Abstract
Context.—Pathologic stage is the main prognostic factor for predicting outcome in renal cell carcinoma (RCC). Because of its unreliability in predicting tumor progression, other factors are needed to provide additional prognostic information.
Objective.—The expression of CD44, cyclooxygenase 2, and platelet-derived growth factor receptor α (PDGFR-α) was evaluated as a potential prognostic factor for survival in patients with RCC.
Design.—Sixty-two patients (42 men and 20 women; median age, 61 years), undergoing partial (10 cases) or radical (55 cases) nephrectomy for RCC were retrospectively analyzed by immunohistochemical analysis for CD44, cyclooxygenase 2, and PDGFR-α expression. Impact of various factors on disease-specific and overall survival was calculated using Cox proportional hazards models.
Results.—There was a gradual increase in CD44 and cyclooxygenase 2 expression with increasing RCC nuclear grade. In contrast, PDGFR-α expression showed no consistent relationship with nuclear grade. On univariate analysis, metastasis at time of surgery (P < .001), tumor size (P = .004), pathologic stage group (P = .001), and nuclear grade (P = .004) were correlated with disease-specific survival. On multivariate analysis, only the presence of metastasis at diagnosis (P < .001) was significant. For overall survival, metastasis (P < .001), tumor size (P = .02), pathologic stage group (P = .01), nuclear grade (P = .003), and PDGFR-α (P = .03) were significant on univariate analysis. Only metastasis (P = .001) and PDGFR-α (P = .03) were significant on multivariate analysis.
Conclusions.—When combined with other variables, PDGFR-α expression in RCC may provide additional predictive value related to the patient's overall survival. However, CD44 and cyclooxygenase 2 do not seem to be independent prognostic indicators in predicting outcomes for patients with RCC.
Renal cell carcinoma (RCC) accounts for 3% of adult neoplasms and is the most common renal malignancy.1 The incidence of RCC has continued to increase. Surgical resection of the primary tumor remains the only known effective treatment for RCC. Currently, the best prognostic factor for RCC is pathologic stage.2,3 Because the outcomes for RCC can still be unpredictable, other prognostic factors are needed. Recent findings support the utility of MIB-1 (Ki-67), p53, and ploidy status as additional relevant prognostic factors for RCC. Individually and combined with p53, MIB-1 and ploidy were of prognostic significance on univariate analysis and are helpful in characterizing tumor aggressiveness in predicting disease-free survival.4–8 Recently, additional new markers including CD44, platelet-derived growth factor receptor α (PDGFR-α), and cyclooxygenase 2 (COX-2) have been proposed as potential prognostic indicators.9,10 Studies to date have been limited and have shown different results as to predicting disease outcome with these 3 markers.
CD44 is a surface transmembrane glycoprotein that was initially identified on lymphocytes. Several isoforms of CD44 have been identified, the result of alternative posttranscriptional splicing modifications of 10 exons within a single gene located on the short arm of chromosome 11.11–13 CD44 is known to participate in lymphocyte homing through interaction with hyaluronic acid on high endothelial venules, and is used by T lymphocytes for migration to selective sites in lymphoid tissues.9 It participates in normal and tumoral cell-cell and cell-extracellular matrix interactions. The extracellular domain of CD44 is the principal receptor of an extracellular matrix molecule known as hyaluronic acid.13 Hyaluronic acid is a nonsulfated glycosaminoglycan that regulates cellular processes of adhesion, migration, and proliferation, and maintains extracellular matrix osmotic balance through its receptors, including CD44. Hyaluronic acid overexpression has been correlated with the metastatic potential of certain human tumors.13–19 It is becoming generally accepted that these cell-matrix interactions of CD44 play a role in tumor cell invasion and metastasis. However, it is still to be determined whether CD44 can be an effective predictor of disease outcome in patients with RCC.
Platelet-derived growth factor (PDGF) is encoded by the proto-oncogene c-sis. Mutations of c-sis can render the gene oncogenic, resulting in overexpression of the β chain of PDGF. Several human tumors including astrocytomas and osteosarcomas have subsequently been found to express PDGF. These tumors also express receptors for PDGF, and are thereby subject to an autocrine stimulation loop. In most tumors exhibiting an autocrine loop, the growth factor gene itself is not altered. More often, products of other oncogenes that lie along the signal transduction pathway cause overexpression of the growth factor genes. This eventually causes the cells to produce large quantities of growth factor.9 The predictive value of PDGF or its receptors in RCC has not been fully elucidated, with only 1 study thus far suggesting an association between PDGFR-α and tumor progression.10
Cyclooxygenase enzymes are known to exist in two different forms, COX-1 and COX-2. The COX-2 form is the mediator of the cyclooxygenase pathway of arachidonic acid metabolism in the production of prostaglandins. Prostaglandins are involved in the human inflammatory response with production of pain, fever, vasodilation, edema, inhibition of platelet aggregation, and others.9 Cyclooxygenase 2 is not expressed in most tissues, but has been shown to increase during the inflammatory response through interactions with agents such as growth factors and cytokines. Few studies have shown a relationship between COX-2 and RCC.20–24.
The goal of the current study was to evaluate the role of CD44, COX-2, and PDGFR-α activity as prognostic factors in patients with RCC.
MATERIALS AND METHODS
From 1995 to 2004, 119 patients underwent partial or radical nephrectomy for RCC at our institution. A retrospective database containing information on these 119 patients was constructed with institutional review board approval. The data in the database obtained from the cancer registry at our institution included tumor type, tumor size, nuclear grade using the Fuhrman grading system, gender, patient age, metastatic status, pathologic stage, and DNA ploidy status at the time of surgery. CD44, COX-2, and PDGFR-α expression was assessed for 67 patients who had tissue available for analysis. Two patients were excluded from the study because of multiple surgeries for von Hippel-Lindau disease. Three patients were excluded because of insufficient follow-up information. Thus, 62 patients were analyzed.
Specimens were fixed in 10% neutral-buffered formalin and processed into paraffin blocks in the traditional manner. Paraffin blocks were sectioned at 4 μm. After deparaffinization, all slides were heat treated in citrate buffer pH 6.0 for 4 minutes using the BioCare Decloaking Chamber (BIOCARE MEDICAL, Concord, Calif) followed by a 10-minute cooldown. Endogenous biotin and avidin were blocked using 20% egg white (15 minutes) and 5% powdered skim milk followed by several distilled water rinses. The COX-2 (1:50, Cayman Chemical Company, Ann Arbor, Mich) and PDGFR-α (1:100, Santa Cruz Biotechnology Inc, Santa Cruz, Calif) staining were performed using DakoCytomation EnVision+ Labeled Polymer on a Dakocytomation Autostainer (DakoCytomation, Carpinteria, Calif) per company protocol. CD44 (1:100, DakoCytomation) staining was performed using iView reagents on the NexES instrument (both from Ventana Medical System Inc, Tucson, Ariz) per company protocol. Primary antibodies, diluted in Dakocytomation Antibody Diluent, were incubated for 30 minutes. The method of Grizzle et al25 was used to determine an immunostain score for each of the markers studied. Slides were scored separately by 2 individuals (O.W.T. and B.S.) and any discrepancy between them was resolved at the double-headed microscope. Cellular staining intensity was graded for each marker on a score from 0 to 4+. The percentage of positive cells at each intensity was multiplied by the appropriate intensity score; these values were summed and then divided by the total number of cells to obtain a weighted average score between 0 and 4. This method has been proven to be more representative than the usual 0, 1+ to 3+ staining intensity as it takes into consideration both antigen presence and changes in antigen expression for the lesion. It also takes into account all relevant areas on the slides.
Statistical analysis was performed with SPSS for Windows (Release 12.0, SPSS Inc, Chicago, Ill). Categorical variables were summarized by frequencies and percentages, and quantitative variables were summarized by medians and ranges. Quantitative variables were compared across groups using the Kruskal-Wallis test. The Wilcoxon rank sum test was used to perform pairwise comparisons on quantitative variables that were globally different among groups. The Fisher exact test was used to compare categorical variables among groups. The duration of follow-up was calculated from the time of surgery until the date of death or last known follow-up. Univariate analysis of time to death (overall or disease-specific) was analyzed by the Kaplan-Meier survival analysis. We compared categorical variables by the log-rank test and continuous variables by Cox proportional hazards analysis. Probability values <.05 were considered to be statistically significant. Because all analyses were considered as exploratory, no corrections for multiple comparisons were made.
RESULTS
Patient data and tumor characteristics are summarized in Tables 1 through 3. The patients' ages ranged from 36 to 81, with a median of 61 years. The median follow-up for the 62 patients was 22 months, with a minimum of 1 day (postsurgical demise) and a maximum of 108 months. Fifty-four of the RCC cases were of the classic clear cell type. The remaining cases included 5 chromophobe and 3 papillary RCCs. Of the 62 patients, 11 had metastasis at the time of surgery and 18 died during the follow-up, with 10 dead of disease (Table 2). Thirty-seven patients were alive without evidence of RCC, and 5 patients were currently alive with evidence of disease. The carcinomas of the majority of the patients were classified as nuclear grade 2 or 3 or as pathologic stage T1 or T3 (Table 3).
Relationship of CD44, Cyclooxygenase 2 (COX-2), and Platelet-Derived Growth Factor Receptor α (PDGFR-α) Expression to Tumor Nuclear Grade and Histologic Cell Type of Renal Cell Carcinoma Cases

Summary of Demographic and Clinicopathologic Data for 62 Patients With Renal Cell Carcinoma (RCC)

CD44, COX-2, and PDGFR-α Expression
Almost all cases were positive for COX-2 (61/62 cases) and PDGFR-α (58/62 cases), whereas positive immunoreactivity for CD44 was noted in only 20 of the 62 cases studied. The final immunostaining score was obtained by calculating the percent positive cells and the staining intensity for each marker, resulting in a weighted average score between 0 and 4. Figure 1 shows representative examples of immunostaining for CD44 (Figure 1, A), COX-2 (Figure 1, B), and PDGFR-α (Figure 1, C). All markers had cytoplasmic localization. When the expression of each marker was compared with the histologic tumor type, chromophobic RCC showed the highest level of expression for all markers (Table 1). Clear cell RCC had the second highest expression for both COX-2 and PDGFR-α, but it had the least expression for CD44 (Table 1). Table 1 and Figure 2 show a gradual increase in CD44 and COX-2 expression with increasing nuclear grade. Higher grade carcinomas (Fuhrman nuclear grade 3 and 4) showed the highest average COX-2 and CD44 values (2 and 0.58, respectively). In contrast, PDGFR-α did not demonstrate a consistent relationship with nuclear grade (Table 1). The variation in CD44, COX-2, and PDGFR-α expression with various demographic and clinicopathologic variables is shown in Table 4. Only for COX-2 expression was there any statistically significant differences for expression between the various groupings, for cell type (P = .05) and pathologic stage (P = .01). Although there were differences in median values (eg, PDGFR-α for tumor size and cell type), these did not approach statistical significance. A definite linear increase was observed for tumor size versus nuclear grade (P = .001) (Figure 2).
Photomicrographs of renal cell carcinoma specimens stained for expression of CD44 (A), cyclooxygenase 2 (B), and platelet-derived growth factor receptor α (C) (original magnifications ×10)
Photomicrographs of renal cell carcinoma specimens stained for expression of CD44 (A), cyclooxygenase 2 (B), and platelet-derived growth factor receptor α (C) (original magnifications ×10)
Relationship of CD44, cyclooxygenase 2 (COX-2), and platelet-derived growth factor receptor α (PDGFR-α) expression and tumor size to tumor nuclear grade for patients with renal cell carcinoma. Nuclear grade using the Fuhrman grading system is shown on the x-axis. Mean intensity index scores for CD44, COX-2, and PDGFR-α and mean tumor size in decimeters are shown for each grade. Figure 3. Overall survival for patients with versus without metastasis at time of diagnosis. The tick marks represent censored patients.
Relationship of CD44, cyclooxygenase 2 (COX-2), and platelet-derived growth factor receptor α (PDGFR-α) expression and tumor size to tumor nuclear grade for patients with renal cell carcinoma. Nuclear grade using the Fuhrman grading system is shown on the x-axis. Mean intensity index scores for CD44, COX-2, and PDGFR-α and mean tumor size in decimeters are shown for each grade. Figure 3. Overall survival for patients with versus without metastasis at time of diagnosis. The tick marks represent censored patients.
Overall Survival
The P values are summarized in Table 5 for the impact of various factors on overall survival, assessed by univariate analysis. Presence of metastasis at diagnosis, tumor size, pathologic stage group, nuclear grade, and PDGFR-α expression were all significant prognosticators of survival. Age, gender, cell type, DNA ploidy, and expression of CD44 and COX-2 were not predictive of survival. On multivariate analysis, metastasis at the time of surgery (P < .001) (Figure 3) and a lower value of PDGFR-α (P = .03) were significantly predictive of an inferior overall survival. Figure 4 shows the survival advantage of having a PDGFR-α value greater than 0.75 (Figure 4, A), whereas similar dichotomization for COX-2 (Figure 4, B) and CD44 (Figure 4, C) did not reveal any effect on survival.
Prognostic Influence of Various Demographic and Clinicopathologic Factors on Overall and Disease-Specific Survival, by Univariate Analysis*

Overall survival for patients with high (>0.75) versus low (<0.75) platelet-derived growth factor receptor α (PDGFR-α) values (A), high (>1.5) versus low (<1.5) cyclooxygenase 2 (COX-2) values (B), and high (>0) versus low (=0) CD44 values (C). The tick marks represent censored patients
Overall survival for patients with high (>0.75) versus low (<0.75) platelet-derived growth factor receptor α (PDGFR-α) values (A), high (>1.5) versus low (<1.5) cyclooxygenase 2 (COX-2) values (B), and high (>0) versus low (=0) CD44 values (C). The tick marks represent censored patients
Disease-Specific Survival
The P values are also summarized in Table 5 for the influence of various factors on disease-specific survival. Presence of metastasis at diagnosis, tumor size, pathologic stage group, and nuclear grade were again significant prognosticators of survival by univariate analysis. However, PDGFR-α did not have a significant impact on survival. Only metastasis at the time of surgery was significantly associated with disease-specific survival on multivariate analysis.
COMMENT
Renal cell carcinoma accounts for 1% to 3% of visceral cancers in humans, and about 85% of primary malignant tumors of the kidney in adults. The majority of these primary malignant renal tumors are of the conventional or clear cell type.3 Other subtypes of RCC include chromophobic, papillary, collecting duct, and RCC unclassified.3 There is a male predominance of about 2 to 3:1. The majority of patients are in the sixth or seventh decade of life at the time of diagnosis.2,3 Survival rates are on the rise, attributable at least in part to earlier detection, with 5-year survival ranging from 60% to 90% for patients with tumors confined to the kidney.11,24 Renal cell carcinoma is known to have a variable clinical course and behavior, often unreliably predicted with the standard prognostic factors such as tumor stage, nuclear grade, and tumor size. Although prognosis is related to stage, a significant number of patients with stage 1 disease develop metastases.11 Some correlation has been shown between tumor subtype and prognosis, with chromophobic RCC having a better prognosis compared with some other subtypes.19 Several studies have attempted to identify potential molecular prognosticators for RCC, with sometimes discordant results. This study looked at the feasibility of using CD44, COX-2, and PDGFR-α as potential biomarkers to predict tumor aggressiveness as related to disease-specific and overall survival.
Studies have shown that the standard form of CD44 or one or more of its variants are overexpressed in many human malignancies, including breast, stomach, colorectal, lung, bladder, non-Hodgkin lymphoma, certain squamous cell carcinomas, melanoma, and RCC.11,15 The enhanced metastatic potential for tumor cells that overexpress CD44 has been well documented. Indeed, an earlier animal model using transfection experiments showed that a fusion protein containing the variant CD44 isoforms could induce metastases in an otherwise nonmetastasizing pancreatic adenocarcinoma in the rat.12,16 In regard to RCC, antibody blockade of CD44 resulted in significant reduction of the invasive capacity of tumor cells.13 Increased CD44 expression has been associated with invasion beyond the renal capsule and distant metastases, as well as higher nuclear grade, thus correlating with stage and grade.11 CD44 overexpression in RCC has shown an association with tumor progression, recurrence, survival, tumor aggressiveness, and perhaps pathogenesis.11–13,19 The mechanism of such pathogenesis has not been clearly defined. Earlier investigators postulated a potential role for the CD44 molecule in tumor metastasis via the homing effects to lymphoid tissues. This function of CD44, however, did not account for the hematogenous spread of many tumors known to overexpress CD44, such as RCC. Current thinking is that CD44 itself may have angiogenic effects, and/or possibly upregulate the expression of other angiogenic factors such as β-fibroblast growth factor and vascular epithelial growth factor.11 One study indicated a possible role of CD44 in reducing the number of microvessels within tumors, stating that decreased microvessel density may lead to proliferation of larger vessels in tumor tissue, thereby increasing the potential for hematogenous metastases.11
The prognostic value of CD44 is uncertain. Most investigators have not been able to show CD44 as a significant prognosticator independent of tumor stage and nuclear grade or size, although results have varied substantially between studies.11,13,14,16,17 One study showed CD44 as an independent prognostic variable correlating with disease-free interval and survival.19 In our study, CD44 did not show significant predictive value for predicting disease-specific or overall survival.
Studies have shown a relationship between COX-2 expression and several human malignancies, including transitional cell carcinoma of the bladder, pancreatic ductal carcinoma, cancers of the stomach, colon, skin, lung, esophagus, gallbladder, breast, larynx, and RCC.20–22 Increased COX-2 levels have been shown to participate in tumor cell proliferation, inhibition of apoptosis with prolonged tumor cell survival, angiogenesis, expression of matrix metalloproteinase-2, and increased invasive potential of cancer cells, regulation of cellular adhesion, and probably others.20–23 Suppression of COX-2 expression was shown to inhibit in vivo tumorigenesis in a COX-2–expressing RCC cell line.20 The mechanism for these interactions of COX-2 with tumor promoters is not entirely known. Cyclooxygenase 2 is known to be a target gene of the ras oncogene. It has also been shown to interact with p53 and possibly to upregulate bcl-2. The COX-2 pathway of prostaglandin production, in particular prostaglandin E2, is involved in carcinogenesis as prostaglandins themselves have been shown to accelerate tumor growth by increasing cell proliferation and suppressing the immune response. Nonsteroidal anti-inflammatory drugs are successfully used in antineoplasia therapy to suppress this pathway.20–24
Several studies have shown COX-2 to be correlated with tumor stage, size, grade, and metastases, but we know of none that have shown COX-2 expression to be an independent prognostic factor for patient survival.20–24 As is the case for CD44, COX-2 has been shown to be expressed in the distal tubules of normal kidney, but not in the proximal tubules.24 Cyclooxygenase 2 also did not show a significant independent prognostic value in the current study. Platelet-derived growth factor exerts its function by binding to 3 different tyrosine kinase receptors. The α form of the PDGF receptor in particular binds PDGF with high specificity, resulting in growth stimulation through interactions with the ras oncogene.9,10 Platelet-derived growth factor and PDGFR are found in the alpha granules of platelets and in mesenchymal and epithelial cells in a variety of tissues.10 The PDGF family of growth factors and receptors includes 5 different isoforms of growth factor and 3 different isoforms of receptor. Receptor-ligand interactions of growth factors, including PDGF, have generated much interest with respect to carcinogenesis, and have been the target of anticancer therapies.
One recent study examined PDGF-AA and PDGFR-αα expression via immunohistochemistry in RCC in an attempt to identify a potential for prognosis and/or therapy related to these molecules.10 Progression-free survival with high PDGFR-αα expression was seen with univariate analysis. Multivariate analysis did not show this association. Platelet-derived growth factor receptor-αα and PDGF-AA were increased in higher grade tumors when compared with those of grades 1 and 2. The PDGFR-αα expression was found to be prognostically associated with outcome in patients with clear cell RCC.10 The current study showed an association with overall survival and PDGFR-α values by both univariate and multivariate analysis.
In a previous analysis of a larger cohort of patients, we had demonstrated that expression of p53 and Ki-67, as well as DNA ploidy, were prognostic for survival of patients with RCC, so in this study we examined whether inclusion of these factors would impact the multivariate analysis conducted for the outcome of overall survival. Specifically, we were interested in whether the finding of prognostic value for PDGFR-α when added to presence or absence of metastasis at diagnosis would be retained. Inclusion of p53, Ki-67, and ploidy status did not alter the previous result, confirming the independent nature of PDGFR-α for predicting outcome.
In the present study, we examined the immunohistochemical staining patterns of CD44, COX-2, and PDGFR-α in archival tissue from patients who had undergone partial or radical nephrectomy for RCC. We know of no previous study that examined these 3 potential biomarkers simultaneously. Platelet-derived growth factor receptor α may provide prognostic value when considered in combination with other factors for the prediction of overall survival in patients diagnosed with RCC.
References
The authors have no relevant financial interest in the products or companies described in this article.
Author notes
Reprints: Ossama Tawfik, MD, PhD, Department of Pathology and Laboratory Medicine, The University of Kansas Medical Center, 3901 Rainbow Blvd, Kansas City, KS 66160 ([email protected])