(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .
Reduced sister chromatid cohesion acts as a tumor penetrance modifier [1]
['Jun Wang', 'Department Of Pharmacology', 'Toxicology', 'University Of Alabama At Birmingham Heersink School Of Medicine', 'Birmingham', 'Alabama', 'United States Of America', 'Holly R. Thomas', 'Yu Chen', 'Department Of Genetics']
Date: 2022-10
Sister chromatid cohesion (SCC) is an important process in chromosome segregation. ESCO2 is essential for establishment of SCC and is often deleted/altered in human cancers. We demonstrate that esco2 haploinsufficiency results in reduced SCC and accelerates the timing of tumor onset in both zebrafish and mouse p53 heterozygous null models, but not in p53 homozygous mutant or wild-type animals. These data indicate that esco2 haploinsufficiency accelerates tumor onset in a loss of heterozygosity (LOH) sensitive background. Analysis of The Cancer Genome Atlas (TCGA) confirmed ESCO2 deficient tumors have elevated number of LOH events throughout the genome. Further, we demonstrated heterozygous loss of sgo1, important in maintaining SCC, also results in reduced SCC and accelerated tumor formation in a p53 heterozygous background. Surprisingly, while we did observe elevated levels of chromosome missegregation and micronuclei formation in esco2 heterozygous mutant animals, this chromosomal instability did not contribute to the accelerated tumor onset in a p53 heterozygous background. Interestingly, SCC also plays a role in homologous recombination, and we did observe elevated levels of mitotic recombination derived p53 LOH in tumors from esco2 haploinsufficient animals; as well as elevated levels of mitotic recombination throughout the genome of human ESCO2 deficient tumors. Together these data suggest that reduced SCC contributes to accelerated tumor penetrance through elevated mitotic recombination.
Tumorigenesis often involves the inactivation of tumor suppressor genes. This often encompasses an inactivation mutation in one allele and loss of the other wild-type allele, referred to as loss of heterozygosity (LOH). The rate at which the cells lose the wild-type allele can influence the timing of tumor onset, and therefore an indicator of a patient’s risk of cancer. Factors that influence this process could be used as a predictive indicator of cancer risk, however these factors are still unclear. We demonstrate that partial impairment of sister chromatid cohesion (SCC), a fundamental component of the chromosome segregation in mitosis and homologous recombination repair, enhanced tumorigenesis. Our data suggest this is through elevated levels of mitotic recombination derived p53 LOH. This study emphasizes the importance of understanding how impaired SCC, mitotic recombination rates, and LOH rates influence cancer risk.
Funding: JMP is supported by the National Cancer Institute R01CA216108 and R21CA219546; as well as National Institute of Health, Office of the Director U54OD030167, a pilot grant from American Cancer Society (ACS IRG-60-001-53-IRG) and the UAB Comprehensive Cancer Center (P30CA013148). SMP is supported by the Cell and Molecular Biology T32 Training Grant (5T32GM008111-28). RCR is supported by the UAB MSTP (NIH-NIGMS 5T32GM008361-21). SJC is supported by the Hudson Alpha Tie the Ribbons Fund and the UAB CCTS (NIH 1UL1TR001417-01). ZC is supported by the National Institute of General Medical Sciences (1R35GM138212). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Li Fraumeni Syndrome (LFS) is an autosomal dominant cancer predisposition syndrome, due to germline heterozygous mutations in p53, in which 50% of patients will succumb to a tumor by age 30[ 18 ]. Part of the etiology of tumor formation in LFS, as well as sporadic tumors, is that the first allele has a p53 mutation, and the second undergoes loss of wild-type allele (Loss Of Heterozygosity, LOH) often through whole chromosome deletion/duplication or mitotic recombination [ 19 ]. Mouse and zebrafish heterozygous p53 null models have recapitulated this early tumor susceptibility, as well as the LOH of the wild-type allele in tumors [ 20 – 24 ]. Further, murine LFS models suggest that within a cohort of p53 heterozygous animals the timing of tumor onset correlates with the prevalence of p53 LOH in tumors [ 25 ]; i.e. the trend is that early-onset tumors often have p53 LOH while late tumors often do not. This suggests that the rate of p53 LOH or prevalence of LOH events influences tumor penetrance. Further, the frequency of tumors with p53 LOH in p53 heterozygous null mice is strain-dependent, in that p53 LOH in a mixed C57/129 strain is ~50%, while in a BALB/c background it is ~96%, suggesting there are undefined genetic factors that can impact the rate of p53 LOH. Note tumor onset was earlier in the BALB/c mice, and the overall higher frequency of mitotic recombination, suggesting difference in DNA repair influence the frequency of p53 LOH and timing of tumor onset [ 26 ]. While not explicitly examined with respect to p53 LOH, inactivation of RecQ helicases, such as WRN or BLM, have been shown to increase the rate of mitotic recombination and therefore act as a LOH modifier [ 27 – 29 ].
Genomic alterations including missegregation, aneuploidy and micronuclei formation, are hallmarks of cancers and associated with poor patient outcomes [ 1 ]. These genomic alterations often occur due to mitotic error; however mutational drivers of these genomic instabilities in tumors are unclear [ 2 – 4 ]. Defects in microtubule attachment and spindle assembly checkpoint have been demonstrated to contribute to genomic instability and cancer predisposition [ 5 , 6 ]. Sister chromatid cohesion (SCC) is another essential process required for equal segregation of chromatids into daughter cells during mitosis and is a probable target for mutations that induce these genomic instabilities. At the core of mitotic SCC, is the cohesin ring, made up of a multiprotein complex that includes SMC1a, SMC3, RAD21, and SA1/2, which lasso the sister chromatids together. ESCO2 establishes SCC around the sister chromatids as they are synthesized during S-phase, whereas Sororin and SGOL1 protect the established SCC until mitosis [ 7 , 8 ]. During metaphase of mitosis the microtubule tension on the kinetochore aligns the sister chromatid into a metaphase plate. During the transition from metaphase to anaphase, the cohesin ring is cleaved by Separase, allowing equal segregation of the sister chromatids to each daughter cell. While SCC is most well-studied with regard to chromosome segregation, it also influences several other biological processes, including DNA repair/homologous recombination [ 9 – 13 ]. The involvement of compromised SCC in cancer has been less well studied since complete loss of SCC triggers cell cycle arrest, apoptosis and/or cellular senescence, which is not conducive to cellular viability, and therefore viewed as disadvantageous to a tumor [ 14 ]. Further, mild SCC dysfunction is also suggested to be toxic to the cell due to the need for centromeric cohesion to establish bi-orientation of kinetochores [ 15 ]. Hence, SCC is considered an all or none process regarding cellular viability. That said, missense mutations in some components of the SCC have been identified in human tumors, however the mechanistic importance of these mutations remains unknown [ 16 , 17 ].
(A) Schematic of determining LOH mechanisms/types in p53 heterozygous tumors. The haplotype of nearby p53 marker in tumor and normal paired samples, cooccurring with p53 wild-type or mutant gene are labeled (left). (B) A table showing the percentage of mitotic recombination (MR) in esco2 +/+ ; p53 zy7/+ and esco2 hi2865/+ ; p53 zy7/+ tumors. (C) Schematic of determining the mitotic recombination with TCGA database. Figure A and C were created with BioRender.com . (D) The number of mitotic recombination in TCGA cancer samples with or w/o ESCO2 mutation/deletion. Genome-wide data was downloaded from NCI Genomic Data Commons. The number of WT and deficient samples in each tumor were indicated: PRAD: 481 WT + 6 Deficient; UCEC: 479 WT + 43 Deficient; OV: 383 WT + 46 Deficient; LUAD: 495 WT + 10 Deficient; COADREAD: 482 WT + 38 Deficient; LUSC: 478 WT + 8 Deficient; BRCA: 935 WT + 29 Deficient and BLCA: 393 WT + 13 Deficient. *, ** and *** indicate p<0.05, p<0.01 and p<0.001 from Mann-Whitney test.
p53 LOH can also be derived from mitotic recombination (MR) in which the region containing p53 becomes homozygous while the region between the MR site and the centromere as well as the opposite arm remains heterozygous ( Fig 6A ) . Since our high and low MN cohort experiment suggested the whole chromosome instability (CIN) is not the driving force behind p53 LOH, we wanted to investigate mitotic recombination as a mechanism of p53 LOH in our esco2 heterozygous zebrafish tumors. For this we identified a polymorphic SNP in rabep1 or cdca5 in normal tissues, that resided on the opposite arm of p53 containing chromosome ( Fig 6A ) . Using these markers, we observed an increased rate (37%vs25%; Fig 6B ) of heterozygosity maintained at rabep1 or cdca5 in tumors derived from esco2 heterozygous animals, indicative of increased MR. Since we observed the largest difference in the survival curves in the population of animals forming tumors later, we decided to analyze the MR rate of tumors that occur in the first half (<T 50 ) and the second half (>T 50 ) of the tumor cohorts. Interestingly, while in the esco2 wild-type cohort the rate of MR was 25% in the first and second half, we observed a much higher MR rate (57%) in the second half of the esco2 heterozygous derived tumor cohort. These data suggest MR-derived LOH is a driving force for the overall earlier tumor onset in the esco2 heterozygous cohort (T 50 = 467 days in esco2 hi2865/+ ; p53 zy7/+ cohort vs. 561 days in the p53 zy7/+ cohort). To determine if human cancers that have ESCO2 deficiencies also have higher MR rates, we analyzed TCGA data to define all LOH loci that maintained a 2N copy number ( Fig 6C ) , this would exclude regional deletions or whole chromosome deletion, as well as amplifications. We then plotted these across the chromosome to define MR events. From this analysis we determined that there were more MR events, in multiple cancer types, in ESCO2 deficient tumors compared to proficient tumors ( Fig 6D ) . Together these data suggest that reduced SCC enhances the rate of MR which drives LOH-sensitive tumorigenesis.
(A) Micronuclei (MN) during interphase were counted in six esco2 +/+ and ten esco2 2865/+ H2A.F/Z-EGFP mRNA injected embryos at 24hpf using two-hour live imaging time-lapse data from each embryo. Each dot represents an embryo measured. Percentage MN/embryo was calculated based on the total number of micronuclei per total nuclei present in the time lapse at t = 0. Total nuclei are indicated in parenthesis. Mean ± SD, ** p-value < 0.01 based on unpaired t-test. Representative black and white images of micronuclei in each genotype are shown below. (B) Proposed model where the proportion of cells with MN would correlate with p53 LOH and timing of tumor formation. (C) Experimental workflow to establish high and low MN cohorts, using confocal living-imaging of 3-dpf p53 +2/+ ; TG PhOTO-N/+ embryos. Figure B and C were created with BioRender.com . (D) Dot-blot of Micronuclei frequency in individual 3-dpf p53 +2/+ ; TG PhOTO-N/+ embryos. Individuals deemed part of the high MN cohort were highlighted with red (MN%>0.9%) and individuals in the low MN cohort were highlighted with green (MN% = 0). (E) Zebrafish Kaplan-Meier curves for tumor-free survival for p53 +2/+ ; TG PhOTO-N/+ unsorted (n = 90), p53 +2/+ ; TG PhOTO-N/+ with high NM ratio(n = 54) and p53 +2/+ ; TG PhOTO-N/+ with high NM ratio (n = 58). P-value is not significance when comparing each other based on Log-rank (Mantel-Cox) test.
Micronuclei (MN) are often observed in cancers, and are derived from ungrouped chromosomes, often lagging chromosomes, during telophase when the nuclear envelope is reformed around chromatid material. MN are an indicator of a mitotic chromosome segregation defect, resulting in one cell gaining chromosomes while the sister cell losing chromosomes which could contribute to p53 LOH. Consistent with our chromosome segregation analysis, we observed a significantly higher number of micronuclei in 1 dpf esco2 heterozygous mutant embryos (Avg. ~6%) versus the wild-type embryos (Avg. ~2%; Fig 5A ). Interestingly, there was strong variability in the percentage of cells with micronuclei within embryos of the same genotype, ranging from 0–3.9% in wild-type embryos and 1.9%-15% in esco2 heterozygous-mutant embryos ( Fig 5A ) . This is also consistent with there being variations in the number of segregation errors observed between embryos of the same genotype ( Fig 4C ; up to 4 missegregation within a scanned field in embryo No.1 verse no events in 4 embryos (No.7-10)). This MN variability and higher levels of MN in esco2 heterozygous animals spurred a hypothesis that animals with higher proportion of cells with MN, would have more frequent p53 LOH events in precancerous somatic cells, therefore a larger pool of cells that could become tumor-forming, which would translate into earlier tumor onset ( Fig 5B ) . To test this hypothesis, we confocal imaged live 3-dpf p53 heterozygous embryos for the MN frequency using a PhOTO-N chromatin labeling transgenic line ( Fig 5C ) . The TG PhOTO-N transgenic lines ubiquitously expressed H2A-dendra fluorescent protein, which allows for imaging chromosome dynamics as well as determining MN frequency. We used the 3-dpf timepoint to improve viability of embryos following the imaging procedure. We imaged 446 embryos and selected the top 15% MN frequency (0.91% to 3.85% frequency) as the high MN cohort and the bottom 15% (no MN observed) as the low MN cohort ( Fig 5D ) . These cohorts as well as an un-imaged p53 heterozygous cohort (with TG PhOTO-N background) were monitored for tumor formation. Surprisingly we did not observe a significant difference in the high vs low MN cohort ( Fig 5E ) . This suggests that chromosome missegregation and elevated MN in embryos are not the driving force behind elevated LOH and earlier tumor onset in esco2 haploinsufficient animals. Note these experiments do not address the impact of chromosome missegregation and elevated MN that occurs in adult somatic tissues.
(A) in-vivo confocal imaging of H2A.F/Z-EGFP and CaaX-mCherry mRNA injected embryos at 24 hours post-fertilization (hpf) for two hours. Representative images of normal and defective mitoses in esco2 +/+ and esco2 m/+ . Arrow in anaphase bridge time-lapse points towards the anaphase bridge formed. Dotted circular in tri-polar time-lapse represents the three future nuclei that will occur. CaaX-mCherry was removed in tri-polar time-lapse for better visualization. t = time in minutes. (B) Division time calculated for each division in six esco2 +/+ and ten esco2 m/+ embryos using two-hour imaging time-lapse data from each embryo. The percentage of cells was calculated for each bin category. N = 73 for esco2 +/+ and N = 132 for esco2 m/+ . (C) Table representing mitotic defects and the associated mitotic timing and cell fate observed in six esco2 +/+ and ten esco2 m/+ embryos.
(A) Representative images of “paired” and “railroad” (RR) metaphase spreads. Inset shows zoomed-in view of each phenotype. (B) The percentage of metaphase spreads in esco2 +/+ and esco2 2865/+ showing paired or RR phenotypes. Two pools of embryos were used for esco2 +/+ and esco2 2865/+ spreads from two independent experiments. A total of 92 esco2 +/+ and 119 esco2 2865/+ spreads were tallied. (C) The percentage of spreads in sgo1 +/+ and sgo1 +/- showing paired or RR phenotypes. Three pools of embryos were used for sgo1 +/+ and sgo1 m/+ spreads. A total of 131 sgo1 +/+ and 142 sgo1 +/- spreads were tallied. ** p-value < 0.01 and *** p-value < 0.001 by paired t-test. (D) Kaplan-Meier curves for tumor-free survival for p53 +/+ , sgo1 +/- ; p53 +/+ , p53 +/- and sgo1 +/- ; p53 +/- cohorts. Cohorts were established by natural single-pair breeding of sgo1 +/- X p53 +2/+ parents (all cohorts were n = 96). P-value = <0.0001 when comparing p53 +/- with sgo1 +/- ; p53 +/- curves based on Log-rank (Mantel-Cox) test.
The most striking observation was that amongst the zebrafish esco2 heterozygous metaphase spreads, 8.3% displayed a”railroad” (RR) phenotype not observed in the wild-type sibling spreads ( Fig 3A and 3B ) . While similar phenotypes have been described in cell culture, it has never been observed in a live organism [ 16 ]. It has been presumed that this phenotype would result in embryonic lethality in a live organism due to the inability to properly orient the kinetochore required for microtubule attachment [ 15 ]. However, in this case potentially the viability is associated with the low proportion of cells that have RR. These observations indicate that there is a gene dose-dependent reduction in cohesion establishment in esco2 heterozygous animals. To address if accelerated tumor onset is specific to esco2 or reduced SCC, we also generated a zebrafish sgo1 null allele using CRISPR/Cas9 technique. SGO1 plays an important role in maintaining SCC after establishment by ESCO2. Homozygous null sgo1 are lethal ( S3 Fig ) , however sgo1 heterozygous mutant animals are viable and display low-level RR (13.4%) in metaphase spreads similar to esco2 heterozygous mutants ( Fig 3C ) . This led us to ask would the reduced SCC in sgo1 mutants also predisposes to early tumor onset in a p53 heterozygous background. Consistent with the esco2 haploinsufficient results, heterozygous loss of sgo1 resulted in statistically significant accelerated tumor onset in a p53 heterozygous background ( Fig 3D , T 50 = 422 days in the sgo1 Δ8/+ ; p53 +2/+ cohort vs. 478 days in the p53 +2/+ cohort; p <0.0001 based on Log-rank test). Together this provides two independent genetic models of reduced SCC that are associated with accelerated tumors in a LOH sensitive background.
(A) Zebrafish Kaplan-Meier curves for tumor-free survival for wild-type, esco2 2865/+ , p53 J19/+ , esco2 2865/+ ; p53 J19/+ , p53 J19/J19 , and esco2 2865/+ ; p53 J19/J19 cohorts. Compared cohorts were established by natural single pair breeding of esco2 m/+ x AB (wild-type strain), esco2 m/+ ; p53 m/m x AB, or esco2 m/+ ; p53 m/m x p53 m/m parents (all cohorts were n = 96). The p-value = <0.0001 when comparing p53 J19/+ with esco2 2865/+ ; p53 J19/+ cohorts based on Log-rank (Mantel-Cox) test. (B) Mouse Kaplan-Meier curves for tumor-free survival for wild-type, Esco2 +/- , p53 +/- , Esco2 +/- ; p53 +/- and p53 -/- cohorts (cohorts with p53 +/+ and p53 +/- background were n>60 and cohorts with p53 -/- background were n>6). The p-value = <0.05 when comparing p53 +/- with esco2 +/- ; p53 +/- curves based on Log-rank (Mantel-Cox) test. (C) Frequency of p53 wild-type loss of heterozygosity (LOH) and esco2 wild-type LOH in zebrafish tumors; as well as the frequency of tumors being Malignant Peripheral Nerve sheath tumors (MPNST). (D) Pie charts showing tumor spectrum in Esco2 +/+ ; p53 +/- (left panel) and Esco2 +/- ; p53 +/- (right panel) mice. No statistically significant in tumor spectrum with Chi-square test. (E) Percentage of patients with LOH on TP53 in TCGA cancer samples with or w/o ESCO2 mutation/deletion. The number of WT and deficient samples in each tumor were indicated. BLCA: 393 WT + 13 Deficient; BRCA: 935 WT + 29 Deficient; COADREAD: 482 WT + 38 Deficient; LIHC: 333 WT + 21 Deficient; LUND: 495 WT + 10 Deficient; LUSC: 478 WT + 8 Deficient; OV: 383 WT + 46 Deficient; PRAD: 481 WT + 6 Deficient; UCEC: 479 WT + 43 Deficient. * indicates p<0.05 from Fisher’s exact test. (F) Percentage of patients with LOH covering in tumor suppressor gene PTEN, BRCA1, BRCA2, RB1, NF1 and APC in the TCGA dataset. N = 4,126 for ESCO2-WT cohort and N = 193 for ESCO2-deficiency cohort. ** and *** indicate p<0.01 and p<0.001 with Fisher’s Exact test. (G) The number of LOH events per tumor sample in TCGA cancer samples with or w/o ESCO2 mutation/deletion. The number of WT and deficient samples in each tumor were indicated: PRAD: 481 WT + 6 Deficient; UCEC: 479 WT + 43 Deficient; OV: 383 WT + 46 Deficient; LUAD: 495 WT + 10 Deficient; COADREAD: 482 WT + 38 Deficient; LUSC: 478 WT + 8 Deficient; BRCA: 935 WT + 29 Deficient; BLCA: 393 WT + 13 Deficient. * and ** indicate p<0.05 and p<0.01 from Mann-Whitney test.
To decipher if ESCO2 loss is a driver of tumorigenesis, we wanted to determine if ESCO2-deficient animals are tumor predisposed. Both mouse and zebrafish homozygous ESCO2 null animals are embryonic lethal[ 44 , 45 ], therefore to determine if reduction in ESCO2 specifically contributes to tumorigenesis, we generated and monitored tumor formation in cohorts (N = 96/cohort) of esco2 +/+ and esco2 hi2865/+ (hi2865 allele is an intron 1 gene trap that results in >95% knockdown of transcript) zebrafish in a tumor sensitized p53 mutant background, p53 zy7/+ (zy7 allele is a thymine to cytosine transition in codon 164 in the p53 DNA binding domain that results in the inability of DNA binding)[ 22 , 44 ]. Within all zebrafish tumor cohort studies to reduce the influence of background genetic variability, zebrafish cohorts to be compared, were generated from a single pair of zebrafish. Therefore, these two cohorts were generated from an esco2 hi2865/+ ; p53 zy7/zy7 crossed to a wild-type AB strain zebrafish. We observed statistically significant acceleration of tumor onset ( Fig 2A ) in the esco2 heterozygous mutant cohort (T 50 = 467 days in esco2 hi2865/+ ; p53 zy7/+ cohort vs. 561 days in the p53 zy7/+ cohort; p<0.0001 based on log-rank test) suggesting loss of esco2 is a tumor driving event. However, these experiments were performed in a tumor-sensitized background and did not indicate if esco2 haploinsufficiency acts as an autonomous tumor suppressor gene. Therefore, we also monitored cohorts of esco2 +/+ and esco2 hi2865/+ zebrafish in a p53 wild-type background, derived from a single pair of esco2 hi2865/+ crossed to wild-type AB. However, over a 2-year period only 2 of 96 esco2 hi2865/+ animals developed a tumor compared to 0 of 96 wild-type animals ( Fig 2A ) , suggesting that if esco2 does act as an autonomous tumor suppressor gene it has low penetrance. These data indicate that haploinsufficiency for esco2 enhances tumorigenesis only in a p53 heterozygous background. Loss of the wild-type p53 allele (loss of heterozygosity (LOH); p53 m/+ ➔ p53 m or m/m ) is an important step during tumorigenesis in p53 heterozygous mutant humans (Li Fraumeni Syndrome) and animals (zebrafish and mouse p53 knockouts) [ 20 , 22 , 46 – 51 ]; therefore we postulated that esco2 heterozygosity may enhance tumor formation in a p53 heterozygous animal by accelerating p53 LOH. If true, esco2 heterozygosity would not enhance tumor formation in a p53 homozygous background (insensitive to p53 LOH). Therefore, we monitored a cohort of p53 zy7/zy7 and esco2 hi2865/+ ; p53 zy7/zy7 zebrafish and found a non-significant difference in tumor enhancement in the p53 p53 zy7/zy7 ; esco2 hi2865/+ cohort compared to p53 zy7/zy7 cohorts created from single pair of esco2 hi2865/+ ; p53 zy7/zy7 x p53 zy7/zy7 ( Fig 2A ; T 50 = 330 days in the esco2 hi2865/+ ; p53 zy7/zy7 cohort vs. 320 days in the p53 zy7/zy7 cohort; p = 0.526 based on Log-rank test). This finding suggests that esco2 heterozygosity influences tumor onset only in a LOH sensitive (p53 heterozygous) background. To determine if these observations were also true in mice, we generated similar cohorts of mice using a p53 null allele and a Esco2 null allele (deletion of exon 4) [ 49 , 52 ]. The mouse cohort data is consistent with the zebrafish data, in that Esco2 haploinsufficient loss accelerates tumor onset only in a LOH sensitive background ( Fig 2B ; T 50 = 516 days in the Esco2 +/- ; p53 +/- cohort vs. 563 days in the p53 +/- cohort; p = 0.0377 based on Log-rank test). To determine if a higher proportion of the esco2 deficient derived zebrafish tumors have p53 LOH, we genotyped paired normal and tumor genomic DNA for the p53 allele. The percentage of tumors in the p53 zy7/+ cohort having LOH of the wild-type p53 allele was 85%, while tumors from the p53 zy7/+ ; esco2 hi2865/+ cohort was ~92% ( Fig 2C ) . While this is trending, it is a non-significant change in the frequency of LOH-containing tumors. However, measuring LOH in a tumor which is the end-product of tumorigenesis may not reflect the process or rate of LOH in somatic precancerous tissues in these animals. We also determined that the esco2 gene did not undergo LOH in any of these tumors, suggesting the phenotype is a consequence of a gene dose haploinsufficient state ( Fig 2C ) . Further, we did not observe a change in tumor spectrum, in that 19 of 20 p53 zy7/+ tumors were malignant peripheral nerve sheath tumors (MPNSTs), and 23 of 23 p53 zy7/+ ; esco2 hi2865/+ tumors were also MPNSTs ( Fig 2C ) . We also evaluated if there was a change in the physical location of the MPNSTs ( S2 Fig ), but we did not observe a statistical difference between the populations. We did observe more undifferentiated sarcomas ( Fig 2D ) in the Esco2 +/- ; p53 +/- mice cohort, however the timing of this tumor type was not statistically different suggesting this is not the driving force for the accelerated tumor onset. Evaluation of human tumors in the TCGA dataset, indicates that there is statistically significant increase in p53 LOH in Liver hepatocellular carcinoma (LIHC), while there is a trend for increased p53 LOH in ESCO2 deficient tumors compared to ESCO2 proficient tumors ( Fig 2E ) . Further, analysis of other tumor suppressor genes in TCGA data indicated that there are statistically significant increases in LOH of BRCA1, PTEN and NF1 genes in ESCO2 deficient tumors. ( Fig 2F ) . To determine if the increased LOH frequency was only associated with tumor suppressor gene or genes under a selective pressure for LOH, we surveyed the entire genome for LOH frequencies in ESCO2 wild-type and deficient tumors ( S1 and S2 Tables). In multiple tumor types we observed significant elevated LOH rates throughout the genome ( Fig 2G ) . Together this indicates that haploinsufficiency in ESCO2 results in a globally higher LOH frequency in tumors.
(A) Oncoprint plot of genetic alterations in ESCO2 and other SCC-associated genes in TCGA dataset. N = 10950. Number of samples with each mutation: N (ESCO2) = 404, N (ESCO1) = 252, N (SMC3) = 216, N (SMC1A) = 322, N (NIPBL) = 743, N (STAG1) = 429, N (STAG2) = 374, N (SGO1) = 165, N (CDCA5) = 159, N (WAPL) = 248, N (PDS5A) = 227 and N (PDS5B) = 375. P-values shown in the figure indicate significance of co-occurrence with ESCO2 from one-sided Fisher Exact test. (B) Kaplan-Meier survival analysis of patients with ESCO2 deficient (N = 1529) versus ESCO2 WT (N = 8317) tumors within TCGA dataset samples. P-value was determined by log-rank test. (C) Stacked bar plot indicating the percentage of patients with an ESCO2 deletion (blue), amplification (red) or mutation (green) and multiple alterations (grey) in uterine corpus endometrioid (UCEC), prostate (PRAD), colorectal (COADREAD), ovarian (OV), bladder (BLCA), liver (LIHC), lung adenocarcinoma (LUAD), lung squamous (LUSC), and breast cancer (BRCA) in the TCGA dataset. Cancer type is ordered based on the genetic alteration frequency. Number of samples with mutation/ total samples in each cancer type: 45/529 UCEC, 37/494 PRAD, 43/594 COADREAD, 41/584 OV, 26/411 BLCA, 23/372 LIHC, 33/566 LUAD, 23/487 LUSC, 50/1084 BRCA. The frequency of deletion in 100 kbp windows throughout Chromosome 8 in 1,111 pan-cancer type (Figure D, N = 11203) and prostate adenocarcinoma (PRAD, Figure E, N = 502) cancer patients. A region is considered as deleted if Log 2 (Copy Number/2) < -1. Blue line shows smoothed deletion frequency. Red vertical line indicates the locus of ESCO2 gene.
We surveyed The Cancer Genome Atlas (TCGA) database for prevalence of genetic alterations in key sister chromatid cohesion (SCC) factors and found that ESCO2 was most frequently deleted or mutated gene in many cancers ( Fig 1A ) . Further patients with ESCO2 alterations have significantly lower probabilities of survival ( Fig 1B ) . We observed that approximately 5–8% of patients within nine of the most common cancers show alterations in ESCO2 and that these alterations are predominantly deletions and a few protein-altering mutations ( Fig 1C ) . The deletions are defined as at least 50% reduction of gene copy number, suggesting that ESCO2 may be a haploinsufficient tumor suppressor gene. However, ESCO2 resides on chromosome 8p21, and in many human cancers, including liver, breast, prostate, ovarian, uterine, colorectal, bladder and lung cancers, recurrent deletion of genes on the p arm of chromosome 8 is observed ( Fig 1D and 1E ) . This recurring deletion has been associated with advanced tumor progression and poor patient survival [ 30 – 32 ]. This region appears rich in tumor suppressor genes [ 33 ]. For example, in breast cancer, genes between 8p21-23 represent five of the top ten most deleted genes [ 34 ]. Within this region, there are established tumor suppressors (DLC1, DOK2 and LZTS1), as well as potential tumor suppressor genes (CSMD1, MTUS1, and MSR1), and many other genes not established in cancers [ 35 – 40 ]. Some of these genes, including ESCO2, may be passengers, deleted purely based on their proximity to tumor suppressors, while others may represent novel genes with a role in cancer formation and progression. ESCO2 is important for the establishment of SCC following new synthesis of the sister strand during S-phase of the cell cycle [ 7 ]. ESCO2 has not previously been implicated in human cancers, but its function certainly suggests it could be important, and it appears to be at one of the sub-peaks of this deletion region (Figs 1D , 1E and S1 ) . In further support of ESCO2 playing a role in tumorigenesis, some Roberts Syndrome patients (RBS) who carry autosomal recessive germline inactivating mutations in ESCO2, display early-onset cancer predisposition [ 41 – 43 ].
Discussion
These data indicate that gene dose-dependent loss of esco2 or sgo1 results in reduction in SCC and enhanced tumor predisposition. Interestingly, amongst viable (majority die in utero or postnatally) Roberts Syndrome (RBS) patients, which are homozygous null of ESCO2, they also display reduced SCC and early onset tumor occurrence [41–43]; supporting the fact that reduced SCC is tumor-promoting across species. Further, our analysis of the TCGA data strongly supports that ESCO2 loss in tumors is associated with elevated mitotic recombination and elevated tumor suppressor inactivation. While this study has focused on ESCO2, SCC is a complex network of pro-cohesion factors and anti-cohesion factors (yeast genetics suggests over 350 genes involved) [58], in which the synergistic or antagonistic effects of combined mutations/polymorphisms in these genes can create varying degrees of SCC dysfunction amongst a population of individuals and potentially define variability of tumor penetrance between individuals. Consistent with this, alterations in other SCC factors such as Separase overexpression, STAG2 loss, SGO1 haploinsufficiency, and others have been associated with tumorigenesis [16,59–61]. Additionally, deficiencies in the tumor suppressor retinoblastoma (RB1) have recently been tied to SCC loss and suggestive of a mechanism by which RB loss promotes tumorigenesis [62]. SCC involvement in tumorigenesis is also interesting in that there is a yin-yang relationship between the cohesion establishment and antiestablishment processes. In yeast, deletion of Eco1, the homolog of ESCO2, is lethal; while concomitant loss of Eco1 and WAPAL, or PDS5, circumvent this lethality [63–66]. This suggests that there could be a window where ESCO2 or another establishment molecule is lost, which would encourage MR and selection, and then a second window where an anti-establishment factor is lost to restore stability of an aneuploid genome. However, this Yin-Yang relationship is yet to be demonstrated during tumorigenesis. Together these data emphasize that defects in SCC can strongly influence tumorigenesis.
While chromosome instability has been the focus of studies looking at the consequence of defects in SCC, some studies have pointed to increased rates of homolog recombination versus sister chromatid recombination [67,68]. In most normal circumstances, due to the cohesin rings, the sister chromatids are in proximity and therefore the preferred source for homologous recombination and produces a “perfect” repair. However, in situations of reduced SCC, chromosome homologs now have better access to each other, and homologous recombination can occur between homolog chromosomes [10,69]. This can result in multiple repair outcomes, including a crossover event and/or gene conversion events. Following mitosis this recombination between homologs can result in chromosome regions becoming homozygous (mitotic recombination (MR)), resulting in a LOH event. This is consistent with our tumor data in which reduced SCC is associated with elevated MR events. Changes in MR rates are most well demonstrated in recombinase deficient individuals and animal models, such as Bloom syndrome proteins (BLMs) or Werner syndrome helicase (WRN), but we demonstrated that this may also be true for SCC reduced models and syndromes.
Interestingly the variability in the age of tumor onset within LFS is great, with some patients developing tumors within the first year of life, while others have been described to be tumor-free at the age of 74. The timing of tumor onset is influenced by many different factors, including environment and genetic heterogeneity. While these are clearly influential, isogenic animal models under controlled environmental conditions can exclude these factors and focus on genetic aspects. In a p53-centric view, the difference between the survival curves of a p53zy7/zy7 and p53zy7/+ cohort (at T 50, a difference of 241 days vs. 561 days respectively) is rationalized as the time required for the 2nd p53 hit to occur (p53 LOH; p53zy7/+ ➔ p53zy7 or zy7/zy7). The difference between p53zy7/+ and p53+/+ curves is the time required for the first p53 hits to occur (p53+/+ ➔ p53zy7/+). However, the reason that the p53zy7/zy7 tumors are not observed immediately at time 0 is due to the time needed for the additional required tumor-promoting mutations to occur (for example Ras and/or other). We demonstrate that reduced SCC through haploinsufficient loss of esco2 or sgo1, decreased the time between p53zy7/zy7 and p53zy7/+ curves (at T 50 , a 147-day difference) suggesting that reduced SCC enhances the rate of p53 LOH events. Further, since there was no change in survival curves in the p53zy7/zy7 background, esco2 haploinsufficiency is not an added oncogenic hit. It is also interesting that in these isogenic and environmentally controlled models that there is a vast difference in the time of tumor onset within these cohorts. For mouse trp53 heterozygous null in a C57BL6/J background, the first tumors are observed at ~150 days, while the later tumors are at >600 days [48]. What influences this difference? We would speculate that the proportion of cells in an organism with stochastically derived p53 LOH in somatic tissues influences the timing of tumor formation. While measuring p53 LOH in tumors is relatively easy due to the clonal source of the genomic DNA, the ability to measure p53 LOH in somatic cells has been close to impossible. However recent advances in single-cell haplotyping might provide a first-time view of p53 LOH rates in somatic tissue and variability between tissues and individuals. If true, individual evaluations of p53, or other tumor suppressor genes, LOH events in somatic tissues of a patient could provide a cancer risk for those individuals.
[END]
---
[1] Url:
https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1010341
Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.
via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/