Chromosomal Instability as a Source of Genomic Plasticity

Summary:

Chromosomal instability (CIN) is a hallmark of the most aggressive malignancies. Features of these tumors include complex genomic rearrangements, the presence of mis-segregated chromosomes in micronuclei, and extrachromosomal DNA (ecDNA) formation. Here we review the development of CIN, and examine CIN in the context of cancer evolution, tumor genomic evolution, and therapeutic resistance. We also discuss the role of whole genome duplications, breakage-fusion-bridge cycles, extrachromosomal DNA (ecDNA) or double minutes (DM) in gene amplification promoting tumor evolution.

Introduction:

Genomic instability is a feature of most tumor cells and is associated with increasing treatment resistance. DNA damage response (DDR) is a core signaling pathway that maintains genomic integrity through facilitation of DNA repair, coordination of cell cycle checkpoint and apoptotic machinery, and signaling to the innate immune response [1–3] [4]. Mutations in the p53 pathway and DDR genes bypass the senescence-associated checkpoint, resulting in genomic instability, including the acquisition of structural and numerical chromosomal abnormalities collectively often referred to as chromosomal instability (CIN) [5,6]. Broadly, CIN is a consequence of replication stress, mitotic errors, and cytotoxic chemotherapy treatment. Furthermore, CIN is a characteristic of aggressive malignancies [7]. Additionally, CIN is associated with an increased risk of metastasis and treatment resistance. The mechanisms by which CIN promotes cancer evolutionary success are likely due to activation of the cGAS/STING pathway and expanding karyotypic diversity [8].

With the advent of whole genome sequencing, we have obtained an improved ability to assess the karyotypic abnormalities within tumor cells. Dissection of the events that lead to somatic copy number alterations (SCNA) has generated models for the origins of chromosomally unstable tumors. These events include early loss of tumor suppressor genes, allowing for tolerance of whole genome duplication (WGD), followed by karyotypic pruning to generate advantageous near triploid karyotypes (Figure 1) [9]. Furthermore, CIN appears to be an ongoing process, as subclonal SCNA are evident both in primary tumors, and metastatic disease [10]. The selective pressures on the cancer cell populations can often lead to subclonal loss of heterozygosity (LOH), gene amplifications which can arise through numerous mechanisms, such chromosome breakage-fusion-bridge cycles (BFB) and complex rearrangements known as chromothripsis (described below). This leads to structures including micronuclei (MN), as well as fold-back inversions, extrachromosomal DNA (ecDNA), and homogenously staining regions (HSRs) [11]. In this review, we will discuss the origins of CIN in tumors with copy number variation and describe genomic events leading to gene amplification and thus drug resistance.

Figure 1: Evolution of Tumor Cells with Copy Number Alterations.

The specific genomic alterations that are characteristic of a subset of tumor cells with copy number alterations. Diploid cells first undergo loss of heterozygosity (LOH) in tumor suppressor genes such as p53 loss and mutations in the DNA damage response (DDR). Cells may undergo different trajectories: 1. (red dotted line) stochastic chromosomal gains and losses at high fitness cost due to the risk of nullisomy 2. (green dotted line) Whole genome duplication results in reduced fitness cost. 3.Stochastic chromosomal gains and losses results in selection of chromosomes with higher expression of oncogenes and lower expression of tumor suppressor genes. 4. This generates a near triploid karyotype, a state better poised for further evolution. 5. Later events in tumor evolution typically, but not exclusively, result in gene amplification events such as extrachromosomal (ecDNA) formation, bridge fusion break cycles (BFB) to form fold back inversions and new, post-whole genome duplication, LOH events to promote immune evasion, treatment resistance and metastasis (some icons made in Biorender).

Early Events in the Evolution of Chromosomally Unstable Cells

The observation that cells can harbor an abnormal chromosome number was by made Boveri and dates over a hundred years ago; in the last twenty years we have defined multiple mechanisms by which chromosomal instability drives tumor cell evolution, with the foundational observatiosn that overexpression of Mad2, a spindle checkpoint gene that regulates mitotic chromosomal segregation resulted in tumor formation within a mouse model [12] and that tetraploidy in combination with p53 loss are sufficient to cause tumorigenesis [13]. More recently, advancing sequencing technologies and computational approaches have resulted in refinement of our understanding of the early and ongoing events in cancer cell evolution.

Numerous studies have utilized patient tumor samples to decipher the timing of SCNA events both on the clonal and sub-clonal levels [14] [10]. It has been proposed that one of the earliest events in tumor evolution is loss of heterozygosity (LOH) of tumor suppressor genes, such as p53 or acquisition of a limited set of driver mutations [10] [14] (Figure 1). Loss of TP53 leads to tolerance of more egregious genomic instability because of abrogation of p53-dependent chromosome missegregation and/or DNA damage checkpoints [6]. LOH events were then followed by WGD in 30–50% of the tumors sequenced [10]; the combination of p53 loss and tetraploidy has long been shown to promote tumorigenesis in mouse models [13]. However, WGD, which yields a tetraploid cell, only provides a partial explanation for the captured genomic state of tumor cells, which exhibit a near triploid karyotype comprised of both LOH and genomic amplifications [13]. To address this discrepancy, Watkins et al utilized a model developed by Laughney et al and Davoli et al, to predict the fitness benefit of stepwise chromosomal gains or losses by taking into account the density of oncogenes and tumor suppressor genes on each chromosome arm [15]. This model was able to partially explain the genomic alterations after WGD whereby cancer cells optimized the number of chromosomes based on their oncogene and tumor suppressor gene densities. Moreover, the authors found evidence for continued SCNA in subclonal populations, indicating ongoing CIN leads to heterogeneity and at times convergent evolution [10], a finding noted in other analyses of patient samples [14]. Interestingly, while LOH has been considered an early event, some LOH events can occur late in a tumor lifetime such as loss of the HLA locus, which facilitates immune evasion [16] [17] (Figure 1). Conversely gene amplification events such as MYC or cyclin D were often found to be subclonal, and in the case of MYC appears to be associated with metastasis.

Why is WGD a common intermediate in the tumors with CIN? Insights from a stochastic computational model of chromosomal evolution taking into account oncogene and tumor suppressor gene density, indicates that WGD attenuates the potential deleterious effects associated with aneuploidy, promoting a less costly evolutionary path (Figure 1) [9]. WGD reduces the fitness penalty as a tetraploid genome allows for whole chromosomes gains and losses while minimizing the risk of nullisomy, ie complete loss of any given chromosome along with the essential genes that it encodes [18]. This permits a tumor to prune away chromosomes with relatively high content of tumor suppressor genes, reaching an evolutionarily advantaged triploid state that also maximizes the number of chromosomes with oncogenic properties [9,10,19]. Moreover, tumors with a WGD events exhibited a higher mutational burden with the exception of tumors with microsatellite instability or tumors secondary to exposures such as lung cancer or melanoma [19] [20]. However, despite the evolutionary advantage of WGD, aneuploidy, and CIN, they come with potential vulnerabilities. Recent work indicates that tetraploid cells are selectively vulnerable to disruption of the spindle assembly checkpoint or loss of the kinesin protein Kif18a due to prolonged mitosis, representing a potential therapeutic opportunity [19] [21] [22].

CIN as a catalyst for ongoing genomic chaos

Bulk whole exome and genome sequencing, as well as single cell WGS technology have revealed numerous genomic alterations in malignant cells, including gene amplifications [17,23–26]. As alluded to above, gene level amplification often occurs later in tumor evolution for which several complimentary models have been proposed. One postulated mechanism underlying gene amplification is the chromosome BFB cycle, which is a mutational process in tumorigenesis that involves chromosomal breakage followed by fusion using non-homologous end-joining forming dicentric chromosomes (i.e. chromosomes with two centromeres – and consequently two kinetochores) (Figure 2A). Given their high likelihood of attaching to microtubules emanating from opposite spindle poles, dicentric chromosomes often end up in a tug of war during anaphase leading to bridge formation and the propagation of breaks, followed by fusion and so on [27,28]. Dicentric chromosomes often form after telomere crisis, yet they can be the result of breaks at any chromosomal location (Figure 2A and 2B). Intriguingly, repeated BFB cycles are thought to yield fold-back inversions and gene amplification, although this is not frequently observed in whole genome studies [28].

Figure 2: Chromosome Bridges Facilitate Gene Amplification.

A. Dicentric Chromosomes form after fusion events such as after (1) critical telomere shortening, exposing chromosome ends and resulting in fusion of chromosomes with two centromeres (2). Upon cell division, dicentric chromosomes are attached to both centromeres (gold) impairing completion of cytokinesis resulting in an elongated connection between the two daughter cells called a chromosome bridge (3), resulting in breakage of the dicentric chromosome. Repair of the broken chromosome may result in (4) fusion of the broken chromosomes creating another dicentric chromosome, containing a duplication. A subsequent cycle of chromosome bridge formation, chromosome breakage (5) and fusion (6) results in fold back inversions and further gene amplification. Break-Fusion Bridge Cycles (7) can then generate large amplifications such as homogenously staining regions (HSR).

B. An alternative model of the gene amplification through chromosome bridge formation. An actomyosin force results in fracture and separation of the dicentric chromosomes into the daughter cells. Broken ends undergo aberrant replication. At the subsequent mitosis, the chromosome that underwent a fusion-break cycle is at risk for further missegregation due to impaired replication. Localization into the micronucleus results in an increased risk of chromothripsis [26], facilitating gene amplification through HSR or ecDNA formation.

To better characterize the genomic consequences of the BFB process, Umbreit et al modified genomically stable RPE-1 cells to generate dicentric chromosomes though four independent mechanisms including transient expression of a dominant negative telomeric repeat binding factor 2, partial knock-down of the condensin SMC2, low dose treatment of the topoisomerase II inhibitor ICRF-193 or utilizing a guide RNA to generate a break near the telomeric region of chromosome 4. Using live cell imaging they observed chromosome bridge formation and breakage in real time [26]. They isolated the observed paired daughter cells and performed single cell sequencing. The authors then demonstrated that a subset of cells had undergone rearrangements, which they postulate is due to error prone DNA repair. Most intriguingly, they used CRISPR/CAS9 to introduce a break near the end of specific chromosomes (e.g. chromosome 4) generating dicentric chromosome 4 fusions. Micronuclei were significantly enriched for this chromosome (Figure 2B). In addition, the authors found that the micronuclei undergo a burst of aberrant and asynchronous replication after nuclear envelope breakdown in mitosis when the replication machinery has access to the MN. They postulated that this abnormal replication, which is associated with DNA damage, facilitates further cycles of BFB and improper segregation. In addition, chromosome bridges led to a catastrophic shattering event followed by random repair leading to a pattern of complex rearrangements, a process known as chromothripsis. This occurred either as a direct consequence of bridge fracture or via missegregation of the resultant chromosome into a MN, followed by spontaneous micronuclear membrane rupture (Figure 2B, Figure 3B) [29–32]. This work elegantly demonstrated that one BFB cycle can propagate catastrophic genomic changes within the span of a single cell cycle resulting in an inherited predilection to continued genomic instability [26]. This also implies that the barriers to tumorigenesis are fragile upon p53 loss and subsequent tolerance to chromosomal damage [33].

Figure 3: Micronuclear Membrane Rupture Results in Chromothripsis Promoting Gene Amplification Through Formation of Homogeneously Staining Region and ecDNA.

A. In a normal mitosis, chromosomes will align at the metaphase plate and progress to anaphase. A merotelic attachment results in a chromosome (light blue) that becomes localized into a spontaneously forming micronucleus (MN)[31].

B. Chromosomes can be localized to the Micronucleus (MN) due to improper mitotic segregation or due to dicentric bridge breakage events. Upon spontaneous MN membrane rupture, a chromosome fracturing event called chromothripsis can occur. Upon Non-homologous end joining (NHEJ) dependent religation, circular fragments can be created of fragments of the oncogene or drug resistance gene [30,31]

C. Based on the model presented in Dunphy et al, circular fragments generated after chromothripsis can undergo circular recombination resulting in gene amplification forming an Homogeneously Staining Region (HSR) or ecDNA [43]

In addition to random integration of chromosome fragments into the genome, another recently appreciated consequence of chromothripsis is the circularization of large chromosomal regions through self-ligation (Figure 3B, 3C). These circular chromosomes have been referred to as double minutes (DM) or more recently extrachromosomal DNA (ecDNA) fragments [11]. ecDNA can be hundreds of kilobases to several megabases in size and typically lacks a centromere [34]. ecDNA is also often present in numerous copies and are randomly segregated to daughter cells promoting heterogeneity rapidly responsive to external conditions and have been vastly underappreciated due to challenges in ecDNA detection using traditional sequencing and analysis methods [34]. Intriguingly, both genetic and epigenetic alterations result in increased expression from ecDNA structures and recent work has shown that beyond the numerical state of ecDNAs contributing to increased gene expression, ecDNA engage in trans interactions with super-enhancers to increase expression of genes located on the ecDNA. Moreover, a subset of ecDNA have been demonstrated to contain super-enhancers themselves and function in trans to upregulate the gene expression of oncogenes localized to chromosomes. [35] [36] (Figure 4B).

Figure 4: Chromosomally Unstable Cells Exhibit Karyotypic Diversity and Evolve Genomic structures such as ecDNA.

A. Chromosomal instability facilitates resistance to oncogene addiction and chemotherapeutic agents likely through initial genomic heterogeneity followed by iterations of genomic instability facilitating treatment resistance through gene amplification of drug-specific resistance genes [40,41]. Upon selective pressure, gene amplification can result in HSR formation, further selective pressure is associated with localization of an HSR into a MN and transition of an HSR to ecDNA [11].

B. ecDNA increase gene expression through increased copy number and through genomic interactions promoting increased gene expression of genes located on the ecDNA through interaction with enhancers localized to chromosomes. Conversely, ecDNAs can function as super enhancers (blue circle) and promote expression of chromosomally localized genes (grey promotor) [35,36]

CIN-induced karyotypic diversity as a source of drug resistance

In addition to causing genomic chaos, CIN has been demonstrated to provide an evolutionary advantage to tumors by bypassing their dependence on truncal driver events. Indeed, experimental models have shown that CIN can promote treatment resistance by overcoming cancer oncogene addiction [37,38]. Salgueiro et al modeled the impact of CIN on oncogene addiction by utilizing a tetracycline inducible model of oncogenic KrasG12D with and without Mad2 overexpression, a spindle checkpoint gene whose loss or overexpression results in CIN. Withdrawal of doxycycline results in loss of oncogenic KrasG12D expression – which leads to regression of the majority of tumors. The authors observed persistence and recurrence of a greater number of KrasG12D/Mad2-overexpressing tumors compared to KrasG12D driven tumors (21.3% versus 6.6%, respectively). Recurrent tumors in both models were chromosomally unstable and harbored cMet amplifications that were not present in the primary tumor. This suggests that CIN promoted escape of KrasG12D oncogene addiction by evolving another oncogenic driver, cMet amplification [37].

How does CIN overcome oncogene addiction? One possible model is that CIN promotes karyotypic diversity and sampling of numerous genomic states. Several studies have demonstrated that CIN promotes chemotherapy resistance. In colorectal cancer (CRC), CIN is associated with a multidrug resistance phenotype, both in inherently CIN+ CRC lines and in a CIN− cell line in which MAD2 was knocked out to induce CIN [39]. In recent work by Lukow et al and Ippolito et al, chromosomally stable cell lines were treated with an Mps1-inhibitor, which increases chromosomal instability, and were then subjected to long term treatment with various chemotherapeutic agents. Interestingly, karyotypic analysis of the resistant clones demonstrated that resistant cells exhibited drug-specific karyotypes consistent with the model that CIN might engender a multidrug resistance phenotype through genomic plasticity on top of amplification of drug-specific resistance gene(s) (Figure 4A) [40,41].

To dissect the genomic mechanisms that facilitate drug resistance, work by Shoshani et al interrogated the genomic mechanisms leading to therapy resistance in cancer. They demonstrated that CIN mediated amplification promoted the evolution of drug resistance. Utilizing Hela cells treated with the anti-metabolite methotrexate (MTX) at three concentrations, they performed paired-end WGS to assess for chromosomal rearrangements in control and drug resistant clones. Cells placed under the stronger selective pressure of higher MTX concentration showed complex rearrangements including a high frequency of ecDNA/DM containing DHFR, a MTX resistance gene [42]. DHFR was localized to MN in 65% of cells, facilitating further cycles of chromothripsis and ecDNA amplification through MN rupture (Figure 4C). Previous studies had found that chromothripsis is dependent on the non-homologous end joining (NHEJ) pathway [31]; indeed, treatment with NHEJ inhibitors resulted in impairment of chromothripsis, which may represent a potential therapeutic target. To explore the mechanisms of gene amplification, the authors then followed the evolution of a clone resistant to an intermediate dose of MTX containing an HSR. Under the selective pressure of a higher dose of MTX, the HSR was lost and an ecDNA/DM containing the DHFR gene was acquired. The authors initially observed increased HSR size (likely due to amplification), followed by increased dicentric HSRs with associated high rates of chromosome bridges and formation of DHFR+ MN. Comparative WGS of one clone as well as selected daughter clones revealed that the HSR likely underwent chromothripsis to produce a highly rearranged DM (Figure 4A). Taken together this work demonstrates that genomic instability can facilitate the evolution of drug resistance through persistent CIN caused by chromosome bridge formation and formation of DM/ecDNA ultimately converging on amplification of drug resistance genes. Moreover, their results drive home the importance of ensuring cancer treatments are delivered at the correct dose, as resistance mechanisms developed at low doses were amplified laying the foundation of resistance to the full, previously lethal, dose.

Beyond experimental systems, there is evidence of CIN promoting resistance mechanisms in patient samples. Comparative sequencing studies on patients with BRAFV600E colorectal cancers who developed resistance to the BRAF inhibitor vemurafenib, showed evidence of BFB cycles in one patient and chromothripsis in another [11]. This indicates that in human tumors, genomic regions that are under positive selective pressure, such as oncogenes or drug resistance genes may become more genomically unstable to facilitate amplification. Similarly, Rosswog et al examined the mechanism of complex rearrangements in multiple tumor types [43]. They too saw evidence of genomic evolution which they dubbed ‘seismic amplifications’ starting with a chromothripsis event in 274/2756 cases across 38 cancer types. Intriguingly, the fold-back inversions associated with BFB cycles was not observed in many tumors with evidence of chromothripsis, but there was evidence of ecDNA formation that either persisted as ecDNA or was reintegrated into the genome as HSRs (Figure 4A). As the genomic structures they observed could not be fully explained by BFB cycles or chromothripsis alone they postulated that ecDNA undergo internal rearrangements. They propose this is due to ecDNA circular recombination that results in greater gene amplification through recombination mediated duplication (Figure 3C).

Conclusions and Future directions:

The incredible advances in sequencing technologies have resulted in a new understanding as well as re-appreciation of the large scale numerical and structural chromosomal changes in cancer. Taken together these studies demonstrate that genomic changes beget further CIN, promoting a self-propagating cycle of chromosomal chaos fueling karyotypic diversity and tumor heterogeneity. We have gained significant new insights into the origins and evolution of tumorigenesis, beginning with loss of heterozygosity of tumor suppressors and subsequent tolerance of whole genome duplication generating a tetraploid state (Figure 1). This is followed by karyotypic pruning to select for chromosomes with high oncogene densities, and tolerance of BFB cycles resulting in gene amplification and ecDNA formation (Figure 2 A, Figure 3B, Figure 4B). Ongoing SCNA alterations and associated CIN promotes karyotypic diversity, which has been shown in mouse models and cell culture to promote treatment resistance (Figure 4A). Future work will refine our understanding of how selective pressure translates into these genomic changes—leading to greater cellular heterogeneity and subclonal drug resistance through gene amplification or gene loss. These events likely represent a source of many treatment failures to both targeted and cytotoxic therapies. Therefore, identifying drugs to target and inhibit genomic rearrangements and CIN, such as NHEJ inhibitors will be an important new direction to facilitate genomic stabilization during treatment thereby preventing resistance. Equally important are targets whose loss is selectively lethal to cells with abnormal karyotypes such as Kif18a. This is an exciting era in cancer treatment as we have an unprecedented insight into the dynamic changes in karyotypes using single-cell sequencing data in both experimental systems and patient samples in parallel with new drugs that target DNA repair pathway. We are hopeful that future treatments will advance beyond DNA damaging cytotoxic agents and move towards combining targeted agents with genome stabilizing drugs to prevent tumor evolution.