Amplification of 1q32.1 Refines the Molecular Classification of Endometrial Carcinoma

Endometrial cancer is the most common gynecologic malignancy, with an incidence that is expected to rise in the next 10 years due to the ageing population and increasing obesity rates (1, 2). Most patients have a favorable disease outcome when diagnosed at early stage; however, approximately 20% of patients still relapse and are incorrectly classified as “low-risk” patients (3). In addition, the survival of patients with metastatic or recurrent disease remains poor, mainly due to limited treatment options. Therefore, better insights into the molecular classification of endometrial cancer are required to improve risk prediction of recurrent disease, assist in adjuvant treatment decisions and develop novel treatment strategies (3–5).

Integrative molecular subtyping efforts for endometrial cancer have first been described by The Cancer Genome Atlas (TCGA; ref. 6), identifying four distinct molecular subgroups. Meanwhile, this molecular classification has been validated by independent studies (5, 7, 8); about 6% to 15% of endometrial cancers harbor POLE mutations, 20% to 30% of endometrial cancers are characterized by microsatellite instability (MSI), and 10% to 30% harbor TP53 mutations. The remaining endometrial cancers (∼40%), representing the largest subgroup, are heterogeneous and currently devoid of any surrogate marker. As such, endometrial cancers are classified as NSMP based on the absence of POLE, TP53 mutations or MSI; these tumors are therefore referred to as “nonspecific molecular profile” (NSMP) endometrial cancers. Endometrial cancers in the NSMP subgroup frequently have mutations in PTEN, CTNNB1, and PIK3CA, but overall a low mutation rate. In addition, these tumors have an intermediate number of somatic copy-number aberrations (SCNAs), but are confined to two distinct copy-number clusters upon unsupervised clustering (6). These two clusters are characterized by a variable disease outcome (5-year relapse-free survival, RFS: 79.4% vs. 63.1% for both clusters), suggesting that SCNAs have the potential to improve molecular stratification of endometrial cancers, in particular in the NSMP subgroup.

Here, we characterized SCNAs in NSMP endometrial cancers to reveal novel prognostic biomarkers that refine the classification of the NSMP molecular subgroup and possibly identify novel therapeutic targets. We first conducted an in-depth analysis of SCNAs in a large cohort of 502 endometrial cancers and then replicated the prognostic significance of relevant SCNAs in 840 endometrial cancers from randomized clinical studies.

For the discovery cohort, fresh-frozen tumors from 141 patients with primary endometrial cancer were collected from the University Hospital Leuven (Leuven, Belgium, n = 99) and the Institut de Recerca Biomèdica (Lleida, Spain, n = 42). Public TCGA data (n = 361) were also used (6). For the validation cohort, formalin-fixed paraffin-embedded (FFPE) tumor material was collected from 973 of the 1,141 patients with endometrial cancer who participated in the PORTEC-1 and -2 studies (9, 10). Details on clinical data collection, DNA extraction, SCNA genotyping, and protein expression analysis are described in the Supplementary Methods (available online). All studies obtained the relevant institutional review board approval in accordance with the principles of the Declaration of Helsinki. Informed consent was obtained according to these approvals.

Human endometrial cancer cell lines HEC-1A and RL95-2 (ATCC), COLO684 (Sigma-Aldrich) and the patient-derived cell line JVE1 were cultured as prescribed. MDM4 was inhibited in vitro using an antisense oligonucleotide (11). Overexpression was induced using pcDNA3.1 vector containing MDM4. Western blot was done using anti-MDM4 (1/5,000 in 5% milk; Bethyl Laboratories), anti-cleaved caspase-3 (1/100 in 5% milk; Cell Signaling Technology) and anti-actin as control (1/1,000 in 5% milk; Santa Cruz Biotechnology). Cell viability was measured using CellTiter-Glo (Promega). Additional details on these and other experimental procedures can be found in the Supplementary Methods (available online).

Associations between clinicopathologic features and SCNA clusters were tested using χ² statistics or Fishers exact test for categorical variables, and t test or ANOVA for continuous variables. Time-to-event analyses were calculated from the date of randomization to the date of endometrial cancer death (cancer-specific survival, CSS), to date of any recurrence (RFS), or date of distant recurrence (distant recurrence, DR).

For the identification of SCNAs associated with patient outcome in the discovery cohort, the penalized Cox proportional hazards model from the R package “glmnet” was used (12). This Cox proportional hazards model with LASSO penalty (Least Absolute Shrinkage and Selection Operator) shrinks small coefficients toward zero and selects a limited set of predictors to prevent overfitting due to either colinearity of the covariates or high-dimensionality. Cox regression included clinical covariates (grade, age, stage, and histology) and SCNAs defined by GISTIC (Genomic Identification of Significant Targets In Cancer; ref. 13), whereby GISTIC peaks were divided into five categories, that is, homozygous deletion, loss, diploid, gain, and amplification based on LogR signal and GISTIC output threshold values (t < −1.3; −1.3 ≤ t < −0.1; −0.1 ≤ t ≤ 0.1; 0.1 < t ≤ 0.9; t > 0.9, respectively). SCNAs spanning >70% of a chromosomal arm were defined as whole-arm SCNAs, while SCNAs spanning <70% of a chromosomal arm were considered focal SCNAs. Additional details on this analysis can be found in the Supplementary Methods (available online).

In the discovery cohort, the impact of relevant SCNAs in the different molecular subgroups was evaluated using an unpenalized Cox proportional hazards model correcting for the clinical factors grade, age, and stage with the R package “survival.”

In the validation cohort, hazard ratios (HRs) were calculated using a Cox proportional hazards model. Briefly, SCNAs were included together with established clinicopathologic prognostic factors: Age as continuous variable, grade (1–2 vs. 3), lymphovascular space invasion (LVSI; substantial vs. none or mild) and adjuvant treatment (vaginal brachytherapy, external beam radiotherapy, or no additional therapy). All reported P values were based on two-sided tests with P < 0.05 considered as significant.

All statistical analyses were performed with R (version 3.1.2) and IBM statistics (version 23.0).

The clinical characteristics of the discovery cohort, which consisted of 502 primary endometrial cancers, 141 of which were profiled in-house and 361 by TCGA, are highlighted in Table 1. SNP array profiles from these tumors were subjected to GISTIC identifying 58 significantly enriched focal amplifications (Supplementary Fig. S1A) and 49 focal deletions (Supplementary Fig. S1B). Consensus clustering was performed using each of these 107 SCNAs resulting in five consensus clusters (Fig. 1A; Supplementary Fig. S1C–S1G; ref. 14).

Clinical features of SCNA-based clusters are described in Supplementary Table S1 and Fig. 1A. As expected, the number of breakpoints and the percentage of the genome affected by SCNAs increased with increasing cluster number: Cluster 1 had no copy-number breakpoints and <1% of the genome was amplified or deleted, whereas cluster 5 had an average of 22.5 [95% confidence interval (CI), 20.3–24.7] breakpoints and 50.7% (95% CI, 47.4–53.9) of the genome was amplified or deleted. The contribution of the established endometrial cancer molecular subgroups (POLE-mutant, MSI, NSMP and TP53-mutant) to each copy-number cluster was as follows: Cluster 1 mainly consisted of endometrial cancers with POLE mutations or MSI endometrial cancers (14.7% and 39.7%, respectively), whereas cluster 5 predominantly consisted of TP53-mutant endometrial cancers (88.7%). The majority of endometrial cancers belonging to the NSMP subgroup fell within clusters 2, 3, and 4 (53.3%, 39.7% and 61.0%, respectively), although MSI endometrial cancers were also present in these clusters (36.7%, 55.6%, and 33.9%, respectively; Supplementary Table S1).

Similar to previous reports, we observed that SCNA clusters predominantly consisting of NSMP endometrial cancers (i.e., clusters 2, 3, and 4) had an intermediate prognosis compared with good prognosis cluster 1 (containing MSI/POLE-mutant endometrial cancers) and worse prognosis cluster 5 (containing TP53-mutant endometrial cancers). However, when comparing RFS between the three clusters only, we noticed additional prognostic differences (5-year RFS: 84.6%, 66.5% and 61.0%, respectively for cluster 2, 3 and 4; Fig. 1B), suggesting that these three clusters harbor-specific SCNAs underlying these differences. We therefore selected all focal amplifications, deletions as well as whole-arm aberrations for each of the 3 SCNA clusters separately (Fig. 1C; Supplementary Fig. S2) and assessed their prognostic effects. Survival analysis with penalized regression revealed four and six chromosomal regions or clinical covariates as stable predictors, respectively, for RFS and CSS (Fig. 2A and B). Among these, amplifications affecting chromosomal region 1q32.1 yielded the highest AUC both for RFS and CSS (Fig. 2C and D), indicating that this region exerted the strongest prognostic effect. In a subsequent univariable analysis, 1q32.1 amplifications correlated with RFS and CSS (1q32.1: HR, 2.27; 95% CI, 1.22–4.22; P = 0.004 for RFS and HR, 3.62; 95% CI, 1.48–8.77; P = 0.002 for CSS; Fig. 2E and F), and also in a multivariable Cox regression 1q32.1 amplifications correlated significantly with RFS and CSS and this independently of other clinical covariates (1q32.1: HR, 1.77; 95% CI, 1.23–2.60; P = 0.002 for RFS and HR, 2.60; 95% CI, 1.48–4.57; P < 0.001 for CSS; Supplementary Table S2).

Overall, we detected amplifications in the 1q32.1 locus in 98 of 268 (36.6%) endometrial cancers belonging to clusters 2 to 4. Interestingly, 1q32.1 amplifications were not present in the good prognosis cluster 2, but at a high frequency in the worse prognosis clusters 3 and 4 (68.2% and 43.7%). The 1q32.1 amplifications were either focal (in 2.6% of endometrial cancers belonging to cluster 2–4) or encompassed the whole-chromosome arm at 1q (34.0%; Supplementary Fig. S3A). Although only a minority of endometrial cancers contained an isolated focal 1q32.1 amplification event, an integrated analysis across different 1q regions revealed that the most significant effects were obtained when focal 1q32.1 events were combined with 1q whole-arm amplifications and considered as one variable (Supplementary Fig. S3B and S3C). Focal amplifications affecting other locations of the 1q arm, e.g., 1q21.1, 1q31.1 and 1q44 (q < 0.25) were also detected, albeit at an even lower frequency (1.5%, 1.5%, and 1.9%, respectively) than 1q32.1 focal amplifications.

Next, the prognostic significance of 1q32.1 amplifications in clusters 2 to 4, which mainly consisted of endometrial cancers classified as NSMP and to a lesser extent as MSI, was explored in these molecular subgroups separately. We considered 1q32.1 to be amplified when this chromosomal region was either focally amplified or part of a whole-arm amplification involving the 1q whole-arm. A multivariable Cox regression analysis confirmed the prognostic value of 1q32.1 amplifications in the NSMP subgroup for RFS (HR, 2.12; 95% CI, 1.26–3.59; P = 0.005) and CSS (HR, 4.02; 95% CI, 1.01–16.0; P = 0.049; Fig. 3A and B), whereas in the MSI subgroup there was no significant correlation between 1q32.1 amplifications and RFS (HR, 1.49; 95% CI, 0.83–2.67; P = 0.18) or CSS (HR, 2.03; 95% CI, 0.79–5.26; P = 0.14; Supplementary Table S3). Similarly, 1q32.1 amplifications were not prognostic in the TP53-mutant subgroup (HR, 0.98; 95% CI, 0.56–1.75; P, 0.96 for RFS and HR, 1.08; 95% CI, 0.51–2.32; P = 0.83 for CSS), whereas POLE-mutant tumors did not develop a relapse. Amplifications in the 1q32.1 locus were found to be either focal (2.3%) or affected the 1q whole-arm (24.1%) in the NSMP subgroup. In the MSI subgroup, however, no focal 1q32.1 amplifications were detected.

Two patient cohorts from the PORTEC-1 and -2 randomized trials (n = 840, Table 1) were used to confirm the prognostic effect of 1q32.1 amplifications in NSMP tumors (n = 487). Overall, 1q32.1 amplifications were frequently identified in endometrial cancers from NSMP (13.3%) and MSI (23.1%) subgroups (Supplementary Tables S4 and S5), albeit at a lower frequency than in the discovery cohort.

Similar to the discovery cohort, the prognostic significance of 1q32.1 amplifications for CSS, RFS, and DR was only observed in endometrial cancers classified within the NSMP subgroup. In the MSI and TP53-mutant subgroups, 1q32.1 amplifications were not prognostic (univariable analysis; Supplementary Table S6). 1q32.1 amplifications were also significant for RFS and DR in a multivariable analysis within the NSMP subgroup (HR, 2.34; 95% CI, 1.17–4.70; P = 0.017 and HR, 3.98; 95% CI, 1.72–9.17; P = 0.001, respectively; Fig. 3C and D). Although focal events were quite rare (0.1%), the prognostic effect of 1q32.1 amplifications was more significant and pronounced than for 1q whole-arm aberrations alone (Fig. 3E and F). Interestingly, stratification of NSMP tumors based on 1q32.1 copy-number status (amplified vs. nonamplified) revealed significant survival effects for RFS and DR. A new molecular classification stratifying NSMP endometrial cancers based on 1q32.1 copy-number status thus improved risk prediction of recurrence (Fig. 4; Supplementary Table S7).

Because 1q32.1 amplifications correlated with worse prognosis in two independent cohorts, we explored whether this region contained potential cancer driver genes. The 1q32.1 region identified by GISTIC had a length of 3.0 Mb (chr1:201175160-204162524) and included 48 genes. Among these genes, MDM4, ELK4 and SLC45A3 have previously been recognized as cancer consensus genes by COSMIC (15).

Differential expression analysis comparing tumors with increased versus normal copy-number in the 1q32.1 locus revealed 19 genes that were significantly overexpressed (FDR < 0.05; Supplementary Table S8). These included the above-mentioned cancer consensus genes MDM4 and SLC45A3. MDM4 showed the highest correlation between 1q32.1 copy number and gene expression compared with other genes in the 1q32.1 region (FDR = 2.51 × 10⁻¹², Spearman R² = 0.69; Supplementary Table S9). Although 1q32.1 amplification strongly correlated with increased MDM4 mRNA abundance, no correlation (P = 0.971) was found with protein expression using immunohistochemistry (n = 48; Supplementary Fig. S4). However, we then calculated the PSI index of MDM4, which represents the ratio of full-length MDM4 (MDM4fl) expression over the sum of MDM4fl and MDM4 short (MDM4s) expression. This ratio is known to correlate with increased MDM4fl protein expression, as detected by Western blotting, when PSI ≥ 0.4 (11). We found that more 1q32.1 amplified NSMP endometrial cancers had a PSI-index ≥ 0.4 compared with nonamplified NSMP endometrial cancers (65% vs. 49%; Supplementary Fig. S5), suggesting that MDM4 protein expression is increased in 1q32.1-amplified endometrial cancers.

To test whether MDM4 is a potential cancer driver of the 1q32.1 region, we explored whether endometrial cancers with 1q32.1 amplifications rely on MDM4 for their growth and/or survival. Unfortunately, among the 36 endometrial cancer cell lines available from various resources (Supplementary Table S10), we only identified cell lines belonging to the MSI or TP53-mutant subgroup. Because 1q32.1 amplifications are slightly less frequent in the TP53-mutant subgroup (Supplementary Table S11) and because TP53 protein expression was not altered in NSMP tumors with or without 1q32.1 amplification (Supplementary Fig. S6), we focused on endometrial cancer cell lines of the MSI subtype. Out of the four endometrial cancer cell lines from which SCNA profiles were obtained (Fig. 5A), two cell lines harboring the 1q32.1 amplification, that is, RL95-2 and COLO684, exhibited increased MDM4 protein levels compared with the other two cell lines without 1q32.1 amplification (HEC-1A and JVE1; Fig. 5B). Inhibition of MDM4, using a recently described antisense oligonucleotide-based strategy that promotes exon 6 skipping (11), resulted in significant growth inhibition of RL95-2 and COLO684, but not HEC-1A and JVE1 (Fig. 5C). qRT-PCR assays confirmed efficient exon 6 skipping and decreased levels of the protein-coding MDM4 full-length transcript at the expense of the nonsense-mediated decay MDM4-short transcript (Fig. 5D). A decrease in MDM4 protein levels was also confirmed by Western blotting (Fig. 5B). Importantly, inhibition of MDM4 was accompanied by an increase in apoptotic cell death, as illustrated by an increase in cleaved caspase-3 levels (Fig. 5E). In contrast, overexpression of MDM4 in HEC-1A and JVE1 endometrial cancer cell lines, which lack the 1q32.1 amplification, resulted in significantly increased cell growth (Supplementary Fig. S7). Overall, these data suggest that MDM4 is one of the putative oncogenic drivers within the 1q32.1 region and also represents a putative therapeutic target for endometrial cancers.

In this comprehensive study, we analyzed SCNAs at a genome-wide scale in 502 endometrial cancers and clustered these into five SCNA-based consensus clusters. Survival analysis of clusters 2, 3, and 4, which mainly consisted of NSMP tumors, revealed 1q32.1 amplification as a SCNA underlying worse prognosis of NSMP endometrial cancers. Indeed, both in the discovery and validation cohort, 1q32.1 amplifications were associated with a significantly worse prognosis in NSMP endometrial cancers, whereas no prognostic effect was found in the MSI subgroup or any other subgroup. Integration of 1q32.1 copy-number status in the molecular classification of endometrial cancer also resulted in stronger recurrence risk assessment. MDM4 mRNA expression showed the highest correlation with 1q32.1 copy-number status and 1q32.1 amplified tumors had a higher PSI index for MDM4, suggesting that also MDM4 protein expression was increased. Inhibition of MDM4 resulted in reduced cell growth in cell lines with an 1q32.1 amplification, whereas vice versa increasing MDM4 expression in nonamplified endometrial cancer cell lines accelerated cell proliferation. This suggests that MDM4 is one of the oncogenic drivers of the 1q32.1 region and represents a putative drug target for NSMP endometrial cancers.

Amplifications affecting the 1q32.1 region occurred more frequently in NSMP endometrial cancers from the discovery than from the validation cohort (26.4% vs. 13.3%). This can be explained by clinical differences between both cohorts: the discovery cohort included all stage and grade endometrial cancers, whereas the validation cohort contained only low-grade and low-stage endometrial cancers. Because tumors with higher grade and stage are known to contain relatively more SCNAs, this might explain the higher frequency of 1q32.1 amplifications in the discovery cohort. Interestingly, 1q32.1 focal amplifications frequently cooccurred with 1q whole-arm amplifications. Two studies previously described (focal) 1q amplifications as a prognostic factor for endometrial cancer, albeit in a much smaller number of tumors (n = 80 and n = 54) without stratification for molecular subgroups (16, 17). In addition, several other groups described a high frequency of (focal) 1q amplifications in endometrial cancers (6, 18, 19), as well as in several other malignancies such as breast cancer (20), multiple myeloma (21), and Wilms' tumors (22). In several of these other malignancies, 1q whole-arm amplifications also correlated with worse survival. The strong correlation between 1q32.1 copy-number and MDM4 mRNA expression suggests that MDM4 may be responsible for the more aggressive clinical behavior of endometrial cancer with 1q amplifications. However, the 1q32.1 region is large and may contain several other driver or passenger genes, such as ELK4, SLC45A3, and PIK3C2B, that could individually or in concert also contribute to the more aggressive clinical behavior observed in NSMP endometrial cancers with 1q32.1 amplifications (15). For instance, PIK3C2B encodes for a PI3K isoform known to contribute to tumor growth (23), ELK4 is an androgen-regulated gene promoting cell growth in prostate cancer (24), and the chimeric transcript ELK4-SLC45A3 is a common event in prostate cancer (25). Besides 1q32.1, other focal amplifications located in the 1q arm can also be associated with tumorigenesis: 1q21.1 (containing NBPF23; ref. 26), 1q21.2 (MCL1; ref. 27), and 1q43 (EXO1; ref. 28). Additional studies will therefore be necessary to unravel how this locus in addition to MDM4 contributes to the more aggressive behavior of NSMP endometrial cancers with an 1q32.1 amplification. An intriguing question is also why 1q32.1 amplifications are prognostic in NSMP endometrial cancers, but not in MSI endometrial cancers. Possibly, the high mutation rate of MSI tumors renders these tumors less dependent on 1q32.1 amplifications.

Nevertheless, the identification of MDM4 as one of the potential driver genes within the 1q32.1 region may lead to new targeted therapy options for patients with endometrial cancer. MDM4 is a well-characterized negative regulator of TP53, and as such, several MDM4-targeted therapies, including small molecule inhibitors are currently tested in clinical trials (reviewed in refs. 29–31). Most of these MDM4 inhibitors act by reactivating the TP53 pathway and reinforcing TP53-dependent tumor suppression. However, we failed to find an association between 1q32.1 amplifications and TP53 protein expression in the discovery and validation cohort. Although 1q32.1 amplifications were slightly less frequent in TP53-mutant than in NSMP or MSI endometrial cancers, both genetic aberrations were not mutual exclusive, suggesting that in endometrial cancer 1q32.1 amplifications act independently of TP53. Additional work is, however, still needed to better understand the mechanism underlying MDM4 inhibition in 1q32.1 amplified endometrial cancers.

Interestingly, a modified molecular classification stratifying NSMP endometrial cancers based on 1q32.1 copy-number status also improved recurrence risk prediction. This suggests that it would be relevant to test for 1q32.1 amplification as an additional marker to the clinically applicable molecular classification proposed by others (5, 8). The addition of a genetic test assessing 1q32.1 amplification could, however, become costly and labor-intensive as it requires different technologic requirements than the other markers. Therefore, the identification of a surrogate marker for 1q32.1 amplification that could be easily implemented would be very valuable. In an effort to find such marker, we performed MDM4 immunohistochemistry, but were unable to correlate protein expression with 1q32.1 copy-number status, at least not with the antibody tested. The lack of correlation between 1q32.1 copy-number and MDM4 expression by immunohistochemistry most likely is due to technical reasons because epitopes of MDM4 are not easily accessible when the protein is in its native state. In addition, up to seven transcripts of MDM4 exist, each with different stability and oncogenic potential (32), it is a possibility that antibodies that detect a specific MDM4 isoform are needed. As such, future studies may still identify such a surrogate marker for 1q32.1 SCNAs, but pending its availability we for now suggest to use targeted SCNA detection to more accurately classify endometrial cancers.

In conclusion, our study assessed the impact of SCNAs on the outcome of endometrial cancers, and more specifically, on a large subgroup of patients with endometrial cancer lacking a specific molecular profile. We show that 1q32.1 amplification is associated with reduced CSS and RFS, and that its addition to the existing classification improves recurrence risk prediction. We identified MDM4 as a plausible candidate gene responsible for the more aggressive behavior of this molecular subgroup and show that cancer cell lines with an 1q32.1 amplification could be inhibited by antisense oligonucleotide-based methods targeting MDM4.

No potential conflicts of interest were disclosed.

Conception and design: J. Depreeuw, E. Stelloo, M. Moisse, F. Amant, D. Lambrechts, T. Bosse

Development of methodology: J. Depreeuw, E. Stelloo, M. Moisse, M. Dewaele, K. Willekens, X. Matias-Guiu, F. Amant, D. Lambrechts, T. Bosse

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Stelloo, C.L. Creutzberg, R.A. Nout, D.A. Garcia-Dios, K. Willekens, X. Matias-Guiu, F. Amant

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Depreeuw, E. Stelloo, M. Moisse, D.A. Garcia-Dios, X. Matias-Guiu, D. Lambrechts, T. Bosse

Writing, review, and/or revision of the manuscript: J. Depreeuw, E. Stelloo, C.L. Creutzberg, R.A. Nout, J.-C. Marine, X. Matias-Guiu, F. Amant, D. Lambrechts, T. Bosse

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Stelloo, E.M. Osse, R.A. Nout, K. Willekens, J.-C. Marine, F. Amant

Study supervision: C.L. Creutzberg, D. Lambrechts, T. Bosse

We would like to acknowledge J.D.H. van Eendenburg for establishing the patient-derived endometrial cancer cell line, and J. Oosting, S. van Veen, T. Van Brussel, E. Vanderheyden, G. Peuteman, and L. Boon for technical support. F. Amant is a senior scientist paid by the Research Fund Flanders (FWO).

This study was supported by a grant of the Belgian Foundation against Cancer (grant number G.0705.14) and by the Dutch Cancer Society (grant number KWF-UL2012-5719). In addition, we thank the Anticancer Fund through the Verelst Uterine Cancer Fund Leuven for scientific and financial support.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.