In the base data set comprised of 5918 malignant cases analyzed by CGH, breast carcinomas and precursor lesions constituted the largest subgroup, followed by neoplasias of the prostate gland and stomach (Table ). The median aberration number per case (number of chromosomes with at least one abnormal segment) showed a wide range (Figure ), from 0 (squamous skin neoplasias, thyroid carcinomas) to 12 (small cell lung carcinomas; SCLC). For some of the entities at the low end of this spectrum, a certain bias through frequent biopsies at early stages can be suspected (e.g. tumors of skin, thyroid, prostate), although pre-maligant tumors had been excluded from the analysis.
Imbalance hot spots in clinico-pathological entities
In the overall aberration profile, gains on chromosomal band 8q24 represented the most common imbalance, followed by gains involving regions on 20q, 1q, 3q, 17q, 7(q) and 5p, as well as other changes (Figure ).
Overall imbalance pattern from all cases. For each chromosomal band (862 bands resolution) the percentage of cases with gains (green, upward) and losses (red, downward) is indicated.
In the following overview, the most frequent regions involved in genomic imbalances are listed, with detailed locus and frequency information for regional hot-spots. Since overall aberration frequencies vary greatly between entities, no absolute cut-off values (e.g. 10% or 20%) were used when selecting these regions. Instead, genomic regions with clear separation from overall background in the imbalance histograms were selected for each entity and sorted in descending frequency of occurrence. Gains and losses were evaluated separately.
For each entity, few outstanding observations are discussed and selected literature is provided. Histologic subtypes are only mentioned in few entities. Because not all involved regions can be discussed here with respect to putative genetic targets, a selection of well characterized oncogenes and tumor suppressor genes with involvement in epithelial neoplasias is provided as part of Table . Detailed aberration profiles for each entity can be found in the Additional file 1
Occurrence of most common imbalances in different epithelial neoplasias by CGH
Breast carcinoma (667 cases)
Gains: 1q31 (50.8%), 8q23 (47.3%), 17q24 (31,2%), 20q (30.9%), 16p, 11q13, 19, 3q
Losses: 16q (30%), 8p23 (26.1%), 17p13 (23.5%), 11q23 (23.1%), 13q21 (22.8%)
In breast carcinomas, gains on 1q and losses on 16q constitute the most frequent copy number changes of each quality. In the literature, a cytogenetic subtype with combination of these imbalances and few other genomic changes has been associated with histology of well differentiated DCIS [40
] and a favorable prognosis [41
]. Interestingly, gains on 11q13 and 12q24 were associated with higher metastasis-free survival in another study [42
], which also associated multiple imbalances with a bad prognosis.
Although ERBB2 has been a known and clinically relevant [43
] amplification target on 17q, distinct amplicons mapping telomeric to the ERBB2 locus have been described [44
Prostate carcinoma (600 cases)
Gains: 8q24 (23%), 7, X
Losses: 8p21 (32.3%), 13q21 (26.5%), 6q16 (18.5%), 16q(23), 5q(21)
In contrast to other adenocarcinomas, and accounting for the overall low copy number variations, prostate ca. usually lack the 1q gains frequent in most carcinoma entities. While Mattfeldt et al
] showed a correlation of 8p loss to higher tumor stage, the negative prognostic impact of gains of 8q and chromosome 7 in was reported from bioptic samples [46
] and early tumors [47
]. A meta analysis of published copy number data recently was provided by Sun et al
Gastric carcinoma (529 cases)
Gains: 20q12 (36.1%), 8q23 (31.7%), 17q21 (21.2%), 7 (up to 20.8%), 13q22q31, 1q, 3q, 5p, 11q13, X
Losses: 17p13 (24.4%), 19p (18%), 18q (16.6%), 1p, 3p, 4, 5q, 9p, 12q, 16
For most chromosomes, an overall background of 4–10% of gains as well as losses can be found. In gastric carcinomas as well as in some other entities, 8q gains have shown their maximum at 8q23, proximal to the c-myc locus [49
], implying a hitherto unidentified target.
Ovarian carcinoma (449 cases)
Gains: 8q24 (43.7%), 3q26 (38.1%), 1q32 (25.3%), 12p12 (19.2%), 2, 5p, 6p, 7q, 11(q13)
Losses: 5q14 (25.2%), 4q (26.3%), 18q21 (19.8%), 17p (19.6%), 8p (19.4%), 6q, 9q, 13q, 16q, X
Gains on 12p have been discussed as early event in ovarian carcinomas with frequent occurrence in borderline tumors [50
], and several other changes have been linked to advanced carcinomas [51
]. Overall, advanced stage tumors showed a higher grade of chromosomal instability.
Colorectal adenocarcinoma (430 cases)
Gains: 20q13 (53%), 13q (38.6%), 8q24 (37.2%), 7(p) (35.6%), X(q21), 1q, 5p, 12(p), 19
Losses: 18q(22) (47.4%), 8p(22) (37.9%), 17p12 (27%), 4 (up to 23.7%), 14, 15, 22
In colorectal carcinoma, gains on 8q23q24 [52
] have been associated with lymph-node positivity. In a recent study, gains on 20q as well as KRAS mutations could bee shown to precede aneuploidy [53
]. An overview is provided in [54
Hepatocellular adenocarcinoma (HCC; 371 cases)
Gains: 1q23q31 (46.6%), 8q24 (44.8%), 6p21 (22.4%), 17q (21.8%), 5, 7, 20
Losses: 4q (up to 31.3%), 8p(21) (31.3%), 13q21 (28.3%), 16q(21) (25.9%), 17p13 (25.3%), 1p, 6q, 14, 18
Losses of 8p have been shown to distinguish HCC from other liver malignancies [55
]. Also, losses on 4q and 13q were associated with poor differentiation [56
]. An overview of genomic changes in HCC with discussion of putative target genes is provided in [57
Head-neck squamous cell carcinoma, excluding nasopharyngeal ca. (HNSCC; 339 cases)
Gains: 3q26 (59.2%), 8q24 (40.8%), 11q13 (31.9%, many specific high-level), 5p (26.5%), Xq, 1q, 7q(21), 12p, 17
Losses: 3p (30.1%), 18q(22) (22.4%), 9p (22.4%), 11q24 (19.2%), 4, 5q, 8p, 13
Genomic imbalances in HNSCC are in line with other squamous cell carcinomas (-3p, +3q, +5p). The high rate of amplifications at 11p13 has been shown to involve the cyclin D1 (CCND1) locus and to be accompanied by high expression of the gene [58
Thyroid carcinoma (314 cases)
Gains: 5(p), 7, 20(q)
Losses: 13q, 22, 1p
Although a number of highly aberrant cases was included (see outliers in box plot, Figure ), thyroid carcinomas, including all variants, showed an overall the lowest level of chromosomal imbalances. The subset included a large study in post-Chernobyl tumors in children, for which imbalances could be found only in 30% of tumors [59
]. Interestingly, anaplastic carcinomas had a low median number but high variability. Although not mapping to regions with frequent genomic gains in the the included cases, previous reports have shown e.g. amplification of PRKCE on 2p21 [60
] and FGF3 [61
Non-small cell lung carcinomas (NSCLC; 314 cases)
Gains: 5p (52.8%), 3q26 (39.4%), 8q24 (35.4%), 1q(21) (28.4%), 3, 6p, 11q13, 12p, 17q, 18p
Losses: 3p (31.9%), 4(q) (26.4%), 5q (26%), 13q21 (25.6%), 8p(21) (24%), 9 (up to 20.5%), 17p (16.5%), 1p, 6q, 10, 18q
NSCLC is a histologically heterogeneous group, consisting of squamous (SCC) and non-squamous (NSCC) cases. Imbalances have been shown to be partially shared between the groups, with differences in the frequency of some regional involvements [62
Cervix carcinoma (226 cases)
Gains: 3q26qter (53.6%), 1q (up to 28.3%), 5p (27.4%), 8q24 (20.3%), 6p, 9(q), 17q, 20q, X
Losses: 2q36 (30.1%), 3p (up to 25.2%), 4 (up to 23.5%), 11q23q24 (23%), 6q, 13q, 17p, 18
The number of chromosomal changes may be influenced by occurrence and type of HPV infection. However, in a large study no correlation between single imbalances and clinical parameters could be identified [63
]. Losses of the telomeric region of 2q were frequent and rarely found in other entities.
Esophagus carcinoma (209 cases)
Gains: 3q(26) (53.2%), 8q24 (49.8%), 7p (34%), 20q13 (32.6), 1q (29.7%), 7q21 (29.7%), 5p (28.7%), Xq (23.4%), 6p12 (22.9%), 12p (22.5%), 11q13 (22.1%), 17q21 (21.6%), 2(q), 9q
Losses: 3p (up to 37.3%), 18q (37.3%), 4 (up to 33%), 5q(21) (29.2%), 9p (26.3%), 8p (23.4%), 13q (up to 23%), 1p (22.5%), 19q (2%), 17p (20.1%), 10, 11, Xp
Esophagus carcinomas showed an overall high "background" of gains and losses, with the 2nd highest number of imbalances per case. This group consists of cases with squamous cell as well as adenocarcinoma or intestinal adeno-ca. histology. Besides specific de-regulation of oncogenes, an effect of the massive genomic changes on chromatin structure has been discussed previously [64
Renal carcinoma (RCC, 195 cases)
Gains: 7 (up to 33.4%), 5q(31) (32.8%), 8q23q24 (19.5%), 20 (up to 18.5%), 17q (up to 17.4%), 1qter (13.1%), 3q, 12, 16
Losses: 3p (up to 43.1%), 14q (up to 26.2%), 6q (21%), 1p (up to 20%), 9 (20%), 8p (19.5%), 13q (19.5%), 17p (19%), 18q (18%), 2, 4, 10
Gains of 5q are a rare occurrence in other carcinomas, and have been identified in papillary as well as non-papillary cases [65
]. An overview of accumulated karyotype data was given in 2004 [66
], proposing different cytogenetic pathways and associating papillary RCC with a hyperdiploid karyotype pattern.
Nasopharynx carcinoma (NPC; 177 cases)
Gains: 12p (up to 33.9%), 1q (24.3%), 3q (22%), 8q22q23 (19.5%), 18p (17%), 11q13 (12.4%), 2(q), 4(q), 6(q), 17q
Losses: 16q (29.9%), 14q24 (27.7%), 11q23 (24.3%), 1pter (23.7%; difficult region), 9, 13q31, 17p, 19p
In contrast to some other carcinoma entities, the maximum of gains mapped clearly proximal to the c-myc
region. A meta-analysis of CGH data in NPC was recently provided by Li et al
]. In concordance with the low number of NPC cases with 8p deletion, the oncogenetically relevant DLC1 on 8p22 has been reported to be inactivated by methylation rather than copy number change [68
Bladder carcinoma (169 cases)
Gains: 20q (21.3%), 8q (20.7%), 17q (20.7%), 11q13 (19.5%), 1q21 (17.8%), 5p, 6p, 7, 10p
Losses: 9 (26%), 13q21 (17.8%), 4q (up to 14.2%), 11p (14.2%), 5q (14.6%), 2q32, 8p, 6q, 18q, X, 17p
While some of the imbalances (e.g. 8q, 17q and 20q gain, 11q13 gain/amplification) were frequent in other entities, losses on chromosome 9 were exceptionally high in transitional cell carcinoma (115 cases). In contrast to other squamous cell carcinomas, gains on 3q were rare (3 of 40 bladder SCC). In one study included in the data, some differences between histological and etiological subtypes of bladder carcinomas were presented [69
Neuroendocrine carcinoma and carcinoid (138 cases)
Gains: 19 (26.1%), 20 (up to 18.1%), 17q (up to 17.4%), 5p (15.9%)
Losses: 11q22q25 (20.3%), 13q21 (18.1%), 3p (up to 15.5%), 6q22 (13%), 10q25q26 (13%)
Possibly due to the heterogeneity of this group, a diffuse background of whole chromosome gains was observed. The genetics of neuroendocrine tumors was recently reviewed with consideration for inherited syndromes as well as molecular cytogenetic results [70
Malignant melanocytic neoplasias (99 cases)
Gains: 8q(21) (36.3%), 6p (32.3%), 1q (22%), 7, 20q, 11q13
Losses: 6q(24) (29.3%), 10 (q22) (22.2%), 3 (up to 21.2%), 9p (21.2%), 13q, 1p, 11q23q24
In contrast to most epithelial malignancies, gains on 6p can be found in a large proportion of malignant melanomas. Above background losses on 10q are found in few other entities, too. Similar to other entities gains on 8q, 1q, 20q and chromosome 7 can be observed. A comprehensive overview about molecular-cytogenetic techniques in the analysis of melanocytic lesions was given by Bauer and Bastian [71
Pancreas adenocarinoma (88 cases)
Gains: 20q (up to 30.7%), 8q(23) (26.2%), 17q (up to 21.6%), 12p11 (19.3%), 7p (18.2%), 3q (up to 17.1%), 5p, 11(q13), 14, 15qter, 16, 19q
Losses: 18q (35.2%), 9p(23) (29.6%), 13q (18.2%), 3p (up to 17.1%), 6q(21) (17.1%), 8p (17.1%), 17p13 (15.9%), 4, 15q, 10q, 12q
This subset only includes pancreas adenoocarcinomas, omitting endocrine tumors.
High level copy number amplifications have been shown e.g. for ERBB2 (17q12) and EGFR (7p12). Interestingly, the maximum for gains on 8q (8q22) is proximal to the c-myc
region. This fact was also mentioned in a recent overview of genomic screening results in pancreas ca. [72
Cholangio-carcinomas (intra- and extrahepatic; 63 cases)
Gains: 8q22 (42.9%), 17q (up to 39.7%), 20q (34.9%), 11q13 (23.8%), 15q (up to 23.8%), 3q(26) (20.7%), 7p (up to 20.7%), 13q (up to 20.7%), 5p (up to 19.1%), 1q, 6p, 12q23q24
Losses: 1p34p36 (up to 22.2%), 4q (up to 20.6%), 17p (19.1%), 5q(14), 6q, 13q, 18qter, X
Gains on 15q were reported in up to 36% (in a study of Korean intrahepatic cholangio-ca. [73
]), but rarely seen in other entities.
Small-cell lung carcinoma (SCLC; 63 cases)
Gains: 19 (up to 44.4%), 3q (up to 41.3%), 17q23 (41.3%), 1p31p34 (38.2%), 8q(23) (36.5%), 20q (up to 30.2%), 14q31q32 (26%), 1q (up to 25.4%), 9q31q34, 6p, 13q32
Losses: 3p (84.1%), 4 (up to 74.1% on 4qter), 13q(14) (69.8%), 10(q) (41.3%), 17p13 (33.3%), 2(q23q24) (31.8%), 8p (25.4%), 9, 11, 15, 6q, 16q
SCLC had the highest number of chromosomal imbalances (median 12 involved chromosomes per case). Also, some of the imbalances constituted the overall highest fractional aberrations in all datasets (e.g. >80% losses on 3p), as well as very frequent losses on 4q, 13q and 10q. Although frequent gains on 7p had been reported from array CGH of SCLC cell lines [74
], this region was rarely involved in the mostly clinical specimen analyzed by chromosomal CGH.
Endometrial carcinoma (56 cases)
Gains: 1q (50%), 8q(22) (39.3%), 3q26 (14.3%), 10 (up to 14.3%)
Losses: 10q (12.5%), 13q (10.7%), 14q, 9q
As in e.g. nasopharynx ca., SCLC and pancreas ca., the maximum of 8q gains was proximal of 8q24. Endometrial ca. show an exceptional high rate of gains on 1q, which were also the most frequent change in a series of 98 cases not available for inclusion here [75
]. Interestingly, chromosome 10 was also frequently involved (both gains and losses).
Carcinomas of the vulva (53 cases)
Gains: 3q (45.7%), 8q(23) (34%), 1 (up to 28.3%), 5p(15) (26.4%), 20q (24.5%), 9q (20.8%), 6, 7, 12, 14, 19, X
Losses: 3p (22.6%), 4p (22.6%), 11q23q25 (20.8%), 13q14 (13.2%), 8p, 10, X
Vulvar carcinomas showed an imbalance pattern characteristic for squamous cell neoplasias, with a high rate of gains on 8q23 and especially 3q. As in cervix ca., the most frequent changes could not be related to HPV status [76
Squamous cell carcinomas of the skin (52 cases)
Gains: 3q (11.5%), 17q (11.5%), Xq (9.6%), 1q, 8q, 14q
Lossses: 9p21 (19.2%), 3p (13.5%), 18q (13.5%), 17p (9.6%)
While containing many cases of borderline malignant behaviour, squamous skin neoplasias had the overall lowest aberration frequencies (median 0). This group did not include the pre-malignant keratoacanthomas, for which a frequent gain of 11q and cyclin D1 overexpression had been shown [77
Disease-specific Involvement of most frequently aberrant chromosomes and relationship of overall aberration profiles
Several chromosomal hot spot regions showed a pronounced disease related variation. The most prominent regions scoring high in specific entities are listed in Table . To reduce sampling bias (e.g. entity specific inclusion of a high proportion of early stage cases) top-scoring aberrations were defined as the most frequent gains or losses in each entity. Some additional imbalances which occured only in few entities at high levels are listed in Table . In an attempt to identify the relation of single case aberration profiles, a cluster analysis of all 5043 informative cases was performed. For that purpose, a band specific matrix with 86 bands resolution was generated, of which the 55 intervals most frequently involved in imbalances were selected (top-scorers in one or more entities, ref. Tables and ; e.g. "8q2"). Hierarchical cluster analysis revealed the complexity of the case-specific aberration patterns, and was able to visualize concordance of imbalance patterns and disease categories in small groups of cases (Figure ).
Selected imbalances with high penetrance limited to one or few entities
Figure 3 Clustering of 5043 malignant epithelial neoplasias by the pattern of gains and losses, using regions previously defined as highly aberrant in one or several entities (ref. tables 2 + 3). For each case (x-axis), gains (green) and losses (red) are indicated (more ...)
To reduce the complexity of the data and simplify the detection of similarities in the aberration patterns of different entities, region specific aberration frequencies of all entities were clustered. For each of the 55 selected intervals, the sum of gains – losses was calculated, resulting in a 22 × 55 matrix (entities × intervals). During hierarchical cluster analysis, values were normalized over each entity's intervals, to account for differences in genomic complexity. Figure depicts the result of the cluster analysis, in which overall similarities and differences in the different entities become apparent.
Figure 4 Clustering of carcinoma entities by their overall imbalance pattern. For each of the selected chromosomal regions, aberrations were summarized (percent gain – percent loss). After normalization of all regions over the respective entity, the color (more ...)
Differences in aberration profiles in some instances will reflect the general histological type, which is substantiated by the close relation of overall profiles e.g. from clinico-pathological groups containing squamous cell carcinoma cases (group NSCLC/esophagus/HNSCC/cervical/vulva in Figure ). To evaluate this effect, the most frequent histopathological entities (adenoca., squamous cell ca., hepatocellular ca., ductal breast ca., transitional cell ca., intestinal adenoca.) were selected. Additionally, adenocarcinomas of the prostate were put in a separate group, due to the previous observation of specificities in the imbalance profile (e.g. overall lack of 1q gains). Clustering was performed analogous to the method described above. Figure visualizes the clustering of those histopathological entities.
Figure 5 Clustering of different histologies in carcinomas by their overall imbalance pattern. Here, the most frequent histological types were automatically grouped for their overall imbalance profiles. Since considerable differences had been found for adenocarcinoma (more ...)
In comparing the most frequent imbalances for their occurrence in different entities, some general observations stood out. Gains involving the terminal part of 8q were ubiquitously found, and often belonged to the most frequent imbalances in the respective entities. Hovever, they were rarely detected in neuroendocrine carcinomas and thyroid neoplasias. Interestingly, in some entities the maximum of detected abnormalities mapped proximal of 8q24 (e.g. 8q21 in melanocytic NPL and 8q22 in NPC and cholangio ca.). How much this reflects a true difference in regional involvement, and therefore may point to differential target gene involvement has to be left open here.
Gains on 1q with maxima on 1q2 were also frequent, and were the most frequent changes in breast, hepatocellular and endometrial carcinomas. However, 1q gains were comparatively infrequent in carcinomas of the prostate and bile ducts as well as in renal and colorectal ca. Overall gains on 3q26q27 combined with losses on 3p were characteristic for squamous cell carcinomas and SCLC, with 3q gains also being frequent in ovarial ca. and other entities.
While 5p gains were found in various entities, gains on 5q were frequent only in ca. of thyroid and kidney, as were gains on chromosome 12. Overall, 13q losses could be detected in most entities. In contrast, 13q gains were predominant in cholangio ca. and also frequent in colorectal and gastric carcinomas.
Region specific gains involving 11q13 were found in a number of entities (bladder ca., HNSCC, pancreas, prostate, skin, ovary, gastric, NSCLC). Interestingly, of the overall 692 cases with gain on 11q13, 18% had a loss of the telomer of 11q by CGH (Figure ). Since a strong selective pressure for such a switch in aberration quality in linked regions can be suspected, this observation may point towards targeting of both genes with oncogene and tumor suppressor functions on 11q in these tumors.
Visualization of single case aberration patterns in 123 carcinomas with gain on 11q13 and concomitant proximal loss. Cases are clustered according to their imbalance patterns (gain/loss status, 320 bands resolution).
When comparing overall aberration patterns, entities containing squamous cell carcinoma cases (HNSCC, NSCLC, cervix and vulva ca., esophagus ca.) appeared related, with a common pattern offrequent losses on 3p, gains on 3q, 8q and varying gains on 1q and 5p. NPC showed a slightly diverging pattern, with predominant 12p gains and 14q losses in addition to slightly lower levels of those imbalances.
Interestingly, in both unselected renal carcinomas as thyroid neoplasias, the average imbalance profiles included frequent gains on chromosomes 5, 7 and 12, while lacking the dominance of gains 1q and 8q found in most other entities.
Apart from these observations, one should refer to the accompanying figures for a general comparison of patterns in different clinico-pathological groups. Also, the use of the Progenetix website tools is encouraged.