**3. Ovarian cancer and copy number alterations**

It seems self-evident that an understanding of "driver SCNAs" is absolutely essential to our capacity to target the biology of the disease. Genetic disorders such as Down's syndrome (trisomy 21) and Cri du Chat (5p monosomy) and DiGeorge Syndrome (loss of only 30–40 alleles on 22q11) clearly indicate the penetrative biology of multiple SCNA. More importantly, such lesions affect only ~2% of the genome, while SCNA in SOC affect 67% of genes. Other subtypes of ovarian cancer vary widely in their SCNA burden, but are typically much lower, and are associated with SNVs.

As most SCNA are "monoallelic" changes resulting in a 1N or 3N genotype, is there any reason to expect a phenotype, given our understanding of recessive alleles? Recurrent patterns in serous ovarian cancer suggest that frequently affected regions may be selected for as the tumor evolves. In high grade SOC, the most prevalent SNVs could have been predicted from literature preceding the genomics era. For decades, the mutation of *TP53,* the "guardian of the genome" has been appreciated due to its master control of multiple DNA repair pathways, cell cycle control, and metabolism. Interestingly, there is selection for SCNA deletion of the chromosome with the wild-type copy of p53, suggesting further suppression or misdirection of p53 furthers SOC development [14]. Inheritance studies have associated the *BRCA1/2* mutation with an increased risk of ovarian cancer, and not surprisingly these mutants contain opposite-chromosome deletions just like p53. *BRCA* genes are necessary to maintain the genome. Like p53, they play a coordinating role in facilitating homology-directed repair of DNA. However, single nucleotide variant mutation is not the most common mechanism of BRCA gene disruption. Only ~6% of patients display non-germline SNVs, while copy number deletions (to 1N) occur in more than 70% of tumors. PTEN, a tumor suppressor commonly mutated in many tumor types but not ovarian, was found as early as 2001 to have reduced expression due to shallow deletions across ~40% of samples [15].

Aside from very infrequent gene losses paired with mutations, there are also a few SCNAs which drive cancer through amplification of oncogenes. The stem-cell transcription factor *MYC* is the most amplified gene in the TCGA cohort (42% with at least a 4N copy number, and an additional 37% with 3N). Myc has been appreciated as a common SOC driver oncogene since 1990 [16]. Homozygous deletions in Rb were discovered around the same time [17, 18], and occur in 9% of tumors. *KRAS* amplifications and gene overexpression were discovered around the same time, but in a smaller minority (13%) of patients [19]. Her2, encoded by *ERBB2*, can be overexpressed but this appears to be a case unrelated to SCNA amplification, which occurs in only 3% of cases [20, 21]. Drug resistance can occur following increases in drug efflux genes, and one of the first identified was *MDR1* (*ABCB1*) [22]. Again, this is only found in a small minority of patients (4%). Comparative genomic hybridization in 2006 identified significantly amplified CCNE1 (cyclin E1) and MDM2 (a negative regulator of p53 from its E3 ubiquitin ligase activity) [23, 24]. The year these studies were published provides

**Figure 2.** Model showing how SCNA changes resulting in differences in protein expression might impact overall function within a single multiprotein complex. The relative function of the complex as a % is shown at left. The right side

fact, only about one third of all genes in primary tumors have a normal 2N gene dosage. Roughly a quarter of the total genes in the tumors show an extra gene copy (to 3N) and just over a third lose a gene copy (to 1N). By contrast, only 0.7% loses both gene copies (0N), while 4.2% are amplified to 4N or greater. In practice, the focus on understanding tumor biology has been only on these last two cases (total deletion and gross amplification, respectively). This has a reasonable basis; the effects of total loss or gross amplification are easiest to study. The common gene changes (i.e., 1N and 3N) have not been the subject of focused study. Many scientists assume that the deletion, or addition, of a single gene copy has limited effect. Recessive genetic alleles are not uncommon in nature, supporting the idea that the loss of a single gene copy can be compensated for. However, the loss of a single gene may not reflect the situation in ovarian cancer, where massive genetic alteration occurs, and compensation may not be possible if the same cellular pathway is repeatedly targeted by SCNAs (**Figure 2**). More than 80% of genes affected by SCNAs show concordant alteration of mRNA levels [9, 10]. For ~70% of genes, this correlates with steady-state protein levels [11]. Thus, SCNAs offer a predictable, but not absolute, indication of protein expression. This is relevant to ovarian cancer, as SCNAs modify on average 67% of the SOC genome, whereas SNVs modify only

114 Ovarian Cancer - From Pathogenesis to Treatment

shows how a sequence of SCNA changes within a pathway could cumulatively impact its function.

a historical context to our knowledge. Were we missing a key driver for SOC? Despite the hundreds of genomes sequenced, few additional single-gene drivers were discovered in the recent "brute-force" landmark studies on SOC from either the TCGA [12] or the AOC [13].

**4. The interplay of p53 mutation with copy number instability**

additional tumor suppressors and oncogenes, which leads to SOC as we know it.

Mre11/ATM-dependent DNA damage responses [44].

and acts permissively to enable SCNA accumulation.

**5. BRCA1/2 mutations and homologous repair defects**

Mechanistically this could occur via the deletion or duplication of entire chromosomes or genomes, followed by many subsequent changes enabled by the extra copies of genes, or via chromosome missegregation event, leading one or more chromosomes to acquire massive damage [41]. Either possibility can explain the high frequency of chromothripsis, a highlydisorganized form of hundreds or thousands of SCNAS, in SOC. Mutant p53 enables such mechanisms of SCNA formation by preventing the death of the cell that bears them, as missegregation directly induces p53-dependent cell-cycle arrest followed by apoptosis [42]. In one well-controlled study, 'dominant negative' p53 reduced the cell cycle delay associated with trisomy in mammalian cells, yet it was rare that gain of any single chromosome in those cells resulted in any proliferative advantage [43]. Thus, partial or gained p53 function may contribute. Many p53 mutations maintain partial function, while mutations such as R273H (the most common variant of TP53 found in SOC) provide a gain of function by directly impairing

It is likely that mutation in *TP53* gene is an enabling event. Lineage tracing using millions of sub-clonal passenger mutations present in SOC tumors suggest that *TP53* mutation arises very early in the proliferation of pre-tumor cells [14]. While few studies focus on normal tissue, a publication on aged skin samples found that islands of cells had developed p53 mutations and achieved local proliferation. Exceptionally few copy number alterations were revealed, with the exception of deletions in *NOTCH1*, and these lesions did not progress to malignancy [45]. In murine models, too, SCNAs follow initiating mutational stimuli [46]. The findings support the idea that p53 is likely to become mutated prior to SCNA accumulation,

Though few SNVs in 'classic' tumor genes are found in SOC relative to other cancer types, BRCA1/2 mutations are among the most frequent at ~10% [12]. BRCA genes work in

Mutation in p53 has a long research history in many cancer types, and ovarian cancer is no exception. Ovarian cancer mutations within *TP53* have been observed since 1991 [34], and have been confirmed in every genetic study since. *TP53* has often been referred to as "the" primary tumor suppressor for its central role in responding to stresses: it can halt the cell cycle, divert metabolism, induce transcription of DNA damage response genes, or if the damage cannot be repaired, induce apoptosis or senescence [35–37]. For serous ovarian cancer, it has been used as a marker for false diagnosis as some studies presume that genuine serous ovarian cancers must contain mutant p53. Similarly, since the beginning of genomic copy number studies using comparative genomic hybridization, SCNAs have been labeled as a ubiquitous event in all types of epithelial ovarian cancer [31]. Not all p53 mutant tumors are high in SCNAs [38]. Nonetheless, tumors with higher than average SCNAs are much more likely to have a facilitating mutation in p53 [26, 39, 40]. This implies a basic premise: ovarian cancer tumors utilize the mutation in *TP53* to enable the proliferation of SCNAs. SCNAs can subsequently occur in

Genomic Copy Number Alterations in Serous Ovarian Cancer

http://dx.doi.org/10.5772/intechopen.72695

117

There are plausible reasons for this. It may be that every SOC tumor is truly unique from a mutational standpoint: that those SNVs found in only one tumor nonetheless are driver genes, collaborating in ways that we understand poorly [25]. It is thus possible that drivers have already been sequenced and annotated by SCNA studies, but due to high "background" or "passenger" SCNAs it remains unclear which SCNAs are critical to the tumor's biology [8]. The implications of this are enormous, and would necessitate an unparalleled level of personalized therapies targeting such extremely rare mutations. A second reason that SNVs have not yielded common drivers may be that further sequencing of whole genomes and epigenomes will reveal additional drivers prevalent across patients which have remained undetected by exome sequencing.

The problem investigators consistently encounter is the ubiquitous heterogeneity in SOC. Heterogeneity exists at all levels of genetics, manifesting as *between-patient* heterogeneity, *between-tumor* (intra-patient) heterogeneity [14], heterogeneity in SNVs within a single tumor [12], heterogeneity in SCNAs [26], and heterogeneity in mRNA expression (correlating with protein expression) [11] or flux in SCNA status [10]. While such problems are not unique to SOC, they are magnified compared to many other tumor types because of the gross incidence of SCNA. Genetic and phenotypic heterogeneity remains the hardest issue to tackle [27, 28], and our own opinion mirrors that of several other groups working in this area: the analysis of affected pathways will offer new approaches to find hidden patterns of tumor suppressors and oncogenes within these heterogeneous data [8, 9, 29].

Fewer genomic studies have been performed on other types of ovarian cancer. Some limited data are available on SCNAs for Clear cell and endometrioid subtypes, which share the amplification of PIK3CA and the MYC-containing 8q24 region with SOC [30–32]. Larger SCNAs encompassing whole chromosome arms rather than smaller changes dominate the clear cell ovarian cancer SCNA landscape [32]. With the exception of 17p loss (containing *TP53*), focal *TPM3* amplification, and focal *ERBB2* amplification, SCNAs are infrequent in mucinous ovarian cancer, suggesting this histotype is SNV or epigenetic driven [30]. Generally, clear cell and endometrioid are intermediate in SCNA quantity between SOC and mucinous subtypes of ovarian cancer.

Despite the limited data on these non-serous subtypes, there is good reason to expect much more data is coming soon. The copy-number arrays employed in the Cancer Genome Atlas studies sell for less than \$100USD per sample, which bests the current, but constantly decreasing, cost of whole-genome sequencing. Eight oncology treatment and research centers are participating in project GENIE, which has just released 19,000 new tumor datasets to the public and will continue to grow [33]. As sequencing becomes a normal part of the treatment strategy for patients, the number of samples will likely outpace scientists' ability to fully analyze and comprehend the complex data. Nonetheless, gathering these data is essential to progressing our understanding of the differences between cancer subtypes, which will facilitate the matching of pharmaceuticals to genotype. For now, the largest datasets exist in SOC, and will be the focus of the remainder of discussion.
