Predictive Markers and Cancer Prevention
Comparison of the prognostic potential of hyaluronic acid, hyaluronidase (HYAL-1), CD44v6 and microvessel density for prostate cancer
Article first published online: 2 JUN 2004
Copyright © 2004 Wiley-Liss, Inc.
International Journal of Cancer
Volume 112, Issue 1, pages 121–129, 20 October 2004
How to Cite
Ekici, S., Cerwinka, W. H., Duncan, R., Gomez, P., Civantos, F., Soloway, M. S. and Lokeshwar, V. B. (2004), Comparison of the prognostic potential of hyaluronic acid, hyaluronidase (HYAL-1), CD44v6 and microvessel density for prostate cancer. Int. J. Cancer, 112: 121–129. doi: 10.1002/ijc.20368
- Issue published online: 9 AUG 2004
- Article first published online: 2 JUN 2004
- Manuscript Revised: 31 MAR 2004
- Manuscript Received: 13 JAN 2004
- U.S. Department of Defense. Grant Number: DAMD 170210005
- National Cancer Institute. Grant Number: RO1 CA 072821-06A2
- American Cancer Society Florida Division
- Scientific and Technical Council of Turkey (NATO Science Fellowship Program)
- prostate cancer;
- prognostic indicator;
- hyaluronic acid;
- microvessel density
Despite the development of nomograms designed to evaluate a prostate cancer (PCa) patient's prognosis, the information has been limited to PSA, clinical stage, Gleason score and tumor volume estimates. We compared the prognostic potential of 4 histologic markers, hyaluronic acid (HA), HYAL-1-type hyaluronidase (HAase), CD44v6 and microvessel density (MVD) using immunohistochemistry. HA is a glycosaminoglycan that promotes tumor metastasis. CD44 glycoproteins serve as cell surface receptors for HA, and the CD44v6 isoform is associated with tumor metastasis. HYAL-1-type HAase is expressed in tumor cells and, like other HAases, degrades HA into angiogenic fragments. Archival PCa specimens (n = 66) were obtained from patients who underwent radical prostatectomy for clinically localized PCa and had a minimum follow-up of 72 months (range 72–131 months, mean 103 months). For HA, HYAL-1 and CD44v6 staining and MVD determination, a biotinylated HA-binding protein, an anti-HYAL-1 IgG, an anti-CD44v6 IgG and an anti-CD34 IgG were used, respectively. HA and HYAL-1 staining was classified as either low- or high-grade. CD44v6 staining and MVD were evaluated quantitatively and then grouped as either low- or high-grade. Using 72 months as the cut-off limit for evaluating biochemical recurrence, HA, HYAL-1, combined HA–HYAL-1, CD44v6 and MVD staining predicted progression with 96%, 84%, 84%, 68% and 76% sensitivity, respectively. Specificity was, 61% (HA), 80.5% (HYAL-1), 87.8% (HA–HYAL-1), 56.1% (CD44v6) and 61% (MVD). Sensitivity and specificity values for each marker did not change significantly in a subset of 45 patients for whom follow-up of longer than 112 months was available. In univariate analysis using the Cox proportional hazards model, preoperative PSA, Gleason sum, margin status, seminal vesicle, extraprostatic extension (EPE), HA, HYAL-1, HA–HYAL-1 and MVD, but not CD44v6, age and clinical stage, were significant in predicting biochemical recurrence (p < 0.05). In multivariate analysis using stepwise selection, only preoperative PSA (hazard ratio/unit PSA change = 1.086, p < 0.0001), EPE (hazard ratio = 6.22, p = 0.0016) and HYAL-1 (hazard ratio = 8.196, p = 0.0009)/HA–HYAL-1 (hazard ratio = 5.191, p = 0.0021) were independent predictors of biochemical recurrence. HA was an independent predictor of prognosis if HYAL-1 staining inference was not included in the multivariate model. In our retrospective study with 72- to 131-month follow-up, EPE, preoperative PSA and HYAL-1 either alone or together with HA (i.e., combined HA–HYAL-1) were independent prognostic indicators for PCa. © 2004 Wiley-Liss, Inc.
Over the last decade, the number of curable PCa cases has significantly increased due to the widespread use of PSA.1, 2 However, despite careful selection of patients, the disease recurs in a substantial percentage of localized PCa cases undergoing curative treatment modalities (i.e., radical prostatectomy and radiotherapy).3, 4, 5, 6 Accurate prediction of the risk of progression would be useful in choosing the type and timing of the most appropriate treatment. Although existing parameters, such as Gleason sum or preoperative PSA, provide some prognostic information, it is difficult to estimate prognosis in PCa patients since two-thirds of them have Gleason sum of 5–7 and serum PSA levels of 4–10 ng/ml.6, 7, 8, 9, 10, 11 Furthermore, all patients with the same pathologic stage and/or grade do not have the same prognosis. Thus, there is a need for accurate prognostic markers to identify the biologic behavior of the tumor. Previously, we showed that HYAL-1-type HAase, either alone or in combination with HA, appears to be an independent predictor of biochemical recurrence among radical prostatectomy patients.12
HA is a nonsulfated glycosaminoglycan made up of repeated disaccharide units, D-glucuronic acid and N-acetyl-D-glucosamine.13 HA maintains the osmotic balance of tissues and regulates cellular processes such as adhesion, migration and proliferation.13 The biologic functions of HA are mediated by different HA receptors, including CD44.14 The concentration of HA is elevated in several tumor types, and in some tumors (e.g., breast, colon), high-level HA expression in tumor-associated stroma and/or tumor cells predicts poor survival.15, 16, 17, 18, 19 Increased urinary HA levels serve as an accurate diagnostic marker for bladder cancer, regardless of tumor grade and stage.20 However, in PCa, HA is not an independent predictor for prognosis.12, 19
HYAL-1-type HAase is present in serum and produced by bladder, prostate and head-and-neck cancer cells.21, 22, 23, 24 The HAase class of enzymes degrades HA into small fragments, some of which (3–25 disaccharide units) induce angiogenesis.25, 26 Angiogenic HA fragments stimulate endothelial cell proliferation, adhesion and migration by activating focal adhesion kinase and mitogen-activated protein kinase pathways.26 We have previously shown the presence of angiogenic HA fragments in PCa tissues and in the urine and saliva of bladder and head-and-neck cancer patients, respectively.22, 24, 27 Given the observations that HYAL-1 is the major tumor-derived HAase expressed and secreted by tumor cells and that it is active at pH ≤ 4.5, it is possible that in the tumor microenvironment, where the pH is acidic, secreted HYAL-1 degrades the HA present in the extracellular matrix into angiogenic HA fragments.22, 24, 27, 28 In a retrospective study with a minimum of 5-year follow-up, we showed that HYAL-1 staining predicts progression with 84% sensitivity and 80% specificity.12 Furthermore, high HYAL-1 staining was an independent predictor for prognosis.
CD44 belongs to a family of cell surface transmembrane glycoproteins involved in cell-to-cell and cell-to-extracellular matrix interactions.29, 30 Alternative splicing of CD44 mRNA in 10 of the 20 exons generates several variant CD44 isoforms.30, 31 The standard form of CD44 (i.e., CD44s) is an HA receptor expressed in a variety of normal and tumor cell types.29, 32 We have previously shown that an isoform of CD44 (ex14/v10) is involved in HA-mediated endothelial cell proliferation.33 The correlation between tumor progression and CD44s and/or its isoforms is unclear. We and others have shown that the androgen-insensitive PCa line PC-3 and primary PCa cells express CD44s and CD44 variants (e.g., CD44v3 and CD44v6); however, the androgen-sensitive, poorly metastatic line LNCaP does not express CD44.34, 35, 36 Contrary to these findings, it has been shown that overexpression of CD44v6 in a rat PCa line decreases metastasis.37 Ekici et al.38 showed that decreased expression of CD44v6 could be a predictor of poor prognosis in clinically localized PCa. Aaltomaa et al.39 reported similar results.
Angiogenesis is an essential process for tumor growth and metastasis.40, 41, 42 The clinical significance of angiogenesis, measured as MVD, has been demonstrated for several tumor types, including gastrointestinal, breast, bladder and renal cell carcinomas.43, 44, 45, 46 Studies that compared various endothelial cell markers (i.e., CD31, CD34 and factor VIII) have shown that CD34 is a sensitive endothelial cell marker for measuring MVD.47, 48 At the present time, the clinical significance of MVD as an independent predictor of pathologic stage and recurrence in PCa remains unclear.47, 48, 49, 50, 51, 52
We compared the prognostic potential of markers HA, CD44v6, HYAL-1 and MVD with regard to clinically localized PCa. Since HYAL-1 degrades HA and generates angiogenic fragments and CD44 acts as a cell surface receptor for both HA and HA fragments, we examined whether these biologically linked molecules are accurate prognostic indicators for PCa and whether they influence each other's prognostic capability.
MATERIAL AND METHODS
Specimens and study patients
Sixty-six randomly selected PCa specimens were obtained from patients who underwent radical retropubic prostatectomy for clinically localized PCa between 1992 and 1995 at the University of Miami. Patients neither received neoadjuvant hormonal therapy nor had metastasis to regional pelvic lymph nodes. The minimum available follow-up on all patients was 72 months. The study was conducted under a protocol approved by the University of Miami's Institutional Review Board. Of the 66 patients, 25 had biochemical or clinical recurrence before 72 months (mean time to recurrence 21.3 months, range 3–61) and 41 were free of disease recurrence (mean follow-up 103 months, range 72–131). Biochemical recurrence was defined as a PSA level ≥ 0.4 ng/ml in 2 successive measurements after the operation, in which case the first date an elevated PSA level was recorded was considered the date of failure. Patient characteristics including age, preoperative PSA and tumor (i.e., Gleason sum, stage, margin, EPE and seminal vesicle invasion) are shown in Table I.
|Progression||Preoperative parameters||Postoperative parameters|
|Age (years)||PSA (ng/ml)||Clinical stage||Gleason sum||EPE||Margin||Seminal vesicle invasion|
|Biochemical recurrence (n = 25)||Median: 64||Median: 9.0||T1c: 10||6 = 2||(+) = 21||(+) = 18||(+) = 14|
|Mean: 65.1||Mean: 14.04||T2a: 5||7 = 14||(−) = 4||(−) = 7||(−) = 11|
|T2b: 10||8 = 6|
|9 = 3|
|No biochemical or clinical recurrence (n = 41)||Median: 65||Median: 6||T1c: 22||5 = 7||(+) = 4||(+) = 9||(+) = 3|
|Mean: 62.98||Mean: 8.1||T2a: 5||6 = 9||(−) = 37||(−) = 32||(−) = 38|
|T2b: 14||7 = 20|
|8 = 5|
IHC and slide grading
For all specimens, paraffin-embedded blocks containing PCa tissues representing the major Gleason score were selected by a pathologist. From each block, 8 slides were prepared. Four slides were used for HA, HYAL-1, CD44v6 and anti-CD34 (for MVD determination) staining. The remaining slides were used either for determining nonspecific staining corresponding to each staining reagent or for repeating the staining to evaluate its consistency. For all staining procedures, specimen slides were deparaffinized, rehydrated and treated with an antigen retrieval solution (Dako, Copenhagen, Denmark).
HA and HYAL-1 staining.
IHC for localizing HA and HYAL-1 in PCa tissues was carried out as described previously.12, 22 HA was localized in PCa tissues using a biotinylated HA-binding protein purified from bovine nasal cartilage, as described previously by Tengblad.53 HYAL-1 was localized using a rabbit polyclonal anti-HYAL-1 IgG, which was generated against a peptide sequence present in the HYAL-1 protein (amino acids 321–338).12, 22
Staining for HA and HYAL-1 was graded as 0 (no staining), 1+, 2+ and 3+. For HA staining, both the tumor-associated stroma and tumor cells were graded in each slide. The overall staining grade for each slide was assigned based on the staining intensity of the majority of the tumor tissue in the specimen. However, if 50% of the tumor tissue stained as 1+ and the other 50% as 3+, the overall staining grade was 2+. If 50% of the tumor stained as 2+ and the remaining as 3+, the overall staining inference was assigned as 3+. The staining scale was further subcategorized into low- and high-grade. For HA staining, low-grade staining included 0, 1+ and 2+ staining and high-grade staining included 3+ intensity. For HA staining, high-grade staining in tumor-associated stroma and/or tumor cells was considered as high-grade. HA staining was graded as low only when both the tumor-associated stroma and tumor cells showed low-grade staining. Therefore, in cases (n = 2) where stromal tissues showed low-grade staining but the tumor cells stained as 3+, the overall HA staining was considered as high-grade. For HYAL-1, high-grade staining represented 2+ and 3+ staining, whereas low-grade staining included 0 and 1+ staining. For the combined HA–HYAL-1 staining, a positive result was indicated only when both HA (stromal, tumor cells or both) and HYAL-1 staining intensities were of high grade. Any other combination was considered negative. All slides were reviewed out of order, to prevent direct comparison of individual cases for HA and HYAL-1. Two readers independently evaluated all slides in a blinded fashion. Of the 132 total slides (i.e., 66 each for HA and HYAL-1), there was discrepancy in 5 HA slides and 4 HYAL-1 slides. These discrepancies were resolved by both readers reexamining the slides simultaneously. In addition, to check for the repeatability of the evaluation system, a third reader randomly picked 35 slides each from the HA and HYLA1 sets and graded them for staining intensity. The discrepancy in slide evaluations by the third reader was <10%.
Following the antigen retrieval step (as described above), slides were incubated with a mouse antihuman hematopoietic progenitor cell CD34 MAb (dilution 1:20; Dako) at 4°C for 15 hr. Slides were then incubated with a biotinylated antimouse antibody and an avidin-peroxidase conjugate solution (Vectastain ABC Kit; Vector, Burlingame, CA). To visualize peroxidase binding sites, slides were incubated with a DAB chromogen substrate solution (Dako) for 10 min. Slides were counterstained with hematoxylin, dehydrated and mounted.
The method described by Weidner et al.49 was used for scoring the microvessels stained with CD34. The area of the highest MVD in each tissue specimen was localized under ×40 magnification and designated as a “hot spot”. Microvessels in the hot spots were counted under ×400 magnification. Any vessel with lumen and endothelial cells or an endothelial cell cluster stained positively for CD34 was considered to be a single countable microvessel. MVD count was defined as the mean value of the counts obtained in 3 separate, contiguous but not overlapping areas within the hot spot. A cut-off value was determined using the ROC curve, and according to this value, 2 groups (low and high MVD) were assigned. Microvessels were examined and counted by the 3 readers (S.E., V.B.L. and W.H.C.) independently and without the knowledge of the clinical and pathologic status of the patients. Sections were reviewed out of order, to prevent direct comparison of individual cases for CD34.
Following antigen retrieval, slides were incubated with a mouse antihuman CD44v6 MAb (dilution 1:50; Bender Med Systems, Vienna, Austria) at 4°C for 15 hr. Sections were then incubated with a biotinylated secondary antibody and an avidin-peroxidase conjugate solution (Vectastain ABC Kit). To visualize peroxidase binding sites, slides were incubated with DAB chromogen substrate solution for 10 min. Slides were counterstained with hematoxylin, dehydrated and mounted.
Slides for CD44v6 were scored as described by Ekici et al.38 All sections included normal prostate tissue and/or benign prostatic hyperplasia glands as internal controls. Intensity of staining was graded as 0 for no staining, 1 for weak, 2 for moderate and 3 for strong. A combined staining score based on an estimate of the percentage of tumor cells stained and the intensity of staining was developed. Areas of tumor cells stained with maximum intensity (primary area) and with lesser intensity (secondary area) were determined in percentage values. The combined score was obtained by adding the scores of the primary and secondary areas. Staining intensities were examined and scored by 2 readers (S.E., V.B.L.) independently and in a blinded fashion. A cut-off value was determined from the ROC curve, and according to this value, 2 groups (low and high CD44v6 staining) were assigned.
Interassay variability regarding staining intensity was determined by Pearson's correlation analysis. Spearman's bivariate correlation coefficients were 0.85, 0.9, 0.98 and 0.95 for HA, HYAL-1, CD34 and CD44v6 staining, respectively. For all markers, high-grade staining was considered to be a true positive if the patient had biochemical recurrence. Consequently, low-grade staining was considered to be a true negative if the patient had no biochemical recurrence. The sensitivity, specificity, accuracy, PPV and NPV for HA, HYAL-1, HA–HYAL-1, CD34 and CD44v6 staining inferences were calculated using a 2 × 2 contingency table (high-grade/low-grade staining and progressed/nonprogressed PCa patients) at 72, 84, 100 and 112 month cut-off limits. For CD44v6 and MVD, ROC curves were developed for determining the optimal cut-off limits that yielded the best possible sensitivity and specificity values. The cut-off limits for CD44v6 and MVD were 180 and 41, respectively. Sensitivity was defined as true positive (i.e., number of recurred patients predicted by a marker/total number of recurred patients). Specificity was defined as true negative (i.e., number of nonrecurred patients predicted by a marker/total number of nonrecurred patients). Accuracy was determined as follows: (number of true positives + number of true negatives)/total number of PCa patients. PPV was determined as follows: number of true positives/(number of true positives + number of false positives). NPV was determined as follows: number of true negatives/(number of true negatives + number of false negatives). Data on various biochemical, surgical and pathologic parameters, as well as HA, HYAL-1, HA–HYAL-1, CD34 and CD44v6 staining inferences, were analyzed by the Cox proportional hazards model, using single-variable analysis (univariate analysis) or step-wise selection analysis. Stratified Kaplan-Meier analyses were performed on the variables found to be significant in the multivariate Cox proportional hazards model. For PSA subset analysis, Mantel-Haenszel χ2 analysis or Student's t-test were used to determine statistical significance. Statistical analysis was carried out using the SAS software program (version 8.02; SAS Institute, Cary, NC).
IHC of tissue markers
The HA, HYAL-1, CD44v6 and CD34 antigens were localized in 66 archival PCa specimens obtained from patients who underwent radical retropubic prostatectomy for clinically localized disease. An increase in PSA levels ≥0.4 ng/ml was taken as an indicator of biochemical recurrence. Figure 1 shows IHC localization of HA, HYAL-1, CD44v6 and MVD in 2 Gleason 7 PCa specimens, one each from a nonrecurred (Fig. 1a,c,e,g) and a recurred (Fig. 1b, d,f,h) patient.
As shown in Figure 1a, very little HA staining was seen in PCa tissue from a patient who did not progress within 72 months. Among the 41 PCa specimens from nonrecurred patients, 25 showed low-grade staining. Figure 1b shows high-grade HA staining in a PCa specimen from a patient who had biochemical recurrence before 72 months (median time to recurrence 19 months, mean 21.3 months). HA staining was seen mainly in tumor-associated stroma. However, high-grade HA staining was also seen in tumor cells in 8 of 25 specimens from patients who had biochemical recurrence. Among these 8 specimens, 6 showed high-grade staining in tumor-associated stroma. Of the 25 patients who had recurred, 24 showed high-grade HA staining and only 1 showed low-grade staining in both tumor-associated stroma and tumor cells.
An anti-HYAL-1 peptide IgG was used to localize HYAL-1. As shown in Figure 1c, little HYAL-1 staining was seen in the PCa tissue from a nonrecurred patient. Of the 41 nonrecurred patients, PCa specimens from 33 had low-grade staining. In the PCa specimen from a patient who later recurred, high-grade HYAL-1 staining was seen (Fig. 1d). HYAL-1 expression was seen exclusively in tumor cells. Of the 25 patients who recurred within 72 months, 21 had high-grade HYAL-1 staining.
CD44v6 was localized using an anti-CD44v6 mouse MAb. Contrary to some earlier reports,38, 39 low-grade CD44v6 staining was observed in the PCa specimen from a nonrecurred patient (Fig. 1e) and high-grade CD44v6 staining, in the PCa tissue from a recurred patient (Fig. 1f). CD44v6 staining was mostly associated with the plasma membrane of tumor cells. We also observed CD44v6 in non-neoplastic epithelial cells in normal prostate and benign prostatic hyperplasia glands. However, the staining intensity of CD44v6 in non-neoplastic cells was less than that in tumor cells. There was a great degree of heterogeneity in CD44v6 staining. For these reasons, we used a semiquantitative method to grade CD44v6 staining.38 Using a cut-off limit of 180 on the scoring scale, 23 of 41 PCa specimens from nonrecurred patients showed low-grade staining, whereas of the 25 patients who recurred, 17 showed high-grade staining.
It has been shown that visualization and scoring of microvessels using anti-CD34 staining are both sensitive and specific.47, 48, 54 We therefore used an anti-CD34 MAb to visualize microvessels in PCa tissues. As shown in Figure 1g, MVD was low in the PCa tissue from a nonrecurred patient. As determined from the ROC curve, a cut-off limit of 41 was set to score low or high MVD. Of the 41 nonrecurred patients, PCa tissues from 25 patients had low MVD. However, MVD was high in 19 of 25 PCa tissues obtained form patients who had a recurrence. Figure 1h shows high MVD in the PCa specimen from a patient who later recurred.
Determination of sensitivity, specificity, accuracy, PPV and NPV
In all patients, a minimum 72-month follow-up was available (mean 103 months, median 104.2 months, range 72–131 months). Therefore, we determined the sensitivity, specificity, accuracy, PPV and NPV of HA, HYAL-1, combined HA–HYAL-1, CD44v6 and MVD at 72, 84, 100 and 112 months of follow-up. As shown in Table II, at 72 months the sensitivity of HA, HYAL-1, combined HA–HYAL-1, CD44v6 and MVD for predicting PCa recurrence was 96%, 84%, 84%, 76% and 68%, respectively. The specificity of HA (61%), CD44v6 (56.1%) and MVD (61%) was lower than that of HYAL-1 (80.5%) and combined HA–HYAL-1 (87.8%). Accuracy was highest for HA–HYAL-1 (86.4%), followed by HYAL-1 (81.8%), HA (74.2%), MVD (66.7%) and CD44v6 (57.6%). Due to higher specificity, the PPV of combined HA–HYAL-1 (80.8%) and HYAL-1 (70%) was high. However, the PPV of CD44v6 was the lowest (48.6%), followed by MVD (54.3%) and HA (60%). Due to high sensitivity, the NPV of HA staining (96.1%) was the highest, followed by HA–HYAL-1 (90%), HYAL-1 (89.3%), MVD (80.6%) and CD44v6 (74.2%) (Table II).
|Parameter||HA, HYAL-1 and HA–HYAL-1||H, HYAL-1 and HA–HYAL-1||CD34 and CD44v6|
|HA (%)||HYAL-1 (%)||HA-HYAL-1 (%)||HA-HYAL-1 (%)||MVD (%)||CD44v6 (%)|
|72 months||84 months||100 months||112 months||72 months||84 months||100 months||112 months||72 months||84 months||100 months||112 months||72 months||84 months||100 months||112 months||72 months||84 months||100 months||112 months|
|Sensitivity||96 (24/25)||96 (24/25)||92.3 (24/26)||92.6 (25/27)||84 (21/25)||84 (21/25)||84.6 (22/26)||85.2 (23/27)||84 (21/25)||84 (21/25)||80.8 (21/26)||81.5 (22/27)||76 (19/25)||76 (19/25)||76.9 (20/26)||77.8 (21/27)||68 (17/25)||68 (17/25)||65.4 (17/26)||62.9 (17/27)|
|Specificity||61 (25/41)||61.1 (22/36)||65.4 (17/26)||80.6 (16/18)||80.5 (33/41)||80.6 (29/36)||84.6 (22/26)||94.4 (17/18)||87.8 (36/41)||88.9 (32/36)||88.5 (23/26)||94.4 (17/18)||61 (25/41)||61.1 (22/36)||65.4 (17/26)||77.8 (14/18)||56.1 (23/41)||52.8 (19/36)||50 (13/26)||61.1 (11/18)|
|Accuracy||74.2 (49/66)||75.4 (46/61)||78.8 (41/52)||91.1 (41/45)||81.8 (54/66)||82 (50/61)||84.6 (44/52)||88.9 (40/45)||86.4 (57/66)||86.9 (53/61)||84.6 (44/52)||86.7 (39/45)||66.7 (44/66)||67.2 (41/61)||71.1 (37/52)||77.8 (35/45)||57.6 (38/66)||59 (36/61)||57.7 (30/52)||62.2 (28/45)|
|PPV||60 (24/40)||63.2 (24/38)||77.4 (24/31)||92.6 (25/27)||70 (21/30)||75 (21/28)||84.6 (22/26)||95.8 (23/24)||80.8 (21/26)||84 (21/25)||87.5 (21/24)||95.7 (22/23)||54.3 (19/35)||57.6 (19/33)||69 (20/29)||84 (21/25)||48.6 (17/35)||50 (17/34)||56.7 (17/30)||70.8 (17/24)|
|NPV||96.1 (25/26)||95.7 (22/23)||89.5 (17/19)||80.6 (16/18)||89.3 (34/37)||87.9 (29/33)||84.6 (22/26)||81 (17/21)||90 (36/40)||88.9 (32/36)||85.2 (23/27)||77.3 (17/22)||80.6 (25/31)||78.6 (22/28)||73.4 (17/23)||70 (14/20)||74.2 (23/31)||70.4 (19/27)||59 (13/22)||52.4 (11/21)|
At 84-month follow-up, the cohort consisted of 61 patients (mean follow-up 107.9 months, median 112 months, range 85–131 months), of whom 36 were in the nonrecurred group. Thus, the sensitivity values of all markers remained unchanged at 84 months compared to 72 months (Table II). There was also no significant change in the specificity, accuracy, PPV and NPV values for HA (61.1%, 75.4%, 63.2%, 95.7%), HYAL-1 (80.6%, 82%, 75%, 87.9%), combined HA–HYAL-1 (88.9%, 86.9%, 84%, 88.9%), MVD (61.1%, 67.2%, 57.6%, 78.6%) and CD44v6 (52.8%, 59%, 50% and 70.4%), respectively. At 100 months, follow-up information was available on 52 patients (mean follow-up 117.1 months, median 117.8 months, range 101.6–131 months). Of these 52 patients, one who had been in the nonrecurred category up to the 84-month follow-up showed biochemical recurrence. Interestingly, this patient was scored as a false positive on HYAL-1 and CD34 at 72- and 84-month follow-up. Thus, the sensitivity of both HYAL-1 (84.6%) and CD34 (76.9%) increased slightly, whereas that of HA (92.3%), combined HA–HYAL-1 (80.8%) and CD44v6 (65.4%) decreased (Table II).
Follow-up information beyond 112 months was available for 45 patients (mean follow-up 121 months, median 120.2 months, range 112–131 months). At 112 months, one patient who was a false positive on HA, HYAL-1, combined HA–HYAL-1 and MVD markers up to 100 months showed biochemical recurrence. Thus, the sensitivity of HA, HYAL-1, combined HA–HYAL-1, MVD and CD44v6 was 92.6%, 85.2%, 81.5%, 77.8% and 62.9%, respectively. At final analysis, both HYAL-1 and combined HA–HYAL-1 had the best specificity (94.4%, 94.4%), accuracy (88.9%, 86.7%), PPV (95.8%, 95.7%) and NPV (81%, 77.3%), followed by HA (80.6%, 91.1%, 92.6%, 80.6%), MVD (77.8%, 77.8%, 84%, 70%) and CD44v6 (61.1%, 62.2%, 70.8%, 52.4%).
Evaluation of the prognostic capability of pre- and postoperative parameters and histologic markers
Since the patients in this cohort had variable follow-up between 72 and 131 months, we used the Cox proportional hazards model and single-parameter analysis to determine the prognostic significance of each of the preoperative (i.e., age, PSA and clinical stage) and postoperative (i.e., Gleason sum, margin, EPE, seminal vesicle invasion) parameters, as well as staining inferences of HA, HYAL-1, combined HA–HYAL-1, CD44v6 and MVD. As shown in Table III, age (p = 0.5104, hazard ratio = 1.019), clinical stage (p = 0.2683, hazard ratio = 1.2620) and CD44v6 staining (p = 0.131, hazard ratio = 1.826) were not significant in predicting biochemical recurrence. However, preoperative PSA (p = 0.0006, hazard ratio/unit PSA change = 1.048), Gleason sum overall (p = 0.0002, hazard ratio = 2.5), margin status (p = 0.0003, hazard ratio = 4.5), EPE (p < 0.0001, hazard ratio = 12.781), seminal vesicle invasion (p < 0.0001, hazard ratio = 6.56), HA staining (p = 0.0008, hazard ratio = 12.091), HYAL-1 staining (p < 0.0001, hazard ratio = 13.192), HA–HYAL-1 staining (p < 0.0001, hazard ratio = 10.749) and MVD (p = 0.0015, hazard ratio = 4.36) significantly predicted biochemical recurrence (Table III). Patients with Gleason sum ≥7 have a greater risk of progression.42 In single-parameter analysis, the hazard of developing biochemical recurrence in Gleason sum ≥7 patients (p = 0.0165, hazard ratio = 5.827) increased 2.3-fold when all Gleason sums were analyzed together (Table III).
|Parameter||χ2||p||Hazard ratio||95% CI1|
|Gleason sum (overall)||14.19||0.00021||2.5||1.552–4.024|
|Surgical margin positivity||13.355||0.00031||4.5||2.008–10.079|
|Seminal vesicle invasion||22.268||<0.0001||6.56||3.002–14.317|
To determine the smallest number of variables that could jointly predict biochemical recurrence in this cohort of patients, we used the Cox proportional hazards model and step-wise selection analysis. When age, preoperative PSA, clinical stage, Gleason sum (overall or ≥7), EPE, seminal vesicle invasion and staining inferences of HA, HYAL-1, CD44v6 and MVD were included in the model, only preoperative PSA (p < 0.0001, hazard ratio/unit PSA change = 1.086), EPE (p = 0.0016, hazard ratio = 6.222) and HYAL-1 (p = 0.0009, hazard ratio = 8.1896) reached statistical significance in predicting biochemical recurrence (Table IV).
|Parameters||HA and HYAL-1 separate||HA–HYAL-1 combined|
|χ2||p||Hazard ratio||95% CI1||χ2||p||Hazard ratio||95% CI|
To demonstrate the joint effect of HYAL-1 and EPE or HYAL-1 and PSA on biochemical recurrence, we performed Kaplan-Meier analysis. As shown in Figure 2a, the probability of biochemical recurrence was highest when HYAL-1 was high and EPE was positive, and a patient had the lowest probability of recurrence when HYAL-1 was low and EPE was negative. Since PSA was a continuous estimate, with values ranging from 0.5 to 62 ng/ml, for the entire cohort (n = 66), we divided the cohort into those with PSA levels <7 and >7 ng/ml, 7 ng/ml PSA being used as the cut-off limit since that was the median value for the entire cohort. As shown in Figure 2b, individuals with HYAL-1 high and PSA > 7 ng/ml had the highest probability of recurrence, followed by those with HYAL-1 high and PSA < 7 ng/ml. Individuals with low HYAL-1 staining and PSA < 7 ng/ml had the lowest probability of recurrence. These data explain why multivariate analysis selected HYAL-1, EPE and PSA as independent prognostic indicators.
Inclusion of the combined HA–HYAL-1 staining inference instead of HA and HYAL-1 staining inferences in the multiple regression model again showed that preoperative PSA (p = 0.0002, hazard ratio/unit PSA change = 1.077), EPE (p = 0.0009, hazard ratio = 6.906) and HA–HYAL-1 (p = 0.0021, hazard ratio = 5.191) were significant in predicting biochemical recurrence (Table IV). None of the other preoperative (PSA, clinical stage) and postoperative (Gleason sum overall or Gleason stratification ≥ or ≤ 7 and seminal vesicle) parameters or CD44v6 and MVD staining inferences reached statistical significance in the multivariate model (p > 0.05 in each case). Kaplan-Meier analysis using HA–HYAL-1 and EPE or HA–HYAL-1 and PSA demonstrated that individuals with high HA–HYAL-1 and positive EPE or high HA–HYAL-1 and PSA > 7 ng/ml had the highest probability of biochemical recurrence (data not shown).
When HYAL-1 was omitted in the model during stepwise analysis, HA (p = 0.0065, hazard ratio = 8.658) together with preoperative PSA (p = 0.0006, hazard ratio/unit PSA change = 1.079) and EPE (p < 0.0001, hazard ratio = 9.073) were significant in predicting biochemical recurrence. Similarly, when PSA was omitted in the model, margin status reached independent prognostic significance together with EPE and HYAL-1 or HA–HYAL-1 (data not shown).
PSA subgroup analysis.
It has been suggested that biochemical recurrence before 24 months indicates systemic disease, whereas biochemical recurrence beyond 24 months suggests local recurrence. To test whether any of the pre- and postoperative parameters as well as IHC markers under study distinguish between these groups, we performed Mantel-Haenszel χ2 analysis (for testing Gleason sum, Gleason sum ≥ 7, EPE, margin status, seminal vesicle invasion, HA, HYAL-1, combined HA–HYAL-1 CD34 and CD44v6) or Student's t-test (for age and preoperative PSA). As shown in Table V, margin status could distinguish between PSA recurrence before and after 24 months (p = 0.269, χ2 = 4.894); however, none of the other parameters or markers reached statistical significance in this comparison.
|Gleason sum (overall)||0.1332||0.453|
|Gleason ≥ 7||0.152||0.699|
|Surgical margin positivity||4.8942||0.0269*|
|Seminal vesicle invasion||0.1222||0.7265|
We compared the prognostic potential of histologic markers HA, HYAL-1, CD44v6 and MVD for predicting biochemical recurrence in PCa patients since their biologic functions are interrelated. For example, HA, an extracellular matrix component, is a high-affinity ligand for CD44.14, 29 HA–CD44 interaction promotes cell adhesion, migration and proliferation.14, 29 HAase degrades HA into small fragments that promote angiogenesis, and MVD is an indicator of angiogenesis.25, 26, 41 In addition to their biologic relatedness, each of these histologic markers has shown potential to predict prognosis for PCa patients.12, 19, 38, 39, 47, 48, 49, 50, 51, 52
The prognostic capability of HA staining in tumors varies depending on the tissue where the cancer originates. HA staining in tumor-associated stroma and/or tumor cells has prognostic capability in breast, colon and gastrointestinal cancers; however, it does not have independent prognostic capability in PCa.13, 14, 15, 16, 17, 18, 19 We found that the HA staining had 92.3% sensitivity and 80.6% specificity to predict biochemical recurrence within 112 months. Interestingly, the specificity of HA staining to predict biochemical recurrence increased from 61% at 72-month follow-up to 80.6% at 112-month follow-up. We have previously shown that at 64-month follow-up, although HA staining had high sensitivity (96%), the specificity of this marker to predict biochemical recurrence was even lower (55.5%) compared to that at 72 months.12 These results indicate that positive HA staining in PCa tissues means that the patient could have a recurrence within 112 months. In this and a previous study, we found that HA staining shows prognostic capability in univariate analysis; however, it is not an independent predictor of biochemical recurrence. Interestingly, however, if HYAL-1 staining inference is not included in the Cox proportional hazards model during stepwise analysis, HA together with preoperative PSA and EPE reaches independent prognostic significance. These results indicate that HYAL-1 provides all of the prognostic information supplied by HA staining, as well as some additional information. However, although there was <10% discrepancy between 3 readers who evaluated the HA and HYAL-1 staining, at present we do not know the repeatability of the staining evaluation system in another laboratory. Therefore, more studies need to be conducted to verify that HYAL-1 is a better prognostic indicator than HA in predicting biochemical recurrence for PCa patients. Nonetheless, the evaluation system, which involves grading of HA staining as high when tumor-associated stroma and/or tumor cells show high-grade staining and as low when tumor-associated stroma and/or tumor cells show low-grade staining, appears to be accurate. This is because the prognostic significance of HA staining, and consequently of HA–HYAL-1 staining, in both univariate and multivariate analyses remained unchanged when the 2 specimens in which HA staining was low-grade in tumor-associated stroma but high-grade in tumor cells were graded as low-grade instead of high-grade (unpublished results).
The prognostic significance of the CD44 standard form and its variant isoforms (e.g., CD44v6) is unclear. Contrary to our finding that CD44v6 expression was elevated in patients who later had biochemical recurrence, 2 reports showed that a decrease in CD44v6 expression correlated with increased Gleason sum and disease progression.38, 39 For example, Ekici et al.38 showed that CD44v6 expression inversely correlates with pathologic stage and disease progression and positively correlates with PSA-free survival. However, in that study, CD44v6 was not an independent predictor of prognosis. Contrary to the findings of Ekici et al.,38 Aaltomaa et al.39 found that CD44v6 is an independent predictor of survival. In the present study, increased CD44v6 expression had reasonable sensitivity to predict prognosis at 72-month (68%) or 112-month (62.9%) follow-up. However, among all of the markers, it had the lowest specificity (56.1% at 72 months and 61.1% at 112 months) to predict biochemical recurrence. This low specificity may explain why CD44v6 staining is not significant in predicting biochemical recurrence in both univariate and multivariate analyses. A likely explanation of why different studies report conflicting results regarding CD44v6 staining and prognosis for PCa is that PCa tissues have a high degree of heterogeneity with respect to CD44v6 staining. Aaltomaa et al.39 reported variability in CD44v6 staining intensity and in the number of tumor cells that were positive for CD44v6. Similarly, Ekici et al.38 reported the heterogeneous nature of CD44v6 expression and developed a semiquantitative method for scoring CD44v6 staining. We used this scoring method to evaluate CD44v6 staining. Nonetheless, it is likely that the heterogeneous nature of CD44v6 will limit its prognostic significance.
Determination of MVD in PCa tissues, using anti-CD34 antibodies, is sensitive and accurate in predicting prognosis.47, 48, 54 Although problems exist in the methods of counting the vessels and in setting a universally accepted cut-off limit for MVD to predict recurrence, MVD has been correlated with Gleason sum, pathologic stage and outcome.47, 48, 49 However, other studies have shown that MVD does not correlate with tumor grade, stage and clinical outcome.50, 51, 52 In our study, MVD at a cut-off limit of 41 showed reasonably high sensitivity at both 72-month (76%) and 112-month (77.8%) follow-up. The specificity of MVD to predict biochemical recurrence increased from 61% at 72-month follow-up to 77.8% at 112-month follow-up, suggesting that in some false-positive patients high MVD may be indicative of biochemical recurrence before 112 months. Although in the univariate analysis MVD showed prognostic significance (Table III), in the multivariate analysis it had no additional prognostic significance. MVD did not reach independent prognostic significance even when HA alone, in the absence of HYAL-1, was included in the Cox model. It is possible that since HA, HYAL-1 and MVD are biologically related, all of the prognostic information provided by MVD inferences is contained in either HA or HYAL-1 staining inferences.
In our study, HYAL-1 staining alone had high sensitivity both at 72-month (84%) and at or beyond 112-month (85.2%) follow-up. Indeed, it had either the same or slightly higher sensitivity to predict biochemical recurrence as the combined HA–HYAL-1 staining inference (84% at 72 months and 81.5% at 112 months). The specificity of HYAL-1 staining inference (80.5%) at 72-month follow-up was slightly lower than that of HA–HYAL-1 (87.8%). However, HYAL-1 and HA–HYAL-1 staining inferences had the same specificity (94.4%) to predict biochemical recurrence at 112 months and beyond. Therefore, HYAL-1 either has the same or slightly better PPV and NPV to predict biochemical recurrence. Thus, contrary to our earlier report that combined HA–HYAL-1 has slightly better prognostic capability than HYAL-1 staining alone at 64 months,12 our present study suggests that HYAL-1 alone is sufficiently accurate to predict biochemical recurrence at 72 months and beyond.
In the multivariate analysis, among all of the pre- and postoperative parameters and histologic markers, only preoperative PSA and EPE had additional prognostic significance if HYAL-1 (or HA–HYAL-1) was included in the analysis. In an earlier study, we found that EPE, margin status and HYAL-1 (or HA–HYAL-1) were independent predictors of prognosis. The difference between that study12 and this is that in the earlier study we used 64-month follow-up as a cut-off limit and performed Wald's forward stepwise regression analysis, whereas in the present study we used the Cox proportional hazards model and stepwise selection to calculate the hazard of biochemical recurrence over the entire period of follow-up (i.e., up to 131 months). Interestingly, when preoperative PSA was not included in the model, margin status reached independent prognostic significance together with EPE and HYAL-1 (or HA–HYAL-1), suggesting that preoperative PSA provides all of the prognostic information related to margin status plus some additional information. Nonetheless, HYAL-1 appears to be an independent prognostic indicator for predicting biochemical recurrence.
As is the case for HA expression in tumor tissues, the prognostic significance of HYAL-1 expression may also vary based on the origin of cancer tissue. For example, in bladder cancer, HYAL-1 expression correlates with tumor grade.21 Also, HAase levels are elevated in brain metastases of carcinomas compared to primary glioblastomas.55 Furthermore, brain metastasis-derived cell lines have 1,000-fold more HAase than glioma-derived cell lines.55 HYAL-1 levels are also elevated in the saliva of patients with squamous cell carcinoma of the head and neck.24 However, HA accumulation without HAase activation has been associated with aggressiveness of ovarian cancer.56 Similarly, HYAL-1 expression suppresses tumor growth in a rat colon carcinoma model.57 However, HYAL-1 expression in PC-3 cells (a PCa cell line) increases their metastatic potential.58 Thus, HYAL-1 expression is associated with PCa progression.
The major dilemma for clinicians in the management of PCa is the identification of the site of disease recurrence, which ultimately guides therapy decisions. It is generally accepted that PSA recurrence within 1–2 years relates to a higher risk of developing metastatic disease.9, 59 However, it is not understood whether existing pre- and postoperative parameters as well as histologic markers can distinguish between patients who will recur within 24 months and those who will not. In our study, PSA subset analysis showed that, except for margin status, none of the markers (i.e., HA, HYAL-1, HA–HYAL-1, MVD and CD44v6) and preoperative (i.e., age, clinical stage and preoperative PSA) and postoperative (i.e., Gleason sum overall or ≥7, EPE, seminal vesicle invasion) was able to distinguish between patients who recurred within 24 months and those who did not. In this cohort, we had 25 patients who recurred within 72 months, of whom 17 (68%) recurred within 24 months and 3 more recurred at 27 months (mean time to recur for the entire cohort 21.3 months). Thus, the ability of various clinical and pathologic parameters as well as histologic markers to predict biochemical recurrence within 24 months may need to be studied in a larger group of biochemically recurred patients.
Nearly two-thirds of PCa patients have preoperative PSA levels of 4–10 ng/ml, stage T1C disease and a biopsy Gleason of 5–7.2, 10, 11 For such patients, one or a combination of accurate prognostic indicators could improve the physicians' ability to identify PCas that are aggressive and will progress so that individualized treatments could be offered. In our study, although preoperative PSA was an independent predictor for prognosis, neither the overall Gleason sum nor Gleason sum ≤ or ≥7 was an independent predictor for prognosis. Among the 4 potential prognostic indicators for PCa that we compared, HYAL-1 appears to be an accurate and independent predictor.
S.E. was a fellow of The Scientific and Technical Council of Turkey (NATO Science Fellowship Program).