Abstract
Tumor-associated macrophages (TAMs) are highly active immune effector cells that may either positively or negatively regulate the growth of various malignant cells, depending on the biological context. However, the role of TAMs in human prostate cancer progression is unclear. TAMs were immunohistochemically labeled using a monoclonal (CD68) antibody in radical prostatectomy specimens derived from 81 prostate cancer patients. CD68-positive cells were counted with the aid of a microscope and expressed as macrophage index (MφI), including TAMs/mm2 total tumor tissue (MφItotal), TAMs/mm2 tumor stroma (MφIstroma), and TAMs/mm2 cancer cell area (MφIcancer). MφIs were analyzed in association with patients’ clinical and pathological stage, recurrence status, and histological grade of the cancer. There were significant inverse relationships between MφItotal and MφIstroma and clinical stage (P = 0.016 and P = 0.006, respectively). Reduced MφItotal was also associated with the presence of positive lymph nodes (P = 0.010). Interestingly, although all of the MφIs differed between Gleason score groups, only MφIcancer was positively associated with Gleason score. Univariate analysis of MφItotal and multivariate analysis of MφItotal with specific pathological markers revealed that MφItotal was an independent predictor for disease-free survival after surgery (Cox proportional hazard model, P = 0.044 and P = 0.007, respectively). For patients with high MφItotal (≥185.8, the mean MφItotal value), the disease-free probability 5 years after surgery was 0.75, which was significantly higher than for those with low MφItotal (0.31, P = 0.0008). Additional immunohistochemical studies that evaluated cytotoxicity-related biomarkers in stroma-associated mononuclear cells suggested reduced functional activities in highly aggressive prostate cancer compared with less aggressive disease. Our results indicate that reduced MφItotal is a novel prognostic marker for prostate cancer.
INTRODUCTION
Macrophages have well-defined roles in inflammation but the biological significance of TAMs 3 remains unclear. TAMs have been shown to display both positive and negative effects with regard to tumor growth, and, therefore, it appears that the biological context plays an important role in determining TAM functions (1, 2, 3) . After differentiation from the pool of monocytes and entry into neoplastic tissues, TAMs can produce various growth factors and cytokines, including fibroblast growth factor 1, platelet-derived growth factor, vascular endothelial growth factor, IL-1, granulocyte-macrophage colony-stimulating factor, TGF-α, prostaglandins, and insulin-like growth factor I (reviewed in Ref. 4 ). Under some conditions, these growth factors have the capacity to stimulate tumor cell growth and/or survival directly and in some cases may promote tumorigenesis indirectly through the induction of tumor vasculature (4, 5, 6) . TAMs are also capable of producing tumor cytotoxicity directly through the production of TNF-α (7, 8, 9) , NO (8 , 10 , 11) , H2O2 (12) , and reactive oxygen intermediates and indirectly through the secretion of cytokines such as IL-12 and IL-18 to stimulate the immune system. In some cases, TAMs appear to function as antigen-presenting cells to further promote antitumor immunity (3 , 13) . Although TAMs have the potential to mediate tumor cytotoxicity and to stimulate antitumor lymphocytes (7) , these activities can be suppressed by tumor-derived cytokines that include IL-4, -6, -10, TGF-β1 (14 , 15) , prostaglandin E2 (16) , and macrophage colony-stimulating factor (17) . TAMs have also been shown to produce and secrete specific proteases, including urokinase plasminogen activator that can have profound effects on tumor cell activities (14 , 18) . The overall complexity of TAM-related biological activities presents a challenging scenario from which to determine the clinical significance of TAMs for specific malignancies.
The results of various immunohistochemical staining analyses using macrophage-specific markers have not led to general consensus regarding the prognostic significance of TAMs. In papillary thyroid cancer, patients with tumors containing TAMs exhibiting neoplastic cell phagocytotic activities had a better prognosis than patients without TAMs (19) . In contrast, the presence of areas of high-density TAMs within “hot spots” was positively correlated with increased vascularity and metastasis and with reduced relapse-free and overall survival in breast cancer (6) .
In the present study we performed quantitative immunohistochemical analysis of TAMs in human prostate cancer tissues using a CD68 monoclonal antibody. We expressed the data as MφI, i.e., TAMs/mm2 total tumor tissue (MφItotal); TAMs/mm2 tumor stroma (MφIstroma); and TAMs/mm2 cancer cell area (MφIcancer). Inverse associations between both MφItotal and MφIstroma and clinical indicators of disease progression were revealed. MφItotal was further shown to be an independent predictor for time to disease progression after radical prostatectomy. Interestingly, a positive association was demonstrated between MφIcancer and Gleason score. Overall, our results clearly demonstrate novel prognostic capacity of reduced TAMs in prostate cancer tissues and reveal the importance of the local tumor microenvironment in regard to the biological and clinical impact of TAMs in prostate cancer.
MATERIALS AND METHODS
Patients and Tissue Processing.
Prostate cancer specimens from 81 patients (aged 45.2–77.9 years; mean, 63.1 years) who had undergone radical prostatectomy were obtained from the Specialized Program of Research Excellence tissue bank, Methodist Hospital, Baylor College of Medicine. Preoperatively, the patients had a cancer staged according to the Union International Contre Cancer clinical staging system (1992) as T1a–c (n = 7), T2a–c (n = 61), and T3a (n = 13). These patients had not received any form of therapy before surgery. After the surgery, the prostate specimens were sliced into 5-mm-thick tissue blocks as described previously (20) . The tissue blocks were fixed in 10% formalin and embedded in paraffin. H&E-stained sections from each cancer were evaluated by a pathologist (T. M. W.) for histological differentiation status as demonstrated by their Gleason score and pathological stage (TNM system); the Gleason score of cancers was also verified by another investigator (G. Y.). Recurrence was defined as a postoperative PSA level >0.4 ng/ml (Hybritech, Inc., San Diego, CA) on two successive measurements.
Immunohistochemistry.
Six-μm-thick sections were made from the tissue blocks fixed in 10% formalin and embedded in paraffin after a routine procedure. The sections were deparaffined and rehydrated. They were then heated in citrate buffer (0.01 m, pH 6.0) with in an 800 W microwave oven for 9 min for antigen retrieval. Endogenous peroxidase in sections was inactivated in 2% H2O2 for 10 min. The sections were then blocked in 3% normal horse serum in 0.2 m PBS (pH 7.4), followed by incubation in monoclonal antibody specific for human CD68 (Dako Corp., Carpinteria, CA). The CD68 antibody was used at a dilution of 1:200 in PBS with 0.5% normal horse serum. The sections were incubated in the primary antibody for 2 h at room temperature and then processed after standard ABC immunostaining procedures with an ABC kit (Vector Laboratories, Burlingame, CA). Immunoreaction products were visualized in a 3,3′-diaminobenzidine/H2O2 solution. To verify the specificity of the immunoreactions, some sections were incubated in normal mouse serum replacing primary antibody. In addition, antibodies to NOS2, TNF-α, and TGF-β1 (from Santa Cruz) were also used to immunostain some paraffin-embedded (for NOS2 and TGF-β1) or frozen (for TNFα) sections of some tumors, using the ABC technique.
Histological Analysis.
TAMs were quantified by systematically screening the entire cancer area and by counting at the hot spot in a cancer area where macrophages accumulated at highest density in the specimens.
For systematic counting, 10 to 15 ocular measuring fields, each composed of 100 grids and having a real area of 0.0625 mm2, were randomly chosen under a microscope at a power of ×400 within a cancer. Sections were positioned such that each measuring field was completely occupied by only cancer tissue. All TAMs in each measuring field were counted and stratified as those localized to cancer stroma and those in contact with cancer cells or penetrating into a cancer lumen. The grids in each measuring field were also stratified by percentage into the two compartments: tumor stroma area or cancer cell/lumen area. The counts of MφItotal, MφIstroma, MφIcancer, and MφIbenign were recorded for each specimen. For counting macrophages in hot spots, the sections were first evaluated at ×100, and the cancer area where TAMs accumulated at higher density was identified. TAMs were then counted at ×400 within the hot spot and expressed as MφIhot spot. All counting was performed by one investigator (S. S.) without knowledge of clinical information. In addition, NOS2-, TNF-α-, and TGF-β1-positive tumor-infiltrating mononuclear cells were also scored separately on tissue sections. For each specimen, the whole cancer area was screened under a microscope at ×200, and positively labeled mononuclear cells were counted from 10 to 20 randomly selected microscopic fields. The densities of these cells were scored as follows: −, no mononuclear cell was labeled; +, ≤1 positive cell per microscopic field; ++, >1 and <5/field; and +++, ≥5/field. NOS2, TNF-α-, and TGF-β1-positive cancer cells were scored according to their relative densities as +, low; ++, moderate; and +++, high.
Statistical Analysis.
The association of MφIs with patients’ clinical stage and Gleason score was evaluated using Kruskal-Wallis test. The association of MφIs with extraprostatic extension, seminal vesicle invasion, lymph node metastasis, and surgical margin was evaluated using the Mann-Whitney test. The Mann-Whitney test was also used to evaluate the difference in MφItotal between benign and malignant cancer patients. Univariate analyses of MφIs as a continuous variable and the postoperative pathological markers as well as multivariate analyses of MφIs adjusting for pathological markers were performed using the Cox proportional hazard regression model. Survival curves of relapse-free survival for high- and low-MφItotal patients were obtained using a Kaplan-Meier analysis and tested with log-rank test. Cox proportional hazard regression model was used to evaluate the difference between these survival curves, adjusting for pathological markers. Associations between tumor histology and immunoreactivity for NOS2, TNF-α, and TGF-β1 were tested using Fisher’s exact test. The frequency of positive cancer cells between the more and less aggressive cancers was grouped and compared as + versus ++/+++. P < 0.05 was considered statistically significant in all of the analyses. All analyses were performed with statistical software (StatView version 5.0, SAS Institute Inc., Cary, NC and/or SPSS 10.0, SPSS Inc., Chicago, IL).
RESULTS
Macrophage Infiltration in Normal Prostatic Tissue and in Tumor Tissue.
CD68-positive cells were observed in all of the specimens tested, and the majority of the CD68-positive cells had morphological features of macrophages. However, some cells with features of a mast cell were labeled by this antibody as well. In normal prostates derived from organ donors or in histologically normal prostatic tissue, CD68-positive cells were present primarily in the stroma, with relatively small numbers present in the lumens of prostatic glands (Fig. 1A) ⇓ .
CD68 immunostaining in human prostates using a monoclonal antibody. Labeled macrophages were present in histologically normal prostatic tissue (A), in prostate cancer stroma (B), and in close contact with cancer cells (C). ×200 (A); ×400 (B and C).
In prostate cancer specimens, CD68-positive cells were seen infiltrating the cancer area. The infiltrating TAMs were distributed in three distinguishable compartments, including tumor-associated stroma, cancer cell region, and the lumens composed of cancer cells. On the basis of our calculations of all of the specimens analyzed, 84% of TAMs were distributed within the stroma surrounding cancer cells (Fig. 1B) ⇓ . The remaining, smaller proportion of TAMs (16%) was in direct contact with cancer cells or penetrated into the lumens of cancer cells (Fig. 1C) ⇓ . This ratio varied with the Gleason grade in individual cancer specimens. As expected, cancers of lower grade (Gleason score, ≤6) had a larger proportion of cancer stroma than cancers of higher Gleason grade, and this increased stroma-cancer cell ratio contributed to an increased number of stroma-infiltrating TAMs. Overall, the macrophage density, i.e., MφItotal, was significantly higher in the malignant/total tumor tissue area (185.8.1 ± 11.8, mean ± SEM; n = 81) relative to adjacent histologically benign prostatic areas (84.1 ± 16.4; n = 10, P = 0.0012; Fig. 2 ⇓ ).
Comparison in macrophage density expressed as MφI between total tumor areas and adjacent benign glandular areas.
Association of MφIs with Clinical and Pathological Characteristics.
Compartment-specific macrophage densities, i.e., MφItotal, MφIstroma, and MφIcancer, were determined and tested for associations with clinical stage, Gleason score, and other pathological parameters (Table 1) ⇓ . Patients with cancers of relatively high TNM stage tended to have a lower MφItotal and MφIstroma. Both MφItotal and MφIstroma were inversely associated with TNM stage (P = 0.016 and P = 0.006, respectively). An inverse association was also found between the MφIhot spot (see “Materials and Methods”) and clinical stage (P = 0.012; Table 2 ⇓ ). All MφIs were significantly different between Gleason score patient groups, although only MφIcancer showed a trend for positive association with the Gleason score (P = 0.006). A significantly lower MφItotal was detected in patients with positive lymph node metastasis (P = 0.010). There was no statistically significant association between any specific MφI and the status of the surgical margins, extraprostatic extension, or seminal vesicle invasion.
Association of clinical and pathological parameters with MφI (systematic counting)
Association of clinical and pathological parameters with MφI (hot spot counting)
MφItotal as a Predictor for Disease-free Survival.
The relationships between specific MφIs and the recurrence in this set of patients were also analyzed using the Cox proportional hazard model. As continuous variables, MφItotal, MφIstroma, MφIcancer, and MφIhot spot as well as other pathological markers were first univariately analyzed for predictive value. The results revealed that the following were significant predictors of time to recurrence: MφItotal, MφIcancer, extraprostatic extension, seminal vesicle invasion, lymph node metastasis, surgical margin status, and Gleason score (Table 3) ⇓ . However, when MφItotal, MφIstroma, MφIcancer, and MφIhot spot were individually combined with the pathological markers in four separate models for multivariate analysis, in the reduced prediction models only MφItotal remained a significant predictor of time to recurrence in the presence of the surgical margin status and Gleason score (Table 4) ⇓ .
Univariate Cox proportional hazard models for time to recurrence
Multivariate Cox proportional hazard models for time to recurrence
Kaplan-Meier Analysis of MφItotal.
The predictive value of low versus high MφItotal was then tested using a Kaplan-Meier actuarial analysis (Fig. 3) ⇓ . The follow-up time ranged from 0.8 to 131.7 months (mean, 46.6 months). Forty-two (52%) of 81 patients had biological recurrences (PSA ≥ 0.4 ng/ml) during follow-up. The MφItotal values for this set of patients averaged 185.8. When using this mean as a cutoff point, 47 cases (58%) were in the low category (<185.8), with 34 cases (42%) falling into the high (≥185.8) category. The recurrence-free probability (disease-free survival) at 60 months in the high-MφItotal category (0.75) was significantly higher than that in the low-MφItotal group (0.31, P = 0.0008; Fig. 3 ⇓ ). The mean recurrence-free survival time for the high-MφItotal category was 96.10 months (95% confidence interval, 78.0–114.2), whereas the low-MφItotal category was 52.2 months (95% confidence interval, 36.7–67.6). The predictive value of the high- versus the low-MφItotal remained significant (P = 0.002) in presence of the surgical margin status and Gleason score, used in the multivariate Cox proportional hazard model.
Recurrence-free survival probability after radical prostatectomy. Comparison between the patients who had a cancer with a MφItotal ≥ 185.8 and those having a MφItotal<185.8.
Compartment-specific Densities of NOS2-, TNF-α-, and TGF-β1-positive Mononuclear Cells in Prostate Cancer.
Because NOS2, TNF-α, and TGF-β1 expression are related to macrophage function, we investigated these activities in mononuclear cells using semiquantitative immunohistochemical analysis and compared well-differentiated cancers from nonrecurrent patients with poorly differentiated cancers from recurrent patients. Overall, the majority of positively labeled cells for each antigen demonstrated macrophage morphology. In the less aggressive cancers, tumor-infiltrating, NOS2-positive mononuclear cells were mainly localized to the stroma of a cancer (Table 5) ⇓ , and their density was significantly lower in the more aggressive cancers (P = 0.001). Compared with the less aggressive cancers, significantly increased densities of NOS2-positive mononuclear cells in close contact with cancer cells (P = 0.001) were observed in the aggressive cancers. Densities of NOS2-positive cancer cells per se within highly aggressive prostate cancer were significantly increased compared with less aggressive prostate cancer (P = 0.001). Significantly fewer TNF-α-positive mononuclear cells were observed in the stroma of highly aggressive cancers compared with the less aggressive ones (P = 0.021), but there was no difference in the densities of TNF-α-positive mononuclear cells in close contact with cancer cells between the two groups (P = 0.354). In the highly aggressive cancers, the densities of cancer cells positive for TNF-α staining appeared to be increased relative to the less aggressive cancers, but no statistically significant difference was detected (P = 0.308). There were similar densities of TGF-β1-positive mononuclear cells infiltrating in the cancer stroma and in close contact with cancer cells in the two groups, but the frequency of TGF-β1-positive cancer cells was significantly increased in the highly aggressive cancers compared with the less aggressive ones (P = 0.005).
Compartment-specific densities of NOS2-, TNF-α- and TGF-β1-positive mononuclear cells in well and poorly differentiated prostate cancer
The NOS2-, TNF-α- and TGF-β1-positive tumor-infiltrating mononuclear cells were scored separately on tissue sections as described in “Materials and Methods.” For well-differentiated (Gleason grade 2) nonrecurrent cancers, four patients for each antigen were scored. For poorly differentiated (Gleason grade 4 or 5) recurrent cancers (n = 11 for NOS2, n = 9 for TNF-α, and n = 6 for TGF-β) were scored. The data included in the table are the ranges of the scores assigned.
DISCUSSION
There have been numerous reports regarding the role of TAMs in malignant progression, yet, in general, there is no clear consensus regarding the functional significance of TAMs in solid tumors and their metastases. In the present study, we demonstrated in prostate cancer that the extent of TAM infiltration within the total tumor tissue (MφItotal) or within the tumor stroma (MφIstroma) is inversely associated with clinical stage. MφItotal was also inversely associated with the presence of positive lymph nodes. We additionally analyzed MφItotal in regard to its capacity to predict the time to recurrence as estimated by biochemical recurrence, i.e., rising serum PSA, after radical prostatectomy using both univariate and multivariate analyses. The results indicated that reduced MφItotal is an independent predictor for time to disease progression. Interestingly, we also found that TAM density within the cancer cell area, i.e., MφIcancer, was positively associated with the Gleason score. In general, our results establish that reduced MφItotal is a novel prognostic marker for prostate cancer progression and that the compartmental localization of TAMs may determine their specific role in prostate cancer progression.
In other tumor systems, the presence of increased lymphocytes within the cancer cell area indicates a favorable prognosis. For example, CD8+ T cells infiltrated within human colorectal cancer have a positive prognostic significance in regard to patient survival (21) . However, in our study MφIcancer was positively associated with Gleason score, a well-established and useful pathological marker of progression. Leek et al. have reported previously that increased presence of macrophages correlates with tumor growth and metastasis in breast cancer (6) . In their report, a significant positive correlation between high vascular grade and increased macrophage index and a strong relationship between macrophage index and reduced relapse-free survival and reduced overall survival were observed. This article used the technique of counting macrophages within hot spot areas within the tumor that are not specified in regard to cellular compartment. In this report, we also used hot spot counting in addition to systematic counting for TAMs. Unlike the results obtained using the systematic counting protocol, the results from hot spot counting, i.e., MφIhot spot, did not show any prognostic significance, but, in general, they were consistent with the results of the systematic counting protocol that demonstrated an inverse association of MφItotal and MφIstroma with clinical stage.
Overall, the results of this study and studies from other laboratories indicate that the cellular microenvironment is critical in determining the biological and clinical significance of specific tumor-associated immunocytes. However, our study also specifies that TAMs within the cancer stroma versus cancer cell compartments may have opposing activities in regard to their effects on cancer progression.
As observed in this study, some of the mononuclear cells in the cancer stroma express NOS2 and TNF-α, which can contribute either indirectly or directly to tumor cell cytotoxicity. The majority of these cells demonstrated macrophage morphology. The frequency of NOS2- and TNFα-positive mononuclear cells was reduced within the cancer stroma in poorly differentiated nonrecurrent prostate cancer compared with differentiated recurrent prostate cancer. In contrast, the densities of NOS2-positive mononuclear cells in contact with the cancer cells in the highly aggressive prostate cancer were increased compared with the less aggressive prostate cancer. Furthermore, significant increases in NOS2- and TGF-β1-positive prostate cancer cells per se were also demonstrated in highly aggressive prostate cancer versus less aggressive disease. The reduction in the densities of NOS2- and TNFα-positive mononuclear cells that infiltrate the stroma in highly aggressive prostate cancers compared with less aggressive disease suggests that the cytotoxic activities of these cells become attenuated during the acquisition of increasing malignant potential of the cancer. The reduced densities of NOS2- and TNF-α-positive mononuclear cells in the cancer stroma does not appear to simply reflect an overall reduction of macrophage density in this compartment, because macrophage densities in high Gleason grade cancers increased relative to low Gleason grade cancers (see Table 1 ⇓ ). Interestingly, in prostate cancer as in many other malignancies, an overall reduction in tumor stroma specifies a more malignant pathological grade. It is conceivable that an overall reduction in specific stromal cells within high-grade lesions lead to reduced potential for specific stromal cell-mononuclear cell interactions that are supportive of antitumor cytotoxic activities. In this regard, our previous studies have demonstrated that the genetic background of prostate stromal cells specifically can dramatically affect the incidence of prostate carcinogenesis in vivo using the mouse prostate reconstitution model system (22) . The increased densities of NOS2-positive mononuclear cells in contact with cancer cells and cancer cells per se as well as TGF-β1-positive cancer cells in highly aggressive prostate cancers relative to less aggressive disease suggest selection for resistance to the cytotoxic/growth-suppressive effects of NOS2 and TGF-β1 activities. Indeed, previous studies have indicated that various malignant cells can overexpress NOS2 (reviewed in Refs. 23 and 24 ), and in particular, TGF-β1 has been associated with prostate cancer progression (25 , 26) . The role of prostate cancer stroma in regard to support of immunocyte-mediated cytotoxic activities is deserving of additional studies based on this report and other observations. In addition, the inverse association between MφItotal and disease-free survival after radical prostatectomy is also deserving of further study, which could provide more insight into prostate cancer progression and potentially establish macrophage density as a clinically useful prognostic marker.
Footnotes
-
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
-
↵1 Supported by National Cancer Institute Grant CA50588 and Specialized Program of Research Excellence P50–58204.
-
↵2 To whom requests for reprints should be addressed, at Scott Department of Urology, Baylor College of Medicine, 6560 Fannin, Suite 2100, Houston, TX 77030. Phone: (713) 799-8718; Fax: (713) 794-7983; E-mail: timothyt{at}www.urol.bcm.tmc.edu
-
↵3 The abbreviations used are: TAM, tumor-associated macrophage; IL, interleukin; TGF-β1, transforming growth factor β1; TNF, tumor necrosis factor; MφI, macrophage index; TNM, tumor-node-metastasis; PSA, prostate-specific antigen; NOS2, nitric oxide synthase 2; ABC, avidin biotin complex.
- Received November 29, 1999.
- Accepted August 18, 2000.
- ©2000 American Association for Cancer Research.
References
Volume 60, Issue 20
