Examining transcriptional changes to DNA replication and repair factors over uveal melanoma subtypes

Comparatively uncontrolled replication carried out by highly evolutionarily conserved multiprotein complexes, is a process shared by all cancers. Although many other processes including immortality, epithelial to mesenchyme transition, telomere metabolism, metastasis etc. contribute to tumorigenesis, the summation of genomic alterations in a tumor must facilitate replication. Duplication in the transformed cell is achieved at the expense of decreased fidelity, making replication a focal point where heterogeneity is created that upon clonal selection leads to tumor expansion and survival. Several individual critical replication genes have been examined in The Cancer Genome Atlas (TCGA) efforts and auxiliary studies, with computational methods placing alterations into known pathways to identify potential targets for precision medicine. However, the behavior of replication factors as a group did not receive analogous systematic investigation, likely because they are thought of as being part of a process rather than a pathway.

TCGA has recently conducted an integrative analysis of uveal melanoma (UVM), and the data providing its molecular foundation are now available to the public as part of the “Rare Tumor Project” [1]. While having a low incidence of 5.1 per million [2] it is the most common intraocular malignancy. UVM is highly lethal, with 50% of patients developing metastatic disease followed by 6–12 month survival from metastatic diagnosis [3]. Datasets for whole exome sequencing (WES), whole genome sequencing (WGS), mRNA miRNA and lncRNA expression, DNA methylation, identification of immune infiltration, and detailed pathology with clinical outcome were generated. Using these data, the current study examines the status of small groups of genes with known biology in replication and repair at the genomic and transcriptional levels in UVM. Selected components from the Pre-replication, Pre-initiation, Replisome, DNA Damage Repair (DDR), and Mismatch Repair (MMR) complexes have been investigated for genomic and transcriptional changes across UVM clusters classified by TCGA.

Alterations indicative of replication stress (RS) correlating with aneuploidy, increased malignancy, and decreased survival have been observed, as well as changes to common replication components involved in bypass of replication stress. Generally RS is thought to be an early and strong driving force in tumorigenesis, and is seen over a broad spectrum of cancer types. It is defined as impediment to the replication fork that causes slowing or stalling of the replication machinery and is brought on by “exogenous” and “endogenous” factors [4]. Examples of exogenous causes are radiation, therapeutic treatment, and diet. Endogenous causes include nucleotide pool availability, DNA structures, protein-DNA complexes, reactive oxidation species, transcription and replication complex collision, mutation and expression alteration in tumor suppressor and oncogenes.

This study specifically identifies changes in expression across UVM subtypes to MCM2, MCM4, MCM5, CDC45, MCM10, CIZ1, PCNA, FEN1, LIG1, POLD1, POLE, HUS1, CHECK1, ATRIP, MLH3, and MSH6. Resulting implications for bio-marker and therapeutics are discussed, and a rationale given for the observed alterations.

Clinical treatments for UVM fall under the broad categories of radiation, surgery, laser, and novel therapies for primary disease [3]. Despite measures, metastatic tumor cells persist in up to half of patients leading to death. It is probable that until therapy can halt the procession of the replication complex in a non-harmful tumor specific manner, that there will always be the potential for the selection of persistent cells. The rationale for focusing on replication and repair genes is to identify changes that directly account for the persistent tumor cells that implicitly make appropriate targets. The replisome is inherently highly processional and is kept under control by complexes of proteins that determine where, when, and how it functions in normal development, homeostasis, and in the tumor phenotype. This study suggests the replisome is increasingly unfettered by the inactivation of controlling complexes such as the origin of replication proteins, DDR, and MMR complexes as UVM progresses in malignancy. Schematic diagrams are presented (Figs. 6, 7 and 8) identifying selected genes of those complexes examined, and how they are thought to function normally.

Expression alteration to genes that create RS was found in untreated UVM, clearly indicating therapeutic radiation was not responsible for its initiation. On the basis of high frequency, the earliest mutations in UVM are thought to be to the guanine nucleotide binding proteins (GTPases) GNAQ and GNA11. It is a plausible hypothesis that alteration to these genes would directly or indirectly result in alterations to nucleotide pool availability in the nucleus, providing an early and endogenous cause of RS. Thought to be an early and driving event in cancer generally, there has been suggestion RS might be therapeutically targetable [23, 24].

This study specifically found changes to the MCM helicase complex. MCM2–7 components effectually act to “license” the firing of origins of replication during G1-phase of the cell cycle, determining the multiple nascent origins that replicate during S phase. A recent comprehensive review has been made by Riera et al. [25]. The complex is thought to load in excess, with dormant origins becoming active when nearby replication forks have stalled functionally serving as backup. Mouse models deficient in MCM components have high rates of cancer and genome instability [26–30], which supports the concept of relative deficiency of MCM factors in the more malignant UVM subtypes. Decrease in MCM2 is thought to reduce replication initiation in gene rich early replicating regions, and to increase genetic damage at a subset of gene rich locations [31] without changing the replication rate. Another putative function for MCM2 due to its histone binding and chaperone capacity may be to orchestrate histone dynamics during replication [32], and it has also been linked conceptually to transcription [33]. UVM appear to have MCM concentration alterations including decreased MCM2, increased MCM4 and decreased MCM5 that correlate with increased malignancy and decreased survival. Overexpression of MCM4 has been found previously in multiple transformed cell lines and numerous tumor types [34, 35]. Multiple kinase signaling pathways are known to target MCM4. It is thought to mediate the repression of late origin firing and provide an intrinsic regulatory domain for signal integration coordinating origin activation and replication fork progression under RS [36]. The unbalanced expression of MCM2–7 individual components likely affects the quantitative amount of functional helicase available, lowering the ability to load in excess and contributing to unregulated replication.

Once the pre-replication complex is augmented with proteins CDC45 and GINS1–4 it is designated as a “pre-initiation complex”, CIZ1 (also known as Cip1-interacting zinc finger protein 1) is thought to function in this context [20]. This interesting gene plays a role in cell cycle control and replication by interaction with Cip1/Waf1 [15], and coordinating cyclin E and A functions in the nucleus [17]. It is required for cells to enter into S phase [20]. CIZ1 physically tethers to non-chromatin nuclear structures in the nuclear matrix, and co-localizes in immune-histochemical studies with PCNA [15]. Two functionally distinct domains exist, an N-terminal domain involved in replication and a C-terminal domain comprising the nuclear matrix anchor [16]. Targeted depletion of transcripts decreases proliferation in vitro [15]. Differential expression plots show the gene is most highly expressed in UVM (Fig. 4). Relative expression across the TCGA SCNA subtypes shows clusters 1, 2, and 3 have CIZ1 expression above the alteration level and cases below are primarily in subtype 4. CIZ1 must be available in enough quantity in subtype 4 to provide the replication function. Relative paucity and alternative splicing likely creates less than optimum conditions for nuclear anchor functioning. Almost all UVM cases examined show evidence of exon 4 alteration previously reported in other cancers as well as alteration to the carboxyl end of the protein. Variant mRNA created by alternative splicing are found in cancers and other disorders that appear to alter protein localization [19]. Variant CIZ1 has recently been published as a circulating biomarker for early-stage lung cancer [22] suggesting the same might be possible for UVM patients.

The findings for CIZ1 led to identification a mononucleotide repeat in Intron 3 with possible MSI. Almost no evidence points to the involvement of MMR components as drivers for UVM, and examination of mutation rate does not show logarithmic elevation typical of MMR deficient tumors. Exon skipping was observed in MLH3 (MLH3 ∆ exon5/∆ exon7), a protein that interacts with MLH1 directly. MLH3 ∆ exon7 has been previously described [37] and in that study the isoform lacked MLH1 binding activity suggesting a mechanism for partial MMR defect in UVM. In addition to playing a role in MMR, MLH3 also has a meiotic phenotype. Evidence of MSI was observed in both tumor tissue and its normal counterpart. Mononucleotide tracks are difficult to sequence and the frequency of MSI found is implausibly high, however, some examples genuinely appear have alterations in the germline that also appear in the tumor. Interestingly, the differential expression plot across tumor types also shows CIZ1 expression is not much different between tumors and their normal tissues quantitatively (Fig. 4), unlike MCM2 for example (Fig. 2) supporting the concept of possible predisposition.

Reviewing relative expression for PCNA (Proliferating nuclear cell antigen), FEN1 (Flap endonuclease 1), and LIG1 (Ligase 1) an increase was found that implied their involvement in overcoming RS. These components of the replisome lacked mutation and SCNA but correlated none-the less with increased expression suggesting transcriptional mechanisms were used to overcome RS and fork collapse in UVM progression. PCNA is a ring shaped homo-trimer that encircles DNA [38]. It interacts competitively with many other factors at the PIP motif [39, 40] and is an essential co-factor for DNA polymerases during replication. It is involved in repair processes and can also be modified post-translationally by phosphorylation, ubiquitination, SUMO proteins, ISGylation, Acetylation, and S-nitrosylation. Each modification has corresponding biological functions that include proliferation, MMR inhibition, translesion synthesis, homologous recombination, genomic stability, and apoptosis, respectively. One of its major roles is to promote tolerance to DNA damage during replication [41]. Because of its role in proliferation, PCNA is a target for cancer therapy. Several small molecules that either block PCNA-Polδ or PCNA trimer formation have been identified with proliferation-inhibitory effects in vitro [38, 42–44].

Two polymerases, POLD1 and POLE, important in the replication of B form DNA were selected for observation. The relative decrease in POLD1 expression in subtypes 3 and 4 and increase in POLE support a recently proposed model [45] in which a switch to POL ɛ and away from POL δ occurs upon DNA damage. The behavior of POL α and specialized polymerases recently hypothesized to assist the replication machinery in the prevention of replication stress [4] has not yet been examined.

Replication stress activates DDR which prevents DNA damage from becoming fixed during replication and passed on in mitosis [46]. It does this in part by coordinating cell cycle control and providing a pause for repair, and in some cases by triggering apoptosis. Dysregulated DDR in a tumor cell contributes to progressive genomic instability that is cancer enabling, and can make the tumor dependent on alternative repair pathways such as base excision repair (BER) that may be targetable (e.g. Poly ADP ribose Polymerase (PARP) inhibitors). In this study we find evidence of dysregulated DDR in the more malignant subtypes. Examining individual components of DDR revealed a significant decrease in relative expression of ATRIP (ATR Interacting Protein) in clusters 3 and 4 [47–51]. PARADIGM and MARINa algorithms previously used to examine the pathway, found protein abundance for DDR components increased suggesting DDR was active in monosomy 3/BAP1 impaired UVM [1]. ATRIP is located at 3p21.31, very close to BAP1 at 3p21.1. Interestingly as mentioned in the TCGA study [1], recent reports suggest that BAP1 itself may function in DDR [52, 53].

In summary, this study examines selected replication and repair factors in UVM with known biology for expression differences with and without mutation across TCGA SCNA subtypes. Differences in expression have been observed that support the concept that these genes and their products are causally involved in UVM. These data suggest further avenues of research for biomarker identification and therapeutic approach.