Minichromosome Maintenance Protein 3 Elicits a Cancer-Restricted Immune Response in Patients with Brain Malignancies and Is a Strong Independent Predictor of Survival in Patients with Anaplastic Astrocytoma Free

Malignant gliomas are the most common primary brain neoplasms in adults, with an annual incidence of 4 to 7 per 100,000. Despite significant advances in standard treatment modalities, such as tumor resection and postoperative radiation and chemotherapy, the prognosis of patients with gliomas remains dismal. The median survival for patients with anaplastic astrocytoma (WHO grade 3) and glioblastoma (WHO grade 4) is <3 and <1 years, respectively (1). Thus, alternative treatment options such as immunotherapy have recently gained increasing attention. Prerequisite for success of immunotherapeutic strategies targeting tumor tissue is the presence of tumor-associated antigens in cancer cells.

The serologic analysis of recombinant cDNA expression libraries has evolved as a powerful strategy to identify tumor antigens, without the need for cultured autologous tumor and T cell lines (2, 3). Autologous or allogeneic sera from cancer patients are used for immunoscreening of cDNA libraries prepared from human tumors or tumor cell lines (4). This method has been applied to a large number of human malignancies, such as melanoma, lung cancer, breast cancer, hepatic cancer, renal cancer, and prostate cancer (4, 5). A broad range of tumor-associated antigens has been identified, including differentiation antigens (e.g., tyrosinase), overexpressed antigens (e.g., carbonic anhydrase XII), splice variant antigens (e.g., restin), retroviral antigens (e.g., HERV-K10), mutational antigens (e.g., p53), and members of the so-called cancer testis antigen family, such as NY-ESO-1 (4). Several glioma-associated antigens detected by screening with autologous or allogeneic human sera have also been described, including transcription factor SOX6 (6), glial fibrillary acidic protein (7), and PHD finger protein 3 (8).

During cell cycling, DNA replication is initiated at several chromosomal “origins of replication.” Cyclin-dependent kinases as key regulators of the cell cycle are thought to control the competence for replication by allowing binding of “prereplicative complexes” to these origins in G₁ phase, thereby rendering the DNA “licensed” for replication. Proteins that have been identified as being essential for prereplication complex formation include the origin recognition complex, the cell division cycle 6/18 proteins, the cell division cycle 10–dependent transcript 1 protein, and the minichromosome maintenance protein (MCM) complex (9).

The family of MCM proteins is highly conserved among eukaryotes and comprises at least six different nuclear proteins (MCM2-7). MCM proteins as a key component of the replication licensing complex ensure that chromosomes are replicated only once per cell cycle. They are recruited onto DNA replication origins in G₁ phase and are released during chromosome replication in S phase. The six proteins are highly similar to each other and form a hexameric complex, which is thought to function as a replicative helicase that moves on chromatin and unwinds DNA as replication proceeds (10). It has been shown in several organisms that quiescent cells in G₀ phase as well as permanently arrested (senescent) and/or terminally differentiated cells down-regulate MCM2-7 expression and are replication incompetent (11, 12). Determination of MCM2-7 expression levels in tumor tissue may therefore serve as a powerful diagnostic tool for accurately assessing the proliferative capacity of malignant cells.

MCM3 encodes a nuclear protein of 808 amino acids and has a molecular weight of 105 kDa (13, 14). Its expression is up-regulated in proliferating cells, whereas intracellular MCM3 levels decrease significantly in differentiated and growth-arrested cells (15).

Ki-67, another proliferation marker, is widely accepted for routine immunohistochemical analysis of cancer tissues, including brain tumors. It is a nuclear protein of yet unknown function, with expression tightly associated with the cell cycle (16). Ki-67 is expressed in (late) G₁, S, G₂, and M phases but is rapidly down-regulated in resting cells. In some tumors (e.g., meningiomas), Ki-67 is of prognostic significance (17), whereas in others, including astrocytomas (18, 19), its relationship to patient survival has been discussed controversially. Its half-life is ∼60 to 90 minutes (16), whereas that of MCM3 seems to be ∼24 hours in the promyelocytic cell line HL-60 (15). Endl et al. (20) showed that Ki-67 and MCM3 were both strongly expressed in the germinal centers of human tonsils but not in the nonproliferating mantle zone. In addition, anti-MCM3 antibodies but not the anti-Ki-67 monoclonal antibody MIB-1 labeled also the intermediate nonproliferating layer of the epithelium, suggesting that MCM3 is expressed in actively cycling cells as well as in cells that have just exited the cell cycle but are still capable of reentering it. MCM3 expression in dysplastic, premalignant, and neoplastic tissues may thus reflect more accurately their proliferative potential than MIB-1 staining, which labels only actively cycling cells.

We show here that the replication licensing factor MCM3 is immunogenic in a cancer-restricted fashion in patients with malignant brain tumors and is an independent negative predictor of survival in patients with astrocytic tumors.

Patients and Tumor Tissues. This study was approved by the Ethics Committee of the Faculty of Medicine, Martin Luther University Halle-Wittenberg. MCM3 immunohistochemistry was done on paraffin-embedded specimens of 142 patients with histologically confirmed newly diagnosed astrocytoma (WHO grades 2–4) and 27 patients with recurrent astrocytoma (WHO grades 2–4). All tissue sections were reviewed by two independent neuropathologists for confirmation of histologic diagnosis. Immunohistochemically stained slides were evaluated by at least two different investigators. The median age of patients with primary tumors was 57 years (range, 19–82 years), and the median age of patients with recurrent disease was 49 years (range, 24–73 years). For further clinicopathologic features, see Tables 1 and 2. Survival data could be obtained from 114 patients with primary astrocytoma, and comparative analysis between MCM3 and Ki-67 expression was done by immunohistochemistry in 91 of these tumors. All astrocytoma patients were treated according to standard protocols, including surgery and adjuvant therapy, such as external fractionated radiation with or without subsequent chemotherapy. Neuroradiological follow-up has been carried out mostly by magnetic resonance imaging and, in some cases, by computed tomography.

cDNA Library Construction. PolyA⁺ RNA was affinity purified from a human glioblastoma using an oligo(dT) column according to the manufacturer's instruction (Qiagen, Valencia, CA). The cDNA was reverse transcribed and inserted directionally into a λ ZAP phage expression vector as outlined by the manufacturer (Stratagene, La Jolla, CA). The cDNA library was packaged into phage particles that were subsequently used for infection of Escherichia coli.

Immunoscreening of the cDNA Library.E. coli harboring the phage cDNA expression library were plated on agar plates at a density of 3,000 to 5,000 plaques per 14 cm ∅ plate. Plates were incubated overnight at 37°C, after which time protein expression was induced with isopropyl-β-d-thiogalactopyranoside. Plaques were transferred to nitrocellulose filters, which were blocked with 5% dry skimmed milk and incubated with allogeneic sera from patients with anaplastic astrocytoma and glioblastoma at 4°C overnight.

Prior to use for immunoscreening, each patient's serum was preabsorbed several times with lysate from untransfected and λ phage-transfected E. coli. Absorbed serum was used at a final concentration of 1:100 to 1:300 for antigen detection. Bound antibodies were detected with alkaline phosphatase–conjugated goat anti-human Fcγ secondary antibodies (Dianova, Hamburg, Germany). Reactive phage plaques were visualized by incubating the filters with 5-bromo-4-chloro-3-indolyl-phosphate and nitroblue tetrazolium. Positive plaques were excised from the plate and submitted to three further rounds of immunoscreening until purity was achieved. Positive plaques representing human immunoglobulins were identified by omitting the serum incubation step in the immunostaining protocol and were excluded from further analysis.

Plaque Assay. The plaque assay was done in the same way as the immunoscreening procedure described in the previous section. Phages with MCM3 cDNA inserts were mixed with control phages (without insert) at a ratio of 1:1 to 1:2 and used to infect E. coli as described above. Filters containing this plaque mixture were used for immunoscreening with sera from patients with brain malignancies and from healthy donors.

Sequence Analysis of Reactive Plaques. Positive clones reacting with human sera were excised to pBK-CMV plasmid forms according to the manufacturer's instructions (Stratagene). Sequencing with dideoxynucleotides labeled with fluorescent dyes was carried out using the ThermoSequenase II Dye Terminator kit (Amersham, Arlington Heights, IL). Sequence analysis was done using BLAST software and the Genbank database.

Rapid Amplification of cDNA Ends Analysis. Rapid amplification of cDNA ends (21) was done using the SMART RACE cDNA amplification kit (Clontech, Palo Alto, CA), and total RNA was extracted from the same glioblastoma that had been used for cDNA library construction.

Immunohistochemistry. MCM3 immunohistochemistry was carried out on formalin-fixed, deparaffinized tissue sections (5 μm) with an affinity-purified polyclonal rabbit anti-MCM3 antibody (kind gift from R. Knippers, University of Konstanz) at a dilution of 1:100. For Ki-67 staining, the monoclonal antibody MIB-1 (DAKOCytomation, Ely, United Kingdom) was used at a dilution of 1:25. Antigen was retrieved by microwaving in 10 mmol/L citrate buffer. Slides were incubated with primary antibody in a humidified chamber for 1 hour at 37°C. Bound antibody was detected using the avidin-biotin complex method (Vector Laboratories, Burlingame, CA) and 3-amino-9-ethylcarbozole (DAKOCytomation) as the chromogen. Slides were counterstained with hematoxylin. Negative controls included substitution for the primary antibody with rabbit polyclonal pre-immune serum as well as “irrelevant” monoclonal immunoglobulin G antibodies (DAKOCytomation). Tissue sections were grouped according to percentage of MCM3-expressing tumor cells (0, negative; 1, 1–10% positive; 2, 11–25% positive; 3, 26-50% positive; 4, 51–75% positive; 5, 76–90% positive; and 6, 91-100% positive). Staining intensities were rated as weak (+), moderate (++), or strong (+++).

Statistical Analysis. Statistical evaluation was done using the SPSS software version 11.5 (SPSS Inc., Chicago, IL). Bivariate correlation of different variables was assessed using the Pearson two-sided χ² test. Survival data were evaluated using univariate and multivariate Cox regression analyses. P < 0.05 was considered statistically significant.

Immunoscreening of a Human Glioblastoma cDNA Expression Library. As autologous serum was not available, mixtures of allogeneic sera from patients with large necrotic glioblastoma (n = 8) or anaplastic astrocytoma (n = 1) were used for immunoscreening of a human glioblastoma cDNA library expressed in E. coli. A total of 1 × 10⁶ plaques were screened with three different serum mixtures, each consisting of three sera. The final dilution of each serum was 1:240 to 1:300. Twenty positively reacting clones were identified, of which two also showed a positive reaction when serum incubation was omitted, indicating that these plaques encoded human immunoglobulin G derived from tumor-infiltrating B lymphocytes. Five plaques that had been recognized by two of the mixtures were identified as the replication initiation factor MCM3 (Genbank accession no. NM_002388, coding region nucleotides 111–2,537). Three of the clones detected by both mixtures had inserts of identical size (nucleotides 483–3,104). In addition, one of the mixtures also reacted with a MCM3 fragment corresponding to nucleotides 1,308–3,104, whereas the other serum sample identified a positive plaque comprising nucleotides 929–3,110. No cDNA insert harboring the entire MCM3 coding sequence (nucleotides 111–2,537, 808 amino acids) was identified. Sequence analysis of all five inserts revealed a correct open reading frame with respect to the vector translation initiation codon ATG residing 5′ of the cDNA insert and no occurrence of mutations. The five MCM3 cDNA inserts that were detected by immunoscreening comprise the major part of the coding sequence, except for amino acids 1 to 124. We thus attempted to clone the 5′ end of MCM3 from the tumor tissue that had been used for cDNA library construction by rapid amplification of cDNA ends (21) to rule out the occurrence of sequence alterations in the amino-terminal part of MCM3 in this specific tumor. Despite intensive efforts, we did not succeed in obtaining additional MCM3 sequences, suggesting that the mRNA used for generation of the cDNA library might have already been degraded due to suboptimal freezing procedures. Other tumor tissues from patients whose sera had reacted positively with MCM3 were not available for RNA extraction and subsequent MCM3 sequence analysis.

Serologic Immune Response to MCM3 in Patients with Brain Tumors (Plaque Assay). Sera from 89 patients with astrocytoma [WHO grade 2 (n = 15 primaries), anaplastic astrocytoma (n = 17 primaries, n = 2 recurrences), and glioblastoma (n = 46 primaries, n = 9 recurrences)] were tested for immunoreactivity with MCM3 in the plaque assay. In addition, eight sera from patients with other intracerebral malignancies were screened [oligoastrocytoma WHO grade 3 (n = 2), oligodendroglioma WHO grades 2 (n = 2) and 3 (n = 1), primary central nervous system lymphoma (n = 1), and brain metastasis from lung adenocarcinoma (n = 2)]. Sera from 30 healthy volunteers were used as controls. The plaque assay was done using the longest MCM3 cDNA insert comprising nucleotides 482 to 3,104 (amino acids 125-808) of the MCM3 sequence. Phages containing no cDNA insert were included in each assay as negative control (Fig. 1). MCM3 immunogenicity clearly showed a cancer-restricted pattern: whereas none of the healthy controls showed a positive reaction with MCM3, antibodies directed against MCM3 were detected in sera from seven patients with glioblastoma, one patient with primary central nervous system lymphoma, and one patient with brain metastasis from lung cancer (Fig. 1; Table 3). Notably, 0 of 34 sera from patients with astrocytoma (WHO grade 2 or 3) reacted with MCM3, whereas 7 of 55 (12.7%) sera from glioblastoma patients contained anti-MCM3 antibodies. To obtain more information about the strength of the anti-MCM3 immune reaction, the nine positively reacting sera were diluted further up to 1:50,000. Whereas seven of these sera showed no reaction with MCM3 when diluted above 1:300, sera of two patients with glioblastoma recognized MCM3 at a serum dilution of 1:1,000 and 1:25,000, respectively (Table 3). Interestingly, both of the patients' tumors showed a high level of MCM3 expression with clearly >50% MCM3-positive tumor cells.

In 51 patients, both plaque assay for MCM3 and MCM3 immunohistochemistry of their tumors were available. In each case, blood had been taken at the time of tumor resection. None of the patients (n = 7) whose tumors contained ≤10% MCM3-positive tumor cells developed antibodies against MCM3, suggesting that a threshold level of intratumoral MCM3 expression might be necessary to elicit a humoral immune response against this antigen. On the other hand, as only 4 of 21 (19%) patients with ≥50% MCM3-positive cells in their tumors had detectable anti-MCM3 antibodies, a high level of intratumoral MCM3 seems not per se to be associated with the occurrence of anti-MCM3 antibodies.

MCM3 Expression in Human Astrocytomas. Proliferative capacity is a hallmark of neoplastic tissues. As the replication licensing factor MCM3 is tightly associated with cell cycling and thus proliferation, we examined by immunohistochemistry whether MCM3 is of prognostic relevance in human astrocytomas. Staining was done with an affinity-purified rabbit anti-MCM3 antibody, known to selectively recognize MCM3 but not other members of the MCM protein family (14). Immunohistochemistry was done on 142 primary astrocytomas [WHO grade 2 (n = 27), anaplastic astrocytoma (n = 50), and glioblastoma (n = 65)] and 27 recurrent astrocytic tumors [WHO grade 2 (n = 4), anaplastic astrocytoma (n = 11), and glioblastoma (n = 12)]. Two primary astrocytomas (WHO grade 1) were also stained for MCM3 but not included in the statistical analysis. Primary and recurrent tumors were evaluated separately. For clinicopathologic features of astrocytoma patients, see Tables 1 and 2.

MCM3 Expression in Primary Tumors. Whereas in normal human brain tissue samples only very few cells stained positive for MCM3, MCM3 expression in general was much more abundant in astrocytomas (Fig. 2). In all tumors, MCM3 labeling was restricted to the nucleus and most tumors showed intense staining for MCM3. The two astrocytomas (WHO grade 1) that were not included in the statistical analysis had ≤5% MCM3-positive tumor cells (Fig. 2). MCM3 expression in primary tumors did not correlate with sex (P = 0.800) but was associated with age (P < 0.001; cutoff ≤50 years). The number of MCM3-positive tumor cells correlated significantly with tumor grade (P < 0.001): 44.4% (12 of 27) of low-grade astrocytomas (WHO grade 2), 66% (33 of 50) of anaplastic astrocytoma, and 86.2% (56 of 65) of glioblastoma contained >10% MCM3-positive tumor cells. More than 25% MCM3-positive tumor cells were found in 18.5% (5 of 27) of astrocytomas WHO grade 2, 50% (25 of 50) of anaplastic astrocytoma and 73.8% (48 of 65) of glioblastoma. More than 50% tumor cells stained positive for MCM3 in 11.1% (3 of 27) astrocytomas WHO grade 2, 24% (12 of 50) anaplastic astrocytoma, and 43% (28 of 65) glioblastoma. More than 75% MCM3-positive tumor cells were only detected in 1 (2%) anaplastic astrocytoma and 7 (10.8%) glioblastoma.

Comparative Evaluation of MCM3 and Ki-67 Expression. MCM3 is part of the replication licensing complex and thus tightly associated with cell proliferation. Therefore, we were interested in comparing the expression levels of MCM3 and the established proliferation marker Ki-67 in diffuse astrocytomas. Ninety-one primary astrocytomas [WHO grade 2 (n = 19), anaplastic astrocytoma (n = 29), and glioblastoma (n = 43)] were available for comparative analysis of both proteins. Ki-67 expression was proven to be significantly associated with the number of MCM3-positive cells (P < 0.001; Fig. 3), histologic grade (P < 0.001), and age (P = 0.017; cutoff for Ki-67 ≤10%, cutoff for age ≤50 years) but not with sex (P = 0.228). MCM3 staining identified significantly more tumor cells in cycle than MIB-1 in astrocytomas of all three histologic grades (Fig. 3). Survival data were available in 71 of these patients (17 alive and 54 deceased). In univariate analysis, Ki-67 was proven to be a strong prognostic marker for shortened survival (P < 0.001; relative risk, 2.90; cutoff ≤10% Ki-67-positive tumor cells). In contrast to our findings with MCM3, Ki-67 lost its predictive power (P = 0.385) when investigated in a multivariate analysis together with sex, age, histologic grade, and MCM3 expression.

Relationship of MCM3 to Survival of Patients with Astrocytomas. Data for survival analysis were available from 114 patients (22 alive and 92 deceased) with astrocytoma (WHO grades 2-4). Cox regression analysis revealed that MCM3 was a significant prognostic marker in patients with astrocytoma (P <0.001; relative risk, 3.67; Fig. 4A; Table 4). Mean survival of patients with tumors harboring ≤10% MCM3-positive cells was 47.7 ± 7.2 months, whereas survival decreased to 21.2 ± 6.7, 11.5 ± 2.1, and 7.5 ± 1.2 months when astrocytomas contained 11% to 25%, 26% to 50%, or >50% MCM3-positive tumor cells, respectively. With respect to patients with high-grade astrocytoma (anaplastic astrocytoma and glioblastoma) that had been followed for >1 year after diagnosis (n = 95), only 31.9% (23 of 72) of the patients with tumors harboring >10% MCM3-positive tumor cells survived >1 year, whereas 73.9% (17 of 23) of the patients with tumors containing ≤10% MCM3-positive tumor cells (P < 0.001) survived >1 year. A significant impact on survival of astrocytoma patients could also be shown in the univariate analysis (Table 4) for histologic grade (P = 0.001), age (P < 0.001), and Ki-67 (P < 0.001; cutoff ≤10% Ki-67-positive cells) but not for sex (P = 0.923).

A striking difference in the impact of MCM3 on survival of patients with high-grade astrocytoma (anaplastic astrocytoma and glioblastoma) was noted when survival analysis was done separately for patients with anaplastic astrocytoma (n = 31) and glioblastoma (n = 54; Fig. 5). A high MCM3 labeling index (LI) was significantly associated with shorter survival in patients with anaplastic astrocytoma (P = 0.006) but not in patients with glioblastoma (P = 0.430), although the mean MCM3 LI was clearly higher in glioblastoma patients (54 ± 3.8%; range, 5–90%) than in anaplastic astrocytoma (30.3 ± 4.7%; range, 0–70%). When patients with anaplastic astrocytoma and glioblastoma were analyzed separately for the impact of Ki-67 on survival, no significant association was seen for either subgroup.

Multivariate analysis including sex, age, and histologic grade showed that MCM3 predicted outcome independent of other variables (P < 0.001; relative risk, 2.86; Fig. 4B; Table 5). When Ki-67 staining was included in the multivariate analysis, MCM3 retained its prognostic significance (P = 0.014), whereas this was not the case for Ki-67 (P = 0.247; cutoff ≤10% Ki-67-positive cells).

MCM3 and Time to Recurrence. In 56 patients with primary high-grade astrocytoma [anaplastic astrocytoma (n = 19) or glioblastoma (n = 37)], who underwent tumor resection and adjuvant therapy, precise data on the time to recurrence (mean, 10.3 ± 1.47 months; range, 1–53 months) were available and were correlated to MCM3 expression in the respective tumors at diagnosis. Recurrence-free survival was proven to be significantly associated with MCM3 expression in univariate (P = 0.010; cutoff ≤50% MCM3-positive cells) as well as in multivariate analysis (including sex, age, and histologic grade; P = 0.011; Fig. 6A). Patients whose tumor recurred within 1 year (n = 42) had 45.8 ± 4.5% MCM3-positive tumor cells, whereas only 26.8 ± 4.5% tumor cells stained positive for MCM3 in patients (n = 14) showing recurrence after >1 year (P = 0.006).

When the impact of MCM3 expression on recurrence-free survival was analyzed separately for anaplastic astrocytoma and glioblastoma, a cutoff of ≤25% MCM3-positive tumor cells was shown to be significantly associated with longer recurrence-free survival in anaplastic astrocytoma (P = 0.020; relative risk, 4.19; Fig. 6B), whereas in glioblastoma a cutoff of 50% was proven to be prognostically significant (P = 0.020; relative risk, 2.47; Fig. 6C). When sex and age were included in the analysis, MCM3 lost its predictive power in anaplastic astrocytoma (P = 0.113; cutoff for MCM3 ≤25%) but not in glioblastoma (P = 0.012; cutoff for MCM3 ≤50%).

MCM3 Expression in Recurrent Astrocytic Tumors. In addition to the 142 primary astrocytomas, 27 recurrent tumors (WHO grades 2–4) from patients who received standard treatment were analyzed for MCM3 expression. As seen with primary tumors, MCM3 expression significantly correlated with age (P = 0.003; cutoff ≤50 years) but not with sex (P = 0.167). The number of patients with recurrent low-grade astrocytoma [WHO grade 2 (n = 4)] is too low for accurate statistical analysis of the relationship between MCM3 LIs and histologic grade. However, an association between histologic grade and MCM3 expression levels could also be observed in recurrent tumors: 1of4 (25%) astrocytomas WHO grade 2, 4 of 11 (36.4%) anaplastic astrocytoma, and 9 of 12 (75%) glioblastoma contained >10% MCM3-positive cells.

We report here for the first time that a member of the prereplicative complex, MCM3, can elicit a cancer-restricted high-titer immunoglobulin G immune response in tumor patients. Our data show that MCM3 expression is strongly associated with histologic differentiation in astrocytic brain tumors as well as with expression of the established proliferation marker Ki-67. In addition, we found that MCM3 is an independent predictor of poor prognosis in patients with astrocytoma.

Recently, a variety of glioma-associated antigens capable of eliciting a humoral immune response in brain tumor patients have been characterized. Among others, these are mutant p53 (22), endostatin (23), the transcription regulator SOX6 (6), PHD finger protein 3 (8), and the differentiation antigen glial fibrillary acidic protein (7). In addition, expression of tumor-associated antigens, such as SSX-2, tyrosinase, or GAGE, known to elicit a humoral immune response in patients with malignancies other than astrocytoma, has also been described in human astrocytic tumors (24, 25). To the best of our knowledge, members of the replication licensing complex have not yet been shown as potential autoantigens eliciting humoral immune responses in patients with astrocytoma or any other type of cancer.

Although mutational events are known to occur relatively often in cancer, mutated antigens have only rarely been detected by immunoscreening of tumor cDNA libraries with sera from cancer patients (4). It is hypothesized that mutated antigens might indeed induce humoral immune responses in cancer patients but that the respective antibodies will also cross-react with the corresponding wild-type protein, as has been shown for anti-p53 antibodies (26). We did not succeed in obtaining sequence information about the amino-terminal 124 amino acids of the MCM3 protein expressed in the tumor tissue, which was used for generation of the cDNA library. Thus, we cannot entirely rule out that an altered sequence in this part of the protein is responsible for the immunogenicity of MCM3 in brain tumor patients.

The cDNA library was generated from a tumor with large necrotic areas and high MCM3 expression (∼70% MCM3-positive tumor cells) in the remaining tumor part. Although no PCR-based amplification step, known to bias transcript representation, was included in the cDNA library construction protocol, it is likely that MCM3 transcripts were highly represented in this library due to abundant MCM3 expression in the tumor. Overexpression of a wild-type protein per se is thought to be one of the main reasons for induction of a humoral immune response in cancer patients (5). As all patients with detectable anti-MCM3 antibodies had tumors that expressed MCM3 above the “threshold level” of 10% MCM3-positive tumor cells, increased antigenic load in conjunction with cell death and presentation of cell debris to the immune system may have triggered the production of MCM3 antibodies in some of the brain tumor patients. The scarce expression of MCM3 in normal brain tissue and the absence of mutations in the five MCM3 cDNA inserts we sequenced further support this hypothesis.

It was recently shown in several cancer cell lines that MCM3 is cleaved by caspase-3 and caspase-7 after induction of apoptosis by different stimuli (27). Cell death–related proteolytic cleavage and protein modifications are known characteristics of molecules with a propensity to become autoantigens (28) and may thus represent additional reasons for the autoantigenic potential of MCM3.

Whether the occurrence of anti-MCM3 autoantibodies is pathogenetically relevant or of prognostic significance can only be answered after analysis of further cases. In this context, it may be interesting to extend these studies also to patients with primary central nervous system lymphomas and brain metastases, as anti-MCM3 antibodies were also detected in sera of such patients (Table 3).

In the astrocytomas we examined, MCM3 stained at least as many but, in the vast majority, clearly more cells than Ki-67 (Fig. 3). This discrepancy can be readily explained by the fact that Ki-67 is only expressed in actively proliferating cells, whereas MCM3 labels such cells as well as cells that are licensed to replicate and thus can potentially proliferate but do not synthesize DNA (20). MCM protein–expressing cells may be of high predictive value for premalignant lesions, as could be shown for dysplastic lesions of the cervix. Williams et al. (29) found 80% to 100% MCM5-positive cells in these lesions, although Ki-67 and proliferating cell nuclear antigen labeled 3% to 8% of the cells.

Regarding high-grade astrocytomas, our data clearly show that MCM3 is of prognostic relevance for predicting survival in patients with anaplastic astrocytoma but not in glioblastoma patients (Fig. 5). Provided that MCM3 expression reflects the proliferative compartment of a given tumor more accurately than Ki-67 (20), these data indicate that proliferative capacity is of higher significance for patients with anaplastic astrocytoma than for patients with glioblastoma. In the latter, biological characteristics, such as invasiveness, migratory capacity, resistance to apoptosis, angiogenesis, and tumor vascularization, are likely to play a more important role for the course of the disease than cell proliferation.

Studies of MCM3 expression in malignant tissues have thus far been limited to the immunohistochemical examination of uterine cervical and endometrial carcinomas (30, 31). Ishimi et al. (30) found in their series of five cervical carcinomas that MCM3 was more frequently expressed in cancer cells than in basal (proliferating) cells of the normal squamous cell epithelial layer. In contrast, Kato et al. (31) reported that MCM3 protein levels were significantly lower in endometrial carcinoma than in normal proliferative endometrium and did not correlate with Ki-67 expression levels, suggesting that the replication licensing complex may be altered in endometrial cancer.

Expression of other MCM proteins, such as MCM2, MCM5, and MCM7, has been examined by immunohistochemistry in a wide variety of normal cells, premalignant (dysplastic) lesions, and tumor tissues derived from breast, lung, stomach, colon, kidney, bladder, and prostate as well as in meningioma and oligodendroglioma (32). With the exception of one study done in T₁ papillary bladder carcinoma, which did not find a significant association between MCM2 and Ki-67 LI (33), the published studies corroborate our findings. In general, MCM protein expression was positively associated with histologic dedifferentiation, tumor grade, Ki-67 LI, mitotic index, and poor prognosis. As observed in our study for MCM3, antibodies directed against different MCM proteins labeled considerably more tumor (and premalignant) cells than anti-Ki-67 antibodies, suggesting again that MCM proteins in general are clearly more sensitive markers for the proliferative compartment of tumor tissues.

Wharton et al. (34) examined expression of MCM2 in a series of oligodendrogliomas (n = 32). As shown in our larger series of astrocytomas for MCM3, expression of MCM2 correlated significantly with histologic grade and Ki-67 LI. MCM2 consistently identified more proliferative cells than Ki-67 in oligodendroglial tumors, and a high MCM2 LI was linked to a shorter survival. This is in accordance with our own findings for MCM3 in astrocytomas and indicates that MCM proteins may indeed be used as informative tumor markers in brain tumors.

The study of Hunt et al. (35) focused on the question whether MCM2 was a predictor of recurrence in benign meningioma. Indeed, MCM2 expression was significantly higher in recurrent than in nonrecurrent tumors. The observation that high MCM2 expression levels were associated with shorter recurrence-free survival was also made by Kruger et al. (33) in bladder carcinoma. Both studies confirm our own findings, which indicate that high MCM3 expression predicts a higher probability of early recurrence in patients with high-grade astrocytoma.

Given the essential role of MCM proteins in cell division and their relatively scarce expression in normal (noncycling) tissues, targeting the replication licensing complex seems an attractive option for a specific anticancer therapy. Indeed, treatment of U20S osteosarcoma and H1299 lung carcinoma cells with geminin, a small protein specifically inhibiting cell division cycle 10–dependent transcript 1, resulted in decreased binding of MCM2 to chromatin, a block in cell proliferation, and apoptosis (36). Feng et al. (37) treated several cancer cell lines as well as normal L-02 liver cells with antisense oligodeoxynucleotides and small interfering RNA molecules targeted to MCM2. In cancer cells, MCM2 gene silencing resulted in inhibition of DNA replication and cell proliferation as well as in apoptosis. In contrast, apoptosis was not induced in normal L-02 cells, which were mainly arrested in G₁ phase and therefore were still functional but stopped to proliferate. With regard to our findings indicating that MCM3 is overexpressed in highly proliferative astrocytic tumors, targeting MCM3 in these tumors may represent a promising strategy to block cycling of glioma cells.

We conclude that analysis of humoral immune responses in patients with diffuse astrocytic brain tumors may help to identify molecular markers of prognostic significance for this disease. The glioma-associated antigen MCM3 most likely reflects the proliferative fraction in astrocytic tumors more accurately than Ki-67 and thus may help to identify subgroups of gliomas with high proliferative capacity.

We thank Matthias Kappler for support with the statistical analysis, Jana Beer for help with immunohistochemistry, and the staff of the Department of Neurosurgery and the HLA Laboratory for providing assistance with collection and storage of tumor and serum specimens.