Gene amplification in neoplasia: A cytogenetic survey of 80 131 cases

Gene amplification is a crucial process in cancer development, leading to the overexpression of oncogenes. It manifests cytogenetically as extrachromosomal double minutes (dmin), homogeneously staining regions (hsr), or ring chromosomes (r). This study investigates the prevalence and distribution of these amplification markers in a survey of 80 131 neoplasms spanning hematologic disorders, and benign and malignant solid tumors. The study reveals distinct variations in the frequency of dmin, hsr, and r among different tumor types. Rings were the most common (3.4%) sign of amplification, followed by dmin (1.3%), and hsr (0.8%). Rings were particularly frequent in malignant mesenchymal tumors, especially liposarcomas (47.5%) and osteosarcomas (23.4%), dmin were prevalent in neuroblastoma (30.9%) and pancreatic carcinoma (21.9%), and hsr frequencies were highest in head and neck carcinoma (14.0%) and neuroblastoma (9.0%). Combining all three amplification markers (dmin/hsr/r), malignant solid tumors consistently exhibited higher frequencies than hematologic disorders and benign solid tumors. The structural characteristics of these amplification markers and their potential role in tumorigenesis and tumor progression highlight the complex interplay between cancer‐initiating gene‐level alterations, for example, fusion genes, and subsequent amplification dynamics. Further research integrating cytogenetic and molecular approaches is warranted to better understand the underlying mechanisms of these amplifications, in particular, the enigmatic question of why certain malignancies display certain types of amplification. Comparing the present results with molecular genetic data proved challenging because of the diversity in definitions of amplification across studies. This study underscores the need for standardized definitions in future work.


| INTRODUCTION
Gene amplification in mammals, first identified in cells developing resistance to methotrexate, 1 plays a crucial role in the development and progression of many types of cancer through excess copy number and consequent overexpression of oncogenes.3][4][5] Genomic amplifications can manifest cytogenetically as extrachromosomal acentric, autonomously replicating circular DNA structures denoted double minutes (dmin) or as large intrachromosomal homogeneously staining regions (hsr) that lack the longitudinal differentiation ordinarily shown by G-banding.dmin were first reported in the mid-1960s in human embryonal malignant tumors 6,7 and in Rous sarcoma virus (RSV)-induced sarcomas of mice, 8 and 10 years later, Biedler and Spengler 9 first described the occurrence of hsr in two cell lines derived from neuroblastoma patients.Subsequently, these structures have been identified in a variety of tumor cells from experimental animals and humans 10 but have so far not been recorded in non-neoplastic tissues.2][13][14] In addition to dmin and hsr, ring chromosomes (r) in neoplastic cells, first reported in the 1960s in acute myeloid leukemia 15 and breast cancer, 16 and later found in a large variety of benign and malignant human neoplasms, have often been shown to contain low to moderate level amplified gene regions. 17Different mechanisms have been proposed to drive dmin formation, including chromosome breakage leading to simple circularization of an excised chromosome segment or co-ligation of multiple fragments induced by chromothripsis.hsr and ring formation can also be an outcome of multiple mechanisms, such as breakage-fusion-bridge events, intra-chromosomal tandem duplications, or neochromosomes. 4,18ne amplification has emerged as an important diagnostic biomarker in cancer, associated with poor prognosis, resistance to therapy, and as a guide to the development of targeted therapies.For example, in breast cancer, amplification of the human epidermal growth factor receptor 2 (ERBB2 alias HER2), found in approximately 20% of the cases, is associated with aggressive tumor behavior and poor clinical outcome; it is a well-established biomarker for the selection of patients for HER2-targeted therapies, such as trastuzumab 19,20 that has significantly improved patient outcomes.Another wellestablished example is lung cancer, where amplification of the MET oncogene, found in 5%-10% of non-small cell lung cancers, is associated with poor prognosis and its detection is critical for the selection of patients who are likely to benefit from MET-targeted therapies. 21plification of specific genes has been observed in multiple cancer types, and some of the most commonly amplified genes include MYC, EGFR, CCND1, and MDM2. 14 far, no comprehensive survey of the cytogenetic markers associated with gene amplification in human neoplasia has been performed.In this study, we examined the prevalence of dmin, hsr, and ring chromosomes in a cohort of 80 131 cytogenetically investigated cases that encompass a range of hematologic disorders and benign and malignant solid tumors.We present a comparative assessment of the characteristics and distributions of these three aberration types across different tumor entities.

| MATERIALS AND METHODS
Clonal chromosome abnormalities reported in the literature were extracted from the Mitelman Database of Chromosome Aberrations and Gene Fusions in Cancer 22 ; https://mitelmandatabase.isb-cgc.org).
When queried on September 19, 2022, it contained 74 466 neoplastic disorders with at least one clonal numerical and/or structural chromosome change.We then added 8148 unpublished cases from our laboratory, making a total of 82 614 cases available for evaluation.For the present investigation, we excluded 2483 cases that could not be classified as benign or malignant.Thus, a total of 80 131 cases remained for analysis, comprising 62 742 hematologic malignancies and 17 389 solid tumors (4642 benign and 12 747 malignant).
All dmin, hsr, and r were ascertained.Descriptions of such aberrations preceded by a question mark (?) were excluded.A total of 4130 cases had at least one dmin, hsr, or ring chromosome.For each case the combinations dmin and hsr (dmin + hsr), dmin and r (dmin + r), hsr and r (hsr + r), dmin or hsr (dmin/hsr), and dmin or hsr or r (dmin/hsr/r) were determined.Gene content per cytoband and band length (number of nucleotides) were retrieved from the Ensembl genome database assembly GRCh38 (http://www.ensembl.org/index.html).
Abbreviations of all tumor morphologies used throughout the text, and in tables and figures are provided in Table 1 and Table S1.

| RESULTS
Table S1 shows detailed information about numbers and frequencies of dmin, hsr, and r as well as combinations of the three aberration types, that is, dmin + hsr, dmin + r, hsr + r, dmin/hsr, and dmin/hsr/r.The occurrences of dmin + hsr, dmin + r, or hsr + r are very rare in all tumor entities and hence these frequencies were not further evaluated.Table 1 presents a summary of the findings in the six major tumor groups (hematologic disorders, benign and malignant epithelial and mesenchymal tumors, and malignant neuroectodermal tumors) comprising 53 specific tumor entities studied in sufficient numbers (>100) to allow meaningful conclusions.The results are presented graphically in Figures 1-4.The major results for the three aberration types are reported below.

| Double minutes
A total of 1036 cases (1.3%) displayed dmin, as the sole anomaly in 36 cases.In the hematologic disorders (HEM), dmin were present tumors in that malignant tumors have higher frequencies of hsr than their benign counterparts, 394/12747 (3.1%) vs. 8/4642 (0.2%), with the highest frequency seen in MET (4.0%).Notably, all types of benign solid tumors have the same low hsr frequency as the hematologic disorders.When considering malignant epithelial tumors, hsr frequencies vary considerably depending on anatomic site, from 0% in liver cancer to 14.0% in head and neck carcinomas.Marked differences were also apparent when METs were subdivided according to morphologic subtypes.For example, the hsr frequency is more than three times higher in squamous cell carcinomas than in adenocarcinomas, 11.6% versus 3.6%.The variation among MMT and MNET is also pronounced; in MMT from 0% in synovial sarcoma to 5.1% in myogenic sarcomas, and in MNET from 0% in peripheral neuroectodermal tumors (PNETs) to 9.0% in NB.In contrast to the findings for dmin, all astrocytomas, irrespective of grade, showed the same very low frequencies of hsr.
The chromosomal localization could be determined for a total of 691 hsr (300 in hematologic disorders, 385 in malignant solid tumors, and 6 in benign solid tumors).Table 2 shows the distribution of the hematologic disorders and malignant solid tumors; additional information may be found in Table S2.All chromosomes, except the Y chromosome, contained an hsr.An intriguing observation was the pronounced involvement of chromosome 11, which was affected in 39.3% of the hematologic disorders and 16.6% of malignant solid tumors (for details about band localizations, see below); otherwise, no conspicuous preferential chromosomal involvement was evident.Nevertheless, a significant association (R 2 = 0.6729) emerged when comparing hsr frequencies per chromosome between hematologic disorders and malignant solid tumors.
There was no discernible relationship between numbers of hsr and chromosome size within the entire dataset, neither in the hematologic disorders nor in malignant solid tumors.
A total of 526 hsr were localized to 169 specified bands (425 in 157 G-band negative bands and 101 in 163 G-band positive bands).No association was found between numbers of hsr/band and numbers of cancer-associated cytogenetic breakpoints/Mb or number of proteincoding genes/Mb (Table S2).The hsr are widely scattered throughout all chromosome bands with no apparent clustering except for 11q13 and 11q23, constituting 8.7% and 16.9% of all hsr, respectively.There is a mutually exclusive involvement of these two bands in that 11q23 is preferentially affected in AML, MDS, and CML (77/89, 87%), whereas 11q13 almost exclusively is affected in head and neck squamous cell carcinomas (37/46, 80%).Conversely, 11q13 is practically never involved in hsr formation in hematologic disorders (2/46, 4%) and 11q23 is only rarely affected in malignant solid tumors (5/89, 6%).

| Ring chromosomes
A total of 2689 cases (3.4%) had at least one r, found as the sole anomaly in 183 cases.The great majority (67%) of all r were seen as supernumerary structures without any information about possible Frequencies (%) of dmin, hsr, r, dmin/hsr, and dmin/hsr/r in the six major tumor groups, that is, hematologic disorders (HEM), benign epithelial tumors (BET), benign mesenchymal tumors (BMT), malignant epithelial tumors (MET), malignant neuroectodermal tumors (MNET), and malignant mesenchymal tumors (MMT).
chromosomal origin.The frequencies vary substantially among different tumor groups, from 2.3% in HEM to 17.1% in MMT, but the variation is even more pronounced between specific tumor entities.
The hematologic disorders in general show fairly similar, low frequencies (0.6%-3.7%) with the exception of 10.7% in acute erythroleukemia (AML M6).The frequency in benign solid tumors is generally higher (4.3%) than in hematologic disorders.The most extensively studied BETs are adenomas with the highest frequency of rings (5.6%) but showing great variation depending on tumor location: 0.9% in the thyroid and 9.9% in salivary glands.Generally, BMT shows a similar frequency (4.1%) as BET, ranging from 0% in fibroblastic/myofibroblastic tumors to 7.4% in myogenic tumors (leiomyoma and rhabdomyoma).It may be mentioned that a subset of lipomas, namely lipoblastoma (Table S1), shows a high frequency of rings (8.9%).
Among malignant solid tumors, the frequencies are in general similar among neuroectodermal and epithelial tumors (3.2%-4.6%),but substantially higher among mesenchymal tumors (17.1%).The frequency of rings in MET is quite similar in different organ locations (0%-7.3%),with one exception: 20.8% in pancreatic carcinomas.There are no apparent differences among different morphologic types, for example, 3.9% in squamous cell carcinomas, 4.8% in adenocarcinomas, and 5.3% in transitional cell carcinomas.Among MMT, which contains tumor types with the highest frequencies of rings, there is an extensive variation from 0.8% in Ewing sarcoma to 47.5% in liposarcoma.Among the three best-studied liposarcoma subtypes, there is a dramatic difference from 3.9% in myxoid/round cell liposarcomas to 68.6% in dedifferentiated and 76.6% in well-differentiated liposarcoma.
The chromosomal distribution of all encountered ring chromosomes is shown in Table S3.Rings were attributed only to whole chromosomes due to the uncertainty in defining breakpoints.A total of 799 ring chromosomes could be localized to individual chromosomes.
All chromosomes were found to be involved in ring formation ranging from 5 affecting the Y chromosome to 115 affecting chromosome Frequencies (%) of dmin and hsr in the 53 tumor subtypes presented in Table 1.Within each of the six major tumor groups (HEM, BET, BMT, MET, MNET, and MMT), tumor subtypes were arranged in decreasing order of the dmin frequency.
F I G U R E 3 Frequencies (%) of dmin/hsr and r in the 53 tumor subtypes presented in Table 1 and Figure 2. Within each of the six major tumor groups (HEM, BET, BMT, MET, MNET, and MMT), tumor subtypes were arranged in decreasing order of the dmin/hsr frequency.
F I G U R E 4 Frequencies (%) of dmin/hsr/r in decreasing order of frequency in the 52 tumor subtypes (Table 1) with at least one of these aberrations.The figure shows the relative contributions of dmin, hsr, and r.
involved in solid tumors, in particular mesenchymal tumors, but not in hematologic disorders.Chromosome 1 is often affected in both hematologic disorders and solid tumors.

| Combinations of dmin, hsr, and r
First, we compared the pairwise frequencies of dmin and hsr in the 53 tumor entities presented in Table 1.No obvious association was found (R 2 = 0.1891).We then compared the frequencies among the six major tumor groups (HEM, BET, BMT, MET, MNET, and MMT).As shown in Figure 1, there were no apparent differences among the hematologic disorders and benign tumors, all having very low frequencies of both dmin and hsr.Among the malignant solid tumors, the frequencies are quite similar in MET and MMT, whereas in MNET dmin are substantially more common than hsr.
Figure 2 shows a detailed comparison between dmin and hsr in the 53 individual tumor entities.Again, no obvious differences were noted among the various subtypes of HEM, BET, and BMT, all having similarly low frequencies.Among the malignant solid tumors, dmin are more common than hsr in most tumor entities; in some instances, most notably in NB, pancreatic carcinoma, astrocytoma grades III-IV, and PNET, the dmin frequencies are dramatically higher.Two tumor typescarcinomas of the head and neck and breast cancerdeviate from the general pattern in that hsr are substantially more common than dmin.When the combined frequencies of dmin and/or hsr (dmin/hsr) are considered (Figure 3), the highest frequencies (>15%) are seen in NB, pancreatic cancer, astrocytoma grades III-IV, and head and neck carcinoma.
Next, we compared the pairwise frequencies of dmin/hsr and r in the 53 tumor entities presented in Table 1.No association was found (R 2 = 0.0728).Figure 1 shows that rings are more common than dmin/hsr in HEM, BET, BMT, and MMT, equally common in MET, but substantially less frequent in MNET. Figure 3 shows that among the hematologic disorders and benign tumors, with a generally low frequency of dmin/hsr, ring chromosomes are somewhat more common, most apparent in salivary gland adenomas and leio-/rhabdomyomas.
Among malignant solid tumors, the general pattern is that dmin/hsr dominate over r in most malignant epithelial and neuroectodermal tumors, in particular in NB, but also in head and neck carcinomas, stomach cancer, astrocytoma grades III-IV, PNET, and retinoblastoma.
In contrast, ring chromosomes dominate over dmin/hsr among most MMTs, in particular in liposarcoma and fibroblastic/myofibroblastic sarcomas.
T Finally, we combined all data on dmin, hsr, and r.In total, 4130 cases (5.2%) had either dmin, or hsr, or r, or a combination of these aberrations (dmin/hsr/r).The frequencies in the major tumor groups are presented in Table 1 and Figure 1.Obviously, amplifications in the form of dmin/hsr/r are decisively more common in all three malignant solid tumor types (MET, MNET, and MMT) as compared to the hematologic disorders and benign epithelial and mesenchymal solid tumors.
Among the hematologic disorders with an overall frequency of 3.2%, the incidence is lowest in CML (0.7%) and highest in MDS (4.9%).The frequencies are fairly similar in BET (5.3%) and BMT (4.6%); among individual tumor types only adenomas of the salivary glands stand out (11.3%).Among MET with 10.4% dmin/hsr/r-containing cases, very large differences were noted with 39.3% in pancreatic cancer and only 2.6% in skin cancer.Irrespective of tumor location, we noted a higher frequency of dmin/hsr/r in squamous cell carcinomas as compared to adenocarcinomas (16.7% vs. 10.0%),due to a substantially higher contribution of hsr in squamous cell carcinomas (11.6% vs. 3.6%).The highest frequency was seen in MMT (21.9%), in which there was an extensive variation from 2.2% in Ewing sarcoma to 52.2% in liposarcoma.Among MNET with a total of 15.9%, the frequency varies from 3.3% in ependymoma to 38.9% in NB. Figure 4 presents the sums of the frequencies of dmin, hsr, and r for the 52 tumor subtypes with at least one of these aberrations.The variation is great, from 0.8% in CML to 55.6% in liposarcoma.The general pattern is that benign solid tumors and hematologic disorders cluster at the lower end of the scale and that MNET and MMT dominate among those with the highest frequencies.In fact, 9 of the 10 tumors with the highest dmin/hsr/r frequencies belong to these two categories.The malignant epithelial tumors are scattered over the entire spectrum from 2.5% in skin cancer to 47.8% in pancreatic cancer.Figure 4 also shows the relative contributions of dmin, hsr, and r.As shown, one aberration type dominates in some tumors but may be negligible or absent in others, whereas all three are equally frequent in some tumors.In total, all three aberrations were found in 35 of the 52 tumor types, and among the remaining 17 tumors, 9 had only dmin and r, 3 had hsr and r, 1 had dmin and hsr, and 4 had only r.

| DISCUSSION
Gene amplification plays a critical role in cancer development and progression by leading to excess copy numbers and subsequent overexpression of oncogenes.7,[23][24][25][26] It has been convincingly demonstrated that dmin and hsr may be seen as two interrelated alternative forms of amplification. 11,13,14Amplified sequences contained in dmin can integrate into chromosomes to form hsr and vice versa, that is, hsr sequences can be excised to form dmin. 12 On the other hand, ring chromosomes do not, or only rarely, form dmin or hsr, but can linearize to rod chromosomes that in contrast to hsr show light and dark bands.It should be mentioned that not all rings found in tumor cells represent gene amplification.Some rings are just composed of a circularized chromosome, or part of a chromosome, without any copy number gain. 17,27Thus, the frequency of rings, which in general dominate over both dmin and hsr, may provide an overestimation of the frequency of amplification.For example, a subset of ring chromosomes involving chromosome 7, which are preferentially found in hematologic disorders, may cause deletions of material from 7q, a characteristic aberration in myeloid neoplasms. 22e current study is the most extensive cytogenetic survey of gene amplification in benign and malignant neoplasms.The findings provide comprehensive data on the frequencies and patterns of gene amplification markers across a wide range of neoplastic disorders and highlight the variable prevalence and distribution of dmin, hsr, and ring chromosomes.Before discussing the salient features of the present study, we would like to draw attention to some advantages and disadvantages related to cytogenetic and molecular identification of gene amplification in neoplasia.Cytogenetics is admittedly a blunt method to detect and define gene amplification in tumor cells.Low numbers of dmin and small-sized hsr may be missed, and the distinction between small ring chromosomes and dmin may sometimes be problematic.Furthermore, no information about the copy numbers or the origin of amplified sequences is obtained; not even hsr mapped to a particular chromosome band always provides a clue to the chromosomal origin of the amplicon.Similarly, the chromosomal origin of the amplicons in rings is usually not possible to identify.In both hsr and r, co-amplified sequences from one or more chromosomes may contain pathogenetically important target genes.Finally, it should be pointed out that amplicons could arise from mechanisms that are either impossible (e.g., episomes 28 ) or difficult (e.g., scattered amplification 29 ) to detect at chromosome banding analysis.Hence, further studies of gene amplification in cancer are likely to benefit from a combination of cytogenetic and molecular genetic approaches.In contrast to cytogenetics, molecular genetic analyses may reveal copy numbers and possible target genes.Indeed, molecular analyses have shown that a large fraction of rings cytogenetically described as +r in, for example, bone and soft tissue tumors in fact contain material from chromosome 12, and sometimes also from other chromosomes. 29,30On the other hand, cytogenetics provides information that is difficult to obtain from sequence-based data about the structural organization of amplified sequences.Furthermore, the strength of the present cytogenetic study is the large numbers of tumors investigated, across all types of human neoplasms which by far surpasses the series of molecularly investigated tumors.
The frequency of dmin varies considerably among different tumor types.dmin were rare in hematologic disorders, occurring in less than 1% of cases in both myeloid and lymphoid disorders.This finding is consistent with previous studies that have shown that dmin are uncommon in hematologic malignancies. 23,31Our data reveal that one particular subtype-acute erythroleukemia (AML M6)-however, shows a frequency well above all other hematologic disorders, in fact, a magnitude exceeding many malignant solid tumors.A relatively high frequency of dmin in AML M6 was also noted in a review of the data in the Mitelman database ascertained already in 2007. 32Also, in benign solid tumors, the frequency of dmin was low, with the highest frequency being 1.9% in salivary gland adenomas.The low incidence of dmin in benign tumors is indirectly supported by the fact that dmin have only been reported previously in single cases of different benign tumors, with the exception of 10 cases of meningioma and seven salivary gland adenomas. 22In contrast, malignant solid tumors display a substantially higher abundance of dmin in most tumor types.Our results corroborate and extend previous findings of a high prevalence of dmin in many of the most aggressive forms of cancer. 5,13In addition to very high frequencies in malignant glioma and neuroblastoma (20%-30%), established several decades ago, [33][34][35][36] we identified high frequencies in pancreatic carcinomas (21.9%), dedifferentiated liposarcoma (17.6%), embryonal and alveolar rhabdomyosarcoma (10.8% and 15.9%, respectively), and osteosarcoma (10.4%).Our data also demonstrate that far from all malignant solid tumors display high frequencies of dmin.In fact, and surprisingly, some show very low frequencies, comparable to the hematologic disorders and benign solid tumors, for example, malignant melanoma (0.3%), kidney cancer (0.4%), myxoid liposarcoma (0.5%), breast cancer (1.0%), synovial sarcoma (1.0%), and Ewing sarcoma (1.0%).As is apparent, there is an extensive variation within epithelial and mesenchymal tumors.We did not find any significant difference among the three major histogenetic types of malignant epithelial tumors, that is, adenocarcinoma, squamous cell carcinoma, and transitional cell carcinomas.
No systematic study has previously been performed on the frequency of hsr in neoplasia.We found distinct patterns in different tumor types.In hematologic disorders, less than 0.5% of cases contained at least one hsr with great variation from 0% in some major entities such as MDS/MPD and CML to 1.4% in erythroleukemia (AML M6).It is of interest that this particular subtype of acute myeloid leukemia also has the highest frequency of dmin among all hematologic disorders.Takeda et al. 37 recently performed a comprehensive molecular study of 124 AML M6 cases and revealed frequent focal gains and/or amplifications of genes implicated in erythroid proliferation and differentiation, particularly EPOR and JAK2, which resulted in enhanced STAT5 signaling and enhanced cell proliferation.
These cases were frequently accompanied by gains and amplifications of ERG/ETS2, associated with a very poor prognosis, and often showed high sensitivity to ruxolitinib in vitro and in xenograft models, highlighting a potential role of JAK2 inhibition in the treatment of acute erythroleukemia.A striking difference among solid tumors was found between benign and malignant tumors within each morphologic entity in that malignant tumors have higher frequencies of hsr than their benign counterparts.Notably, all types of benign solid tumors have the same low hsr frequency as the hematologic disorders.
Among malignant solid tumors, distinct differences were noted among specific entities.In MET, hsr frequencies vary from 0.2% in kidney cancer to 14% in head and neck carcinomas.Marked differences are also apparent when malignant epithelial tumors are subdivided according to morphologic subtypes.For example, the hsr frequency is more than three times higher in squamous cell carcinomas (11.6%) than in adenocarcinomas (3.6%) and transitional cell carcinomas (2.4%).Among MNET and MMT, the overwhelming majority of specific subtypes have no or very few hsr-containing cases.The exceptions among those studied in reasonable numbers (>50 cases) are neuroblastoma (9.0%), leiomyosarcoma (7.6%), dedifferentiated liposarcoma (5.9%), and osteosarcoma (5.0%).In contrast to the findings for dmin, all astrocytomas, irrespective of grade, showed the same very low frequencies of hsr.
Overall, hsr appeared to be scattered throughout the genome with no apparent clustering except for a marked preference for chromosome 11 and a mutually exclusive involvement of bands 11q23 in AML/MDS/CML and 11q13 in head and neck squamous cell carcinomas.The presence of hsr in 11q23 with amplification of KMT2A is well established in subsets of AML/MDS patients, 31 and hsr in 11q13 with CCND1 amplification is a characteristic feature of head and neck squamous cell carcinomas. 38The scattered distribution of hsr may be related to the re-integration of dmin into chromosomes with doublestrand breaks. 4 identified 2689 cases with ring chromosomes with frequencies varying from 2.3% in the hematologic disorders to 8.3% in malignant solid tumors.We found fairly similar low frequencies among different subtypes of hematologic disorders with one exception, 10.7% in AML M6, that is, the same AML subtype that displays the highest frequencies of both dmin and hsr.The frequency in benign solid tumors is generally higher (4.3%) than in hematologic disorders.Malignant solid tumors have two times as high frequency of ring-containing cases as benign solid tumors.The frequencies are in general similar among epithelial tumors, germ cell/gonadal stromal cell tumors, and neuroectodermal tumors (2.9%-4.6%),but substantially higher among mesenchymal tumors (9.1%).All chromosomes were found to be involved in ring formation.Chromosomes 7 and 11 are preferentially involved in hematologic disorders but rarely among solid tumors.Chromosome 12 on the other hand is often involved in BMTs and MMTs (preferentially lipomas and liposarcomas), but rarely in epithelial tumors and in hematologic disorders.Chromosome 1 is frequently affected in subsets of both hematologic disorders and solid tumors.The results of the only review available on rings in neoplastic disorders, 24 based on a survey of about 800 malignant neoplasms studied by chromosome banding or fluorescence in situ hybridization, are in fairly good agreement with our findings in malignant tumors.Thus, both studies identified an overrepresentation of chromosomes 7 and 11 in HEM, chromosome 1 in MET, chromosomes 1 and 22 in MNET, and chromosomes 1 and 12 in MMT.
However, some discrepancies were noted in that chromosome 8 was more frequently involved in MET in Gebhart's review, whereas in our study chromosomes 3 and 17 were overrepresented in MET, chromosome 9 in MNET, and chromosome 10 in MMT.
When combining all three cytogenetic signs of possible gene amplification, a general pattern emerges (Table 1, Figure 1) that amplifications in the form of dmin/hsr/r are decisively more common in all three malignant solid tumor types (epithelial, mesenchymal, and neuroectodermal) as compared to the hematologic disorders and benign solid tumors.Among the malignant solid tumors, MMT and MNET had the highest frequencies.In fact, as seen in Figure 4, 9 of the 10 tumors with most dmin/hsr/r belong to these two categories.As is also apparent from Figure 4, an extensive variation is seen among the malignant solid tumors irrespective of their histogenetic origin, ranging from 2.2% (Ewing sarcoma) to 55.6% (liposarcoma) among MMT, from 3.2% (ependymoma) to 42.5% (neuroblastoma) among MNET, and from 2.5% (skin cancer) to 47.8% (pancreatic cancer) among MET.The relative contribution of dmin, hsr, and r varied in the three malignant solid tumor groups with a dominance of r in MMT and dmin in MNET, whereas the frequencies of dmin, hsr, and r are quite similar in MET.The relative contribution of dmin, hsr, and r varies considerably among individual tumor types (Figure 4) and no obvious pattern was discernible.One aberration type dominates in some tumors but may be negligible or absent in others, whereas in some tumors, all three are equally frequent.The reasons for this variation are unknown.
The structural characteristics of dmin, hsr, and r have different consequences that may impact the stability of these aberrations over successive cell divisions.dmin are highly flexible as regards copy numbers.During mitosis, these structures lacking a centromere are randomly distributed at anaphase, resulting in daughter cells harboring different numbers of dmin.Thus, tumor cells can easily achieve an optimal copy number status for tumor development enabling adaptation to different therapeutic modalities.This is more difficult for the more stable hsr that follow a regular mitotic scheme.Rings have a fairly good plasticity, allowing fine-tuning of optimal copy number of oncogenes and other linked genes included in the amplicon, through breakage-fusion-bridge (BFB) cycles. 39 the other hand, longitudinal studies of liposarcomas with rings show little or no change of core amplicons. 40y gene amplification in the form of dmin/hsr/r play a role as an initiating or early event in the process of tumorigenesis?This cannot be excluded because all three aberration types may be found as a sole anomaly, most often as ring chromosomes and occasionally as hsr.Nonetheless, this is rare, only encountered in 222 cases (0.3%), and even in those cases where they actually are found as the sole anomaly, there remains the possibility that cytogenetically invisible but pathogenetically important abnormalities serve as initiating factors.The only convincing example is ring formation in liposarcomas, in which neither recurrent gene fusions, nor single nucleotide variants or other copy number changes have been detected. 40In the majority of cases, they co-exist with various numerical and/or structural abnormalities, often as part of complex karyotypes, suggesting that dmin, hsr, and r usually are secondary events that emerge during tumor evolution, possibly impacting important clinical parameters during tumor progression, like invasiveness, metastatic potential, and resistance to chemotherapy.Notably, the substantially higher frequency of dmin/ hsr/r in high-grade astrocytomas-the only tumor type in which the occurrence could be assessed in different malignancy grades-lends strong support to the assumption that they are acquired over the course of tumor evolution.A thought-provoking question is then if certain primary inducing events can somehow shape or perhaps necessitate the subsequent emergence of such amplification mechanisms.Remarkably, several tumor entities (e.g., chronic myeloid leukemia, acute promyelocytic leukemia, myxoid liposarcoma, synovial sarcoma, and Ewing sarcoma), characterized by well-established primary fusion genes, 41 were found to have notably low frequencies of dmin/hsr/r, suggesting a possible link between distinct primary oncogenic events and different patterns of gene amplification during tumor progression.An interesting observation in this context is that among gene fusion-positive alveolar rhabdomyosarcomas, the fusion gene PAX7::FOXO1 was found to be amplified in 93% of cases, contrasting sharply with the mere 9% occurrence in tumors harboring the gene fusion variant PAX3::FOXO1. 42Even more puzzling, some tumors with the exact same fusion gene display copy number differences depending on clinical context: the COL1A1::PDGFB fusion that characterizes the dermatofibrosarcoma protuberans family of tumors is almost always amplified in ring chromosomes in tumors from adults, but found in only one or two copies in tumors from children. 43Apart from such intrinsic factors that could influence if and how gene amplification occurs, probably also exogenous agents could shape the amplicons.For instance, MYC amplification is seen in the vast majority of radiation-induced angiosarcomas, especially when located in the breast, but only rarely in de-novo lesions. 44 As outlined above, dmin/hsr/r frequencies are in general much higher in malignant neoplasms than in benign ones, which is well in line with molecular data as well as with the general notion that gene amplification is important for tumor progression. 25What remains enigmatic, however, is why some malignancies preferentially display a certain type of amplification.For instance, in neuroblastomas dmin/ hsr is 15 times more common than ring chromosomes, whereas ring chromosomes are 11 times more common than dmin/hsr in well-differentiated liposarcomas (WDLS), and hsr is twice as common as dmin/r in head and neck carcinomas.As dmin, hsr, and ring chromosomes likely arise through different chromosomal mechanisms, one could speculate that the frequency variation is associated with exposure to different clastogenic agents and/or different DNA damage repair deficiencies.However, a recent study on copy number signatures in cancer did not detect any such correlation. 49A more likely explanation would be that the variation in amplicon structures among malignancies is related to the fact that neoplastic transformation/progression usually requires alterations of several signaling pathways, and that a particular amplicon could have more or less strong impact on these pathways.For instance, the ring chromosomes in WDLS have been extensively studied and it has been shown that they always include the MDM2 gene, encoding a negative regulator of TP53. 50However, several other genes from other parts of chromosome 12, for example, CDK4, FRS2, HMGA2, are always co-amplified, and it has been experimentally shown that the concomitant overexpression of multiple genes is critical for WDLS tumorigenesis 51 ; such co-amplification of genes/segments from distinct parts of one or more chromosomes is typical in ring chromosomes, but rare and difficult to achieve in the relatively small amplicons found in dmin or hsr.In a similar vein, the breakage-fusion-bridge-associated formation of an hsr in 11q13 in head and neck carcinomas can simultaneously result in three pathogenetically relevant events in carcinogenesis: amplification of CCND1 in 11q13, loss of critical tumor suppressor genes (e.g., ATM) distal to the hsr in 11q13, and low-copy gain of an oncogene (e.g., PIK3CA) in the segment providing the telomere for the derivative chromosome 11. 52,53In contrast to these two examples of co-operative events made possible by ring-or hsr-formation, it has been shown in transgenic models that over-expression of MYCN alone is sufficient for transforming murine neural crest progenitor cells into neuroblastoma cells, 54,55 which fits the frequent finding of small amplicons in the form of dmin in neuroblastoma, with MYCN as the only consistently included gene. 12Whether such aspects are of general relevance for the variation in dmin/hsr/r frequencies will require combined molecular and cytogenetic efforts.

ACKNOWLEDGMENTS
This study is supported by the Swedish Cancer Society (grant/award numbers CAN2021-1398 and CAN2020-0671) and the Swedish Childhood Cancer Foundation (grant/award numbers PR2020-0021 and PR2020-0008).
T A B L E 1 Frequencies of dmin, hsr, r, dmin/hsr, and dmin/hsr/r in 80 131 neoplasms.a a Tumor types studied in insufficient numbers to allow conclusions (<100 cases) are not presented in the table but are included in the totals of the major tumor groups to which they belong as well as in the grand total.
A B L E 2 Chromosomal distribution of 685 hsr in malignant neoplasms.
Chromosomes preferentially involved in ring formation in major tumor entities.
T A B L E 3 While our study and Rosswog's data agreed on liposarcomas and pancreatic cancer having elevated frequencies of "other amplifications," similarly high frequencies were observed in Rosswog's study across several tumor types that had minimal dmin/hsr/r occurrences in our study.Similarly, no meaningful correlation emerged when comparing our dmin/hsr and dmin/hsr/r data with the data compiled by Harbers et al.It is important to note in this context that no correlations were seen between amplification frequencies across the 17 tumor types that were studied by both Harbers et al. and Rosswog et al.Clearly, comparing gene amplification frequencies across studies of different tumor types necessitates uniform definitions of amplification.
18fortunately, there is no cytogenetic information on secondary angiosarcomas, and it is thus unknown if the MYC amplification in secondary angiosarcomas is due to dmin, hsr, or r.Still, these observations underscore the intricate interplay between specific genes, the tumor microenvironment, carcinogens, and the subsequent amplification dynamics in cancer progression.A comprehensive investigation into the presence of dmin/hsr/r within tumors harboring characteristic gene amplifications,18presented data on "seismic amplification," characterized by multiple rearrangements and discontinuous copy number levels, along with "other amplifications" in 38 cancer types, 23 of which corresponded to our study.In general, we found limited apparent correlations between our data and the amplification data from these two molecular studies.Occasional agreements emerged; for instance, liposarcomas exhibited the highest dmin/hsr/r frequency in our study (52%) and also the highest "seismic amplification" frequency in Rosswog et al.(100%).Both studies also concurred on the low amplification frequencies in hematologic disorders (0%-5%).However, many discrepancies were evident, such as in pancreatic cancer.This tumor had the second-highest dmin/hsr/r frequency in our study (39%), but a very low "seismic amplification" frequency (4%), similar to the hematologic disorders.