Disruption of the TP53 locus in osteosarcoma leads to TP53 promoter gene fusions and restoration of parts of the TP53 signalling pathway

TP53 is the most frequently mutated gene in human cancer. This gene shows not only loss‐of‐function mutations but also recurrent missense mutations with gain‐of‐function activity. We have studied the primary bone malignancy osteosarcoma, which harbours one of the most rearranged genomes of all cancers. This is odd since it primarily affects children and adolescents who have not lived the long life thought necessary to accumulate massive numbers of mutations. In osteosarcoma, TP53 is often disrupted by structural variants. Here, we show through combined whole‐genome and transcriptome analyses of 148 osteosarcomas that TP53 structural variants commonly result in loss of coding parts of the gene while simultaneously preserving and relocating the promoter region. The transferred TP53 promoter region is fused to genes previously implicated in cancer development. Paradoxically, these erroneously upregulated genes are significantly associated with the TP53 signalling pathway itself. This suggests that while the classical tumour suppressor activities of TP53 are lost, certain parts of the TP53 signalling pathway that are necessary for cancer cell survival and proliferation are retained. In line with this, our data suggest that transposition of the TP53 promoter is an early event that allows for a new normal state of genome‐wide rearrangements in osteosarcoma. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.


Introduction
Osteosarcoma is the most common primary bone malignancy.It is usually detected in children and adolescents, often close to the active growth plate of long bones [1].Osteosarcoma is characterised by complex genetic alterations, including hundreds of numerical and structural chromosome aberrations [2][3][4][5][6][7][8].
Thus far, however, no osteosarcoma-specific genetic alteration has been identified.With the exception of the heavily rearranged genome, a consistent genetic pattern between patients is lacking.There is no genetic biomarker that can be used for risk stratification or as a treatment target, which hinders the development of individualised treatment protocols and new treatment strategies.
Consistent with the genetic complexity, most osteosarcomas harbour mutations in TP53 [3,5].A portion of these are hotspot missense mutations.Their pathogenetic significance is not fully understood, but reports from other tumour types suggest that such single nucleotide variants constitute gain-of-function or separationof-function mutations [9][10][11][12][13][14][15].The latter implies that while the tumour suppressor activity of wild-type TP53 is lost, mutant p53 reactivates parts of the TP53 pathway that are important for cancer cell survival and proliferation.Paediatric osteosarcomas are unique among highgrade malignancies in that at least half of the cases show structural instead of single nucleotide variants in TP53 [2,3,5].These structural rearrangements usually affect TP53 intron 1 and thereby separate the promoter region from the coding parts of TP53, often resulting in loss of the latter.Interestingly, the promoter region is not lost, but instead relocated.The fusion of the TP53 promoter region to other parts of the genome is an expected response to the DNA double-strand break, to avoid further deterioration of the affected chromosome.However, the preservation of a relocated TP53 promoter region also enables the erroneous activation of genes other than those originally under TP53 promoter control.Such transfer of regulatory elements is a known driver in neoplasia, commonly denoted promoter swapping/ switching or enhancer hijacking [16,17].A classic example is found in lymphoma, where the coding parts of the transcription factor MYC are put under the control of the regulatory elements of the immunoglobulin gene loci, resulting in transcriptional upregulation of MYC [18,19].Promoter swapping has been shown to operate in bone tumours other than osteosarcoma, for example in chondromyxoid fibroma and aneurysmal bone cyst where strong promoters are juxtaposed to the entire coding sequences of the GRM1 and USP6 genes, respectively [20,21].Previously reported promoter substitutions have typically involved supposedly strong promoters, assumed to be constitutively active in the cell of origin [16].Here, we show that the relocated promoter region of a tumour suppressor gene can induce the expression of other genes erroneously placed under its control.Our data demonstrate that the TP53 promoter region is constitutively active in TP53mutated cells and its activity is accentuated by further genetic damage.Furthermore, genes induced by the TP53 promoter region restore parts of the lost TP53 pathway.Thus, our findings show that TP53 separation-of-function can be achieved through TP53 promoter gene fusions.

Subject information and tumour material
Fresh-frozen tumour biopsies from 148 conventional osteosarcomas were subjected to genomic analyses.The clinical features were typical of conventional osteosarcoma patients.The age of the patients ranged from 3 to 81 years, with a median age of 15 years and a mean age of 20 years, and there were 68 females and 80 males.The discovery cohort consisted of osteosarcomas from 36 patients.Most samples in this cohort were chemotherapy-treated resection specimens.The validation cohort consisted of 108 treatment-naïve diagnostic biopsies.Detailed information is displayed in supplementary material, Tables S1-S3.For comparison, we included osteoblastomas (n = 13), a chondromyxoid fibroma, a phosphaturic mesenchymal tumour of bone, and a parosteal osteosarcoma.All tumour material was obtained after informed consent, and the study was approved by the Regional Ethics Committee of Lund University and the Ethikkommission beider Basel (reference 274/12).

DNA and RNA extractions
Fresh-frozen tumour biopsies were dismembered and homogenised using a Mikro-Dismembrator S (Sartorius AG, Goettingen, Germany).The material was optimally split into two fractions: one was used for immediate DNA extraction and the other, when available, was stored in Qiazol at À80 C for later RNA extraction (Qiagen, Hilden, Germany).DNA was extracted using the DNeasy Blood & Tissue Kit including the optional RNase A treatment (Qiagen).DNA quality and concentration were measured using a NanoDrop ND-1000 and a Qubit 3 Fluorometer (both from Thermo Fisher Scientific, Waltham, MA, USA).The material stored in Qiazol was heated at 65 C for 5 min and RNA was extracted using the RNeasy Lipid Tissue Kit including the optional DNase digestion (Qiagen).RNA quality and concentration were assessed using a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and a NanoDrop ND-1000.

Whole-genome mate pair sequencing for detection of structural variants
To detect structural chromosomal abnormalities, mate pair libraries were prepared for sequencing using the Nextera Mate Pair Library Preparation Kit (Illumina, San Diego, CA, USA).This was done according to the manufacturer's instructions except for the number of shearing cycles, which was increased to 3 cycles.Paired-end 76 base pair reads were generated using an Illumina NextSeq 500 sequencing instrument (Illumina, San Diego, CA, USA).Sequencing depth was on average 3.2Â (mapping coverage 2.13Â) and the mean insert size was 3.0 kbp, resulting in a median spanning coverage of 63.2Â of the human genome (mean 63.1Â, range 5.2Â-119.1Â).All samples were sequenced with high quality and yield; between 12.4 and 115.5 million read pairs were obtained per sample and the average quality scores were 31.3-34.1.Sequencing reads were trimmed using NxTrim v 0.4.2 and subsequently aligned against the GRCh37/hg19 build using the Burrows-Wheeler Aligner v0.7.15 [22,23].To identify structural rearrangements, the sequence data were analysed using the Integrative Genomics Viewer [24,25], as well as the structural variant callers TIDDIT v2.12.1, Delly2 v0.7.8, and Manta v1.2.2 [26][27][28].Structural alterations were considered true when identified by at least two of the three variant callers.
Whole-genome paired-end sequencing of multisampled osteosarcomas Whole-genome paired-end sequencing was performed using the Agilent SureSelect v3 Library Preparation Kit (Agilent Technologies, Santa Clara, CA, USA).Paired-end 150 base pair reads were generated using an Illumina HiSeq 2500 sequencing instrument.Sequencing depth was on average 13.4Â (mapping coverage 14.1Â) and the mean insert size was 0.34 kbp, resulting in a median spanning coverage of 14.5Â of the human genome (mean 14.3Â, range 5.2Â-40.9Â).Sequencing reads were aligned against the GRCh37/ hg19 build using the Burrows-Wheeler Aligner v0.7.15.To identify structural rearrangements, the sequence data were analysed as described above.It is important to stress that whole-genome paired-end sequencing is a less optimal technique to detect structural variants, compared with mate pair sequencing, and therefore requires a higher sequencing depth.The reason for this is the higher spanning coverage of the human genome obtained by mate pair sequencing, due to the analysed DNA fragments being approximately one order of a magnitude larger.In the present study, the median spanning coverage for mate pair data was 63.2Â compared with 14.5Â for paired-end data.
Genome-wide DNA copy number and loss of heterozygosity analyses SNP array analysis was used for combined DNA copy number and loss of heterozygosity investigation.DNA was extracted according to standard procedures from fresh-frozen tumour biopsies and hybridised to CytoScan HD arrays, following protocols supplied by the manufacturer (Thermo Fisher Scientific).Somatic copy number alterations in a proportion of the cases were published by Smida et al in 2017 [7].Data analysis was performed using the Chromosome Analysis Suite v4.1.0.90 (Thermo Fisher Scientific), detecting imbalances by visual inspection, and by segmenting log 2 values using the R package 'copynumber', available via Bioconductor (https://bioconductor.org/packages/release/bioc/html/copynumber.html).The inbuilt pcf function was used with a strict gamma value of 100 to create copy number segments, and the plotFreq function was used to create the frequency plot of losses and gains on chromosome 17.The threshold for gain was set as a log 2 value of 0.2 and the threshold for loss as À0.2.SNP positions were based on the GRCh37/hg19 sequence assembly.TP53 promoter gain is defined as copy number loss, or copy number neutral loss of heterozygosity, of whole or parts of the TP53 coding region coupled to concurrent relative copy number gain of the TP53 promoter region.

Visualisation of structural and copy number variants using Circos plots
Circos plots were generated using the R packages Circlize or RCircos [29,30], by integrating genomic copy number data obtained from either SNP array analysis or whole-genome sequencing and structural variant data based on whole-genome sequencing and the TIDDIT algorithm described above.Copy number segments based on SNP array data were generated as described above.Copy number segments based on sequencing data were generated using CNVkit [31].

Whole-genome low-pass sequencing of single cells
Whole-genome sequencing of cryopreserved primary osteosarcoma cells was performed as described in detail previously [32].In brief, library preparation was performed using a modified single-cell wholegenome sequencing protocol and 77 base pair single reads were generated using a NextSeq 500 sequencing instrument (Illumina).From each assessed tumour, 93 individual cells were sequenced at an average depth of 0.01Â.Copy number analysis was performed using AneuFinder [33], and bin positions were based on the GRCh38/hg38 sequence assembly.

RNA sequencing for detection of gene fusions, expression levels, single nucleotide variants, and indels
Total RNA was enriched for polyadenylated RNA using magnetic oligo(dT) beads.Enriched RNA was prepared for sequencing using the TruSeq RNA Sample Preparation Kit v2 according to the manufacturer's protocol (Illumina).Paired-end 151 base pair reads were generated from the cDNA libraries using an Illumina NextSeq 500 instrument.Sequencing reads were aligned to the GRCh37/hg19 build using STAR v2.5.2b [34].For comparison of relative gene expression levels, data were normalised using Cufflinks with default settings [35] and visualised using the Qlucore Omics Explorer version 3.5 or 3.8 (Qlucore AB, Lund, Sweden).As an outgroup in gene expression analysis, we included osteoblastomas (n = 13) in addition to the 68 osteosarcomas.FusionCatcher v1.0 and STAR-Fusion v1.4.0 were used to identify candidate fusion transcripts from the sequence data [36,37].Single nucleotide variants and indels in the TP53 gene were detected using VarScan v2.4.1 [38] and MuTect v1.1.7 [39].Constitutional variants were excluded based on information from the Genome Aggregation Database (gnomAD v2.1.1)[40].The detected variants were finally confirmed by manual inspection using the Integrative Genomics Viewer.Unsupervised correlation-based principal component analysis was performed on all 68 osteosarcomas using the Qlucore Omics Explorer.

Pathway enrichment analyses
Genes identified as being partners to the TP53 promoter region in cases with TP53 promoter gain were queried as a list against (1) the Pathway Analysis with ANUBIX (PathBIX) database [42], using the Reactome pathway database with a network cut-off set at 0.99 without applied clustering; and (2) the Molecular Signatures Database (MSigDB) [43][44][45], where the H and C1-6 gene sets were used to compute gene overlaps.

RT-qPCR verification of TP53 promoter partner gene upregulation
The relative expression levels of ROR2 (Hs00896174_m1), ELF1 (Hs01111181_m1), and E2F3 (Hs00605457_m1) were investigated using RT-qPCR and TaqMan Gene Expression assays (Thermo Fisher Scientific).The TBP (Hs99999910_m1) gene was used as an endogenous control.Calculations were performed using the comparative Ct method (i.e.ΔΔCt).The experiment included technical triplicates per sample.Samples were assayed using a QuantStudio™ 3 Real-Time PCR System (Thermo Fisher Scientific).
Cell models harbouring TP53::ROR2, TP53::SAT2, or TP53::ADGRE1 A promoter-less vector (pSMPUW Universal Lentiviral Expression Vector; Cell Biolabs, Inc., San Diego, CA, USA) containing the TP53::ROR2 fusion was constructed (GenScript, Piscataway, NJ, USA).The TP53 promoter was represented by the first 2,000 bp upstream of TP53 together with exon 1 and the first 500 bp of intron 1 of TP53.These TP53 sequences were fused to the last 500 bp of ROR2 intron 1 and the coding sequences of ROR2 exons 2-13.This hybrid sequence is denoted TP53::ROR2 and thus contains the complete coding sequence of ROR2 transcript variant ENST00000375715.1 under the control of the TP53 promoter, mimicking the fusion event found in case 9.A vector containing the same ROR2 sequences but lacking TP53 sequences was used as a control.
CRISPR-mediated knockout of TP53 in one mesenchymal and one epithelial cell line was performed as described elsewhere [46].In brief, hCas9 and a guide RNA for TP53 exon 6 were transduced into the TERT-immortalised cell line human foreskin fibroblast BJ-5ta and retinal pigment epithelial cell line RPE-1 (ATCC CRL-4001, CRL-4000, LGC Standards, Middlesex, UK).The cell lines were used in the experiments immediately after purchase and were tested negative for mycoplasma.Antibiotic resistance-selected cells were single cell-cloned and analysed for mutations with the Surveyor Mutation Detection Kit (Integrated DNA Technologies, Inc., Coralville, IA, USA).Clones with detected mutations were validated for homozygous or compound heterozygous mutations using Sanger sequencing or Nextera sequencing (Illumina).This confirmed a 19 bp deletion in TP53 exon 6 in a BJ-5ta clone.Large genomic copy number alterations in this clone were investigated by CytoScan HD array analysis (Thermo Fisher Scientific), revealing a hemizygous deletion of distal 17p, with a break in WRAP53, in all cells.Thus, one TP53 allele was deleted, and the remaining allele harboured a frame-shift mutation, resulting in complete knockout of this gene.In an RPE-1 clone, two separate heterozygous mutations affecting TP53 exon 6, where one allele harboured a 1 bp insertion and the other a 13 bp deletion, were detected.Genomic copy number analysis revealed a normal copy number of chromosome arm 17p, indicative of a complete knockout of TP53 via a compound heterozygous mutation.Both the BJ-5ta TP53 À/À and RPE-1 TP53 À/À clone were transduced with the TP53::ROR2 and ROR2 only vectors, respectively.The osteosarcoma cell lines Saos-2 and MG-63 were acquired, as they are known to harbour the TP53 promoter fusions TP53::SAT2 and TP53::ADGRE1 (previously EMR1), respectively [5] (ATCC HTB-85, ATCC CRL-1427).
Bj5ta TP53 À/À and RPE-1 TP53 À/À harbouring either TP53::ROR2 or ROR2 only as well as Saos-2 and MG-63 were exposed to the DNA-damaging agent cisplatin at concentrations ranging from 1 to 5 μM.The osteosarcoma cell line OSA was used as a control for SAT2 and ADGRE1 expression, respectively.Cells were harvested for RNA extraction 4 days following cisplatin treatment.The relative expression levels of ROR2 (Hs00896174_m1), SAT2 (Hs00374138_g1), and ADGRE1 (Hs00892590_m1) were investigated using RT-qPCR and TaqMan Gene Expression assays (Thermo Fisher Scientific).The TBP (Hs99999910_m1) gene was used as an endogenous control.Calculations were performed using the comparative Ct method (i.e.ΔΔCt).The experiment was performed in biological triplicates with each replicate including technical triplicates per sample.Samples were assayed using a 7500 RT-PCR system (Thermo Fisher Scientific).

Statistical analyses
Statistical analyses were performed using two-tailed Mann-Whitney U-tests.

Results
Ectopic localisation of the TP53 promoter is associated with osteosarcoma in the young By whole-genome mate pair sequencing, TP53 structural variants were detected in 13/36 and 16/36 osteosarcomas of our discovery and validation cohorts, respectively (Figure 1A and supplementary material, Tables S1, S2, S4, and S5).Based on DNA copy number profiles from SNP array analysis, we identified a subset of cases with a copy number profile of chromosome arm 17p that we termed TP53 promoter gain.We defined this pattern as copy number loss, or copy number neutral loss of heterozygosity, of whole or parts of the TP53 coding region coupled to concurrent relative copy number gain of the TP53 promoter region (Figure 1B).This was found in 16/108 cases (Figure 1C and supplementary material, Table S2).Both TP53 promoter gain, determined by SNP array analysis, and TP53 structural variation, determined by whole-genome mate pair sequencing, were associated with young age of onset (Figure 1D,E and supplementary material, Tables S1 and S2).In an additional 24 of the 108 tumours analysed by SNP arrays, we detected a copy number shift within the TP53 locus but lacking at least one criterion for TP53 promoter gain (see supplementary material, Table S2).In all, we identified TP53 structural variants in 40% of osteosarcomas (see supplementary material, Tables S1 and S2).Cases that lacked a TP53 structural variant displayed either wild-type TP53, larger copy number losses, or gains of chromosome arm 17p that did not directly affect TP53 integrity, a complete homozygous loss of the gene, and/or single nucleotide variants or indels leading to missense or frameshift mutations (see supplementary material, Tables S1 and  S2).Osteosarcomas with TP53 missense mutations were mutually exclusive from those with TP53 structural variants.The latter cases showed a higher number of chromosome breaks genome-wide than those without a structural variant affecting TP53, with a clustering of breakpoints to chromosome arms 6p, 8q, 12q, 17p, and 19q (Figure 1F,G and supplementary material, Figure S1).The genomic breakpoint pattern was highly similar to the copy number pattern for chromosome 17 (Figure 1C,H).In summary, TP53 missense mutations were mutually exclusive from TP53 structural variants, and the latter correlated with young age, a high number of genome-wide chromosome rearrangements, and preservation of the TP53 promoter region.
The recurrent pattern of a transposed TP53 promoter region suggested that it regulates other genes in a fashion that favours tumour development, through gene fusion or promoter swapping events [16].To test this hypothesis, we assessed gene fusions and expression levels by RNA sequencing, extended the DNA copy number analysis, and evaluated TP53 status in diagnostic biopsies, resection specimens, metastases, and/or individual tumour cells from TP53-rearranged osteosarcomas as outlined below.
Transposition of the TP53 promoter is a single early event that can spark genome-wide rearrangements and oncogene amplification In a subset of osteosarcomas, DNA sequencing supported intra-and inter-chromosomal events (inversions, insertions, or translocations) that transposed the TP53 promoter without compromising chromosome stability (Figure 2A-C).In these cases, no further rearrangement involving the TP53 promoter, or its partner region, was detected.In another subset of osteosarcomas, transposition of the TP53 promoter was the initiating event that generated unstable, most likely dicentric, derivative chromosomes (Figure 2D,F and supplementary material, Figures S2 and S3).Such derivative chromosomes are known to repeatedly break and re-join with multiple partner chromosomes in osteosarcoma [48,49].This amplifies both the TP53 promoter fusion gene and additional genomic regions of potential importance for osteosarcoma progression, such as regions on chromosomes 6, 12, and 17 (Figure 2F).Notably, this sequence of events is different from chromothripsis and multi-way translocations, which in other subtypes of bone tumours are known to generate gene fusions (Figure 2G,H) [50,51].We found no evidence for the generation of TP53 structural variants or TP53 promoter gene fusions through one massive burst of genome rearrangements in osteosarcoma.Instead, the genomic footprint of TP53 promoter gene fusions mimics that of oncogene amplification through breakage-fusion-bridge cycles, found in, for example, low-grade osteosarcoma with ring chromosomes and MDM2 amplification (Figure 2I).Thus, according to our model, transposition of the TP53 promoter is an early spark for genome-wide rearrangements in osteosarcoma, including whole-genome doubling [52][53][54].Results from whole-genome sequencing of multisampled bulk and single-cell tumour DNA supported this model.TP53 fusion-positive osteosarcomas harboured their respective fusions in all investigated diagnostic biopsies, post-chemotherapy resection specimens, and metastases, as well as in all investigated individual neoplastic cells (Figure 2J,K and supplementary material, Figures S3A-F, S4-S6, and Tables S1-S7).

The transposed TP53 promoter induces the expression in vivo of partner genes which are themselves part of the TP53 signalling pathway
To assess if the TP53 promoter induces the expression of its respective partner genes in vivo, we analysed the expression levels for the partner genes and the WRAP53 gene.The TP53 promoter is bidirectional and TP53 promoter gene fusions in osteosarcoma normally induces both TP53 and WRAP53 [55], enabling elevated expression levels of the latter to be used as a proxy for adequate representation of neoplastic cells.Figure 3A-F displays three representative osteosarcomas that harbour whole or parts of ROR2, MAP4K4, and E2F3, respectively, placed under the  control of the TP53 promoter.TP53 exon 1 and partner gene exons placed under the TP53 promoter showed higher expression levels than exons excluded from the fusions.Other notable genes that were placed under the control of the TP53 promoter included, for example, ELF1, CDC5L, H3-3B, YTHDF1, ZNF780A, SUZ12, NDEL1, CDKN1A, and NFYA.In whole-transcriptome sequenced cases with an adequate representation of neoplastic cells, an upregulation of the 3' partner gene was seen compared with other osteosarcomas and osteoblastomas (see supplementary material, Figures S7  and S8).Although none of the 3' partners were recurrent in themselves, their respective functions seemed to merge on common pathways.For example, several partner genes have been shown to regulate cartilage and growth plate development, osteoclast formation, or to be implicated in cancer development in other tumour types [56][57][58][59][60][61][62][63][64][65].Intriguingly, pathway enrichment analyses showed that TP53 promoter partner genes in cases with TP53 promoter gain are themselves part of the TP53 signalling pathway (p < 0.001, FDR < 0.05, Figure 3G and supplementary material, Table S8).In line with this, the global gene expression pattern of individual osteosarcomas did not correlate with their TP53 mutation status (see supplementary material, Figure S9).Thus, in TP53-mutated cases, be it by structural or single nucleotide variants, a compensatory mechanism could restore part of the TP53 wild-type phenotype.In supplementary material, Tables S1-S3, the matched genomic and transcriptomic data for all detected TP53 promoter gene fusions are displayed.A schematic diagram showing which gene(s) (or parts of genes) are placed under the control of the TP53 promoter and associated transcriptomic data in all cases harbouring a TP53 structural variant is provided in supplementary material, Figure S7.Taken together, these data demonstrate that the transposed TP53 promoter is active in osteosarcoma and that it induces the expression of genes important for tumour and bone development which, at least in some instances, are themselves involved in the TP53 signalling pathway.

Cisplatin evokes gene expression through the TP53 promoter in vitro
As a proof-of-concept, we modelled the above findings in vitro.First, we knocked out TP53 in one mesenchymal (BJ-5ta) and one epithelial (RPE-1) cell line by CRISPR genome editing and single-cell cloning.Second, we constructed a vector containing the TP53 promoter region fused to the coding DNA sequence of ROR2 (TP53::ROR2, based on the fusion event found in case 9).As a control for the absence of the promoter, we used the same vector but without the TP53 promoter region (ROR2 only).Third, we exposed TP53 À/À cells harbouring either TP53::ROR2 or ROR2 only to the DNA-damaging agent cisplatin (Figure 4A,B).We found that the TP53 À/À background, even in the absence of cisplatin, was sufficient to activate the TP53 promoter and elicit expression of a gene placed under its control (Figure 4C,D).Induced DNA damage through cisplatin treatment further increased the expression level of the TP53 promoter partner gene.The same phenomenon was detected in the osteosarcoma cell line Saos-2, known to harbour the TP53::SAT2 fusion [5] (Figure 4E).In the highly proliferative osteosarcoma cell line MG-63 harbouring a TP53::ADGRE1 (formerly EMR1) fusion [5,66], there was a clear upregulation of the partner gene compared with control (Figure 4F).However, these cells were heavily affected by the cisplatin treatment and the ADGRE1 expression level decreased with increasing cisplatin concentration.A probable explanation for this effect could be that their high proliferation rate renders them more susceptible to cisplatin.Thus, in a TP53 À/À background, a constitutively active TP53 promoter can induce expression of a gene transposed into its vicinity in a fashion that can be accentuated by additional genetic damage.

Introduction of TP53::ROR2 in TP53-null cells reverts their global gene expression profile back towards the wild-type phenotype
Finally, we used whole-transcriptome sequencing to assess the global gene expression patterns in cultured wild-type BJ-5ta, BJ-5ta TP53 À/À , and BJ-5ta TP53 À/À harbouring TP53::ROR2 cells.Unsupervised principal component analysis showed that knockout of TP53 clearly affected the global gene expression profile, i.e. there was a clear distinction between the wild-type cells and those with a complete knockout of TP53 (Figure 4G).Intriguingly, in TP53 promoter gene fusions in osteosarcoma 153 TP53 À/À cells that harboured the TP53::ROR2 fusion, the global gene expression profile reverted back towards that of wild-type cells.This is in line with the above findings that TP53 promoter swapping events counterbalance loss of wild-type TP53 function through separation-of-function mutations.
following decades, efforts from several research groups confirmed these rearrangements, and genomic patterns similar to what we term here as TP53 promoter gain were reported in osteosarcoma and subtypes of soft tissue sarcomas [69].In parallel, somatic structural variants affecting TP53 were also found in subsets of leukaemia and carcinomas, including chronic myelogenous leukaemia [70][71][72], lung cancer [73], and prostate cancer [74][75][76][77].Such variants inevitably silence the TP53 gene, but evidence for concomitant gain-of-function or separation-of-function mechanisms through structural variation has not been described.This is likely because essentially all studies investigating TP53 mutations have focused on single nucleotide variants [9][10][11]14,15].Here, we propose that dislocating the TP53 promoter region from the coding parts of the gene represents an effective means to accomplish separation-of-function mutations, particularly in young osteosarcoma patients who have not lived the long life thought necessary to accumulate a high number of somatic single nucleotide variants.By abrogating normal TP53 function while simultaneously inducing cancer-related signalling pathways, structural variants of the TP53 locus likely mirror the mode of action of many TP53 missense mutants [11,14].For TP53 structural variants and missense mutations to come into play as separationof-function mutations, the cancer cell must likely be deprived of the production of normal TP53 through losses, nonsense, or frameshift mutations affecting the other allele.
There may be at least two probable reasons for not recognising TP53 separation-of-function mutations through structural variation in previous studies.First, even though TP53 promoter gene fusions have been reported in osteosarcomas and a handful of prostate adenocarcinomas, stray cases of other carcinomas and melanoma, and in high-grade sarcomas, no recurrent 3' partner has been identified [3,5,76,[78][79][80][81].The TP53 promoter is thereby a promiscuous fusion partner that has the capacity to induce the expression of many different genes.This, however, does not exclude an important functional outcome.There are numerous examples of interchangeable partners of gene fusions that are disease-specific, strongly indicating that activation of a specific pathway, in one way or the other, is the key feature for transformation [16,80].Second, the TP53 gene fusions in osteosarcoma involve transfer of promoter activity.Although a well-recognised concept in neoplasia, its detection requires access to matched highquality genomic and transcriptomic data.We generated a unique combined dataset for a large series of paediatric and adult osteosarcomas, sampled across several regions and time points.This enabled us to show for the first time that a promoter activated by genetic damage can induce cancer-driving genes transposed into its vicinity.Importantly, we found this recurrent phenomenon to occur in all neoplastic cells of TP53-rearranged osteosarcomas, rendering it a particularly meaningful mechanism to explore further for therapeutic applications.Whether this also holds true for select cases of carcinomas, melanoma, and other high-grade sarcomas remains to be investigated.In this context, it is worth stressing that TP53 promoter gene fusions have been detected in only a minor percentage of reported prostate adenocarcinomas and almost exclusively co-occur with the pathognomonic TMPRSS2::ERG fusion [76,79].
We have focused on the TP53 promoter region as the driver of ectopic gene expression.It is possible that the complex rearrangements in osteosarcoma result in relocation and copy number changes of enhancer regions as well [82].This may accentuate the effect of the TP53 promoter depending on which other genomic regions are also translocated into its vicinity.For instance, in lineage-ambiguous stem cell leukaemia and liposarcoma, copy number amplifications can amplify the effect of enhancers and even create so-called 'super enhancers' [83,84].In acute lymphoblastic leukaemia, deletions can instead lead to an overexpression of FLT3 due to enhancer repositioning [17].In either case, enhancer repositioning and amplification can have a profound effect on transcriptomic deregulation, necessitating a comprehensive search beyond promoter regions for such phenomena in future studies.
DNA sequencing data and interpreted the results.KHS, VD, MK, MB, TJ, DB and KHN performed SNP array analyses and interpreted the results.KHS, JN and LM performed RNA extractions and RNA sequencing experiments.KHS, VD and KHN carried out bioinformatic analyses of RNA sequencing data and interpreted the results.HvdB, DCJS and FF applied low-pass whole-genome sequencing on single cells.KHS, LC, LM and JN conducted genomic PCR, RT-PCR, Sanger sequencing and RT-qPCR experiments, designed lentiviral vectors and performed the in vitro experiments.KHS and KHN prepared the manuscript, with contributions from all other authors

Figure 1
Figure 1 Legend on next page.
152 KH Saba et al © 2023 The Authors.The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.www.pathsoc.orgJ Pathol 2024; 262: 147-160 www.thejournalofpathology.com

Figure 1 .
Figure 1.Structural variants in TP53 are associated with young age at onset and a high number of chromosomal breaks genome-wide.(A) Schematic representation of TP53 structural variation in a discovery (n = 36) and a validation osteosarcoma cohort (n = 36).The TP53 promoter region is marked in yellow [47].The region used to represent the TP53 promoter region in in vitro experiments is marked by a dashed box.Arrowheads and arrows represent structural variants involving sequences 5' and 3' of the breakpoint, respectively.(B) DNA copy number profile of 17p in a representative osteosarcoma with gain of the TP53 promoter region.(C) Frequency plot of genomic copy number gain (red) and loss (blue) for chromosome 17 across conventional osteosarcomas (n = 108).(D) Age distribution of osteosarcoma patients without (n = 92) and with (n = 16) TP53 promoter gain as determined by SNP array analysis.**p < 0.01, two-tailed Mann-Whitney U-test.(E) Age distribution of osteosarcoma patients without (n = 43) and with (n = 29) TP53 structural variants as determined by DNA mate pair sequencing.*p < 0.05, two-tailed Mann-Whitney U-test.(F) Breakpoint burden distribution of osteosarcomas without (n = 43) and with (n = 29) TP53 structural variants as determined by DNA mate pair sequencing.***p < 0.001, two-tailed Mann-Whitney U-test.(G) Circos plot showing genome rearrangements in a representative osteosarcoma with structural variation in TP53.Light blue and dark grey lines denote intra-and inter-chromosomal events, respectively.The dark blue line represents the specific structural variant relocating the TP53 promoter region.(H) All detected breakpoints affecting chromosome 17 across conventional osteosarcomas (n = 72).Red histograms represent total counts within a specific genomic window.

Figure 2 .
Figure 2. Transposition of the TP53 promoter is a single early event that can spark genome-wide rearrangements and oncogene amplification.(A-C) Intrachromosomal events resulting in TP53 gene fusions (green lines).(D-F) Interchromosomal events resulting in TP53 gene fusions (dark blue lines).The derivative dicentric chromosomes repeatedly break and re-join with multiple partner chromosomes.(G-I) The genomic footprints of (G) chromothripsis in a chondromyxoid fibroma, (H) a multi-way translocation in a phosphaturic mesenchymal tumour of bone, and (I) breakage-fusion-bridge cycles in a parosteal osteosarcoma are exemplified.(J) Genomic copy numbers in a representative individual cell from an osteosarcoma with a TP53::MAP4K4 fusion.(K) Heat map of genomic copy numbers across all 43 sequenced individual neoplastic cells of the TP53::MAP4K4 fusion-positive case.Each row of copy number states represents a single cell.

Figure 3
Figure 3 Legend on next page.

Figure 3 .
Figure 3.The bidirectional TP53 promoter induces the expression of WRAP53 and oncogenes in vivo.(A) Exon expression levels in case 9, in which TP53 intron 1 is fused to ROR2 exons 2-9.(B) Normalised gene expression levels.(C) Exon expression levels in case 22, in which TP53 intron 1 is fused to MAP4K4 exons 1-15, including coding regions for the kinase domain, in the opposite direction.(D) Normalised gene expression levels, including all exons of MAP4K4 in case 22. (E) Exon expression levels in case OS046, in which TP53 intron 1 is fused to regions upstream of the complete coding sequence of E2F3.(F) Normalised gene expression levels.Different colours mark individual exons, with the darkest red signifying exon 1 and the darkest blue signifying the terminal exon of the given gene.The exact colour scale varies depending on the number of exons in a gene.Dotted lines indicate the fusion points.Red dots mark the case under investigation.OB, osteoblastoma; OS, osteosarcoma.(G)The TP53 promoter partners ELF1, NFYA, E2F3, CDKN1A, CDC5L, H3-3B, SNRPC, and MAP4K4 are connected to the TP53 signalling pathway, either as direct downstream effectors of TP53 or several steps downstream.Genes marked in blue are part of the TP53 pathway, while those marked in green are the TP53 promoter partners that were queried against the MSigDB and PathBIX database.Overlaps are marked in both blue and green.p < 0.001, FDR < 0.05.