- Email: [email protected]
Quantitative DNA analysis and proliferation in breast carcinomas
Path. Res. Pract. 188, 428-432 (1992)
Quantitative DNA Analysis and Proliferation in Breast Carcinomas 1 A Comparison between Image Analysis and Flow Cytometry A. K. C. Lee, B. Wiley, J. M. Dugan, W. H. Hamilton, M. Loda, G. J. Heatley, L. Cook and M. L. Silverman Department of Anatomic Pathology and the Sias Surgical Research Unit, Lahey Clinic Medical Center; Burlington, Massachusetts; the Department of Pathology, New England Deaconess Hospital and Harvard Medical School, Boston, Massachusetts, U.S.A.
SUMMARY
The DNA content and proliferation in 100 invasive breast carcinomas were evaluated by computerized image analysis (IA) and flow cytometry (FCM). For DNA content, image analysis of Feulgen-stained slides of fresh tumor imprints were compared with flow cytometry of propidium iodide-stained disaggregated fresh tumor tissue. The DNA indices obtained by the two methods showed close correlation by linear regression analysis (r=0.89, p <.001). There were 44 (44 %) diploid and 56 (56 %) aneuploid tumors. There was agreement between the two methods in detection of aneuploidy in 81 % of tumors. Image analysis required smaller tissue samples, permitted direct visualization and selection of tumor cells, and was more sensitive in detecting tetraploid and highly aneuploid cell populations. In contrast, flow cytometry histograms provided better resolution, and were more effective in detecting multiploid tumors and near-diploid aneuploid tumors. Aneuploidy was significantly related to various adverse prognostic parameters, namely, negative estrogen receptor, high mitotic rate, high histologic and nuclear grades. Proliferation was evaluated by measuring the FCM S phase fraction (SPF), and by image analysis quantitation of immunohistochemical staining using Ki-67 monoclonal antibody. SPF and Ki-67 count showed modest correlation (r=0.42). Both SPF and Ki-67 count were significantly related to the mitotic rate, histologic and nuclear grades. Our results indicate that the two methods provide comparable results, but offer individual advantages and are complementary techniques in analyzing DNA ploidy and proliferation in breast carcinomas.
Introduction
Quantitative DNA analysis and proliferation of breast carcinomas may provide useful prognostic information 4 . 1 Supported in part by National Cancer Institute Grant CA 40395 and the Eleanor Naylor Dana Charitable Trust, New York, N.Y.
0344·0338/92/0188-0428$3.50/0
The advent of computer-assisted image analysis has provided an alternative way to flow cytometry in measuring DNA content and proliferative rate l , 2, 5, 6, 7,15. This prospective study compared the results of flow cytometry and image analysis in evaluating the DNA content and proliferation rate of 100 invasive breast carcinomas, and expanded our previous observations on the DNA content of a smaller series of breast cancer patients 12 • © 1992 by Gustav Fischer Verlag, Stuttgart
DNA Content & Proliferation . 429
Material and Methods Fresh tissue obtained from surgical specimens of invasive breast carcinomas was used for flow cytometry and image analysis studies. Routine histologic parameters were determined on H&E sections of the carcinomas. There were 79 ductal, 18 lobular, 2 colloid and 1 metaplastic spindle cell carcinomas.
Image Analysis For DNA content, slides of fresh tumor imprints were fixed in 10 % buffered neutral formalin (30 minutes), washed in distilled water (10 minutes) and stained with the Cell Analysis System Feulgen staining kit (CAS Inc., Elmhurst, IL., U.S.A.). Image analysis was performed by the CAS-200 Image Analyzer (CAS Inc.) using a quantitative DNA software program. In each case, between 150-250 well preserved, non-overlapping cells, were analyzed. The DNA content was determined by the analyzer based on the premise that the total nuclear optical density of each nucleus was proportional to the amount of DNA present. A DNA histogram was generated and the DNA indices of the primary and secondary peaks were determined. The histograms were classified as diploid and aneuploid, and the latter was further subclassified as hypodiploid, hyperdiploid, tetraploid, multiploid or hypertetraploid, depending on the number and position of the abnormal peaks. Proliferation was evaluated by immunohistochemical staining of acetone-fixed fresh frozen tissue sections in 61 tumors using the Ki-67 monoclonal antibody (Dako Corp. Santa Barbara, CA., U.S.A.) (Fig. 1). The percentage of stained nuclei was quantitated with the image analyzer using the Cell Proliferation Index software program.
Flow Cytometry Flow cytometric analysis was performed o,n fresh tumor tissue, which was mechanically and enzymatically dissociated, treated with detergent to obtain bare nuclei, and
Fig. 1. Immunoperoxidase staining of fresh frozen tissue section of breast carcinoma using Ki-67 monoclonal antibody, showing the nuclear staining and accentuation of nucleolar areas.
stained with propidium iodide. Non-tumor tissue from the same specimen or sex-matched lymphocytes were used as controls. The nuclear suspensions were analyzed on a Coulter EPICS-C flow cytometer (Coulter Electronics, Hialeah, FL). A minimum of 10,000 events were measured for each sample. The histograms were classified as in image analysis. SPF was determined using the SOBR model or the SFIT model.
Statistical Analysis Linear regression analysis to compare results obtained by flow cytometry and image analysis, and univarate analysis to evaluate the relationship of ploidy and proliferation to other variables, were performed. Continuous variables were analyzed for normality using the ShapiroWilk analysis of variance. Statistical differences between variables were analyzed using unpaired t-tests or Wilcoxon rank sum analysis as appropriate (BMDP3D). Contingency tables were analyzed by Miettinen's modification of the Fisher exact test. Probability values were two-tailed with p <0.05 regarded as statistically significant. Correlations between variables were analyzed by simple linear regression using BMDP6D Statistical Software.
Results
Quantitative DNA Analysis There was good correlation between the DNA indices obtained with image analysis and flow cytometry by linear regression analysis (r=0.89, p <0.001, Fig. 2A). For the detection of aneuploidy, there was complete agreement in the results obtained by both methods in 81 carcinomas (81 %) (Fig. 3), with discordance observed in 6 carcinomas (6 %) while 10 tumors were equivocal by image analysis and 3 by flow cytometry (Table 1). The equivocal tumors by flow cytometry displayed borderline increase in the
430
A. K. C. Lee et aI.
3 2.8
Table 1. Quantitative DNA analysis of 100 breast carcinomas
r- 089 pdl01
Flow cytometry
Image analysis
Cordordant cases (n=81)
diploid aneuploid
diploid aneuploid
36 45
Discordant cases (n=6)
aneuploid aneuploid"
diploid aneuploid'c
1 5
Equivocal cases (n=13)
diploid aneuploid ?
? ? tetraploid
8 2 3
)(
~ ~ c
.
~
2
18 16
~
f
total ,. discordance in aneuploidy patterns ? equivocal cases
Flow Cytometry: DNA Index
A
100
50,--------------------------------------,
40
~
o
0
o
30
Q)
(;
"
C/l
"'l'
20
;;::
10 o
o~O~O~O~0r_~§~O~O~O~._--,---_,--_.----._--~~ o
4
8
12
20
16
28
24
32
S Phase Fraction (%)
B
Fig. 2. Linear regression analysis comparing A) DNA indices measured by image analysis and flow cytometry, and B) proliferation measured by image analysis of Ki-67 staining and S-phase fraction measured by flow cytometry.
G2M fraction but a definite tetraploid peak was seen on image analysis. In general, even when FCM detected tetraploid tumors, the tetraploid peaks were smaller and more subtle than observed by image analysis. The equivocal tumors observed by image analysis presented difficult histograms in the near-diploid range usually because the small number of analyzed cells did not provide adequate
J
!!2 Qi ()
'0 Q;
E
::J
o
8
16
24
DNA (Picograms) Image Analysis
o
5
10
Proliferation Rate
The range of S phase fraction measured by flow cytometry was 1-40 %, (median 8.5 %, mean 11.1 +/- 8.2 %), while the Ki-67 score of stained nuclei was 1-55 %, Table 2. Relationship of aneuploidy and proliferation to other variables Aneuploidy
.D
Z
resolution. In such equivocal cases by image analysis, the coefficients of variaton (C.V.) of total optical density of the main peak and the DNA index of internal lymphocyte control were helpful in differentiating diploid from neardiploid aneuploid rumors. Diploid tumors had smaller C.V., with the cut-off point being 8 % in this series. Flow cytometry provided histograms with better resolution and was more sensitive in detecting multiploid rumors and near-diploid aneuploid rumors. In contrast, image analysis permitted morphologic correlation and selection of tumor cells, and was more effective in detecting tetraploid carcinomas and the presence of highly aneuploid cell populations to the right tail of the histogram. By flow cytometry it was frequently difficult to determine whether the signals present in the right tail of the histogram represented highly aneuploid cells or rumor cell aggregates. The presence of aneuploidy was significantly related to various adverse prognostic parameters (Table 2).
15
20
25
DNA (Fluorescence) Flow Cytometry
Fig. 3. Comparison of DNA histograms obtained by image analysis and flow cytonletry. In this case, both methods demonstrated the presence of a hyperdiploid peak.
Estrogen receptor(-) 0.002 Progesterone receptor(-) 0.12 High mitotic rate <0.001 High histologic grade 0.002 High nuclear grade <0.001 Tumor size 0.88 Lymphatic & blood vessel 0.26 mvaSIOn Menopausal status 0.79 Lymph node status 0.43
p Value S Phase Ki-67 count fraction 0.007 0.67 <0.001 0.002 <0.001 0.76 0.43
0.4 0.87 <0.001 <0.001 <0.001 0.04 0.84
0.11 0.53
0.75 0.46
DNA Content & Proliferation . 431 55 50 45 40 35
C
:; 0 ()
~
s::
30
....
25 20
Mo
15 10 MO
Low
High Mitotic Rate
Range 1-13% Mean
A
5%
Range 1-47% Mean 17%
50 45 40 C
0
."5
u:
til
Cl>
35 30 25
en til .c 0...
15
CfJ
10
20 M• • • •
M• • • • • • • • • •
....•.....
......... Low
B
Mitotic Rate
Range 1-31% Mean 7%
High
Range 3-40% Mean 14%
Fig. 4. Relationship of mitotic rate to A) Ki-67 antibody count (p<0.001) and B) S-phase fraction (p
(median 10.0 %, mean 14.5 +/-13.1 %). Linear regression analysis showed only moderate correlation between the methods (r=0.42, p=0.002, Fig.2B). Both S phase fraction and Ki-67 staining were significantly related to high mitotic rate (Fig. 4), and high histologic and nuclear grades. Aneuploid tumors demonstrated significantly higher S phase fractions (p
aneuploid cells beyond the tetraploid region. One drawback appears to be its occasional difficulty in resolving equivocal histograms in the diploid/near-diploid range, as a result of the relatively smaller number of cells analyzed. In contrast, FCM measures a larger number of cells and offers histograms with better resolution. Consequently, it is superior to IA in detecting multiploid tumors and in distinguishing diploid from near-diploid aneuploid qtmors. Hence, the potential sources of errm with FCM and IA lie in the interpretation of abnormalities in the tetraploid and diploid regions, respectively2, 7, 12. Some possible reasons for FCM to underestimate tetraploid and highly aneuploid cell populations include the dilution effect by non-neoplastic cells, relative fragility of aneuploid cells, and sampling problems. Also, FCM may not be able to distinguish between highly aneuploid tumor cells and tumor cell aggregates. Proliferation in solid tumors may be measured by diverse methods, namely mitotic count, thymidine labeling index, bromodeoxyuridine uptake, argyrophilic nucleolar organizer region count, SPF measurement and quantitation of proliferation-associated antigens such as Ki-67 and Cyclin. SPF estimation may include non-neoplastic cells, and may sometimes be difficult, especially in tumors with complex histograms displaying overlapping peaks and containing substantial debris. Also, reproducibility may be poor because of interobserver variations and the many sophisticated mathematical models being used to interpret S phase fractions. Ki-67 and other proliferation associated antigens require frozen sections although there are now monoclonal antibodies applicable to paraffin sections. SPF measures cells in the S phase of the cell cycle, whereas proliferation-associated antigens measure different phases of the cell cycle, for instance Ki-67 identifies all cells in the cell cycle9 and Cyclin correlates more with the SPF. Our SPF values are within the range reported by others 16 , whereas our Ki-67 scores are generally lower than other series3 , 7,10,13,14 but similar to that ofIsoli et al. 10 • Our data indicate that the Ki-67 count and SPF bear moderate correlation by linear regression analysis. In concordance with other studies 3, 7, 8, 10,13,14 they both correlate well with the mitotic rate and other measures of proliferation. IA and FCM appear to be complementary techniques in evaluating DNA content and proliferation in breast cancers. These two parameters are related to various adverse prognostic parameters. It remains to be seen if they represent independent prognostic parameters that may be useful in prognostication and in guiding management of breast cancer patients.
References 1 Aner GU, Askensten U, Erhardt K, Fallenius A, Zetterberg A (198'7<) Comparison between slide and flow cytophotometric DNA measurements in breast tumors. Anal Quant Cytol Histol9: 138 2 Bauer TW, Tubbs RR, Edinger MG, Suit PF, Gephardt GN, Levin HS (1990) A prospective comparison of DNA quantitation by image and flow cytometry. Am J Clin Pathol 93: 322
432 . A. K. C. Lee et a1. 3 Charpin C, Andrac L, Habib M, Vacheret H, Xerri L, Devictor B, Lavaut MN, Toga M (1989) Immunodetection in fine-needle aspirates and multiparametric (SAMBA) image analysis. Receptors and growth fraction evaluation in breast carcinomas. Cancer 63: 863 4 Clark GM, Dressler LG, Owens MA, Pounds G, Oldaker T, McGuire WL (1989) Prediction of relapse or survival in patients with node-negative breast cancer by DNA flow cytometry. N Engl J Med 320: 627 5 Claud RD 3d, Weinstein RS, Howeedy A, Straus AK, Coon JS (1989) Comparison of image analysis of imprints with flow cytometry for DNA analysis of solid tumors. Modern Pathol 2: 463 6 Cornellise q, Van Driel-Kulker AM (1985) DNA image cytometry on machine-selected breast cancer cells and a comparison between flow cytometry and scanning cytophotometry. Cytometry 6: 471 7 Dawson AE, Norton JA, Weinbert CS (1990) Comparative assessment of proliferation and DNA content in breast carcinoma by image analysis and flow cytometry. Am J Pathol136: 1115 8 Dervan PA, Gilmartin LG, Loftus BM, Carney DN (1989) Breast carcinoma kinetics. Argyrophilic nucleolar organizer region counts correlate with Ki-67 scores. Am J CI Pathol 92: 401 9 GerdesJ, Lemke H, Baisch H, Wacker H, Schwab M, Stein H (1984) Cell cycle analysis of a cell-proliferation associated nuclear antigen defined by the monoclonal antibody Ki-67. J Immuno1. 133: 1710
10 Isola 11, Helin HJ, Jelle MJ, Kallioniemi 0 (1990) Evaluation of cell proliferation in breast carcinoma. Comparison of Ki-67 immunohistochemical study, DNA flow cytometric analysis, and mitotic count. Cancer 65: 1180 11 Koss LG, Wersto RP, Simmons DA, Deitch D, Herz F, Free SZ (1989) Predictive value of DNA measurements in bladder washings. Comparison of flow cytometry, image cytophotometry and cytology in patients with a past history of urothelial tumors. Cancer 64: 916 12 Lee AK, Dugan J, Hamilton WM, Cook L, Heatley G, Kamat B, Silverman ML. Quantitative DNA analysis in breast carcinomas. A comparison between image analysis and flow cytometry. Modern Pathology 4: 178 13 McGurrinJF, Doria MI, Dawson TJ, Karrison T, Stein HO, Franklin WA (1987) Assessment of tumor cell kinetics by immunohistochemistry in carcinoma of breast. Cancer 59: 1744 14 Walker R, Camplejohn RS (1988) Comparison of monoclonal antibody Ki-67 reactivity with grade and DNA flow cytometry of breast carcinomas. Br J Cancer 57: 281 15 Uyterlinde AM, Smeulders AW, Baak JP (1989) Reproducibility and comparison of quantitative DNA histogram features obtained with a scanning micro densitometer and a flow cytometer in breast cancers. Anal Quant Cytol Histol 11: 353 16 Visscher DW, Zarbo RJ, Greenawald KA, Crissman JD (1990) Prognostic significance of morphological parameters and flow cytometric DNA analysis in carcinoma of the breast. In: Pathology Annual, Rosen PP, Fechner R (Eds) vol. 25, part 1, p 171. Norwal, CT: Appleton & Lange
Received September 29; 1991 . Accepted October 28, 1991
Key words: Breast cancer - DNA - Proliferation - Flow cytometry - Image analysis Dr. Arthur K. C. Lee, Department of Anatomic Pathology, Lahey Clinic Medical Center, 41 Mall Road, Burlington, Massachusetts, 01805 U.S.A.
Quantitative DNA Analysis and Proliferation in Breast Carcinomas 1 A Comparison between Image Analysis and Flow Cytometry A. K. C. Lee, B. Wiley, J. M. Dugan, W. H. Hamilton, M. Loda, G. J. Heatley, L. Cook and M. L. Silverman Department of Anatomic Pathology and the Sias Surgical Research Unit, Lahey Clinic Medical Center; Burlington, Massachusetts; the Department of Pathology, New England Deaconess Hospital and Harvard Medical School, Boston, Massachusetts, U.S.A.
SUMMARY
The DNA content and proliferation in 100 invasive breast carcinomas were evaluated by computerized image analysis (IA) and flow cytometry (FCM). For DNA content, image analysis of Feulgen-stained slides of fresh tumor imprints were compared with flow cytometry of propidium iodide-stained disaggregated fresh tumor tissue. The DNA indices obtained by the two methods showed close correlation by linear regression analysis (r=0.89, p <.001). There were 44 (44 %) diploid and 56 (56 %) aneuploid tumors. There was agreement between the two methods in detection of aneuploidy in 81 % of tumors. Image analysis required smaller tissue samples, permitted direct visualization and selection of tumor cells, and was more sensitive in detecting tetraploid and highly aneuploid cell populations. In contrast, flow cytometry histograms provided better resolution, and were more effective in detecting multiploid tumors and near-diploid aneuploid tumors. Aneuploidy was significantly related to various adverse prognostic parameters, namely, negative estrogen receptor, high mitotic rate, high histologic and nuclear grades. Proliferation was evaluated by measuring the FCM S phase fraction (SPF), and by image analysis quantitation of immunohistochemical staining using Ki-67 monoclonal antibody. SPF and Ki-67 count showed modest correlation (r=0.42). Both SPF and Ki-67 count were significantly related to the mitotic rate, histologic and nuclear grades. Our results indicate that the two methods provide comparable results, but offer individual advantages and are complementary techniques in analyzing DNA ploidy and proliferation in breast carcinomas.
Introduction
Quantitative DNA analysis and proliferation of breast carcinomas may provide useful prognostic information 4 . 1 Supported in part by National Cancer Institute Grant CA 40395 and the Eleanor Naylor Dana Charitable Trust, New York, N.Y.
0344·0338/92/0188-0428$3.50/0
The advent of computer-assisted image analysis has provided an alternative way to flow cytometry in measuring DNA content and proliferative rate l , 2, 5, 6, 7,15. This prospective study compared the results of flow cytometry and image analysis in evaluating the DNA content and proliferation rate of 100 invasive breast carcinomas, and expanded our previous observations on the DNA content of a smaller series of breast cancer patients 12 • © 1992 by Gustav Fischer Verlag, Stuttgart
DNA Content & Proliferation . 429
Material and Methods Fresh tissue obtained from surgical specimens of invasive breast carcinomas was used for flow cytometry and image analysis studies. Routine histologic parameters were determined on H&E sections of the carcinomas. There were 79 ductal, 18 lobular, 2 colloid and 1 metaplastic spindle cell carcinomas.
Image Analysis For DNA content, slides of fresh tumor imprints were fixed in 10 % buffered neutral formalin (30 minutes), washed in distilled water (10 minutes) and stained with the Cell Analysis System Feulgen staining kit (CAS Inc., Elmhurst, IL., U.S.A.). Image analysis was performed by the CAS-200 Image Analyzer (CAS Inc.) using a quantitative DNA software program. In each case, between 150-250 well preserved, non-overlapping cells, were analyzed. The DNA content was determined by the analyzer based on the premise that the total nuclear optical density of each nucleus was proportional to the amount of DNA present. A DNA histogram was generated and the DNA indices of the primary and secondary peaks were determined. The histograms were classified as diploid and aneuploid, and the latter was further subclassified as hypodiploid, hyperdiploid, tetraploid, multiploid or hypertetraploid, depending on the number and position of the abnormal peaks. Proliferation was evaluated by immunohistochemical staining of acetone-fixed fresh frozen tissue sections in 61 tumors using the Ki-67 monoclonal antibody (Dako Corp. Santa Barbara, CA., U.S.A.) (Fig. 1). The percentage of stained nuclei was quantitated with the image analyzer using the Cell Proliferation Index software program.
Flow Cytometry Flow cytometric analysis was performed o,n fresh tumor tissue, which was mechanically and enzymatically dissociated, treated with detergent to obtain bare nuclei, and
Fig. 1. Immunoperoxidase staining of fresh frozen tissue section of breast carcinoma using Ki-67 monoclonal antibody, showing the nuclear staining and accentuation of nucleolar areas.
stained with propidium iodide. Non-tumor tissue from the same specimen or sex-matched lymphocytes were used as controls. The nuclear suspensions were analyzed on a Coulter EPICS-C flow cytometer (Coulter Electronics, Hialeah, FL). A minimum of 10,000 events were measured for each sample. The histograms were classified as in image analysis. SPF was determined using the SOBR model or the SFIT model.
Statistical Analysis Linear regression analysis to compare results obtained by flow cytometry and image analysis, and univarate analysis to evaluate the relationship of ploidy and proliferation to other variables, were performed. Continuous variables were analyzed for normality using the ShapiroWilk analysis of variance. Statistical differences between variables were analyzed using unpaired t-tests or Wilcoxon rank sum analysis as appropriate (BMDP3D). Contingency tables were analyzed by Miettinen's modification of the Fisher exact test. Probability values were two-tailed with p <0.05 regarded as statistically significant. Correlations between variables were analyzed by simple linear regression using BMDP6D Statistical Software.
Results
Quantitative DNA Analysis There was good correlation between the DNA indices obtained with image analysis and flow cytometry by linear regression analysis (r=0.89, p <0.001, Fig. 2A). For the detection of aneuploidy, there was complete agreement in the results obtained by both methods in 81 carcinomas (81 %) (Fig. 3), with discordance observed in 6 carcinomas (6 %) while 10 tumors were equivocal by image analysis and 3 by flow cytometry (Table 1). The equivocal tumors by flow cytometry displayed borderline increase in the
430
A. K. C. Lee et aI.
3 2.8
Table 1. Quantitative DNA analysis of 100 breast carcinomas
r- 089 pdl01
Flow cytometry
Image analysis
Cordordant cases (n=81)
diploid aneuploid
diploid aneuploid
36 45
Discordant cases (n=6)
aneuploid aneuploid"
diploid aneuploid'c
1 5
Equivocal cases (n=13)
diploid aneuploid ?
? ? tetraploid
8 2 3
)(
~ ~ c
.
~
2
18 16
~
f
total ,. discordance in aneuploidy patterns ? equivocal cases
Flow Cytometry: DNA Index
A
100
50,--------------------------------------,
40
~
o
0
o
30
Q)
(;
"
C/l
"'l'
20
;;::
10 o
o~O~O~O~0r_~§~O~O~O~._--,---_,--_.----._--~~ o
4
8
12
20
16
28
24
32
S Phase Fraction (%)
B
Fig. 2. Linear regression analysis comparing A) DNA indices measured by image analysis and flow cytometry, and B) proliferation measured by image analysis of Ki-67 staining and S-phase fraction measured by flow cytometry.
G2M fraction but a definite tetraploid peak was seen on image analysis. In general, even when FCM detected tetraploid tumors, the tetraploid peaks were smaller and more subtle than observed by image analysis. The equivocal tumors observed by image analysis presented difficult histograms in the near-diploid range usually because the small number of analyzed cells did not provide adequate
J
!!2 Qi ()
'0 Q;
E
::J
o
8
16
24
DNA (Picograms) Image Analysis
o
5
10
Proliferation Rate
The range of S phase fraction measured by flow cytometry was 1-40 %, (median 8.5 %, mean 11.1 +/- 8.2 %), while the Ki-67 score of stained nuclei was 1-55 %, Table 2. Relationship of aneuploidy and proliferation to other variables Aneuploidy
.D
Z
resolution. In such equivocal cases by image analysis, the coefficients of variaton (C.V.) of total optical density of the main peak and the DNA index of internal lymphocyte control were helpful in differentiating diploid from neardiploid aneuploid rumors. Diploid tumors had smaller C.V., with the cut-off point being 8 % in this series. Flow cytometry provided histograms with better resolution and was more sensitive in detecting multiploid rumors and near-diploid aneuploid rumors. In contrast, image analysis permitted morphologic correlation and selection of tumor cells, and was more effective in detecting tetraploid carcinomas and the presence of highly aneuploid cell populations to the right tail of the histogram. By flow cytometry it was frequently difficult to determine whether the signals present in the right tail of the histogram represented highly aneuploid cells or rumor cell aggregates. The presence of aneuploidy was significantly related to various adverse prognostic parameters (Table 2).
15
20
25
DNA (Fluorescence) Flow Cytometry
Fig. 3. Comparison of DNA histograms obtained by image analysis and flow cytonletry. In this case, both methods demonstrated the presence of a hyperdiploid peak.
Estrogen receptor(-) 0.002 Progesterone receptor(-) 0.12 High mitotic rate <0.001 High histologic grade 0.002 High nuclear grade <0.001 Tumor size 0.88 Lymphatic & blood vessel 0.26 mvaSIOn Menopausal status 0.79 Lymph node status 0.43
p Value S Phase Ki-67 count fraction 0.007 0.67 <0.001 0.002 <0.001 0.76 0.43
0.4 0.87 <0.001 <0.001 <0.001 0.04 0.84
0.11 0.53
0.75 0.46
DNA Content & Proliferation . 431 55 50 45 40 35
C
:; 0 ()
~
s::
30
....
25 20
Mo
15 10 MO
Low
High Mitotic Rate
Range 1-13% Mean
A
5%
Range 1-47% Mean 17%
50 45 40 C
0
."5
u:
til
Cl>
35 30 25
en til .c 0...
15
CfJ
10
20 M• • • •
M• • • • • • • • • •
....•.....
......... Low
B
Mitotic Rate
Range 1-31% Mean 7%
High
Range 3-40% Mean 14%
Fig. 4. Relationship of mitotic rate to A) Ki-67 antibody count (p<0.001) and B) S-phase fraction (p
(median 10.0 %, mean 14.5 +/-13.1 %). Linear regression analysis showed only moderate correlation between the methods (r=0.42, p=0.002, Fig.2B). Both S phase fraction and Ki-67 staining were significantly related to high mitotic rate (Fig. 4), and high histologic and nuclear grades. Aneuploid tumors demonstrated significantly higher S phase fractions (p
aneuploid cells beyond the tetraploid region. One drawback appears to be its occasional difficulty in resolving equivocal histograms in the diploid/near-diploid range, as a result of the relatively smaller number of cells analyzed. In contrast, FCM measures a larger number of cells and offers histograms with better resolution. Consequently, it is superior to IA in detecting multiploid tumors and in distinguishing diploid from near-diploid aneuploid qtmors. Hence, the potential sources of errm with FCM and IA lie in the interpretation of abnormalities in the tetraploid and diploid regions, respectively2, 7, 12. Some possible reasons for FCM to underestimate tetraploid and highly aneuploid cell populations include the dilution effect by non-neoplastic cells, relative fragility of aneuploid cells, and sampling problems. Also, FCM may not be able to distinguish between highly aneuploid tumor cells and tumor cell aggregates. Proliferation in solid tumors may be measured by diverse methods, namely mitotic count, thymidine labeling index, bromodeoxyuridine uptake, argyrophilic nucleolar organizer region count, SPF measurement and quantitation of proliferation-associated antigens such as Ki-67 and Cyclin. SPF estimation may include non-neoplastic cells, and may sometimes be difficult, especially in tumors with complex histograms displaying overlapping peaks and containing substantial debris. Also, reproducibility may be poor because of interobserver variations and the many sophisticated mathematical models being used to interpret S phase fractions. Ki-67 and other proliferation associated antigens require frozen sections although there are now monoclonal antibodies applicable to paraffin sections. SPF measures cells in the S phase of the cell cycle, whereas proliferation-associated antigens measure different phases of the cell cycle, for instance Ki-67 identifies all cells in the cell cycle9 and Cyclin correlates more with the SPF. Our SPF values are within the range reported by others 16 , whereas our Ki-67 scores are generally lower than other series3 , 7,10,13,14 but similar to that ofIsoli et al. 10 • Our data indicate that the Ki-67 count and SPF bear moderate correlation by linear regression analysis. In concordance with other studies 3, 7, 8, 10,13,14 they both correlate well with the mitotic rate and other measures of proliferation. IA and FCM appear to be complementary techniques in evaluating DNA content and proliferation in breast cancers. These two parameters are related to various adverse prognostic parameters. It remains to be seen if they represent independent prognostic parameters that may be useful in prognostication and in guiding management of breast cancer patients.
References 1 Aner GU, Askensten U, Erhardt K, Fallenius A, Zetterberg A (198'7<) Comparison between slide and flow cytophotometric DNA measurements in breast tumors. Anal Quant Cytol Histol9: 138 2 Bauer TW, Tubbs RR, Edinger MG, Suit PF, Gephardt GN, Levin HS (1990) A prospective comparison of DNA quantitation by image and flow cytometry. Am J Clin Pathol 93: 322
432 . A. K. C. Lee et a1. 3 Charpin C, Andrac L, Habib M, Vacheret H, Xerri L, Devictor B, Lavaut MN, Toga M (1989) Immunodetection in fine-needle aspirates and multiparametric (SAMBA) image analysis. Receptors and growth fraction evaluation in breast carcinomas. Cancer 63: 863 4 Clark GM, Dressler LG, Owens MA, Pounds G, Oldaker T, McGuire WL (1989) Prediction of relapse or survival in patients with node-negative breast cancer by DNA flow cytometry. N Engl J Med 320: 627 5 Claud RD 3d, Weinstein RS, Howeedy A, Straus AK, Coon JS (1989) Comparison of image analysis of imprints with flow cytometry for DNA analysis of solid tumors. Modern Pathol 2: 463 6 Cornellise q, Van Driel-Kulker AM (1985) DNA image cytometry on machine-selected breast cancer cells and a comparison between flow cytometry and scanning cytophotometry. Cytometry 6: 471 7 Dawson AE, Norton JA, Weinbert CS (1990) Comparative assessment of proliferation and DNA content in breast carcinoma by image analysis and flow cytometry. Am J Pathol136: 1115 8 Dervan PA, Gilmartin LG, Loftus BM, Carney DN (1989) Breast carcinoma kinetics. Argyrophilic nucleolar organizer region counts correlate with Ki-67 scores. Am J CI Pathol 92: 401 9 GerdesJ, Lemke H, Baisch H, Wacker H, Schwab M, Stein H (1984) Cell cycle analysis of a cell-proliferation associated nuclear antigen defined by the monoclonal antibody Ki-67. J Immuno1. 133: 1710
10 Isola 11, Helin HJ, Jelle MJ, Kallioniemi 0 (1990) Evaluation of cell proliferation in breast carcinoma. Comparison of Ki-67 immunohistochemical study, DNA flow cytometric analysis, and mitotic count. Cancer 65: 1180 11 Koss LG, Wersto RP, Simmons DA, Deitch D, Herz F, Free SZ (1989) Predictive value of DNA measurements in bladder washings. Comparison of flow cytometry, image cytophotometry and cytology in patients with a past history of urothelial tumors. Cancer 64: 916 12 Lee AK, Dugan J, Hamilton WM, Cook L, Heatley G, Kamat B, Silverman ML. Quantitative DNA analysis in breast carcinomas. A comparison between image analysis and flow cytometry. Modern Pathology 4: 178 13 McGurrinJF, Doria MI, Dawson TJ, Karrison T, Stein HO, Franklin WA (1987) Assessment of tumor cell kinetics by immunohistochemistry in carcinoma of breast. Cancer 59: 1744 14 Walker R, Camplejohn RS (1988) Comparison of monoclonal antibody Ki-67 reactivity with grade and DNA flow cytometry of breast carcinomas. Br J Cancer 57: 281 15 Uyterlinde AM, Smeulders AW, Baak JP (1989) Reproducibility and comparison of quantitative DNA histogram features obtained with a scanning micro densitometer and a flow cytometer in breast cancers. Anal Quant Cytol Histol 11: 353 16 Visscher DW, Zarbo RJ, Greenawald KA, Crissman JD (1990) Prognostic significance of morphological parameters and flow cytometric DNA analysis in carcinoma of the breast. In: Pathology Annual, Rosen PP, Fechner R (Eds) vol. 25, part 1, p 171. Norwal, CT: Appleton & Lange
Received September 29; 1991 . Accepted October 28, 1991
Key words: Breast cancer - DNA - Proliferation - Flow cytometry - Image analysis Dr. Arthur K. C. Lee, Department of Anatomic Pathology, Lahey Clinic Medical Center, 41 Mall Road, Burlington, Massachusetts, 01805 U.S.A.