Comprehensive Analysis of the Expression and Prognosis for the DREAM Complex in Human Cancers

The DREAM complex is an evolutionarily conserved cell cycle regulating multi-protein complex. In addition to playing an essential function in the cell cycle, it also plays a vital role in various survival activities. Accumulating evidence suggests that the DREAM complex plays a crucial role in oncogenesis. However, the regulatory mechanism of the DREAM complex in cancer remains unclear. This study used multi-omics data from Cancer Genome Atlas and Cancer Cell Line Encyclopedia to comprehensively identify the DREAM complex in tumor samples from 33 cancer types. In the genomic landscape, we identified the missense mutation as the dominant alteration events. Expression analysis showed that the expression of methylation-mediated the DREAM complex was downregulated. In addition, we found that the expression of the DREAM complex can be performed to predict the survival of various cancer patients. Pathway activation analysis showed that the DREAM complex is related to apoptosis inhibition, cell cycle, DNA damage response, RAS/MAPK, and RTK signaling pathway activation. Importantly, through a comprehensive analysis of drug sensitivity genomics in cancer databases, we identified a number of potential drugs that may target the DREAM complex. In summary, this study revealed the genomic changes and clinical features of the DREAM complex in 33 cancers, which may also provide new insights for cancer treatment and may offer alternative options for the treatment of clinically refractory cancers.


INTRODUCTION
The DREAM complex, also named the LINC complex, is an evolutionarily conserved cell cycle regulation polyprotein complex (Lewis et al., 2004;Harrison et al., 2006;Litovchick et al., 2007;Schmit et al., 2007). The DREAM complex, associated with p130 (retinoblastoma-like protein 2/ RBL2), E2F4, and DP1 (E2F dimerization partner 1), exists on the promoters of about 1,000 cell cycle-related genes, inhibiting their transcription and inducing cell silencing (Litovchick et al., 2007;Sadasivam and DeCaprio, 2013;Gal et al., 2021). The DREAM complex including a retinoblastoma family member, an E2F transcription factor and its dimerization partner, and five proteins related to products of MuvB genes LIN-9, LIN -37, LIN -52, LIN -53, and LIN -54 (comprising the MuvB core) have been described in distinct organisms (Pilkinton et al., 2007;Sadasivam and DeCaprio, 2013;Walston et al., 2021). In different stages of the cell cycle, the composition of the core complex does not change. Still, it does interact with various proteins to change its function from transcription repressors of G0 and G1 to activators of S, G2, and M. In the G0 phase and the early G1 phase, the MuvB core interacts with E2F4/ 5, DP1/2, and RBL2 to inhibit the G1/S and G2/M genes to form the inhibitor DREAM. In the late stage of G1, DREAM-specific protein separates from the MuvB core and then binds to B-MYB in the S phase, which transactivates cell-cycle genes related to the S/G2/M phase (Korenjak et al., 2004;Litovchick et al., 2007;Sadasivam et al., 2012;Fischer et al., 2016).
The DREAM complex plays a vital role in orchestrating the cell cycle, deregulation of this complex is connected with several cancers (Sadasivam and DeCaprio, 2013;MacDonald et al., 2017). However, further research is still needed to elucidate the genomic and clinical characteristics of the DREAM complex in cancer.
In this study, we systematically assessed the genomics and clinical characteristics of the DREAM complex in 33 solid tumors. In addition, we studied the impact of these gene expression changes on clinical outcomes and drug sensitivity. We also explored possible downstream signaling pathways, which the DREAM complex may mediate.
mRNA Differential Expression Analysis mRNA Seq data and clinical data were collected from the TCGA database. In the mRNA differential expression analysis, normalized mRNA expression data based on RNA-seq [RNA-Seq V2 RSEM (RNA-Seq by Expectation-Maximization)] was obtained from the TCGA data portal. The number of samples for each cancer type ranged from 48 to 1,098. We performed the analyses across 14 cancer types with sufficient normal tissues (at least 10 samples with matching tumor adjacent normal tissue), including BLCA, BRCA, COAD, ESCA, HNSC, KICH, KIRC, KIRP, LIHC, LUAD, LUSC, PRAD, STAD, and THCA. The RSEM normalized mRNA expression values were used. Fold change is expressed by mean (Tumor)/mean (Normal); the p-value was determined using a t-test and was adjusted by the false discovery rate (FDR). Genes with fold change (F.C.) >2 and false discovery rate (FDR)-adjusted p-value < 0.05 were retained for further analysis. If there was no significant gene in a cancer type, then that cancer type will be excluded from the final analysis.

Survival Analysis
The mRNA expression data of the DREAM complex in 33 cancers were combined with the corresponding clinical survival data for expression survival analysis. The mRNA expression data of the DREAM complex in 33 cancers were combined with the related clinical survival data for expression survival analysis. According to the median RSEM value, the tumor specimens were divided into high group and low group. Then, we used the R package "survival" to estimate the two groups' survival time and survival status. The log-rank test was performed in a univariate Cox regression analysis using a proportional hazards model to examine the association between each gene and patient survival separately, and the determined log-rank p-values were adjusted after the Benjaminii-Hochberg FDR. Genes with adjusted log-rank p-values (also known as Q-values) <= 0.05 are preserved. (Nguyen and Le, 2020).

Subtype Analysis
Genes may have different expression levels in cancer subtypes. Therefore, we performed the analysis of gene expression in tumor subtypes. We analyzed subtypes in the following cancers: BRCA, KIRC, LUAD, STAD, HNSC, LUSC, BLCA, the number of subgroup of subtype must have at least 10 samples. The method used for clinical correlation analysis depends on the number of subgroups of the subtype: t-test and ANOVA t-test. The final results included genes and cancer types whose p-values were less than 0.05 after statistical testing.

Copy Number Variation Analysis
The raw data of CNVs of 33 cancer types were obtained from the TCGA database and processed using GISTICS2.0 (Mermel et al., 2011). In this CNV module, we calculated the percentage of CNVs in each cancer type, the correlation of CNVs with gene mRNAs. CNV can be divided into two subtypes, namely heterozygous and homozygous, that is, CNV occurs on only one chromosome or on both chromosomes at the same time. The CNV data processed by GISTIC was used for percentage statistics based on CNV subtypes, and the correlation was calculated using the original CNV data and mRNA RSEM data. We only discussed the genes with 5% CNV in cancer. According to the method employed by Schlattl et al. (2011), the mRNA expression and CNV raw data were merged according to their TCGA barcode. We tested the association between paired mRNA expression and paired CNV percent samples on Person's product-moment correlation coefficient, followed by a T distribution. The FDR adjusted the p-value.

Methylation Analysis
Methylation data were available in the TCGA database. Only 14 cancer types had matching tumor-adjacent normal data. Therefore, differential methylation analysis was based on these 14 cancer types. Cancer types that contain at least 10 paired TCGA normal samples were selected to be calculated, but only paired samples were incorporated. The methylation difference between tumor and normal samples was identified using Student's t-test at an FDR-adjusted p-value <= 0.05. In theory, methylation can lead to the abnormal expression of a gene. We matched methylation and mRNA expression data by patient header I.D. using the TCGA barcode. We further collected the genes whose gene methylation was significantly associated with expression, combined the methylation data with clinical overall survival data, and divided the genes' methylation level into two groups based on the middle methylation level. A high methylation risk rate was defined as high risk. Otherwise, it was described as low risk. Differences between groups were measured using a log-rank test, and two-tailed p-values of less than 0.05 were considered statistically significant.

Pathway Activity Analysis
Based on the reversed-phase protein array (RPPA) data in the TCPA database, the activity of the downstream signaling pathway was analyzed. According to the method described by Rehan et al. (Akbani et al., 2014). Pathway activity score (PAS) was estimated by the approach adopted by Akbani et al. and Ye et al. (2018). According to the median expression level, the gene expression level was divided into high and Low. The Student's test determined the PAS difference between the groups. The p-value was adjusted for the FDR, and p < 0.05 was considered to indicate significant enrichment. When PAS (A high gene group)>PAS (A low gene group), we assessed gene A might have an activating effect on this pathway; otherwise, it had an inhibitory effect on this pathway.

Drug Sensitivity Analysis
The small molecules were obtained from The Genomics of Drug Sensitivity in Cancer (GDSC) database and The Cancer Therapeutics Response Portal (CTRP). According to Rees et al., we analyzed the correlation between gene expression and drug sensitivity (Basu et al., 2013;Seashore-Ludlow et al., 2015;Rees et al., 2016). To explore the correlation of the DREAM complex and drug sensitivity, the Pearson correlation coefficients of transcript levels and AUCs were used and normalized based on Fisher's Z transformation. Gene set resistance analysis was performed on GDSC IC50 drug data. Spearman correlation coefficient is used to indicate whether gene expression is related to drug sensitivity.

Statistical Analysis
The Student's t-test was used to compare the differences in the DREAM complex expression and the differences in methylation between tumors and corresponding normal tissues. The Cox proportional hazard model was used to calculate survival risk, and H.R. The Kaplan-Meier survival curve was used to estimate the prognostic significance of each variable. The log-rank test was used for comparison. Unless otherwise specified, the rank-sum test detects two sets of data, and a p-value < 0.05 is considered statistically significant.

Gene Set Expression and Subtype Analysis of the DREAM Complex
Based on the GTEx data set, we explored the differential expression of the DREAM complex in normal tissues. As shown in Figure 1A, the expression of TFDP1 in the Blood, Skin, and Testis, the expression of RBL2 in Bladder, Cervix Uteri and the Uterus, the expression of RBBP4 in the Testis, the expression of MYBL2 in Blood, Spleen and the Testis, and the expression of E2F4 in Spleen, Uterus, and Pituitary were significantly upregulated. At the same time, we found that the DREAM complex was abnormally expressed in 14 solid tumors. (p < 0.05, Figure 1B). The expression levels of RBL2 and LIN52 were significantly downregulated in numerous cancers. The expression levels of MYBL2, RBL1, LIN9, and E2F5 in various cancers were significantly upregulated. However, there was no significant difference in the expression of the DREAM complex in THCA. In addition, to identify clinically relevant genes that affect tumor subtypes, we evaluated the expression of subtypes of the DREAM complex in cancers. We found that LIN9, MYBL1, and MYBL2 in BRCA; RBL1, MYBL2, and RBL2 in KIRC; RBL1, MYBL2 and RBL2 in LUAD; LIN9, LIN54, MYBL2, and TFDP1 in STAD; LIN9 and RBBP4 in HNSC; LIN9, RBBP4, and LIN54 in LUSC; LIN9 and MYBL1, RBL1 and LIN52 in BLCA were significantly differentially expressed (p < 0.05, Figure 1C). Expression survival analysis showed that MYBL2, which showed increased expression in KIRC, LGG, KICH, ACC, MESO, SARC, ESCA, KIRP, LIHC, and LUAD; MYBL1 in KIRC, LGG, KICH, ACC, MESO, KIRP, and PAAD; LIN9 in LGG, KICH, and ACC; LIN37 in KIRC and MESO; was related to poor survival. The genes with decreased expression, TFDP1, RBL2, LIN52, and LIN54 in KIRC; TFDP1 in LGG; RBL2 and E2F4 in KICH; RBL2 in ACC; RBBP4, LIN52 and E2F5 in GBM; RBL1 and TFDP2 in READ; and TFDP2 in UVM, were associated with poor survival (p < 0.05, Supplementary Figure S1A). We compared the DREAM complex transcription levels through the Oncomine database (https://www.oncomine.org/resource/login. html, an online cancer microarray database) in cancer and normal samples. Compared with the normal samples, the expression level of the DREAM complex mRNA in cancer was significantly upregulated (Supplementary Figure S1B). Survival analysis of six significantly differential genes showed that high expression of E2F4, LIN9, MYBL1, MYBL2, RBL1, and TFDP1 were associated with poor overall survival in the GEPIA2 database (p < 0.05, Supplementary Figure S1C). The above results indicated that the abnormal expression of the DREAM complex might was involved in the occurrence of tumors.

Somatic Mutations of the DREAM Complex
To understand and compare the frequency and variant types of the DREAM complex in each cancer subtype, we analyzed data from The Cancer Genome Atlas Network (TCGA) studies. As shown in Figure 2A, SNV frequencies of UCEC, SKCM, and COAD were 1%-43% in these cancers. The SNV frequency of the DREAM complex was 93.02% (600 of 645 tumors). SNV analysis indicated that the main frequent variant classification of the DREAM complex gene mutation was missense mutation. SNV percentage analysis demonstrated that the top 10 mutated genes were RBL1, RBL2, MYBL2, MYBL1, TFDP1, LIN9, TFDP2, LIN54, E2F5, and RBBP4, of which the mutation percentages were 23%, 20%, 14%, 14%, 13%, 13%, 11%, 11%, 9%, and 8%, respectively. The SNV frequency of the DREAM complex was increased in UCEC, SKCM, KIRC, and LUSC ( Figure 2B).

Copy Number Variation of the DREAM Complex
To identify alteration in CNV, we analyzed the CNV data of the DREAM complex in the TCGA database. The CNV pie chart distribution showed that these genes' main copy number variants include heterozygous amplification and deletion.  Figure S2A). In addition, we also found that mRNA expression is positively correlated with CNV, especially E2F4 and TFDP1 in BRCA, and RBBP4 in LGG (p < 0.05, Supplementary Figure S2B). These results indicated that the DREAM complex had heterozygous amplification and heterozygous deletion, which mediated its abnormal expression and might play an essential role in the occurrence and development of tumors.

Methylation Analysis of DREAM Complex
We explored the methylation of the DREAM complex to determine epigenetic regulation. The methylation of the DREAM complex in different tumors was highly heterogeneous. There were more hypermethylated genes than hypomethylated genes in KIRC, PRAD, KIRP, COAD, PAAD, HNSC, UCEC, LUAD, BRCA, LIHC, and LUHC. Whereas there were more hypomethylated than hypermethylated genes in BLCA. The MYBL2, MYBL1, LIN54, E2F4, RBBP4, LIN37, RBL2, LIN52, E2F5, and TFDP1 were hypermethylated in most cancers (p < 0.05). RBL1 was hypomethylated in LUSC, BRCA, LUAD, BLCA, and KIRC (p < 0.05). LIN9 was hypomethylated in LIHC, HNSC, PAAD, and COAD. TFDP2 was hypomethylated in KIRC, PRAD, and BRCA ( Figure 4A). We then analyzed the correlation between methylation and mRNA expression, and the results showed that the expression level of most genes was negatively correlated with the methylation level. Only the methylation of TFDP2 in SKCM and CESC was positively associated with gene expression (p < 0.05, Figure 4B). Survival analysis showed that hypermethylation of TFDP2, RBL1, and MYBL2 was associated with poor survival in most cancers. In addition, the hypomethylation of RBL2, MYBL1, TFDP1, and LIN37 was mainly related to poor survival (p < 0.05, Figure 4C).
The above analysis showed that abnormal DNA methylation may regulate the odd expression of the DREAM complex and was related to the prognosis of cancer patients.

Drug Sensitivity Analysis
To elucidate the effect of the DREAM complex on the therapeutic effect of chemotherapeutics, we integrated the drug sensitivity and gene expression profile data of tumor cell lines provided by GDSC and CTRP. Spearman's correlation analysis showed that drug sensitivity toward DMOG (Dimethyloxalylglcine) correlated with the expression of MYBL1 (negative correlation with IC50). However, drug resistance toward Trametinib, CI-1040 associated with the expression of E2F5, TFDP2, LIN52, LIN9, and MYBL2 (positive correlation with IC50) (p < 0.05, Figure 6A). Drug resistance toward all drugs in Figure 6B correlated with the expression of MYBL1 (positive correlation with IC50). Drug sensitivity toward all drugs in Figure 6B correlated with the expression of RBL2, TFDP2, TFDP1, RBL1, RBBP4, MYBL2, LIN9, LIN54, LIN52, E2F5, and E2F4 (negative correlation with IC50). These results indicated that the abnormal expression of the DREAM complex might mediate resistance to chemotherapy and targeted drug therapy.

DISCUSSION
In the cell cycle process, the coordinated expression of periodic related genes can maintain the integrity of the genome and ensure cell proliferation so as not to be blocked by environmental signals. The DREAM complex, also known as the LINC complex, is an evolutionary conserved cell cycle regulatory protein complex. More and more evidence shows that the DREAM complex has a potential inhibitory effect in tumorigenesis (Ho et al., 2009;Kim et al., 2021). In different organisms, the destruction of the DREAM complex can lead to developmental defects, genomic instability, tumorigenesis, and increased mortality (Malumbres and Barbacid, 2009;Reichert et al., 2010;Hauser et al., 2012). Therefore, studying the related mechanisms of the DREAM complex in cancer is necessary to understand tumorigenesis and explore potential targets for clinical treatment. We have performed a comprehensive and systematic characterization of the DREAM complex in multiple samples of 33 cancers by mining multiple sets of analysis data. Our research results reveal a variety of potential mechanisms of the DREAM complex in the cancer environment and reveal common signal pathways related to cancer pathways, thereby elucidating the overall regulation of the DREAM complex in cancer.
Our genetic analysis shows that the copy number of the DREAM complex gene changes frequently. We also found that mRNA expression is positively related to CNV, especially E2F4 and TFDP1 in BRCA and RBBP4 in LGG. The expression analysis of the DREAM complex confirms that the change of copy number is positively correlated with the expression, indicating that the shift in copy number can affect the expression of clock genes, thereby promoting the occurrence of tumors. Recent studies have also shown that MYBL2 is significantly upregulated in endometrial cancer (E.C.) and esophageal squamous cell carcinoma (ESCC), accompanied by significant copy number changes (CNA). Copy number amplification strengthens the expression of MYBL2 mRNA, leading to poor prognosis and severe E.C. pathology (Qin et al., 2016;Le et al., 2021).
Epigenetic analysis showed that the methylation of the DREAM complex in different tumors is highly heterogeneous. The MYBL2, MYBL1, LIN54, E2F4, RBBP4, LIN37, RBL2, LIN52, E2F5, and TFDP1 were hypermethylated in most cancers. At the same time, RBL1 was hypomethylated in LUSC, BRCA, LUAD, BLCA, and KIRC. The correlation between methylation and mRNA expression showed that the expression level of most genes was negatively correlated with the methylation level. Meanwhile, survival analysis showed that hypermethylation of TFDP2, RBL1, and MYBL2 was associated with poor survival in most cancers. A new DNA methylation 10-CPG disease-free survival prognosis shows that MYBL2 is associated with a high risk of prostate cancer (Hou et al., 2020). Homomorphic diffuse gliomas occur in children and adults and have concise morphologies with frequent MYBL1 and MYB changes and specific DNA methylation features (Wefers et al., 2020). Therefore, we speculated that the genetic and epigenetic modifications of the DREAM complex might promote tumors in some cases. The DREAM complex gene inhibits cell apoptosis, cell cycle, and DNA damage response, which have been confirmed in several cancers (Musa et al., 2017;Iness and Litovchick, 2018;Bayley et al., 2020;Kim et al., 2021;Liu et al., 2021). These results indicate that the DREAM complex plays a crucial role in regulating cancer-related pathways and is consistent with our analysis results. In addition, our results show that MYBL2 and MYBL1 genes are also involved in the activation of the EMT pathway to promote tumor metastasis, which has also been confirmed in some tumors (Togashi et al., 2018;Xie et al., 2020;Li et al., 2021;Sakuma et al., 2021).
These results indicate that the DREAM complex constitutes an interactive network of tumor-related signaling pathways and may be involved in inhibiting various tumor progression and improving survival rates. Our drug sensitivity analysis showed that drug resistance toward Trametinib, CI-1040 was associated with the expression of E2F5, TFDP2, LIN52, LIN9, and MYBL2. However, these drugs' potential mechanism of action on the DREAM complex expression and tumor progression remains to be further studied.
In summary, we have elucidated the genomics and clinical characteristics of the DREAM complex in 33 solid tumors. Our research found that the ectopic expression of the DREAM complex mediated by genome changes is involved in the activation of cancer-related pathways. Targeting the DREAM complex may be a novel and meaningful way to treat cancer. Our findings are significant and provide new insights into the regulation of the DREAM complex in tumors. The DREAM complex has various regulatory levels such as genetic and epigenetic changes, mRNA expression, and pathway correlation. These differences may lead to differences in drug efficacy, treatment response, and patient survival. Although our research provides new insights into the changes in the DREAM complex regulation, this analysis still has some limitations. First of all, based on the information of TCGA, we can only assess the overall genetic changes of tumor tissues and cannot analyze the genetic differences of tumor cells. Second, we only explored some potential changes that may affect the expression and function of the DREAM complex.
Post-transcriptional and post-translational modifications, such as changes in mRNA splicing, m6A methylation, and protein stability, still require further consideration. Third, since our research is mainly based on gene expression, we cannot explore the role of CNV and SNV changes in tumors. Therefore, future research to comprehensively investigate cancer heterogeneity and individualized treatment are necessary.

AUTHOR CONTRIBUTIONS
XL conceived and designed the study and drafted the manuscript. LW prepared the data and revised the manuscript. All authors read and approved the final manuscript. The association between paired mRNA expression and CNV percentage in samples was based on a Person's product moment correlation coefficient. The size of the point represents the statistical significance, where the bigger the dot size, the higher the statistical significance. FDR, false discovery rate.