Increased Vascularization Predicts Favorable Outcome in Follicular Lymphoma Free

Follicular lymphoma is the second most common lymphoma and is usually characterized by an indolent course. Most patients present with advanced-stage disease and at some point require therapy that typically results in only a temporary remission. Follicular lymphoma is currently treated with chemotherapy, radiotherapy, IFN-α, and several forms of specific immunotherapy (using anti-CD20 monoclonal antibodies), but no curative treatment is available with the possible exception of allogeneic stem cell transplantation (1). Although a median survival of 8 to 10 years in chemotherapy-treated patients is reported in several studies (2, 3), a subset of patients have a considerable worse prognosis, and prognostic indices have been generated to identify these patients (4, 5). With the expansion of treatment modalities, the choice of when and how to treat a patient requires a better identification of patients with a poor prognosis or certain biological characteristics fitting specific therapies. Angiogenesis is considered a crucial phenomenon in the pathogenesis of both solid tumors and hematologic malignancies of myeloid and lymphoid origin (reviewed in refs. 6, 7). Several studies suggest a role for angiogenesis in malignant non-Hodgkin's lymphoma. High levels of both proangiogenic and antiangiogenic molecules in peripheral blood and lymphoma specimens have been found to be associated with adverse prognosis (8–12). Some authors report a positive correlation between tumor vascularization and malignancy grade according to several lymphoma classification systems and higher vessel counts in malignant lymph nodes compared with reactive nodes (13–16). Others report increased vascularization of reactive lymph nodes compared with follicular lymphoma (17) or an equal vessel density in florid follicular hyperplasia and follicular lymphoma (13). Therefore, the clinical significance of increased vessel density is not clear (18) and difficult to establish because most of the studies describe heterogeneous populations, including a wide selection of histologic subtypes of non-Hodgkin's lymphoma and different treatment regimens. Phenotypic differences among blood vessels of reactive lymph nodes, follicular lymphoma, and diffuse large B-cell lymphoma (17) indicate that the clinical significance of lymph node vascularization might vary in different histologic entities, which underlines the importance of studying the relation between vascularization and clinical outcome in one single histologic lymphoma subtype.

We investigated the clinical significance of tumor vascularization and vascular endothelial growth factor (VEGF) expression in previously untreated patients with follicular lymphoma receiving uniform first-line treatment.

Patients.In December 1994, the Southeast Netherlands Comprehensive Cancer Centers Cooperative Group (Interzol) started a prospective, open, nonrandomized study in patients with previously untreated follicular non-Hodgkin's lymphoma of low-grade malignancy in Ann Arbor stages II bulky (>5 cm), III, or IV. The histologic diagnosis of the original lymph node biopsy samples was reviewed centrally according to the WHO classification. All patients were treated with eight four-weekly cyclophosphamide-vincristine-prednisone (CVP) courses [cyclophosphamide 750 mg/m² i.v. at day 1, vincristine 1.4 mg/m² (with a maximum of 2 mg) i.v. at day 1, and prednisone 60 mg/d p.o. at days 1-5] in combination with IFN-α2b (5 × 10⁶ IU s.c. thrice a week) as induction therapy. Thereafter, responding patients received maintenance treatment with IFN-α2b (5 × 10⁶ IU s.c. thrice a week) until intolerable toxicity, relapse, or progressive disease, whichever came first (19). An initial wait-and-see period was allowed before induction therapy was started. When relapse or progressive disease occurred, further treatment was left to the discretion of the treating physician.

Immunohistochemistry and Microvessel Count. Microvessel density (MVD) was assessed in samples of lymph node biopsies taken at the time of lymphoma diagnosis. The samples had been fixed in formalin and embedded in paraffin by routine methods. As endothelial marker, an anti-CD34 monoclonal antibody was used. Unlike anti-CD31, this antibody does not cross-react with lymphoid cells, and better quality of staining than with anti–factor VIII was obtained. Paraffin-embedded lymph node biopsies were cut into 4 μm sections. After deparaffination and inhibition of endogenous peroxidase, the sections were incubated overnight at 4°C with anti-CD34 (mouse anti-human CD34 Clone Qbend/10, dilution 1:750, Neomarkers, Fremont, CA). Subsequently, the sections were incubated with biotinylated horse anti-mouse (1:200, Vector Laboratories, Inc., Burlingame, CA) and peroxidase-conjugated avidin (1:100, Vector Laboratories). Visualization was done with 3,3′-diaminobenzidine substrate, and counterstaining was done with hematoxylin.

Initially, MVD was assessed by semiquantitative grading into three categories (low, intermediate, and high) by two investigators (A.K. and H.v.K.). MVD was separately estimated in the interfollicular areas and the follicles. The presence or absence of hotspots was registered. This method showed poor reproducibility, and no relation with patient outcome was found (data not shown). Then, we decided to perform microvessel count interactively using a KS400 image analysis system (Carl Zeiss Vision GmbH, Hallbergmoos, Germany). This system consists of a Zeiss Axioplan 2 Imaging microscope with a Zeiss AxioCam MRc camera attached to it. Microscopic images of specimens were digitized using a 20× objective (Plan Apo, numerical aperture 0.6, resulting in square pixels with size 0.275 μm² in the specimen). The size of a single measurement image is 525 × 368 μm (0.19 mm²). Tumor sections were first scanned at low magnification (5× or 10× objective) to determine three areas with the most intense vascularization (hotspots). In most cases, obvious areas with more intense vascularization were observed, indicated by a higher density of brown label against the background staining. Subsequently, in these selected hotspots, individual microvessels were interactively counted at higher magnification (20× objective) in the measurement image, which was projected on a computer screen. Microvessels were marked in the image by hand with a small square using a mouse-controlled cursor to prevent from double counting. The software collected data automatically. Any brown staining endothelial cell or endothelial cell cluster clearly separated from adjacent microvessels, tumor cells, normal cells, and other connective tissue elements, with or without the presence of a lumen, was considered a single countable microvessel. Vessels within sclerotic areas were not included (20–22). The microvessel counts in the three hotspots were averaged and denoted MVD. The observer (A.K.) was blinded to the clinical data of the patient. The intraobserver variability was assessed by repeating this procedure after a time interval of 2 months, including the identification of the hotspots. The investigator (A.K.) was blinded to the outcome of the first assessment.

VEGF In situ Hybridization. Sections of 10 patients were selected for detection of VEGF RNA expression. The selection took place according to the MVD, containing five sections with the lowest MVD (MVD = 10, 15, 16, 25, and 26) and five sections with the highest MVD (MVD = 61, 63, 66, 69, and 70). These sections were subjected to in situ hybridization with a digoxigenin-labeled VEGF antisense RNA probe. Actin hybridizations were also done to check RNA integrity in the sections. Sense probes for both RNAs were used as negative controls. Kidney tissue was used as a positive control.

Statistical Analysis. Spearman rank correlation and coefficient of variation were used to evaluate the intraobserver variability. Patient characteristics between groups were compared with the χ² test, the Fisher exact test, or the Mann-Whitney U test, when appropriate. Progression-free survival was measured from the start of induction therapy with CVP and IFN-α until the time of disease progression or until the end of the observation period in patients without progressive disease. Overall survival was measured from the start of induction therapy until death from any cause or until the end of the observation period. In a patient who had an initial wait-and-see policy, both progression-free survival and overall survival were also measured from the date the induction therapy was started. Survival curves were calculated according to the Kaplan-Meier method, and the log-rank test was used for analyzing differences between curves. The relative influence of different variables on progression-free and overall survival was studied by multivariate analysis using the nonparametric hazards model of Cox. Statistical analysis was done using the Statistica 4.5 release for MS Windows (Statsoft) with a statistical significance level of 0.05. All tests were two sided.

Patient Characteristics. Between December 1994 and January 2000, 62 patients with follicular lymphoma entered the study. From 58 patients, sufficient biopsy material was available to perform anti-CD34 staining. In 12 cases, the staining was of insufficient quality to perform reliable MVD assessment, thus leaving 46 cases suitable for analysis.

The clinical characteristics of these patients are summarized in Table 1. An initial wait-and-see period was adopted in nine patients, with a median duration of 13 months (range, 2-32 months). Thirty-six patients responded to induction therapy with CVP courses combined with IFN-α2b and were subsequently treated with IFN-α maintenance therapy, whereas 1 patient had stable disease and 9 had progressive disease. During the follow-up interval with a median duration of 70 months, 19 patients died, 4 patients in the group with a complete remission after induction therapy, 8 in the group with a partial remission, and 7 in the group with progressive disease. Sixteen patients did not receive all eight CVP courses, eight of whom because of progressive disease during induction therapy. IFN-α was stopped due to toxicity in two patients during the induction phase and in seven during the maintenance phase (Fig. 1). At revision, the lymphomas of 31 patients were classified as Berard grade 1 follicular lymphoma, 10 Berard grade 2, and 5 Berard grade 3.

Progression-Free Survival and Overall Survival in Relation to MVD. A typical pattern of distribution of the blood vessels was observed, showing prominent vascularization in the interfollicular areas and a relatively low number of vessels within the follicles (Fig. 2). Hotspots of microvessels were therefore usually situated between the follicles. When measured by interactive quantification, MVD ranged from 10 to 70 per measurement field of 0.19 mm² (median, 38; mean, 42; range, 51.8-362.7/mm²). The coefficient of variation of the two MVD assessments was 4.56; the correlation coefficient was 0.90 (P < 0.0001).

After a median follow-up of 70 months (range, 33-94 months), the median progression-free survival for all patients was 24 months, whereas the median overall survival was not reached. There was no difference in progression-free and overall survival between the 46 cases used in the analysis and the 16 cases who also entered the study but for whom no anti-CD34 stained material was available. The median MVD was 38 in the Berard grade 1 cases, 49 in the grade 2 cases, and 30 in the grade 3 cases (P = 0.18). The progression-free and overall survival did not differ between patients with different Berard grades. The survival curves of patients with MVD in the lowest tertile (MVD < 34; n = 15) and intermediate tertile (34 ≤ MVD ≤ 51; n= 16) showed significant overlap for both progression-free survival and overall survival and deviated remarkably from the curves of the patients with MVD in the highest tertile (MVD > 51; n = 15; Fig. 3A and C). Therefore, the outcome of patients with MVD in the lowest and intermediate tertiles were grouped together and compared with that of the patients with MVD in the highest tertile. The characteristics of the two groups of patients are summarized in Table 1. Patients within the highest tertile of vessel counts had significantly better progression-free survival than the patients with MVD in the combined lowest and intermediate tertile (median, 47 versus 13 months; P = 0.02; Fig.3B). The median overall survival was >94 months in the patients with high MVD and 59 months in the patients with lower MVD (P = 0.03; Fig. 3D). In univariate analysis, a statistically significant negative association with overall survival was found for an intermediate or high international Prognostic Index (ref.23; P = 0.004), male sex (P = 0.03), presence of bulky disease at presentation (P = 0.03), and absence of bone marrow involvement at presentation (P = 0.05). When MVD was tested in multivariate analysis together with the other significant variables from the univariate analysis, the association of high MVD with better progression-free survival and overall survival remained unchanged.

VEGF In situ Hybridization. There was almost a complete lack of expression of VEGF RNA in the lymphoma specimens in both sections with low MVD and high MVD (Fig.4). Positive control sections showed good VEGF expression. Actin in situ hybridization as a control for the quality of the specimens was also positive.

In this study, we focused on the clinical significance of tumor vascularization in pretreatment pathologic lymph nodes of patients with follicular lymphoma who received uniform first-line treatment with CVP chemotherapy in combination with IFN-α2b followed by IFN maintenance therapy. We found that high MVD, as a variable of increased tumor vascularization, was associated with a significantly more favorable outcome in terms of both progression-free and overall survival.

The initially used method of semiquantitative grading of MVD showed poor reproducibility. This is probably due to inaccuracy because a lower magnification (100×) was used, the use of a semiquantitative scale instead of a continuous scale, and the evaluation of the entire section instead of hotspots as well as the highly subjective nature of estimating vascularization grade instead of actual vessel counting. Interactive quantification resulted in highly reproducible counts, although there was a long time interval between the assessment and renewed selection of hotspots in the second assessment. Because of the observed distribution of blood vessels around the follicles in the interfollicular areas, it is important to perform counts in hotspots. This means that all counts have been done in the interfollicular areas. Randomly chosen fields, without regard to this typical distribution, will almost certainly include follicles. This will result in lower counts and lower reproducibility if a limited number of fields are counted and also ignores a possible biological meaning of the difference in MVD in follicles and the interfollicular areas.

The positive association between MVD and patient outcome is in contrast with the notion that, in hematologic malignancies as well as in solid tumors, an increase of angiogenesis-associated variables is related to adverse prognosis (24, 25). There is one report on ovarian cancer in which high intratumoral MVD is an independent predictor of complete response to paclitaxel/platinum–based chemotherapy and is associated with improved progression-free survival and overall survival (26). However, this has not been confirmed by several other studies (27–29). In bladder carcinoma, a positive relationship between MVD and prognosis has been described as well, but here increased tumor vascularity was associated with inflammation in the tumor and was not an independent predictor of outcome (30). Our observations seem to be in contradiction also with studies showing an association of increased serum concentrations of VEGF, basic fibroblast growth factor, and endostatin and adverse outcome in lymphoma (9, 12). However, the populations used in these studies were heterogeneous regarding lymphoma subtype, and only a minority had follicular lymphoma (15.5% and 16.8%, respectively). In addition, treatment regimens differed, and patients were not treated in prospective studies. No subanalysis for follicular lymphoma has been done. This means that the results of these studies do not have to apply for each lymphoma subtype. In addition, although basic fibroblast growth factor, VEGF, and endostatin clearly are important regulatory factors in angiogenesis, a direct correlation between circulating levels of these factors in the serum and intratumoral MVD does not have to be present. Indeed, several studies refute such a correlation (31–33).

Several options are available to explain our findings. One concerns the role of IFN-α. All patients in our study were treated with IFN-α2b, both concomitant with CVP chemotherapy in the induction phase and subsequently as monotherapy in the maintenance phase in patients responding to induction therapy. The role of IFN in the treatment of follicular lymphoma is controversial. Several studies indicate its efficacy (34, 35), but this is not confirmed by others (36, 37). A meta-analysis of randomized trials concluded that, if IFN was combined with more intensive initial therapy, this resulted in a prolonged survival in responding patients (38). The exact mechanism of action of IFN-α in lymphoma is unknown, but besides its antiproliferative and immunomodulatory properties, IFN-α also has antiangiogenic effects (39–44) and as such is applied in the treatment of highly angiogenic diseases, such as hemangioma of infancy and giant cell tumors (45–48). A possible explanation for our findings is that patients with follicular lymphoma with increased angiogenesis characterized by a high MVD are more susceptible to the antiangiogenic effects of IFN-α2b therapy than those with low vascularity.

Another explanation for our findings may consist of differences in pathways of angiogenesis and in characteristics of blood vessels between lymphomas of various histologic subtypes and solid tumors. In concordance with our data, several authors report that blood vessels in lymph nodes in follicular lymphoma are predominantly in the interfollicular areas, with relative low vascularity in the neoplastic follicles (13, 14, 17, 49, 50). This structured pattern of vascular distribution is similar to that found in reactive nodes and is lost in aggressive lymphomas (50). If these blood vessels were a direct result from pathologic tumor cell–mediated angiogenesis, a closer spatial association with the tumor cells in the follicles and a different pattern of distribution would be expected. In solid tumors, hypoxia is believed to be a prime stimulator of angiogenesis. Hypoxia-inducible factor-1 has been identified as a key mediator in the hypoxia-driven increased expression of the important proangiogenic molecule VEGF (51). It is, however, questionable whether hypoxia is a factor of major importance in low-grade follicular lymphoma, a disease characterized by indolent behavior and a slow growth rate. This is supported by Stewart et al. (50) who observed less expression of hypoxia-inducible factor-1, hypoxia-inducible factor-2, and VEGF in a group of indolent lymphomas—consisting mainly of follicular lymphoma—than in aggressive lymphomas, although this difference did not reach statistical significance. Ho et al. (52) also compared VEGF-A expression in indolent lymphoma with that in aggressive lymphoma and found that only a minority of the cases with indolent lymphoma expressed VEGF-A in contrast to aggressive lymphomas, which all expressed the protein. Most of the indolent lymphomas that did express VEGF-A showed histologic transformation to aggressive lymphoma, and VEGF-A was expressed in the large lymphoid cells. Foss et al. (53) found prominent expression of VEGF-A in aggressive T-cell lymphomas and inflamed lymphoid tissue but not in indolent B-cell lymphomas. We also failed to find any VEGF-A expression in the lymph node sections used in this study whether in the ones with low or in those with high vessel density. We conclude from these data that VEGF is not relevant in angiogenesis in follicular lymphoma.

Finally, the phenotype of blood vessels in different subtypes of lymphoma seems to vary. Overrepresentation of disorganized immature blood vessels, characterized by a lack of pericytes, has been found in aggressive lymphomas and Hodgkin's disease compared with reactive lymph nodes and indolent, mostly follicular, lymphomas (54). Using different markers to assess blood vessel maturity, Passalidou et al. (17) showed a higher number of immature blood vessels in diffuse large B-cell lymphoma than in reactive nodes or follicular lymphoma. These phenotypic variations in blood vessels and differences in pathways of angiogenesis between follicular lymphoma and aggressive lymphomas and certain nonhematologic malignancies might contribute to the explanation of the unexpected relationship between MVD and patient outcome in our study for several reasons. Important functional differences in tumor blood flow could exist between chaotic, immature vascularization, as found in solid tumors and in aggressive lymphomas, and structured mature vascular beds, as observed in follicular lymphoma. Increased vascularization may therefore be an advantage in follicular lymphoma treatment, because here it might provide better tumor perfusion; hence, higher concentrations of chemotherapeutic agents might be reached. In immature, disorganized vascular beds, an increase in vascular density probably does not translate into better perfusion (55).

Another explanation is that the observed differences in vascularization are merely a reflection of underlying biological features responsible for the observed effect of high MVD on patient outcome and that the blood vessels themselves are not functionally involved and are thus only a surrogate marker.

In any case, the importance of performing studies concerning the relationship of tumor vascularity and clinical outcome in homogeneous groups of patients, with respect to both the histologic entity of the studied population and the administered therapy, is stressed. If the findings of this pilot study are confirmed in a larger number of patients, MVD in follicular lymphoma might be part of a prognostic index and a possible target of new treatment modalities. Future investigations should be directed at the underlying mechanisms of angiogenesis in follicular lymphoma as well as functional aspects of vascularization. The role of IFN-α needs to be reassessed when those patients treated with chemotherapy alone do not display this remarkable impact of vascularization on prognosis.