Markedly elevated levels of
macrophage colony-stimulating factor (CSF-1) in the serum and
ascites of
epithelial ovarian cancer patients have been previously associated with a poor prognosis. However, measurements of circulating
CSF-1 cannot separate
CSF-1 originating in the
cancer cell from that originating in stromal macrophage or fibroblast. To study the prognosis related to expression of
CSF-1 and its receptor in primary and metastatic
ovarian cancers and to compare the significance of epithelial versus stromal
CSF-1 expression, an immunohistochemical study of 130 ovarian
carcinomas was performed. Twenty-two stage I and II and 108 stage III and IV primary
tumors were studied. Metastatic lesions were also studied in 96 of these 130 cases, 90 of which came from those cases with advanced-stage disease. The intensity and extent of staining for
CSF-1 in epithelium and stroma and for epithelial
CSF-1 receptor was scored. Kaplan-Meier curves of survival were compared with the log-rank test. The Cox regression model was used for multivariate analysis. In the primary
tumors, there was strong expression of
CSF-1 receptor in 65%, epithelial
CSF-1 in 36%, and stromal
CSF-1 in 22%. In the
metastases, there was strong staining for
CSF-1 receptor in 65%, epithelial
CSF-1 in 41%, and stromal
CSF-1 in 15%; strong staining for both
CSF-1 receptor and epithelial
CSF-1 was noted in 26% of the cases. When the
metastases expressed both
CSF-1 receptor and epithelial
CSF-1 strongly, a significant decrease in disease-free survival in stage III invasive
ovarian cancers was observe (P = 0.043), which was found to be an independent prognostic factor (P = 0.007), with an increased relative risk of recurrence of 2.3-fold. Although strong staining for stromal
CSF-1 in the primary
tumor was not found to have prognostic value, for all stages and for the subsets of stages III and IV and for stage III alone, the finding of any degree of stromal
CSF-1 expression in the ovary was a favorable prognostic factor for disease-free (P = 0.046) and overall (P = 0.015) survival. This finding was associated with younger patients (P = 0.007) and low-grade
tumors (P = 0.033) and was not an independent prognostic factor on multivariate analysis. Among the primary
tumors, there was a significant association (P = 0.022) between stromal
CSF-1 staining and lack of strong coexpression of
CSF-1 receptor and epithelial CSF-1; 67 of 94 cases shared these features in the primary
tumors. In the
metastases of invasive stage III cases, strong staining for stromal
CSF-1 was a favorable prognostic factor for overall survival in the absence of strong
CSF-1 receptor staining (P = 0.033) and was associated with low-grade
tumors (P = 0.0002). We report that strong expression of epithelial
CSF-1 along with its receptor in the
metastases of
ovarian cancer patients appears to be a strong independent poor prognostic factor for outcome. We find that expression of the same
cytokine (CSF-1) in the stroma of the primary
tumors is associated with low-grade
tumors and lack of strong coexpression of
CSF-1 receptor and epithelial
CSF-1, leading to an improved long-term outcome. This study may help explain the previous observations that elevated levels of
CSF-1 in serum and
ascites are associated with a worse prognosis in advanced
ovarian cancer patients; the results suggest that the source of secreted
CSF-1 may largely be the epithelium. The results of this study suggest that paracrine effects of stromal
CSF-1 on
tumor behavior contrast with those demonstrated when the
tumor cell is capable of autocrine intracellular or extracellular interactions between
CSF-1 and its receptor.