| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Tumor Biology |
Department of Clinical Genetics, University Hospital, SE-221 85 Lund [M. H., D. G., G. B. H., F. M.], and Department of Genetics, Lund University, SE-223 62 Lund [T. S.], Sweden
 |
ABSTRACT |
|---|
 Top ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
MATERIALS AND METHODS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
. Each karyotype was then assessed for the presence or absence of the selected imbalances. The NIPT4
was then calculated, and the 538 tumors with at least one imbalance were selected. Because the identification of single band deletions is unclear, these were excluded when calculating NIPT. Tumor grade was recorded when reported. This information was present for 204 tumors; 27 were grade I, 98 grade II, and 79 grade III.
|
Principal Component Analysis.
To search for possible patterns of correlations between the imbalances, PCA was performed using the Statistica software package (Statsoft, Tulsa, OK). PCA is a standard multivariate method used frequently to search for underlying structures in data sets (10)
. In short, principal components are linear combinations of the original variables, orthogonal, and ordered with respect to their variance so that the first principal component has the largest variance. To analyze imbalances, these were used as variables and the individual tumors as the observations; this will group imbalances seen frequently in the same tumors. To determine the appropriate number of factors to extract, the scree test was used (11)
. The factor models arrived at were evaluated at three levels, total variance accounted for, communalities, and by the residual correlation matrix. The fraction of the original variance accounted for by the specific factor model is given by the cumulative percentage of the explained variance for each component in the factor model. The communality of a given variable (imbalance) is an estimate of how well the factor model predicts the behavior of the given variable. The communality ranges from 0 to 1, where a value of 1 indicates full explanation of the variance. The residual matrix is obtained by subtracting the original correlation matrix with the one produced by the given factor model. This analysis indicates how well the factor model accounts for relationships between specific pairs of imbalances.
To test the robustness of the correlation-based PCA, the relationships between the imbalances were also investigated by using 
2 statistics. The 2 value for independence was calculated for each combination, and weighted 2 distances were computed using the Correspondence Analysis Module of the Statistica software.5
Factors (dimensions) were then extracted by means of PCA to maximize the distances between the variables (12)
.
To analyze the tumor population, the tumors were used as variables, and the imbalances were used as the observations; this will group tumors with similar sets of imbalances. The NIPT for each individual tumor was used as an observation in addition to the imbalances (6) . Because this parameter shows the greatest variation, the first principal component will be almost completely dominated by this variable, and the position of the tumors along the first principal component will be ordered according to karyotypic complexity. Classifying imbalances (6) were identified by testing each imbalance for the presence in all members in a given cluster produced by the PCA. Communalities and the residual correlation matrix were not considered when evaluating the PCA of the tumors.
|
RESULTS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
. The most frequent were +1pq, 1p-, 3p-, 6q-, and -16q. Deletions of the most terminal bands were seen at 2q, 5p, 7q, 8q, 10q, 12q, 15q, 16q, 19p, and 19q with frequencies spanning from 12 to 15%. The average number of such terminal deletions/tumor increased linearly from 0.2 to 4.4, with NIPT = 1–13 (r = 0.96). The imbalances 4q-, 6q-, and 8p- were negatively correlated with grade I (P = 0.042, 0.020, and 0.047, respectively), whereas +7 was positively correlated with grade III (P = 0.024). The distribution of the NIPT showed a sharp decrease in frequencies of tumors with increasing NIPT (Fig. 1)
. Very few tumors had >10 imbalances. The distribution is close to a geometrical distribution with some overrepresentation of tumors with 1, 2, and 3 imbalances. The monomodal nature of the distribution indicates that the tumor population was homogeneous with respect to the mode of karyotypic evolution; no subpopulation of tumors with increased karyotypic instability was present.
|
. Gain of 1pq and chromosome 7 and loss of 1q and 3p were identified as early imbalances because their distributions showed modal values of 1. To establish the relative order of appearance of the imbalances, tumors containing either of the early imbalances +1pq, 1q-, 3p-, and +7 were selected for analysis. This amounted to 348 cases. In Fig. 3
, the means of the bootstrapped TO values using this set of tumors are plotted together with the 2.5, 25, 75, and 97.5 percentiles. Early imbalances (TO = 1–2) were +1pq, 1q-, 3p-, +7, and -16q. These are followed by the moderately early imbalances (TO = 4.5–6) 1p-, 3q-, and +8q. Intermediate imbalances (TO = 8.5–10) were 6q-, 11q-, and -22, followed by the late imbalances (TO = 10.5–12) 10p-, -13, -17, 18q-, -X and the very late imbalances (TO = 12–14) 4p-, 4q-, 9p-, 11p-, -15p, 14p-, and 21p-. The simulation study revealed the imbalances +1pq, 1q-, 3p-, and +7 to occur earlier than expected from random events.
|
|
explained 32% of the variability and delineated two major cytogenetic pathways. One originated with +1pq and 16q-, and possibly +7, and a second with 1q- and 3p-. The moderately early imbalances 1p-, 3q-, and 6q- were predominantly associated with losses of 1q and 3p. The late imbalances 8p-, 9p-, 10p-, -17, 18q-, -21, -22, and -X, positioned approximately between the +1pq/16q- and the 1q-/3p- clusters, were followed by the very late imbalances 4p-, 4q-, -13, -14p, and 15p-. The three-factor solution (Fig. 4b)
explained 37% of the original variance and revealed a dense cluster of predominantly late imbalances, whereas the early imbalances were at some distance and well separated. The third factor separated +1pq and +7, and 1q- and 3p-, respectively. This further partitioning of the early imbalances suggests the presence of four different pathways initiated by +1pq, 1q-, 3p-, or +7. To evaluate how well each imbalance was represented in the three-factor model, the communalities were calculated. These ranged from 0.24 to 0.63, except for +8q, which was 0.07. Thus, all of the imbalances but +8q were well represented in the three-factor model. By including a fourth factor, the communality for +8q increased to 0.25. Similarly, the communality for +7, which showed the lowest value in the three-factor solution, increased from 0.24 to 0.41. No significant changes in communality were seen for the remaining imbalances.
|
2 distances. The two- and three-dimensional solutions explained 32 and 37% of the total 2, respectively (Fig. 4, c and d)
. The loadings on the first dimension for the imbalances correlated with the obtained TO values, r = 0.91. In Fig. 4, c and d
, it may be seen that the results obtained using the 2 metric are almost identical to the ones obtained by the correlation coefficients.
Correlations between Imbalances.
Because the obtained PCA model represents a compromise of all correlations and may lack important information on specific imbalances, the part of the correlation matrix that included the imbalances +1pq, 1q-, 3p-, +7, and +8q were analyzed in more detail (Table 2)
. Gain of 1pq and 3p- showed significant (P < 0.01) correlations with most of the other imbalances, whereas +8q and 1q- only showed association with a limited number of imbalances, and +7 behaved in an intermediate fashion (Table 2)
. All of the remaining, and later, imbalances showed significant multiple correlations with coefficients ranging from 0.12 to 0.57 (data not shown). Several of the imbalances that showed correlation with +1pq also show significant correlation with 3p-, i.e., 8p-, 10p-, 11p-, -13, 14p-, 15p-, and 18q-. The imbalances +7, +8q, 11q-, 16q-, 21p-, and -22 were, however, more associated with +1pq, whereas 1p-, 3q-, 4p-, 4q-, and 6q- were more associated with 3p-. Imbalances occurring in both pathways were predominantly late, as determined by the temporal analysis. Gain of 8q showed significant (P < 0.01) correlation with +1pq and +7 but was independent of the other imbalances. Similarly, 1q- showed, apart from negative correlation with +1pq, significant correlation with 1p- and 3q- but was otherwise independent of the other imbalances. Trisomy 7 showed associations with fewer imbalances than did +1pq and 3p-, and these were also of mixed character. Three were in common with the +1pq pathway (+8q, 21p-, and -22), two with the 3p- pathway (4p- and 4q-), and two were in common with both pathways (10p- and 18q-). No significant correlation specific for +7 was seen. As a measure for how autonomous +7 was relative to the +1pq, 1q-, and 3p- imbalances, the proportion of +7 cases not showing concomitant presence of +1pq, 1q-, or 3p- was compared with the equivalent proportions for +1pq, 1q-, and 3p-, respectively. The proportion for +7 (14 of 102) was significantly smaller (P < 0.012) than for +1pq (68 of 153), 1q- (47 of 86), and 3p- (54 of 139), respectively. Cases with +1pq and +7 showed a significantly higher proportion of grade III tumors (P = 0.02) than did +1pq cases without +7. A similar pattern, albeit at borderline significance (P = 0.06), was seen for cases with either 1q- or 3p with or without +7.
|
. The tumors could thus be classified as 1p- positive but -16q negative, 1p- and -16q positive, 16q- positive but 1p- negative, and tumors negative for both 1p- and -16q. These groups had NIPT averages of 5.5, 8.7, 4.5, and 5.9, respectively. The PCA of the +1pq-negative tumors also revealed four clusters, determined by the presence or absence of 1q- and 3p-, respectively (Fig. 5b)
. The 1q- group of tumors had an average NIPT of 3.5, the 3p- group 5.4, the 1q- and 3p- group 9.3, and the group of tumors showing absence of both imbalances an average NIPT of 2.7. The latter group, which also showed absence of +1pq because of the selection criteria, constituted 41% (223 cases) of all tumors. The majority of these (>95%) had fewer than seven imbalances. The proportion of grade III/grade II tumors in this group (43 of 43) was significantly higher than the equivalent proportion for cases having one of the early imbalances +1pq, -1q and 3p-, and fewer than seven imbalances, 13 of 30 (P = 0.033).
|
|
DISCUSSION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Deletions of single terminal bands were seen at unexpectedly high frequencies. Furthermore, they were nonrandom with respect to the chromosome arms involved. The deletions were present in both low- and high-complexity karyotypes, and the frequency increased linearly with NIPT. This suggests that the mode of acquisition of these imbalances is the same as for the more extensive losses. The correlation of terminal deletions to the overall cytogenetic complexity could indicate that telomeric crisis (15) plays a role in the cytogenetic evolution of breast cancer. Recent investigations have shown that shortening of telomeric TTAGGG sequences can cause widespread chromosomal instability in neoplastic cells (15) . It is thus possible that at least some of the chromosomal imbalances result from telomeric dysfunction, leading to chromosomal missegregation and breakage during mitosis (16) . On the other hand, the high frequency of telomeric deletions could merely be a side effect of cellular proliferation beyond the normal replicative life span of breast epithelium, with little if any pathogenetic significance. However, only a limited number of chromosome arms showed high frequencies of terminal rearrangements. Because short telomeres could make the arms more liable to deletion of terminal bands, a comparison was made with the reported length of telomeres in normal chromosomes (17) , but no association was seen. Whether this preference of certain chromosome arms is caused by a selective advantage from deletion of specific genes in these regions or whether the terminal chromosome material is dispensable and may be lost with no affect on cellular fitness remains an open question.
To evaluate the temporal order of the imbalances, we used the karyotypic complexity at which a given imbalance was most frequent as a measure for lateness (6) . We have shown previously that this variable correlates well with tumor progression (13) , i.e., imbalances predominantly seen in high-grade tumors have large TO values. To acquire more reliable estimates of the TO, bootstrapped estimates of both TO values and confidence intervals for the mean TO were calculated for each imbalance. The obtained results showed a clear order of appearance. Four major early imbalances (TO = 1), +1pq, 3p-, 1q-, and +7, were identified; moderately early imbalances were 16q-, 1p-, 3q- and +8q, whereas the remaining may be classified as late, among which 9p-, 21p,- and 4q- were the latest. The simulation study showed that TO values for +1pq, 1q-, 3p-, and +7 were significantly smaller in the observed data as compared with the simulated, and thus that these imbalances occurred earlier than expected. This suggests that +1pq, 1q-, 3p-, and +7 have a larger selective value at earlier stages than other imbalances, underscoring the presence of a certain temporal specificity among the imbalances. However, because the bootstrapped TO values show substantial overlap, no strict order of imbalances may be arrived at; the data rather point to a preferred order of events from which deviations occur frequently in individual cases.
The temporal analysis and the three-factor PCA solution emphasized the importance of +1pq, 1q-, 3p-, and +7 in breast cancer development; all were classified as early events and were well separated by the PCA. Of these, +7 was atypical. The proportion of +7 cases that did not also show the presence of +1pq, 1q-, or 3p- was significantly smaller than the equivalent proportions for +1pq, 1q-, and 3p-. This makes +7 more of a secondary imbalance than the temporal analysis would suggest. Although gain of chromosome 7 was correlated with at least eight other imbalances, it also demonstrated a substantial independent behavior, seen by the sharp increase in communality when including a forth dimension, an increase that was only paralleled by +8q. Thus, +7 behaves as a promiscuous secondary imbalance. Furthermore, +7 was associated with grade III tumors, irrespective of the preceding imbalance. This feature of +7, to be associated with high-grade tumors, was also seen in a recent investigation of urinary bladder cancer (13) . However, in this tumor type +7 seems to represent a separate cytogenetic pathway.
Gain of 8q was the most atypical of the investigated imbalances. The three first components could not account for the behavior of this imbalance. Not until the fourth principal component was added could some of the variance of the +8q be captured. All other imbalances were sufficiently described by the three-factor model, and adding a fourth factor provided no significant increase in communalities, with the possible exception of +7. Gain of 8q was, however, significantly correlated with +1pq and +7. A second feature of +8q was the dependence on preceding imbalances, seen by the very low frequency of +8q tumors with 1–4 imbalances (Fig. 2c)
. Thus, the data indicate that +8q shows a great deal of independent behavior but may only have a selective value at very specific stages of tumor progression.
The PCA clearly delineated two major cytogenetic pathways, one characterized by +1pq and one by 3p- and possibly 1q-. This was particularly well seen in the two-factor PCA solution, where these imbalances occupied two opposite ends of the second principal component. Similar pathways were seen irrespective of the matrices used, either being based on product moment correlation or on 2 statistics. Gain of 1pq showed positive correlation with 13 of the imbalances, of which +7, +8q, -11q-, -16q, -21p, and -22 were specific for +1pq, and the remaining, mostly late, imbalances were shared with 3p-. The PCA of the 1pq-containing tumors resulted in four clusters, each determined by the presence or absence of 1p- or -16q. This could indicate four cytogenetic subtypes. However, these imbalances are mechanistically related to +1pq through i(1q) or der(1;16), and hence the clusters may rather reflect mutational mechanisms than pathways promoted by different types of selection. The fact that 1p- was not significantly correlated with +1pq but with 3p- underscores the mechanistic aspect of the +1pq subgroups.
The most frequent imbalances in the 3p- pathway were 1p-, 3q-, 4p-, 6q-, -13, 15p-, and -17. Imbalances that were significantly more associated with 3p- than with +1pq were 1p-, 3q-, 4p-, 4q-, and 6q-. Loss of 1q was positioned close to 3p- by the PCA, and consequently, the 1q- and 3p- pathways demonstrated several similarities; both showed lower frequencies of +7, +8q, and 16q- than did tumors with +1pq. The major importance of 3p- and 1q- in the tumors without 1pq was also indicated by the fact that the PCA identified 1q- and 3p- as classifying imbalances.
The analysis of the tumor population revealed a group of tumors that lacked the early imbalances +1pq, 1q-, and 3p-. This group had a lower average NIPT but had, despite this, progressed further because they showed a significantly higher proportion of high-grade tumors. Because the temporal analysis displayed a highly ordered appearance of genetic events, directed toward the more advanced stages, this suggests that these tumors may have cryptic changes equivalent to the early cytogenetic events, therefore showing a lower NIPT. Thus, given a multistep model for tumorigenesis, the initiating events may not always be detected by cytogenetic banding methods. Conversely, high-grade tumors with few and early imbalances may be expected to harbor cryptic changes equivalent to late imbalances.
In conclusion, the accumulated cytogenetic data on breast cancer reveal a highly complex pattern of chromosomal aberrations. This not withstanding, we were able to show that imbalances appear in a temporally ordered fashion with distinctively early and late imbalances. Furthermore, two major cytogenetic pathways were identified, one characterized by the early imbalance +1pq and one by the similarly early 3p- (Fig. 6)
. Frequent secondary imbalances in the +1pq pathway were 16q-, 1p-, and 11q-, followed by -22, whereas the loss of 3p was followed by 1p-, 3q-, and 6q-. A variant of the 3p- pathway was initiated by 1q-. This imbalance was also associated with 1p- and 3q- but did not otherwise show any significant associations. Gain of chromosome 7 was associated with +1pq, +8q, and with higher tumor grades, whereas gain of 8q showed a high level of independence. When the tumors progressed to more complex karyotypes, the pathways converged into a single route. The tumors thus appear to progress toward a common genetic end-state. The data also show that complex karyotypes may not be understood simply by the presence or absence of specific imbalances but rather as an expression of multifaceted interactions of tumor-promoting changes. It would thus seem impossible to reduce tumor phenotypes to single genetic events and expect these to fully explain the behavior of tumor cells, a problem often encountered when using the status of single genes as prognostic markers. Hence, the complex behavior of the chromosomal changes in breast cancer may give further insight into the nature of the processes leading to cellular transformation.
|
|
FOOTNOTES |
|---|
1 Supported by the Swedish Cancer Society, the Crafoord Foundations, and the Nilsson Family Foundation. 
2 To whom requests for reprints should be addressed, at Department of Clinical Genetics, University Hospital, SE-221 85 Lund, Sweden. Phone: 46-46-173739; Fax: 46-46-131061; E-mail: mattias.hoglund{at}klingen.lu.se
3 Internet address: http://cgap.nci.nih.gov/Chromosomes/Mitelman.
4 The abbreviations used are: NIPT, number of imbalances per tumor; TO, time of occurrence; PCA, principal component analysis.
5 Internet address: http://www.statsoft.com/textbook/stathome.html.
Received 1/10/02. Accepted 2/27/02.
|
REFERENCES |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Gerstung, M. Baudis, H. Moch, and N. Beerenwinkel Quantifying cancer progression with conjunctive Bayesian networks Bioinformatics, November 1, 2009; 25(21): 2809 - 2815. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Melchor, S. Alvarez, E. Honrado, J. Palacios, A. Barroso, O. Diez, A. Osorio, and J. Benitez The Accumulation of Specific Amplifications Characterizes Two Different Genomic Pathways of Evolution of Familial Breast Tumors Clin. Cancer Res., December 15, 2005; 11(24): 8577 - 8584. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. V. Roschke, G. Tonon, K. S. Gehlhaus, N. McTyre, K. J. Bussey, S. Lababidi, D. A. Scudiero, J. N. Weinstein, and I. R. Kirsch Karyotypic Complexity of the NCI-60 Drug-Screening Panel Cancer Res., December 15, 2003; 63(24): 8634 - 8647. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Frigyesi, D. Gisselsson, F. Mitelman, and M. Hoglund Power Law Distribution of Chromosome Aberrations in Cancer Cancer Res., November 1, 2003; 63(21): 7094 - 7097. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Hoglund, D. Gisselsson, G. B. Hansen, T. Sall, and F. Mitelman Ovarian Carcinoma Develops through Multiple Modes of Chromosomal Evolution Cancer Res., June 15, 2003; 63(12): 3378 - 3385. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Hoglund, D. Gisselsson, G. B. Hansen, T. Sall, F. Mitelman, and M. Nilbert Dissecting Karyotypic Patterns in Colorectal Tumors: Two Distinct but Overlapping Pathways in the Adenoma-Carcinoma Transition Cancer Res., October 15, 2002; 62(20): 5939 - 5946. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Meeting Abstracts Online |
