Immunohistochemical Detection of Ki-67 and PCNA ... - SAGE Journals

J Vet Diagn Invest 10:237–246 (1998)

Immunohistochemical detection of Ki-67 and PCNA in canine mammary tumors: relationship to clinical and pathologic variables Laura L. Penã, Ana I. Nieto, Dolores Pe´rez-Alenza, Pedro Cuesta, Maria Castanõ Abstract. The objectives of this study were to measure the proliferation indices in canine mammary tumors using immunohistochemical detection of Ki-67 and proliferating cell nuclear antigen (PCNA), to determine the relationship of these antigens to clinical and pathologic variables, and to investigate the usefulness of these antigens as prognostic indicators. Ninety-six female dogs with 115 primary nonmetastasized spontaneous mammary tumors and dysplasias were included in the study. Immunostaining was performed using MIB-1 and PC10 monoclonal antibodies against Ki-67 and PCNA, respectively. Ki-67 and PCNA proliferation indices were determined. Dogs were followed for 18 months, with clinical examinations every 3–4 months. There was a significant correlation between Ki-67 and PCNA indices in the dogs with dysplasias and benign tumors but not in the dogs with malignant tumors. The clinical stage at first presentation was related to the proliferative index measured with Ki-67 but not to that measured with PCNA. Proliferation indices were significantly lower in the nonmalignant tumors and dysplasias than in the malignant tumors. In malignant tumors, the PCNA index had a positive correlation with the histologic malignant grade and the nuclear grade. High index values of Ki-67 were positively correlated with metastasis, death from neoplasia, low disease-free survival rates, and low overall survival rates. PCNA displayed no significant association with these variables. Multivariate analyses concerning metastasis, disease-free survival, and overall survival revealed that the Ki-67 index had prognostic value.

The proliferation index (PI), determined by cell cycle-related markers such as Ki-67 and proliferating cell nuclear antigen (PCNA), has prognostic value in human mammary carcinomas.2,7,25,36,42 Some researchers have noted the value of these nuclear antigens in the prediction of disease-free survival and overall survival as independent prognostic factors in multivariate analyses.38,41,43 Other authors have reported inconsistent results.16,39 Immunohistochemical methods for detecting nuclear antigens such as Ki-67 and PCNA are inexpensive, quick, and technically simple and are a good alternative to flow cytometric methods employed previously. The correlation between the neoplastic growth fraction, determined by cytometric S-phase analysis, and PCNA and Ki-67 immunohistochemistry results appears close in human tumor studies.15,40 PCNA is an auxiliary protein of DNA polymerased expressed in dividing cells during G1. PCNA increases in late G1, peaks in S-phase, and remains high during G2 and M-phase because of its long half-life.8 Ki-67 is a nuclear nonhistone protein with a molecular mass of 345 and 395 kD that is expressed in cycling cells during late G1, G2, S-phase, and M-phase. Expression of Ki-67 disappears rapidly after mitosis.18 From the Department of Animal Pathology II, Veterinary School, and the Processing Data Center, Complutense University, 28040 Madrid, Spain. Received for publication March 27, 1997.

PCNA immunostaining with PC10 antibody was reported in canine and feline formalin-fixed, paraffinembedded mammary tumor specimens and appears to be correlated with the histologic type (benign/malignant) and the mitotic index but not with nuclear grade.30 Neoplastic proliferation has been related to prognosis in follow-up studies detecting the S-phase fraction in canine mammary tumors (CMT) by flow cytometry.22 A novel antibody (MIB-1) developed against the Ki67 antigen and effective in formalin-fixed, paraffinembedded material17 has simplified the retrospective investigation of the PI of archived specimens. In this study, PI in CMT using immunohistochemical labeling of Ki-67 and PCNA were measured to investigate the relationship of these antigens to clinical and pathologic characteristics and to check their validity, via multivariate analyses, as independent and reproducible prognostic parameters. Materials and methods Animals and clinical examinations. The 96 female dogs included in this study were presented between January 1992 and July 1993 with primary spontaneous mammary tumors and no pulmonary metastasis. Clinical variables and data on the reproductive history and hormonal status were obtained at first presentation and included breed, body size, ovariohysterectomy, prevention of estrus (with estrogens/progestagens combinations), number of estrous cycles, regularity and duration of estrus, age at first estrus and at first partu-

237

238

Penã et al.

rition, number of gestations, and history of pseudopregnancy or pseudolactation. The animals with at least 1 malignant tumor were categorized by clinical stage using the TNM system (T 5 size of primary tumor; N 5 regional lymph node involvement; M 5 distant metastasis)28 and surgically treated. All grossly affected lymph nodes (N1) were histopathologically or cytologically examined after surgery to asses lymph node metastases. Clinical stage was established as local (without lymph node involvement 5 N0) or regional (with histologically confirmed lymph node involvement 5 N1). The rate of growth was defined as slow, medium, or fast based upon owner information. Tumor size was described in 3 dimensions, of which the largest was used for the statistical analysis. Taking into account this last measurement, and in accordance with the World Health Organization (WHO) system, the tumors were classified as T1 (,3 cm), T2 (3–5 cm), or T3 (.5 cm). Likewise, adherence to underlying tissues and skin and ulceration was recorded. All dogs included in the study were clinically and radiologically free from distant metastases (M0). Histopathology. For histopathologic study, tumor specimens were fixed in buffered formalin, embedded in paraffin, cut in 4-mm sections, and stained with hematoxylin and eosin. The histopathologic exam was done following the WHO classification system of canine mammary tumors.21 Using a previously established3,31 histologic grading system of human breast carcinomas and considering the criteria regarding malignancy in canine mammary tumors,24 3 degrees (I, II, III) of histologic malignant grade (HMG) tumors were recognized. HMG I included well-differentiated tumors with no vascular or lymphatic invasion, HMG II included moderately differentiated neoplasms with no vascular or lymphatic invasion, and HMG III included poorly differentiated tumors with or without vascular or lymphatic invasion. The nuclear grade of malignancy (NG) was also assessed separately in each specimen and categorized as I, II, and III.3,24,31 Immunohistochemistry. Ki-67 and PCNA immunostaining was performed on deparaffined sections (poly- L-lysine slides)a after microwave processing.34 The sections were placed in a microwave-resistant container and immersed in 10 mM sodium citrate buffer (pH 6.0) and heated 3 times (5 min each) at maximum power (790 W),b replacing the evaporated buffer with distilled water after each heating session. The sections were left to cool at room temperature (RT) for 20 min. Slides were then rinsed 3 times (5 min each) in Tris-buffered saline (TBS, pH 7.4). A peroxidase–antiperoxidase (PAP) technique was employed using the following antibodies: monoclonal mouse anti-human Ki-67c (clone MIB-1, working dilution 1:25, incubated 1 hr at RT), monoclonal mouse anti-human PCNAd (PC10 clone, working dilution 1:200, incubated overnight at 4 C), rabbit anti-mouse IgGe (working dilution 1:100, incubated 30 min at RT), and mouse PAPa (working dilution 1:150, incubated 30 min at RT). Corresponding control slides were prepared by replacing the primary antibody with TBS. Chromogen developing was performed with diaminobenzidinea and 30% hydrogen peroxide and counterstained with hematoxylin. All the washings and dilutions were performed with TBS. Evaluation of immunostaining. Positive cells were counted by 2 observers simultaneously with a double-headed light

microscope. In each specimen, positive and negative nuclei (minimum 1,000 neoplastic cells) in 8–10 representative high-power fields were counted. Every immunostained nucleus was considered positive. The PI, or proportion of positive neoplastic cells in each sample, was calculated. Follow-up study. Dogs were followed after surgical treatment for 18 mo. Clinical examinations were performed every 3–4 mo and included a radiologic evaluation of the thorax. Tumor recurrence (local) and distant metastases were noted, as well as time and cause of death. Disease-free survival and overall survival times were recorded for each dog. These variables were categorized as low (0–6 mo), medium (7–17 mo), and high (18 mo). Statistical analysis. Statistical tests were performed with Biomedical Data Processing Software.10 Pearson and Yates chi-square tests were used for studying categoric variables, and Pearson correlation coefficients of the continuous variables were calculated by pairs. Analyses of variance were used to study the differences in means of continuous variables in the groups determined by categoric variables. These programs use Levene F-tests for analyzing the homogeneity of variances. If variances are assumed to be equal, then Ftests or pooled t-tests are chosen to study the variables. If variances are not equal, then a Welch test or separate t-test is selected. Tukey and Student–Newman–Keuls (SNK) tests were complementary tests to analyze the specific differences among the HMG and NG categories and the PIs. In each statistical analysis, the cases in which there are missing data for the variables studied are excluded. A segmentation analysis (decision tree) of the Ki-67 PI to predict metastasis was done.11 Multivariate logistic regression analyses of the prognostic factors also were performed.10 These analyses evaluate the influence of sets of variables on a dependent variable (recurrences, metastasis, disease-free survival, overall survival), calculating its odd ratio (OR) according to the formula OR 5 esp (a0 1 a1x1 1 . . . 1 an), where a0 is the constant of the dependent variable and a1, a2, . . ., an are the coefficients affecting the sets of related independent variables included in the model (x1, x2, . . ., xn). The independent variables are selected according to their statistical significance in several steps. Those variables that are significant at 1 step are then considered constants to study the combined effect of the other variables on the dependent variable. The dogs were classified in 2 groups: those with only dysplasias and/or benign tumors (group DA) and those with at least 1 malignant tumor (group DB). The tumors were also divided in 2 groups; dysplasias and benign tumors (group TA) and malignant tumors (group TB). The statistical tests examined dog-related variables and tumor-related variables separately and in combination. To analyze the tumor and dog data together in dogs with .1 mammary lesion, values of Ki-67 and PCNA were assigned to each dog. In dogs bearing only benign lesions, the higher Ki-67 and PCNA indices were chosen. In those cases in which the dog had .1 malignant neoplasm, variables of the malignant tumor (including Ki-67 and PCNA indices) with the worst clinical and pathologic prognosis were chosen for the statistical study. A probability of error (P) of #0.05 was considered significant.

Ki-67 and PCNA immunohistochemistry in canine mammary tumors

239

Figure 1. Solid canine mammary carcinoma with a high Ki-67 index (26%). Many nuclei (some of them indicated with white arrows) and mitosis (black arrow) are positive. Ki-67 immunostaining.

Results The 96 female dogs ranged from 5 to 13 years of age (x¯ 6 SD 5 9.90 6 2.22 years). Most of them were intact bitches (n 5 86, 89.58%) with no (n 5 53, 55.20%) or only 1 (n 5 23, 23.95%) parturition. Only 17 bitches had been prevented from estrus by the administration of estrogens/progestagens combinations. The anamnesis revealed 36 previous pseudopregnancies and 23 pseudolactations. Fifty-one dogs had dysplasias and/or benign mammary gland tumors but not malignant neoplasms (39 with benign tumors, 10 with benign tumors and dysplasias, 2 with only dysplasias). Forty dogs had 1 malignant tumor and 5 had multiple malignant tumors. Clinical stage for malignant neoplasia was local in 36 cases and regional (with histologically confirmed lymph node involvement) in 9 cases (1 with HMG I, 2 with HMG II, 6 with HMG III). The total number of tumors collected was 128 (17 dysplasias, 60 benign tumors, 51 malignant tumors). Dogs with complete clinical history and clinical data on the tumors, which had been previously histologically diagnosed, were included in the follow-up study, although some of the corresponding tumors could not be analyzed further because of a fire in the pathology laboratory that destroyed some of the slides and tissue blocks. Excluding these destroyed tissues, 115 mammary gland tumors (15 dysplasias, 60 benign tumors, 40 malignant neoplasms) were used in the study. Fifteen of the 40 malignant tumors recurred, and 12 metastasized, 6 of which also recurred. From the 45 animals with at least 1 malignant tumor (group DB), 21 were alive and 20 were dead or euthanized due to neo-

plasia at the end of the follow-up period. Three animals died from other causes and 1 was lost to followup. Twenty-seven dogs had low to medium diseasefree survival (DFS), 17 had a high DFS, and 1 was lost to follow-up. The overall survival (OS) was high in 21 dogs, low to medium in 21, and unknown in 3. No recurrences or metastases were presented in dogs with benign tumors, and OS was high for all of these dogs. Ki-67 and PCNA immunostaining was nuclear and varied in intensity. Mitotic cells exhibited chromosomal Ki-67 reactivity, and PCNA was expressed in the cytoplasm of mitotic cells (Fig. 1). The Ki-67 index ranged between 1% and 31% (n 5 93, x¯ 5 7.33%), and the PCNA index ranged between 1% and 90% (n 5 89, x¯ 5 27.27%). There was a significant correlation between the Ki-67 and PCNA indices in the group of dysplasias and benign tumors (group TA) (n 5 57, r 5 0.62, P , 0.001) but not in malignant tumors (group TB). Univariate analyses

Clinical data. To determine the possible influence of different factors such as age or hormonal history on tumor proliferation, their relationship with Ki-67 and PCNA scores were investigated. No significant relationship was found between the Ki-67 and PCNA scores and the breed, size of the dog, or most of the reproductive variables (age at first parturition, number of parturitions, ovariohysterectomy, prevention of estrus, number of estrous cycles, estrous cycle duration and regularity, pseudolactation, pseudopregnancy).

240

Penã et al.

Table 1. Ki-67 in canine mammary tumors: significant clinical variables and pathologic variables.

Table 2. PCNA in canine mammary tumors: significant pathologic variables.

Ki-67 index n

x¯

SEM

Clinical stage* Local (N0) Regional (N1)

36 9

10.17 16.80

1.41 2.74

Histology Benign/dysplasias† Malignant*

57 36

4.25 12.17

0.61 1.50

Histologic malignant grade HMG I HMG II HMG III

6 10 13

10.50 8.40 16.07

4.68 1.82 2.23

Nuclear grade NG I NG II NG III

2 13 14

15.50 8.38 15.42

14.50 1.71 2.16

PCNA index P

n

x¯

Histology Benign/dysplasias* Malignant†

57 32

18.66 42.59

2.44 4.87

0.06 (NS)

Histologic malignant grade HMG I HMG II HMG III

5 10 12

21.20 31.00 55.17

8.88 6.09 8.28

0.08 (NS)

Nuclear grade NG I NG II NG III

2 12 13

20.50 28.33 53.61

19.50 5.71 7.77

0.02

,0.0001

* Malignant tumors, group TB. † Group TA.

The PCNA score was correlated with the age of the dog (P 5 0.023, r 5 0.345), and the Ki-67 score appeared correlated with the age at first estrus (P 5 0.008, r 5 0.59) only in malignant tumors. The relationship between the clinical parameters of the tumor and the proliferation rate was investigated. Ki-67 and PCNA were not significantly associated with tumor size, rate of growth, ulceration, or adhesion to underlying tissues. The clinical stage at first presentation in dogs with at least 1 malignant tumor was related to the PI measured with Ki-67 but not with PCNA. Low levels of Ki-67 displayed a significant association with the local stage (N0), and high levels of Ki-67 were associated with the regional stage (N1, with histologically confirmed lymph node involvement) (P 5 0.02) (Table 1). Histopathologic findings. Data corresponding with the PIs measured with PCNA and Ki-67 in the tumors are displayed in Tables 1 and 2. PCNA indices were higher than Ki-67 indices in both benign and malignant neoplasms (Tables 1, 2; Figs. 2 vs. 3, 4 vs. 5). With respect to benign (group TA) versus malignant (group TB) neoplasms, both PIs were significantly lower for the nonmalignant tumors and dysplasias than for the malignant tumors (P , 0.0001). In malignant tumors, the PCNA index displayed a positive relationship with the histologic malignant grade (P 5 0.042) and the nuclear grade (P 5 0.032) (Table 2). With the Tukey studentized range method, the PCNA index discriminated between HMG I and HMG III (P , 0.05). SNK tests separated HMG I and HMG II from HMG III (P 5 0.01 and P 5 0.05, re-

SEM

P

,0.0001

0.042

0.032

* Group TA. † Group TB.

spectively). NG II and NG III were significantly separated by Tukey’s test (P , 0.05) but not by the SNK test. Ki-67 and PCNA levels did not significantly distinguish between dysplasias (Ki-67 index x¯ 5 5.41, PCNA index x¯ 5 21.09) and benign tumors (Ki-67 index x¯ 5 3.98, PCNA index x¯ 5 18.70). Both markers were independent of the histopathologic malignant type, although the Ki-67 index in solid carcinomas and sarcomas (x¯ 5 12.94) was higher than the Ki-67 index in tumors with tubular formations (x¯ 5 11.31). Follow-up study. There were no tumor- or dog-related variables statistically associated with recurrence of malignant neoplasms. A significant association (P 5 0.022) existed between ulceration and metastasis; 53.8% of the ulcerated tumors metastasized as compared with only 18.5% of the nonulcerated neoplasms. High values of Ki-67 were associated with metastasis in the follow-up study (P 5 0.002) but not with recurrences (Table 3). PCNA failed to predict metastases or recurrences. With the Ki-67 index segmentation analysis, it was possible to establish Ki-67 cutoff points with respect to the probability of metastasis. If the Ki-67 index was ,11.5, the probability of metastasis was 5%; if the Ki-67 index was between 11.5 and 24, the probability increased to 25%, and when the Ki67 level was .24, the probability of metastasis was 99%. Considering only dogs with at least 1 malignant tumor, and only those who died from neoplasia or were alive, high levels of Ki-67 were related to death from neoplasia (P 5 0.047). Animals with a low Ki-67 PI had a significantly higher DFS (P 5 0.002) and OS (P , 0.001). PCNA displayed no significant association with these variables (Table 3). Old age (x¯ 6 SD 5 10.95 6 1.51 years, P 5 0.04),


Figure 2.

241

Epithelial area in a canine benign mixed tumor with a PCNA index of 40%. PCNA immunostaining.

short estrous duration (x¯ 6 SD 5 13.22 6 5.92 days, P 5 0.007), low number of estrous cycles (P 5 0.015), death by neoplasia (P 5 0.007), number of malignant tumors/animal (P 5 0.017, 100% of the animals with .1 malignant tumor had a low DFS), large tumor size (P 5 0.048), and ulceration (P 5 0.024) were also related to a shorter DFS. Advanced age of the animal (P 5 0.003), short estrous duration (P , 0.011), low number of estrous cycles per year (P 5 0.050), short DFS (P , 0.001), large tumor size (P 5 0.043), HMG

(P 5 0.040), and NG (P 5 0.025) were also related to a low OS. Multivariate analyses

To study the combined influence of groups of variables on a given variable (recurrences, metastasis, DFS, OS), several multivariate analyses were performed in the group of animals with at least 1 malignant tumor (group DB, n 5 45). For these analyses, complete data for all the variables was necessary;

Figure 3. Same tumor and area of Fig. 2. Ki-67 index is only 5%. Few positive nuclei are present (some marked with arrows). Ki-67 immunostaining.

242

Figure 4.

Penã et al.

Canine tubular adenocarcinoma with a medium PCNA index (40%). PCNA immunostaining.

therefore, some dogs for which data were incomplete were not included. Thus, several multivariate analyses using different sets of variables were performed. Those with the most number of cases included are described below. No single variable was selected with regard to recurrences. Taking metastasis as a dependent variable and selecting a set of variables (age, pseudolactation, tumor size, ulceration, lymph node involvement, histologic type of malignant tumor [malignant 1 or 2], Ki-67 index, PCNA index), multivariate analysis con-

Figure 5.

cerning metastasis (n 5 32 cases, 23 negative and 9 positive for metastasis) revealed Ki-67 in step 1 as an independent prognostic factor (P 5 0.003). When Ki67 was the selected variable (step 2), no other variable appeared significantly related to metastasis (Table 4). Taking DFS as a dependent variable (n 5 31; low to medium DFS , 18 months, n 5 20; high DFS . 18 months, n 5 11) and using age and Ki-67 as related variables, Ki-67 (P 5 0.001) emerged as an independent and significant variable in step 1. When Ki-67 was considered constant (step 2), age was the next

Same tumor as that in Fig. 4. Ki-67 index is high (15%). Ki-67 immunostaining.


Table 3.

Ki-67 index (P 5 0.03) and age (P 5 0.024) affected Os independently and were selected in steps 1 and 2, respectively (Table 4, OS multivariate analysis 1). When other data were considered (Ki-67 index, ulceration, lymph node involvement), 28 cases remained, and ulceration, clinical stage, and the Ki-67 index were independent prognostic factors selected in steps 1, 2, and 3, respectively. Ulceration represented an almost perfect mathematical model to predict OS (all the cases with ulceration had a low to medium OS; SE is not shown) (Table 4, OS multivariate analysis 2).

Ki-67: significant follow-up variables.* Ki-67 index n

x¯

SEM

Range

Death None Caused by tumor

11 17

8.54 15.00

1.88 2.45

1–20 1–31

Disease-free survival (DFS) Low–medium High

20 11

15.45 5.36

2.16 1.26

2–31 1–16

Overall survival (OS) Low–medium High

20 13

17.62 6.82

2.74 1.37

4–31 1–20

Metastasis No Yes

25 9

9.56 19.44

1.19 3.79

1–23 1–31

P

0.047

0.002

0.001

0.002

* Only dogs with malignant tumors are included.

variable affecting DFS but not at significant level (P 5 0.1) (Table 4, DFS multivariate analysis 1). In a second multivariate analysis of DFS, including other variables (age, lymph node involvement, tumor size, adherence to underlying tissues, rate of growth, ulceration, Ki-67 index, PCNA index), the number of cases dropped to 27 (low–medium DFS, n 5 20; high DFS, n 5 7) and age was selected in step 1. No other variable was selected in step 2, although tumor size and the Ki-67 index were close to being selected by the analysis (P 5 0.052, P 5 0.058, respectively) (Table 4, DFS multivariate analysis 2). With OS as a dependent variable and using age and Ki-67 as related variables (n 5 32; low–medium OS , 18 months, n 5 13; high OS . 18 months, n 5 19), Table 4. tumors.

243

Discussion Ki-67 and PCNA immunostaining was successful using MIB-1 and PC10 monoclonal antibodies after microwave processing. In some cases, after repeating the immunochemical process several times, the quality of the immunostaining was not considered good; only tumors with quality immunohistochemistry staining were included in the study (Ki-67, n 5 93; PCNA, n 5 89) to avoid underestimated PIs. In nonmitotic cells, Ki-67 and PCNA expression was restricted to the nuclei, but in mitotic cells (once the karyotheca has disappeared and only chromosomes were visible), Ki-67 immunostaining was chromosomal, and PCNA expression was cytoplasmic. This pattern of Ki-67 and PCNA expression is commonly reported in human tissues, with some exceptions.2 In canine mammary tumors, similar PCNA immunostaining has been reported.29 There are no previous reports concerning Ki-67 and canine mammary tumors. The cytoplasmic expression of PCNA in mitotic cells could be related to its association with cytoskeletal elements.40 For unknown reasons, PC10 did not stain the cytoplasm of a few mitotic cells.

Multivariate analysis of metastasis, disease-free survival (DFS), and overall survival (OS) in dogs with malignant mammary

Studied variable

Metastasis Constant a0 5 22.876 Disease-free survival Analysis 1, constant a0 5 25.291 Analysis 2, constant a0 5 25.416 Overall survival Analysis 1, constant a0 5 29.725 Analysis 2, constant a0 5 10.21

Step no.

Selected variable

Coefficient an

SE

Exponential Coefficient/SE coefficient

Confidence interval

P

1

Ki-67

0.139

0.054

2.57

1.15

1.03–1.28

0.003

1 2 1

Ki-67 age age

0.178 0.398 0.626

0.081 0.264 0.263

2.20 1.51 2.39

1.20 1.49 1.87

1.01–1.41 0.86–2.56 1.09–3.22

0.001 0.103 0.005

1 2 1 2 3

Ki-67 age ulceration* clinical stage† Ki-67

0.155 0.669 115.8 212.74 0.153

0.062 0.354 ··· 1.21 0.084

2.51 1.89 ··· 210.6 1.83

1.17 1.95 ··· ··· 1.17

1.03–1.33 0.94–4.03 ··· ··· 0.98–1.39

0.003 0.024 0.004 0.008 0.026

* Because the category ‘‘ulceration yes’’ predicts perfectly the variable low–medium overall survival, no other coefficients were calculated. † Because regional clinical stage was encountered in most of the cases with low to medium overall survival, exponential coefficient and confidence interval were not calculated.

244

Penã et al.

In this study, as in many others in human breast tumors, PCNA expression is more intense than that of Ki-67 probably because Ki-67 is restricted to the cell cycle whereas PCNA is also expressed in noncycling cells in association with DNA repair.35 PCNA also has a long half-life and persists after the end of mitosis. Mitotic cells have very low PCNA expression,23 which suggests that although PCNA is comparable to Ki-67, it is influenced by unknown factors and loses its usefulness for quantifying proliferative activity.37 Ki-67 and PCNA were significatively correlated in benign tumors (group TA) but not in malignant neoplasms (group TB). The lack of correlation between Ki-67 and PCNA in malignant tumors has also been observed in human breast carcinomas25 and is attributable to the high values of PCNA and its presence in noncycling cells. Ki-67 labeling in benign and malignant CMT was similar to that reported in human breast tumors.14,25,42,43 The PCNA values in our study were higher than most of the values reported for human breast carcinomas and higher than other published values for CMT.29 The microwave technique used in the present study could allow the immunostaining of PCNA epitopes masked by formalin fixation, which would explain the higher PCNA values. Nevertheless, PCNA index in malignant tumors was quite similar to the PCNA labeling obtained by others25 with the same PC10 antibody. Previously reported correlations between Ki-67 and PCNA indices and clinical and pathologic variables have been contradictory. In the present study and in accordance with some investigations in women, Ki-672 and PCNA38,39 indices were not unequivocally correlated with age, whereas Ki-67 labeling exhibited a significant association with lymph node status.42,43 In the present series of tumors, similar to some human studies, clinical variables such as tumor size, rate of growth, or adherence to underlying tissues were not associated with Ki-672,7,25,42 or PCNA.36,38,39 Nevertheless, an association between tumor size and Ki-6743 and PCNA20 staining has been reported. The lack of a relationship between PIs and tumor size could be attributed to neoplastic changes; tumor size is the result of previous growth and these PIs measure the growth rate at the time of surgery. In addition, the rate of growth reported in this study was given by the owner and was therefore subjective. In spite of the controversy concerning the association between Ki-67 and PCNA indices and clinical variables, the data supported that tumor grade is associated with both the Ki-67 index and the PCNA index. The PCNA index, but not Ki-67 index, displayed a significant association with HMG and NG. Similar correlation between PCNA and NG has not been found in CMT.29 These results concerning PCNA in CMT are

not fully comparable because the antigen microwave retrieval technique we used produced PCNA levels that were higher than those obtained by others without this pretreatment.29 According to previous studies in human breast lesions,38 PCNA and Ki-67 indices were significantly different in benign and malignant tumors (P , 0.001). In spite of the fact that the labeling of these nuclear markers increased with malignancy from benign neoplasms to papillary/tubular adenocarcinomas to solid carcinomas, there was no significant association with histologic type. Benign tumors were not discriminated from dysplasias nor were malignant neoplasms with tubular/papillar formations differentiated from those without such structures. Similar results have been reported in CMT stained with PC10 anti-PCNA29; the PCNA index in benign tumors was significatively higher than that in malignant tumors (P , 0.001). These authors also indicated a significant correlation between PCNA index and mitotic index, but the comparison between PCNA values and other pathologic, clinical, or follow-up variables were not studied. Prediction of the postsurgical behavior of CMT is not possible through histopathologic examination alone, and different pathologic and clinical parameters have been proposed to assure accurate prognosis in univariate follow-up studies. These parameters include the age of the animal, rate and type of growth, tumor size, ulceration, lymph node status at time of surgery, histologic malignant type, nuclear grade, presence of lymphoid infiltrates, and mitotic index, although those prognostic parameters that affect survival differ in the various follow-up studies.1,4,12,13,19,24,27,30 In recent years, the presence of estrogen or progesterone receptors has been included as a better prognostic factor in CMT.26 Some reproductive variables such as age at first estrus, abnormal estrous cycles, or pseudopregnancy do not play a significant role in the progression of malignant tumors,12,33 although they are recognized as important factors in the appearance of these neoplasms. An increased argyrophilic nucleolar organizer region (ribosomal DNA related to the cell cycle and proliferative activity) count is associated with an increasing degree of malignancy and tumor-related death in bitches with invasive mammary carcinomas.6,9 One of the most important factors is, undoubtedly, tumor size.1,5 In the present study, it was not possible to predict recurrences with the variables analyzed. The appearance of distant metastases in the follow-up period was significantly associated with high Ki-67 labeling, ulceration, and poorly differentiated NG. The SPAD-S program was used to find significant intervals of Ki67 values that could be used to predict metastases; when Ki-67 was .24%, the probability of developing distant metastases increased to 99%.


Low OS and DFS were both correlated with increased age (10.95 and 11.00 years, respectively), short estrous cycle, large tumor size, and a high Ki67 index in the univariate study. Old age and tumor size are robust prognostic factors in CMT,22 and the present univariate analysis indicates that Ki-67 should also be considered. The PCNA index was not related to any of the follow-up variables investigated. Many studies in human breast cancer indicate the usefulness of Ki-67 in predicting DFS and OS in univariate analyses.7,14,16 The grading system of HMG used here is based mainly in the Scarff-Bloom-Richardson (SBR) grading system commonly used in human breast cancer and is considered a good prognostic factor.3,31 This grading system is based on 3 characteristics of the tumor, the grade of tubular differentiation, grade of nuclear pleomorphism (NG), and mitotic index, and incorporates the criterium of histologic vascular invasion for CMT.19 There has not been a specific grading system for CMT with follow-up validation. Our results indicate that this combination of human SBR and criteria of invasiveness can be prognostic in the dog because HMG and NG were related to low OS in univariate analyses. Multivariate analyses to study the joint effect of the variables on distant metastases, DFS, and OS have revealed that high Ki-67 labeling, increased age, large tumor size, ulceration, and regional clinical stage are reliable prognostic parameters of malignant behavior in CMT. Some authors have found that DFS14,41,43 and OS41,43 multivariate analyses demonstrated Ki-67 as a significant independent predictor in human breast cancer, whereas other authors have not obtained those results.16 In contrast to the demonstrated validity of Ki67 as a prognostic factor in the majority of the studies, the PCNA index does not seem to be a good indicator of OS and DFS in univariate analyses16,36,38,39 or multivariate analyses.16,39 Thus, PCNA may not be a good proliferation marker in human breast carcinomas.32 In contrast, in a few studies PCNA appeared to be an independent factor affecting DFS and OS in univariate or multivariate analyses.20,36,38 The only published multivariate analysis of survival of dogs with malignant CMT22 indicated that increased age, elevated S-phase fraction, and ploidy status were independent predictors of survival. These findings indicate that increased age and the proliferative ratio obtained with flow cytometry are important and independent prediction factors in CMT and coincide with our results concerning the Ki-67 index. Moreover, Ki-67 immunodetection presents several advantages over flow cytometric methods because the Ki-67 detection technique is less expensive, is easier to perform, is done in situ, and can be performed in a normal his-

245

topathology laboratory. Immunodetection of the Ki-67 antigen using the monoclonal antibody MIB-1 is an easy technique because microwave antigen retrieval allows the improved staining of the antigen. Because of the heterogeneous nature of CMT, it is important to count the Ki-67 stained nuclei in poorly differentiated selected areas that usually have a high proliferation rate. Most authors agree that the poorly differentiated areas should be selected and that all the nuclei (weakly and strongly stained) must be counted, so that all the cells in the cell cycle can be considered. The multivariate analyses presented here are complementary to the univariate analyses. The low number of cases that could be included in the study implies that the prognostic factors such as age, tumor size, or Ki-67 index, that emerged as significantly related to metastasis, DFS, and OS are strong independent indicators of prognosis. If there had been more cases and variables to include in the multivariate analyses, other prognostic indicators such as HMG, NG, and histological type of malignant tumor may have been significant. Nevertheless the multivariate studies are of interest because they show almost invariably that the Ki67 index is a good prognostic indicator. In the present study, increased Ki-67 immunostaining was an objective prognostic parameter in predicting the postsurgical behavior of canine mammary carcinomas. The Ki-67 index was an independent and objective prognostic factor in predicting metastases, DFS, and OS. PCNA, however, was not. Nevertheless, further studies with a greater number of malignant tumors are necessary to confirm the relationship between Ki-67 and PCNA indices and the clinical and pathologic variables and to confirm the validity of the Ki67 index as a prognostic factor. Acknowledgements We thank Dr. Javier Codesal for his collaboration in the immunohistochemical detection of PCNA and Pedro Aranda for his technical assistance.

Sources and manufacturers a. b. c. d. e.

Sigma Chemical Co., St. Louis, MO. Semi-tech Electronic, Ltd., Hong-Kong, China. Immunotech, Marselle, France. Zymed Laboratories, San Francisco, CA. DAKO, Glostrup, Denmark.

References 1. Allen SW, Mahaffey EA: 1989, Canine mammary neoplasia: prognostic indicators and response to surgical therapy. J Am Anim Hosp Assoc 25:540–546. 2. Barnard NJ, Hall PA, Lemoine NR, Kadar N: 1987, Proliferative index in breast carcinoma determined in situ by Ki-67 immunostaining and its relationship to clinical and pathological variables. J Pathol 152:287–295.

246

Penã et al.

3. Bloom HJG, Richardson WW: 1957, Histological grading and prognosis in breast cancer. Br J Cancer 1:359–377. 4. Bostock DE: 1975, The prognosis following the surgical excision of canine mammary neoplasms. Eur J Cancer 11:389–396. 5. Bostock DE: 1986, Canine and feline mammary neoplasms. Br Vet J 142:506–515. 6. Bostock DE, Moriarty J, Cocker J: 1992, Correlation between histologic diagnosis mean nucleolar organizer region count and prognosis in canine mammary tumors. Vet Pathol 29:381–385. 7. Bouzubar N, Walker KJ, Griffiths K, et al.: 1989, Ki-67 immunostaining in primary breast cancer: pathological and clinical associations. Br J Cancer 59:943–947. 8. Bravo R, Frank R, Blundell PA, Macdonald-Bravo H: 1987, Cyclin/PCNA is the auxiliary protein of DNA polymerase-d. Nature 32:515–517. 9. Destrexhe E, Vanmanshoven P, Coignoul F: 1995, Comparison of argyrophilic nuclear organizer regions by counting image analysis in canine mammary tumors. Am J Vet Res 56:185–187. 10. Dixon W: 1993, BMDP statistical software, release 7.0. University of California Press, Los Angeles, CA. 11. Donne´s: 1994, Syste`me pour son analyse (SPAD-S), CISIA ed. Saint-Mande, France. 12. Else RW, Hannant D: 1979, Some epidemiological aspect of mammary neoplasia in the bitch. Vet Rec 104:296–304. 13. Fowler EH, Wilson GP, Koestner A: 1974, Biologic behavior of canine malignant mammary neoplasms based on a histogenetic classification. Vet Pathol 11:212–229. 14. Gaglia P, Bernardi A, Venesio T, et al.: 1993, Cell proliferation of breast cancer evaluated by anti-BrdU and anti-Ki-67 antibodies: its prognostic value on short-term recurrences. Eur J Cancer 29:1509–1513. 15. Garcia RL, Coltrera MD, Gown AM: 1989, Analysis of proliferative grade using anti-PCNA/cyclin monoclonal antibodies in fixed, embedded tissues. Comparison with flow cytometric analysis. Am J Pathol 134:733–739. 16. Gasparini G, Boracchi P, Verderio P, Bevilacqua P: 1994, Cell kinetics in human breast cancer: comparison between the prognostic value of the cytofluorimetric S-phase fraction and that of the antibodies to Ki-67 and PCNA antigens detected by immunocytochemistry. Int J Cancer 57:822–829. 17. Gerdes J, Becker MHG, Key G, Catoretti G: 1992, Immunohistological detection of tumour growth fraction (Ki-67 antigen) in formalin-fixed and routinely processed tissues. J Pathol 168:85– 87. 18. Gerdes J, Lemke H, Basch H, et al.: 1984, Cell cycle analysis of a cell proliferation associated human nuclear antigen defined by the monoclonal antibody Ki-67. J Immunol 133:1710–1715. 19. Gilbertson SR, Kurzman D, Zachraw RE, et al.: 1983, Canine mammary epithelial neoplasms: biologic implications of morphologic characteristic assessed in 232 dogs. Vet Pathol 20:127– 142. 20. Haerslev T, Jacobsen GK: 1994, Proliferating cell nuclear antigen in breast carcinomas. An immunohistochemical study with correlation to histopathological features and prognostic factors. Virchows Arch A Pathol Anat Histopathol 424:39–46. 21. Hampe JE, Misdorp W: 1974, Tumors and dysplasias of the mammary gland. Bull World Health Org 50:111–133. 22. Hellmen E, Bergstro¨m R, Holmberg L, et al.: 1993, Prognostic factors in canine mammary tumors: a multivariate study of 202 consecutive cases. Vet Pathol 30:20–27. 23. Kurki D, Ogata K, Tan EM: 1988, Monoclonal antibodies to proliferating cell nuclear antigen (PCNA)/cyclin as probes for proliferating cells by immunofluorescence microscopy and flow cytometry. J Immunol Methods 109:49–59.

24. Kurzman ID, Gilbertson SR: 1986, Prognostic factors in canine mammary tumors. Semin Vet Med Surg Small Anim 1:25–32. 25. Leonardi E, Girlando S, Serio G, et al.: 1992, PCNA and Ki67 expression in breast carcinoma: correlations with clinical and biological variables. J Clin Pathol 45:416–419. 26. Martin PN, Cotard M, Mialot JP, et al.: 1984, Animal models for hormone-dependent human breast cancer. Cancer Chemother Pharmacol 2:13–17. 27. Misdorp W, Hart AAM: 1976, Prognostic factors in canine mammary cancer. J Natl Cancer Inst 56:779–786. 28. Owen LN: 1980, Mammary neoplasia in the dog and cat. III. Prognosis and treatment of mammary tumors in the bitch. J Small Anim Pract 7:708–710. 29. Preziosi R, Sarli G, Benazzi C, Marcato PS: 1995, Detection of proliferating cell nuclear antigen (PCNA) in canine and feline mammary tumors. J Comp Pathol 113:301–313. 30. Sarli G, Benazzi C, Preziosi R, Marcato PS: 1995, Assessment of proliferative activity by anti-PCNA monoclonal antibodies in formalin-fixed, paraffin embedded samples and correlation with mitotic index. Vet Pathol 32:93–96. 31. Scarff RW, Torloni H: 1968, Histological typing of breast tumors, pp. 13–20. World Health Organization, Geneva, Switzerland. 32. Schmitt FC, Pereira EM, Andrade LM, et al.: 1994, The proliferating cell nuclear antigen index in breast carcinomas does not correlate with mitotic index and estrogen receptor immunoreactivity. Pathol Res Pract 190:786–791. 33. Schneider R, Dorn CR, Taylor DON: 1969, Factors influencing canine mammary cancer development and postsurgical survival. J Natl Cancer Inst 43:1249–1261. 34. Shi RR, Key ME, Kalra KL: 1991, Antigen retrieval in formalin fixed paraffin-embedded tissues, an enhancement method for immunohistochemical staining based on microwave oven heating of tissue sections. J Histochem Cytochem 39:741–748. 35. Shivji MKK, Kenny MK, Wood RD: 1992, Proliferating cell nuclear antigen (PCNA) is required for DNA excision repair. Cell 69:345–367. 36. Shrestha P, Yamada K, Wada T, et al.: 1992, Proliferative cell nuclear antigen in breast lesions: correlation of c-erbB-2 oncoprotein and EGF receptor and its clinicopathological significance in breast cancer. Virchows Arch A Pathol Anat Histopathol 421:193–202. 37. Sullivan RP, Mortimer G: 1993, Value of PCNA and Ki-67 staining in breast cancer. J Clin Pathol 46:188–191. 38. Tahan SR, Neuberg DS, Dieffenbach A, Yacoub L: 1993, Prediction of early relapse and shortened survival in patients with breast cancer by proliferating cell nuclear antigen score. Cancer 71:3552–3559. 39. Thomas M, Noguchi M, Kitagawa H, et al.: 1993, Poor prognostic value of proliferating cell nuclear antigen labelling index in breast carcinoma. J Clin Pathol 46:525–528. 40. Van Dierendonck JH, Wijsman JH, Keijzer R, et al.: 1991, Cellcycle-related staining patterns of anti-proliferating cell nuclear antigen monoclonal antibodies. Comparison with BrdUrd labeling and Ki-67 staining. Am J Pathol 138:1165–1172. 41. Veronese SM, Gambacorta M, Gottardi O, et al.: 1993, Proliferation index as a prognostic marker in breast cancer. Cancer 71:3926–3931. 42. Vielh P, Chevillard S, Mosseri V, et al.: 1990, Ki-67 index and S-phase fraction in human breast carcinomas. Comparison and correlations with prognostic factors. Am J Clin Pathol 94:681– 686. 43. Wintzer HO, Zipfel I, Schulte-Mo¨nting J, et al.: 1991, Ki-67 immunostaining in human breast tumors and its relationship to prognosis. Cancer 67:421–428.