH as a part of the protein complex interacting with DNA modified by thioguanine incorporation, indicating a role of this protein in a broader spectrum of cytotoxic agents.

Western blot analysis of nuclear extracts from six human leukemia cell lines, as well as confocal microscopy of intact cells, demonstrated marked heterogeneity in the nuclear levels of GAPDH (Fig. 4B). Basal intranuclear levels of GAPDH differed between cell lines, as evidenced by Western blot analysis (Fig. 4B). Confocal microscopy experiments also revealed that the human ALL cells studied have different ratios of nuclear to cytosolic GAPDH (Fig. 4C). Moreover, constitutive levels of nuclear GAPDH in these human leukemia cells before MP treatment as measured by Western blot analysis, significantly correlated with their sensitivity to 48 h (p = 0.002) or 72 h (p = 0.001, Fig. 5) treatment with 10 μM mercaptopurine. 697 Cells were used in this study as a negative control because of their very low level (more than 10-fold decreased) of thioguanosine incorporation into DNA, because of low hypoxanthine phosphoribosyltransferase activity (Krynetskaia et al., 1999) and correspondingly low sensitivity to MP. Our data indicate that basal GAPDH intranuclear localization per se is not sufficient to trigger cell death but that a high level of nuclear GAPDH is associated with cytotoxicity when accompanied by sufficient DNA modification.

The newly identified protein complex containing GAPDH binds with similar affinity to DNA duplex containing either a G^S·T or the more prevalent G^S·C mismatch, in contrast to MutSα, which binds primarily to DNA with a G^S·T mismatch. Earlier reports linking GAPDH to apoptosis (Sheline and Choi, 1998), together with the present finding that GAPDH is part of a protein complex that interacts with dG^S- DNA, points to an important role for this protein in cellular responses to DNA modifications. We hypothesize that the complex containing GAPDH serves as a sensor of structural alterations in DNA and links relatively small perturbations in DNA structure to the cellular apoptotic machinery.