GENES, CHROMOSOMES & CANCER 2:19/-197 (1990)
Genomic Alterations in Human Breast Carcinomas Catharina Larsson, Camilla Bystrorn, Larnbert Skoog, Sam Rotstein, and Magnus Nordenskjold Departments of Clinical Genetics (C.L., C.B., Stockholm, Sweden
M.N.),Tumor
Pathology (L.S.), and Oncology (S.R.), Karolinska Hospital,
All human chromosomes were screened in 52 human breast Carcinomas for the occurrence of allele losses, in order t o identify genomic alterations involved in initiation and progression of the disease. Loss of chromosome 22 alleles was detected in 6 out of 8 lobular carcinomas, while chromosome I 7 losses were most frequent in ductal carcinomas. Furthermore, patients who developed advanced disease after many years of mild clinical course showed significantly higher frequencies of allele losses in their primary tumors, compared t o patients with a persistently mild disease course. Finally, in one case, molecular examination suggested a translocation t( 10; 17) with coamplification of the ERBB2 oncogene and chromosome 10 sequences present in the two tumors from this patient, consistent with one of the tumors being a metastasis originating from a subclone of cells in the other tumor. INTRODUCTION
Breast cancer is a heterogeneous disease, occurring in several different clinical forms. T h e course of the disease is also variable and unpredictable. Some patients initially present with apparently low malignant tumors, follow a mild clinical course for several years, and still develop advanced disease with distant metastases many years later. It is not known what causes such unexpected changes in the clinical course, nor is it known what initiates the neoplastic transformation. However, since breast cancer is a major component of many different heritable syndromes, it is likely that genetic changes are important for initiation of tumorigenesis. One way of approaching both these questions would be to identify tumor specific mutations that occur in early stages of the disease. Initiation of tumorigenesis may involve chromosomal deletions unmasking recessive mutations. Such deletions are detected as losses of constitutional alleles in tumors from both hereditary and sporadic forms of the same type of tumor, and indicate the chromosomal localization of the disease locus. This mechanism has been proven for retinoblastoma (De Mars, 1970; Knudson, 1971; Cavenee et al., 1983, 1985) and MEN 1 associated tumors (Larsson et al., 1988) and is strongly implicated for renal cell carcinoma (Zbar et al., 1987; Seizinger et al., 1988; Bergerheim et al., 1989) and acoustic neuroma (Rouleau et al., 1987; Seizinger et al., 1987). In other types of tumors, allele losses reflect chromosomal mutations acquired during tumor progression. Thus, accumulation of allele losses in colorectal carcinoma is associated with a more aggressive course of the disease (Vogelstein et al., 1989). Similarly, progression of astrocytic 0 1990 WILEY-LISS, INC.
tumors to the most advanced stage specifically involves loss of one chromosome 10 complement (James et al., 1988). In primary breast carcinomas allele losses have been reported for the short arm of chromosome 3 (Devilee et al., 1989), l l p (Theillet et al., 1986; Ali et al., 1987), 13q (Lundberg et al., 1987), and 17p (Mackay et al., 1988). These studies have only included a small number of tumors or focused on single chromosomes, and the relationship between the described losses is therefore unclear. In addition, amplification of the ERBBZ oncogene has been suggested to be associated with poor prognostic indicators or poor outcome (Slamon et al., 1987). These observations suggest that chromosomal mutations are involved in initiation and progression of breast carcinoma. In this study we attempted to characterize early stages of breast carcinoma according to losses of constitutional alleles and amplification of ERBBZ and to evaluate their influences on the initial clinical presentation as well as the outcome of the disease. MATERIALS A N D METHODS Human Tissue Samples
Human primary breast cancer samples were obtained from 49 females and 2 males. All tumors were removed surgically prior to irradiation and chemotherapy, and fresh specimens were frozen at -70°C for 0.5-9 years before isolation of DNA.
Received March 12, 1990; accepted May 28, 1990. Address reprint requests to Dr. Catharina Larsson, Dept. of Clinical Genetics, Karolinska Hospital, S-104 01 Stockholm, Sweden.
I 92
LARSSON ET A L
T h e fraction of tumor cells in tissues was visually estimated by scoring representative histological sections at low magnification. Peripheral venous blood was obtained 0.3-9 years after surgery. Relevant clinical details are listed in Table 1. Southern Hybridizations
High molecular weight DNA from breast cancer samples was prepared as follows: Precut pieces of the tumors were homogenized for 3 seconds in 1 ml of a buffer containing 75 mM of NaCl and 24 mM of EDTA, pH 8.0. T h e Polytrone (Kinematica) probe was then washed twice with the same solution and the total volume of 3 ml was incubated for 10-16 h in room temperature with 0.5% SDS and 200 pg/ml of proteinase K. After purification of DNA (Bergerheim et al., 1989) the DNA concentrations were determined by spectrophotometry and quantitative gel analysis. Isolation of DNA from peripheral leukocytes and examination of restriction fragment length alleles were as previously described. RESULTS Allele Losses
Constitutional and tumor genotypes were compared for 52 primary breast carcinomas (Table 1). Fifty-five different polymorphic loci from all human chromosomes were examined to identify losses and related rearrangements of constitutional alleles in the tumor tissue (Table 2). In some cases a loss was detected as a total absence of one of the two parental alleles while in others there was only a reduced signal intensity. T h e last situation has several possible explanations. First, the mutation might be an early event in tumor development, reflecting unmasking of a recessive mutation, but the tumor contains a large proportion of normal cells. Second, the mutation occurred during tumor progression and is only present in a subclone of the tumor cells. A third possibility is that the chromosome of interest is duplicated, and when one homologue is lost the tumor will contain two copies of one allele and one of the other. Therefore, the detected losses were divided into two categories, total and partial. For this purpose the fraction of tumor cells in each sample was estimated on histopathological slides. This was found to vary between 30% and 100%. A loss was then defined as a complete or partial disappearance of one of the constitutional alleles that could be detected by the naked eye and confirmed by densitometry analysis showing a t least a one-third decrease of the signal
intensity. Losses of the same magnitude as the fraction of tumor cells were scored as total, unless other losses of a greater magnitude could be detected in the same tumor. All other losses were scored as partial (Fig. 1). A total of 159 allele losses were identified in 44 of the 52 tumors studied. In some tumors the losses only affected single chromosomes, while in others the rearrangements were multiple and complex. Losses were detected on all chromosomes except Y, and were for most chromosomes both of total and partial types (Table 2). T h e individual tumors also frequently showed both total and partial losses (Table 1, Fig. 1). For each tumor the fraction of losses was determined, i.e., the number of chromosomes on which allele losses were observed divided by the total number of chromosomes for which a polymorphic marker was informative in the patient's constitutional tissue. T h e tumors were also divided into different groups on the basis of histological type (McDivitt et al., 1969; Scarff and Torloni, 1968), clinical stage at diagnosis (UICC, 1982), and outcome after at least 5 years of follow-up. For each of these variables, the medium fraction of allele losses, as well as losses or duplications of specific loci were compared between the different groups. Tumors representing six different histological groups were examined (Fig. 1). Lobular carcinomas
T h e lobular carcinomas showed the lowest median fraction of losses, but there was no clear correlation between the fraction of losses on the one hand and prognostic markers or clinical outcome on the other. However, chromosome 22 was frequently involved, showing either total or partial losses in 6 out of 8 informative cases. T h e two tumors with retained genotype for chromosome 22 also retained heterozygosity for all informative markers on other chromosomes. Ductal carcinomas
T h e 34 tumors of ductal type had a higher level of the median fraction of allele losses. Chromosome 17 was most frequently affected, and all together, 17 total or partial losses could be identified for this chromosome in 28 informative cases (Table 1). In contrast, losses on chromosome 22 were only detected in 2 out of 28 informative cases. This indicates different genetic mechanisms involved in tumorigenesis of ductal and lobular breast carcinomas.
I 93
HUMAN BREAST CARCINOMAS
Other histological types
Five cases of mucinous carcinomas and two cases of papillary carcinomas showed varying fractions of total and partial losses that were spread over the genome with no preferential involvement of a certain chromosome (Table 1). T h e one case of tubular carcinoma had a total loss for chromosome 16, while the single medullary carcinoma showed complex rearrangements with multiple total and partial losses as well as duplications (Table 1). Allele Losses and Clinical Outcome
Seventeen of the patients with ductal carcinomas were included in the study at least three years after surgical treatment, and followed for five years or more. T h e other 17 tumors were included immediately after surgery and the patients followed for two years. One would therefore expect that among the latter cases there would be more patients with a less favorable prognosis. There was neither a significant difference in the median fraction of allele losses between the two groups nor a clear difference when the tumors were divided according to clinical stage at diagnosis. However, three patients in the first group developed distant metastases about five years after diagnosis (cases 47, 51, and 57). For these tumors the fractions of allele losses (23%, 27%, 27%) were higher than the median (10%) and the mean (13%) of the whole group (0.01 < P < 0.05, Wilcoxon’s Two Sample Rank Test). In addition, these three cases all had partial losses of chromosome 20 without duplication of the retained allele. Four patients with shorter followup period (cases 100, 103, 120, and 121) developed advanced disease soon after diagnosis, and three of these also had comparatively high fractions of allele losses (43%, 7%, 31%, 64%, respectively). Amplification of ERBB2
Amplification of the ERBBZ oncogene (3-fold or more) was found in 6 tumors from 5 different patients (cases 13, 48, 104, 113, and 120). These constituted a heterogeneous group with respect to age at diagnosis, histological type, presence of estrogen receptors, and clinical stage of the primary tumor. T h e tumors also showed varying fractions of allele losses, with no specific chromosome preferentially lost or duplicated. On the other hand, all patients had unilateral disease and multiple tumors at diagnosis (Table 1).
In one of these patients (case 104) molecular examination provided a model for development of the two tumors of this patient. T h e amplification of ERBBZ was detected as 3-fold in tumor A and more than 15-fold in tumor B (Fig. 2A). Both tumors also showed similar relative signal intensity of two 1Oq markers (DlOS1 and D10S25). For DlOS25 the constitutional allele was amplified, but for DlOSl the “amplification” was detected as an abnormal restriction fragment not present in the constitutional tissue (Fig. 2A). Additional chromosomal mutations were also found in the two tumors, including a partial loss of one X chromosome locus in tumor A, but not in tumor B and a partial loss of a 17q-marker in tumor B but not in tumor A (data not shown). T h e most likely interpretation of these data is that tumor A contains at least two subclones. One is characterized by a loss a t the DXYSl locus, and the other has a chromosomal rearrangement, probably a translocation t(lOq;l7q) (Fig. ZB), with the breakpoint on chromosome 10 close to the DlOSl locus, resulting in an abnormal junction restriction fragment (Fig. 2A). T h e same translocation was present to a greater extent in tumor B, which therefore most likely is a metastasis originating from the described subpopulation in tumor A (Fig. 2A). DISCUSSION
T h e present study shows that allele losses occur frequently in human breast carcinomas. Some losses are present in all cells of a tumor, while others represent subclones. T h e latter are interpreted as reflecting secondary events during tumor progression. In accordance with the two mutation hypothesis, allele losses that occur early during tumorigenesis can be regarded as candidate regions for localization of genes involved in oncogenesis of breast carcinomas. Such allele losses would be total, and the retained allele should be derived from the affected parent in inherited cases. Thus, such regions may be identified by detailed examination of the parental origin of the lost alleles in inherited cases of the disease (Cavenee et al., 1985; Larsson et al., 1988). T h e results also indicate that allele losses found in early stages of breast carcinomas reflect mutations that can affect the clinical course. Candidate regions for mutations that predispose to a more aggressive course of the disease can probably be found on chromosomes that are preferentially affected only in subpopulations of the tumors. On
I 94
LARSSON ET A L
TABLE I. Allelotypes of 52 Human Breast Carcinomas Chromosome number' Case no.
Age Survivalc (yrs) Sexa Stageb (yrs) ERd I
Lobular carcinomas 40 64 F 44 5 9 F 59 53 F 55 60 F 39 43 F 43 70 F 41 47 F Ill 64 F Ductal carcinomas 50 F 45 49 65 F 56 55 F 55 M 26 38 52 F 73 F 52 54 55 F 57 64 F 58 66 F 7 4 0 F I2 37 F 35 F 13 36 F 33 47 40 F 51 F 51 45 F 28 53 27 F 48 41 F 50 63 F 35 69 F 100 83 F 102 65 F 103 71 F 104A 51 F 1048 51 F 107 76 F 108 86 F 110 71 F 116 44 F 117 68 F 119 46 F 68 F I20 32 F I21 109 61 F 34' M
I I ll-B II
II I-B II 111 11-M I-M II I I I I1 I-A I I II Il-M I-B 11-A 11-A 11-B
I Il-M II I
Il-A I 11-A I-M I-M 111 I Il-M I1 II II Ill-A Ill-A I
10 10 10 6 6 5 3 2 II II II 10 10 10 10 10 10 9 9 9 9 8+ 8+ 6 5
3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
++ + +
2
0 .
(+)
+ . . .
+ ++ +
3
*
*
4
5 6
7
8
9
. . . .. .
II 6 6 2 2
Papillary carcinomas 46 67 F II 105 62 F II
10 2
12 13 14 15 16 17 18 19 20 21 22 XY
. .. .. ... ... .. . . . .. .. . . . . .... . .. . . . o . . .+ .. .. . . . .. . . . . . . *+ . . o+. ... 0
*+
. ... . .. . . . ..
*
++
0
+ * ND
0 0 . .
(+)
(+).
(+) (+)
*+
0
.
.
(+) ND
+ ++ *+ ++
0 .
O+ 0
*+
0
o+
0
.
0
.
0
' 0
0
B . . .
-
0
0
0
. 0 . . . . .
'
0 0 . g . . .
9
.
.. . . . .. .. . .. .. o+ .. .. .. . . . o . . . . . . . ..+ 0
0 .
'
*+
O
W
+
0
0
'
0
0 . '
*+
0
. a + .
0 .
0
0
0 "
o+
o+ o+
*+ o+
0
++ ++* ++ +
0 . 0
* . + . . .. .. . . . . . + . . . . @ * .. . . . '+ . .. . .+. .+. *+
0
(+) (+)
.
0 9 0 ' 0
0
.
0 .
.
.
a+
. .. .. ... . .+. - . .. .. . .. .. .. .. . .. -.. ..... . ++ + . . . .. .. . . . . - . . . . . . . 0 + ++ . . 0 . .. .. .. + . . . . .+.. .. . . *+. 0. o+. ' . (+I ' -... .... ...
(+)
.
.. .. .. . . . . . . . . . . . . . . . . . . ++ + . . . . . . . . . . . . .. .. + . . . o . o+ o+ o+ o+ o+ -+ oo+ . .. . .. ..+. . .. .... . . + . . . . . . . . o+ ... ++ o+ o+ '+ o o +**+ 0+
0 0
0
0
0
0
.
.
.
0
0
0
*
. . . 0 . . . 0 . . . . . . . 0 0
0
0 0
(+)
-
0 0
0
0 .
t+
(+I
++
++ o+ ++*
0 0
0 0 . 0
* 0 + .
.
0
.... .. . . . o+. *+
0
0
0+ 0+
ND
Mucinous carcinomas 42 59 F I 11 78 F 62 118 75 F I 113 46 F 11-M 114 68 F I
10 I I
. . a , . .
*
g
0 o
' f
0
.
0
....
*+
0+
0
. + . .
0
0 0 0 + . .
(continued)
I 95
HUMAN BREAST CARCINOMAS
TABLE I. Allelotyper of 52 Human Breast Carcinomas (continued) Chromosome numbere Case no.
Age
Survivalc
(yrs) Sex” Stageb
(yrs)
ERd I
115
66
F
a
‘
II
2
+
II
10
-
3
4
5
6
7
*+
0
8
9
10 I I
12 13 14 15 16 17 18 19 20 21 22 XY
... .. . .
. . .
Tubular carcinoma Medullary carcinoma 36 46 F
2
0
0
0
o+
0+
0 .
0 0 + 0
0
.
.
.
0
.
M = male, F = female. Stage I-IV, M = unilateral disease with multiple tumors; B = bilateral disease: A = advanced disease with distant metastasis. + = Mortal outcome. ER = presence of estrogen receptors; - = negative: (+) = weakly positive; + = positive; + + = strongly positive; ND = not determined. Allelotypes in the tumor tissues: = retained heterozygosity: 0 = total loss (early in tumor development); 0 = partial loss (in a subclone of cells): t = duplication. Uncharacterizedtumor that was included because that patient is a male.
TABLE 2. Summary of Allele Losses Fraction informative chromosomes with losses(%)
No. of chromosomes with losses
I 11-15 2 16-27 3 4 >28 3-1 I Total No. of total losses Total No. of partial losses Total No. of losses Total No. of informative cases
No. of tumors with allele losses on chromosomes”
I
Genomic Alterations in Human Breast Carcinomas Catharina Larsson, Camilla Bystrorn, Larnbert Skoog, Sam Rotstein, and Magnus Nordenskjold Departments of Clinical Genetics (C.L., C.B., Stockholm, Sweden
M.N.),Tumor
Pathology (L.S.), and Oncology (S.R.), Karolinska Hospital,
All human chromosomes were screened in 52 human breast Carcinomas for the occurrence of allele losses, in order t o identify genomic alterations involved in initiation and progression of the disease. Loss of chromosome 22 alleles was detected in 6 out of 8 lobular carcinomas, while chromosome I 7 losses were most frequent in ductal carcinomas. Furthermore, patients who developed advanced disease after many years of mild clinical course showed significantly higher frequencies of allele losses in their primary tumors, compared t o patients with a persistently mild disease course. Finally, in one case, molecular examination suggested a translocation t( 10; 17) with coamplification of the ERBB2 oncogene and chromosome 10 sequences present in the two tumors from this patient, consistent with one of the tumors being a metastasis originating from a subclone of cells in the other tumor. INTRODUCTION
Breast cancer is a heterogeneous disease, occurring in several different clinical forms. T h e course of the disease is also variable and unpredictable. Some patients initially present with apparently low malignant tumors, follow a mild clinical course for several years, and still develop advanced disease with distant metastases many years later. It is not known what causes such unexpected changes in the clinical course, nor is it known what initiates the neoplastic transformation. However, since breast cancer is a major component of many different heritable syndromes, it is likely that genetic changes are important for initiation of tumorigenesis. One way of approaching both these questions would be to identify tumor specific mutations that occur in early stages of the disease. Initiation of tumorigenesis may involve chromosomal deletions unmasking recessive mutations. Such deletions are detected as losses of constitutional alleles in tumors from both hereditary and sporadic forms of the same type of tumor, and indicate the chromosomal localization of the disease locus. This mechanism has been proven for retinoblastoma (De Mars, 1970; Knudson, 1971; Cavenee et al., 1983, 1985) and MEN 1 associated tumors (Larsson et al., 1988) and is strongly implicated for renal cell carcinoma (Zbar et al., 1987; Seizinger et al., 1988; Bergerheim et al., 1989) and acoustic neuroma (Rouleau et al., 1987; Seizinger et al., 1987). In other types of tumors, allele losses reflect chromosomal mutations acquired during tumor progression. Thus, accumulation of allele losses in colorectal carcinoma is associated with a more aggressive course of the disease (Vogelstein et al., 1989). Similarly, progression of astrocytic 0 1990 WILEY-LISS, INC.
tumors to the most advanced stage specifically involves loss of one chromosome 10 complement (James et al., 1988). In primary breast carcinomas allele losses have been reported for the short arm of chromosome 3 (Devilee et al., 1989), l l p (Theillet et al., 1986; Ali et al., 1987), 13q (Lundberg et al., 1987), and 17p (Mackay et al., 1988). These studies have only included a small number of tumors or focused on single chromosomes, and the relationship between the described losses is therefore unclear. In addition, amplification of the ERBBZ oncogene has been suggested to be associated with poor prognostic indicators or poor outcome (Slamon et al., 1987). These observations suggest that chromosomal mutations are involved in initiation and progression of breast carcinoma. In this study we attempted to characterize early stages of breast carcinoma according to losses of constitutional alleles and amplification of ERBBZ and to evaluate their influences on the initial clinical presentation as well as the outcome of the disease. MATERIALS A N D METHODS Human Tissue Samples
Human primary breast cancer samples were obtained from 49 females and 2 males. All tumors were removed surgically prior to irradiation and chemotherapy, and fresh specimens were frozen at -70°C for 0.5-9 years before isolation of DNA.
Received March 12, 1990; accepted May 28, 1990. Address reprint requests to Dr. Catharina Larsson, Dept. of Clinical Genetics, Karolinska Hospital, S-104 01 Stockholm, Sweden.
I 92
LARSSON ET A L
T h e fraction of tumor cells in tissues was visually estimated by scoring representative histological sections at low magnification. Peripheral venous blood was obtained 0.3-9 years after surgery. Relevant clinical details are listed in Table 1. Southern Hybridizations
High molecular weight DNA from breast cancer samples was prepared as follows: Precut pieces of the tumors were homogenized for 3 seconds in 1 ml of a buffer containing 75 mM of NaCl and 24 mM of EDTA, pH 8.0. T h e Polytrone (Kinematica) probe was then washed twice with the same solution and the total volume of 3 ml was incubated for 10-16 h in room temperature with 0.5% SDS and 200 pg/ml of proteinase K. After purification of DNA (Bergerheim et al., 1989) the DNA concentrations were determined by spectrophotometry and quantitative gel analysis. Isolation of DNA from peripheral leukocytes and examination of restriction fragment length alleles were as previously described. RESULTS Allele Losses
Constitutional and tumor genotypes were compared for 52 primary breast carcinomas (Table 1). Fifty-five different polymorphic loci from all human chromosomes were examined to identify losses and related rearrangements of constitutional alleles in the tumor tissue (Table 2). In some cases a loss was detected as a total absence of one of the two parental alleles while in others there was only a reduced signal intensity. T h e last situation has several possible explanations. First, the mutation might be an early event in tumor development, reflecting unmasking of a recessive mutation, but the tumor contains a large proportion of normal cells. Second, the mutation occurred during tumor progression and is only present in a subclone of the tumor cells. A third possibility is that the chromosome of interest is duplicated, and when one homologue is lost the tumor will contain two copies of one allele and one of the other. Therefore, the detected losses were divided into two categories, total and partial. For this purpose the fraction of tumor cells in each sample was estimated on histopathological slides. This was found to vary between 30% and 100%. A loss was then defined as a complete or partial disappearance of one of the constitutional alleles that could be detected by the naked eye and confirmed by densitometry analysis showing a t least a one-third decrease of the signal
intensity. Losses of the same magnitude as the fraction of tumor cells were scored as total, unless other losses of a greater magnitude could be detected in the same tumor. All other losses were scored as partial (Fig. 1). A total of 159 allele losses were identified in 44 of the 52 tumors studied. In some tumors the losses only affected single chromosomes, while in others the rearrangements were multiple and complex. Losses were detected on all chromosomes except Y, and were for most chromosomes both of total and partial types (Table 2). T h e individual tumors also frequently showed both total and partial losses (Table 1, Fig. 1). For each tumor the fraction of losses was determined, i.e., the number of chromosomes on which allele losses were observed divided by the total number of chromosomes for which a polymorphic marker was informative in the patient's constitutional tissue. T h e tumors were also divided into different groups on the basis of histological type (McDivitt et al., 1969; Scarff and Torloni, 1968), clinical stage at diagnosis (UICC, 1982), and outcome after at least 5 years of follow-up. For each of these variables, the medium fraction of allele losses, as well as losses or duplications of specific loci were compared between the different groups. Tumors representing six different histological groups were examined (Fig. 1). Lobular carcinomas
T h e lobular carcinomas showed the lowest median fraction of losses, but there was no clear correlation between the fraction of losses on the one hand and prognostic markers or clinical outcome on the other. However, chromosome 22 was frequently involved, showing either total or partial losses in 6 out of 8 informative cases. T h e two tumors with retained genotype for chromosome 22 also retained heterozygosity for all informative markers on other chromosomes. Ductal carcinomas
T h e 34 tumors of ductal type had a higher level of the median fraction of allele losses. Chromosome 17 was most frequently affected, and all together, 17 total or partial losses could be identified for this chromosome in 28 informative cases (Table 1). In contrast, losses on chromosome 22 were only detected in 2 out of 28 informative cases. This indicates different genetic mechanisms involved in tumorigenesis of ductal and lobular breast carcinomas.
I 93
HUMAN BREAST CARCINOMAS
Other histological types
Five cases of mucinous carcinomas and two cases of papillary carcinomas showed varying fractions of total and partial losses that were spread over the genome with no preferential involvement of a certain chromosome (Table 1). T h e one case of tubular carcinoma had a total loss for chromosome 16, while the single medullary carcinoma showed complex rearrangements with multiple total and partial losses as well as duplications (Table 1). Allele Losses and Clinical Outcome
Seventeen of the patients with ductal carcinomas were included in the study at least three years after surgical treatment, and followed for five years or more. T h e other 17 tumors were included immediately after surgery and the patients followed for two years. One would therefore expect that among the latter cases there would be more patients with a less favorable prognosis. There was neither a significant difference in the median fraction of allele losses between the two groups nor a clear difference when the tumors were divided according to clinical stage at diagnosis. However, three patients in the first group developed distant metastases about five years after diagnosis (cases 47, 51, and 57). For these tumors the fractions of allele losses (23%, 27%, 27%) were higher than the median (10%) and the mean (13%) of the whole group (0.01 < P < 0.05, Wilcoxon’s Two Sample Rank Test). In addition, these three cases all had partial losses of chromosome 20 without duplication of the retained allele. Four patients with shorter followup period (cases 100, 103, 120, and 121) developed advanced disease soon after diagnosis, and three of these also had comparatively high fractions of allele losses (43%, 7%, 31%, 64%, respectively). Amplification of ERBB2
Amplification of the ERBBZ oncogene (3-fold or more) was found in 6 tumors from 5 different patients (cases 13, 48, 104, 113, and 120). These constituted a heterogeneous group with respect to age at diagnosis, histological type, presence of estrogen receptors, and clinical stage of the primary tumor. T h e tumors also showed varying fractions of allele losses, with no specific chromosome preferentially lost or duplicated. On the other hand, all patients had unilateral disease and multiple tumors at diagnosis (Table 1).
In one of these patients (case 104) molecular examination provided a model for development of the two tumors of this patient. T h e amplification of ERBBZ was detected as 3-fold in tumor A and more than 15-fold in tumor B (Fig. 2A). Both tumors also showed similar relative signal intensity of two 1Oq markers (DlOS1 and D10S25). For DlOS25 the constitutional allele was amplified, but for DlOSl the “amplification” was detected as an abnormal restriction fragment not present in the constitutional tissue (Fig. 2A). Additional chromosomal mutations were also found in the two tumors, including a partial loss of one X chromosome locus in tumor A, but not in tumor B and a partial loss of a 17q-marker in tumor B but not in tumor A (data not shown). T h e most likely interpretation of these data is that tumor A contains at least two subclones. One is characterized by a loss a t the DXYSl locus, and the other has a chromosomal rearrangement, probably a translocation t(lOq;l7q) (Fig. ZB), with the breakpoint on chromosome 10 close to the DlOSl locus, resulting in an abnormal junction restriction fragment (Fig. 2A). T h e same translocation was present to a greater extent in tumor B, which therefore most likely is a metastasis originating from the described subpopulation in tumor A (Fig. 2A). DISCUSSION
T h e present study shows that allele losses occur frequently in human breast carcinomas. Some losses are present in all cells of a tumor, while others represent subclones. T h e latter are interpreted as reflecting secondary events during tumor progression. In accordance with the two mutation hypothesis, allele losses that occur early during tumorigenesis can be regarded as candidate regions for localization of genes involved in oncogenesis of breast carcinomas. Such allele losses would be total, and the retained allele should be derived from the affected parent in inherited cases. Thus, such regions may be identified by detailed examination of the parental origin of the lost alleles in inherited cases of the disease (Cavenee et al., 1985; Larsson et al., 1988). T h e results also indicate that allele losses found in early stages of breast carcinomas reflect mutations that can affect the clinical course. Candidate regions for mutations that predispose to a more aggressive course of the disease can probably be found on chromosomes that are preferentially affected only in subpopulations of the tumors. On
I 94
LARSSON ET A L
TABLE I. Allelotypes of 52 Human Breast Carcinomas Chromosome number' Case no.
Age Survivalc (yrs) Sexa Stageb (yrs) ERd I
Lobular carcinomas 40 64 F 44 5 9 F 59 53 F 55 60 F 39 43 F 43 70 F 41 47 F Ill 64 F Ductal carcinomas 50 F 45 49 65 F 56 55 F 55 M 26 38 52 F 73 F 52 54 55 F 57 64 F 58 66 F 7 4 0 F I2 37 F 35 F 13 36 F 33 47 40 F 51 F 51 45 F 28 53 27 F 48 41 F 50 63 F 35 69 F 100 83 F 102 65 F 103 71 F 104A 51 F 1048 51 F 107 76 F 108 86 F 110 71 F 116 44 F 117 68 F 119 46 F 68 F I20 32 F I21 109 61 F 34' M
I I ll-B II
II I-B II 111 11-M I-M II I I I I1 I-A I I II Il-M I-B 11-A 11-A 11-B
I Il-M II I
Il-A I 11-A I-M I-M 111 I Il-M I1 II II Ill-A Ill-A I
10 10 10 6 6 5 3 2 II II II 10 10 10 10 10 10 9 9 9 9 8+ 8+ 6 5
3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
++ + +
2
0 .
(+)
+ . . .
+ ++ +
3
*
*
4
5 6
7
8
9
. . . .. .
II 6 6 2 2
Papillary carcinomas 46 67 F II 105 62 F II
10 2
12 13 14 15 16 17 18 19 20 21 22 XY
. .. .. ... ... .. . . . .. .. . . . . .... . .. . . . o . . .+ .. .. . . . .. . . . . . . *+ . . o+. ... 0
*+
. ... . .. . . . ..
*
++
0
+ * ND
0 0 . .
(+)
(+).
(+) (+)
*+
0
.
.
(+) ND
+ ++ *+ ++
0 .
O+ 0
*+
0
o+
0
.
0
.
0
' 0
0
B . . .
-
0
0
0
. 0 . . . . .
'
0 0 . g . . .
9
.
.. . . . .. .. . .. .. o+ .. .. .. . . . o . . . . . . . ..+ 0
0 .
'
*+
O
W
+
0
0
'
0
0 . '
*+
0
. a + .
0 .
0
0
0 "
o+
o+ o+
*+ o+
0
++ ++* ++ +
0 . 0
* . + . . .. .. . . . . . + . . . . @ * .. . . . '+ . .. . .+. .+. *+
0
(+) (+)
.
0 9 0 ' 0
0
.
0 .
.
.
a+
. .. .. ... . .+. - . .. .. . .. .. .. .. . .. -.. ..... . ++ + . . . .. .. . . . . - . . . . . . . 0 + ++ . . 0 . .. .. .. + . . . . .+.. .. . . *+. 0. o+. ' . (+I ' -... .... ...
(+)
.
.. .. .. . . . . . . . . . . . . . . . . . . ++ + . . . . . . . . . . . . .. .. + . . . o . o+ o+ o+ o+ o+ -+ oo+ . .. . .. ..+. . .. .... . . + . . . . . . . . o+ ... ++ o+ o+ '+ o o +**+ 0+
0 0
0
0
0
0
.
.
.
0
0
0
*
. . . 0 . . . 0 . . . . . . . 0 0
0
0 0
(+)
-
0 0
0
0 .
t+
(+I
++
++ o+ ++*
0 0
0 0 . 0
* 0 + .
.
0
.... .. . . . o+. *+
0
0
0+ 0+
ND
Mucinous carcinomas 42 59 F I 11 78 F 62 118 75 F I 113 46 F 11-M 114 68 F I
10 I I
. . a , . .
*
g
0 o
' f
0
.
0
....
*+
0+
0
. + . .
0
0 0 0 + . .
(continued)
I 95
HUMAN BREAST CARCINOMAS
TABLE I. Allelotyper of 52 Human Breast Carcinomas (continued) Chromosome numbere Case no.
Age
Survivalc
(yrs) Sex” Stageb
(yrs)
ERd I
115
66
F
a
‘
II
2
+
II
10
-
3
4
5
6
7
*+
0
8
9
10 I I
12 13 14 15 16 17 18 19 20 21 22 XY
... .. . .
. . .
Tubular carcinoma Medullary carcinoma 36 46 F
2
0
0
0
o+
0+
0 .
0 0 + 0
0
.
.
.
0
.
M = male, F = female. Stage I-IV, M = unilateral disease with multiple tumors; B = bilateral disease: A = advanced disease with distant metastasis. + = Mortal outcome. ER = presence of estrogen receptors; - = negative: (+) = weakly positive; + = positive; + + = strongly positive; ND = not determined. Allelotypes in the tumor tissues: = retained heterozygosity: 0 = total loss (early in tumor development); 0 = partial loss (in a subclone of cells): t = duplication. Uncharacterizedtumor that was included because that patient is a male.
TABLE 2. Summary of Allele Losses Fraction informative chromosomes with losses(%)
No. of chromosomes with losses
I 11-15 2 16-27 3 4 >28 3-1 I Total No. of total losses Total No. of partial losses Total No. of losses Total No. of informative cases
No. of tumors with allele losses on chromosomes”
I
