Magnitude of Stromal Hemangiogenesis Correlates with Histologic Subtype of Non–Hodgkin's Lymphoma

Despite the fact that multimodality treatment, including combination chemotherapy, radiation, and target-specific monoclonal antibodies, such as rituximab, can induce high rate of remission in many subtypes of non–Hodgkin's lymphoma (NHL), significant proportions of patients relapse with incurable disease (1, 2). Several clinical and diagnostic variables have been developed to improve the prognostic stratification of clinical outcome, including the international prognostic index (IPI; ref. 3), tumor-specific markers, such as BCL-2 (4), BCL-6 (5), and p53 (6), and gene expression profiling signatures (7–9). However, very few studies have addressed the contribution of the stromal and vascular microenvironment to the pathogenesis and clinical behavior of lymphoma.

Neoangiogenesis is essential for tumor growth and progression (10–12). Blockade in angiogenesis can lead to tumor regression in solid tumor models (13, 14). Targeting vascular endothelial growth factor (VEGF)-A with bevacizumab in combination with chemotherapy promotes progression-free survival in human patients with metastatic colorectal cancer, advanced stage non–small cell lung cancer, as well as breast cancer (15–17). Surprisingly, very few studies have addressed the role of angiogenesis in the growth of human lymphomas, although several reports have implicated up-regulation of VEGF-A and VEGF receptors (VEGFR) in mediating lymphomagenesis (18–20).

In search of the cellular and molecular mechanisms involved in the assembly of tumor neovessels, our group and others have shown that recruitment of marrow-derived VEGFR-1⁺ proangiogenic hematopoietic progenitors to the vascular microenvironment contributes instructively and structurally to neoangiogenesis in mouse lymphoma models (21–23). Other studies have also set forth the concept that hematopoietic progenitors and differentiated myeloid cells, including mature monocytes/macrophages, can home to tumor tissue and promote angiogenesis of the growing tumor via possibly two mechanisms. One is through the production of proangiogenic growth factors, such as VEGF-A, VEGF-C/VEGF-D, brain-derived neurotrophic factor, and platelet-derived growth factor, as well as matrix metalloproteinases, such as matrix metalloproteinase-9, to support endothelial proliferation and stromal remodeling (24–26). Secondly, monocytes/macrophages seem to be capable of coexpressing vascular markers both in vitro and in vivo (27, 28), raising the possibility of direct participation in neoangiogenesis and vasculogenesis (29). In human subjects, tumor-associated macrophages have been implicated in aggressive disease and inferior outcome in several tumors, including breast, prostate, ovarian, cervical cancers and follicular lymphoma (FL; ref. 30, 31). However, the contribution of hemangiogenic progenitors to lymphoma angiogenesis in human subjects has not been studied.

Other stromal components can be activated during tumorigenesis and express myofibroblastic markers, including α-smooth muscle actin (α-SMA; ref. 32). In breast epithelial carcinoma model, carcinoma-associated α-SMA⁺ fibroblasts have been shown to promote tumor angiogenesis and growth via recruitment of endothelial progenitors from the bone marrow (33). Pericytes/smooth muscle cells have also been suggested to play a role in regulation of endothelial survival and vessel stability (34). However, it remains unclear whether the degree of the α-SMA⁺ stromal cell incorporation contributes to neovessel formation in human lymphoma as well.

Based on these studies, we hypothesized that the magnitude of recruitment of marrow-derived VEGFR-1⁺ hematopoietic precursors and incorporation of proangiogenic CD68⁺ myelomonocytic cells expressing VEGF-A may correlate with histologic subtypes of human lymphoma. Additionally, stromal production of the α-SMA may also define angiogenic properties of the respective subtypes. To test these hypotheses, we characterized the spatial localization of vascular and stromal cells on different subtypes of human NHL samples from patients with previously untreated diseases and compared them with benign follicular hyperplasia (FH).

Lymphoma specimens. Clinical diagnostic grade frozen and paraffin sections (5-μm thick) of untreated human NHL tissue specimens were obtained from the pathology department at Weill Cornell Medical College (New York, NY) according to Institutional Review Board–approved protocols and Health Insurance Portability and Accountability Act regulations. The specimens used in this study included the indolent subtypes of FL (n = 7) and chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL; n = 5), and aggressive lymphomas, including diffuse large B-cell lymphoma (DLBCL; n = 28) and Burkitt lymphoma (BL; n = 2). Tonsil and FH (n = 5) were used as controls. Clinical information was obtained for prognosis and survival correlations.

Immunohistochemistry. Paraffin tissue sections were stained with the following antibodies: anti-CD34 (QBEnd10, DAKO, Carpinteria, CA), CD68 (KP1 and PGM1, DAKO), α-SMA (1A4, Sigma, St. Louis, MO), BCL-6 (PG-B6, DAKO), MUM1 protein (MUM1p, DAKO), CD10 (56C6, Novocastra Laboratories, Newcastle upon Tyne, United Kingdom), CD138 (BB4, Serotec, Raleigh, NC), podoplanin (AngioBio Co., DelMar, CA), and VEGF-A (VG1 and Z-CVF3, Zymed, Carlsbad, CA). Frozen tissue sections were stained with anti-CD68 (KP1), CD34 (QBEnd10), VEGFR-1 (FB5, ImClone, New York, NY), VEGFR-2 (1121, ImClone), and VEGF-A (VG1 and Z-CVF3). Double staining was done on paraffin and frozen tissue sections with the following antibodies: α-SMA/CD34, CD34/CD68, VEGFR-1/CD68 (frozen section only), VEGFR-2/CD68 (frozen section only), and VEGF-A/CD68. The secondary antibodies were biotinylated using ARK kit (DAKO) followed by alkaline phosphatase ABC kit. Alkaline phosphatase reaction was developed using BT Red from ChemMate kit (Ventana, Tucson, AZ) or Vector Blue Substrate kit II (Vector Laboratories, Burlingame, CA). Frozen sections were double stained using immunofluorescence method. VEGFR-1/CD68 double staining was done with anti-VEGFR-1 followed by Alexa Fluor 594 goat anti-mouse IgG2b and anti-CD68 (KP1) followed by Alexa Fluor 488 goat anti-mouse IgG1. VEGFR-2/CD68 double staining was done first with FITC-conjugated anti-VEGFR-2 and then with mouse anti-FITC (DAK-FITC4, DAKO) followed by Alexa Fluor 488 goat anti-mouse IgG, and biotinylated anti-CD68 (KP1) followed by Alexa Fluor 594–conjugated streptavidin (Molecular Probes, Carlsbad, CA). VEGF-A/CD68 double staining was done with anti-VEGF-A followed by Alexa Fluor 488 goat anti-rabbit IgG and anti-CD68 (KP1) followed by Alexa Fluor 594 goat anti-mouse IgG. All staining experiments were done with appropriate positive and negative controls.

Imaging data analysis. A total of 4 to 10 fields with the highest vessel density in the lymphoma-involved area was selected for analysis. For lymphoma cases with relatively preserved follicular structures, only areas with neoplastic involvement (i.e., follicular center in FL) were studied. For CD34 and CD68 double-stained lymphoma sections, individual CD34-positive vessels and CD68-positive cells were counted manually. For VEGFR-1 and CD68 double-stained frozen sections, the percentage of the pixels covered by the VEGFR-1 and CD68 staining at ×200 magnification was defined as the hemangiogenic index of VEGFR-1 and CD68 and scored using the NIH ImageJ program. Similarly, for α-SMA and CD34 double-stained sections, the percentage of the pixels covered by these markers was defined as the α-SMA index. The coexpression of VEGFR-1 and CD68 was studied using laser scanning confocal microscopy. For readout of the B-cell activation markers, the germinal center B-cell pattern is defined if only CD10 and/or BCL6 are expressed without the expression of the activation markers MUM1 or CD138. ABC pattern is assigned if either MUM1 or CD138 is expressed regardless of the expression status of CD10 and BCL6 (35).

Statistical analysis. All results are expressed as mean and SD. ANOVA models were done for comparing outcome variables (CD34, CD68, and VEGFR-1/CD68 hemangiogenic index) across the various subtype groupings. The nonparametric Kruskal-Wallis test was also used to confirm the ANOVA analyses. For specific pairwise comparisons of interest, post hoc analyses were done with the Tukey's honestly significant difference adjustment. Ninety-five percent confidence intervals for mean differences between subtype groupings were calculated to assess the precision of the obtained estimates. All p-values are two sided with statistical significance evaluated at the 0.05 α level.

Clinical characteristics of the lymphoma cases. Forty-two cases of NHL were studied (Table 1). The median age of the DLBCL patients at diagnosis was 60 years, with median follow-up of 40 months (range, 2-108 months). The gender distribution was evenly divided. Fourteen percent of DLBCL patients had low risk disease according to the IPI, whereas 28% and 57% of patients had low-intermediate and high-intermediate risk diseases, respectively. Twenty-five of 28 (89%) DLBCL patients had stage III or IV disease at diagnosis, and 36% of DLBCL patients were deceased at the time of last clinical follow-up. Five of the 28 cases of DLBCL were transformed from prior existing indolent subtypes, including FL, marginal zone lymphoma, and CLL/SLL, whereas the remainders of the DLBCL cases were de novo (n = 23). B-cell surface markers CD10, BCL6, MUM1, and CD138 were analyzed by immunohistochemistry and used as surrogate markers to stratify the DLBCL samples into either germinal center B-cell (GBC) expression pattern or activated B-cell (ABC) expression pattern (35). Consistent with the clustering of high-risk clinical features in our patient cohort, 79% (22 of 28) of DLBCL cases had the ABC pattern, which portends a less favorable prognosis, and only 11% (3 of 28) of cases had the germinal center B-cell pattern. The remaining three DLBCL cases were not evaluable due to suboptimal tissue procurement. For the FL cases, only grade 1 or 2 diseases were included. For CLL/SLL, three of five cases were ZAP-70 positive by immunohistochemistry.

Microvessel density as measured by CD34⁺ vessels does not correlate with histologic subtype of lymphoma. We first examined the vascular component within the tumor stroma by CD34 staining (Fig. 1A). The average CD34 counts from individual DLBCL cases displayed a wide range of distribution (15.60-84.50 per 200× field). Surprisingly, the means of the CD34⁺ microvessel density (MVD) did not differ significantly between DLBCL, indolent subtypes of FL and CLL/SLL, and the benign control of FH (37.10 ± 16.24 versus 41.88 ± 18.65 versus 47.92 ± 13.05; Fig. 1B; Table 2). Although not statistically significant, the de novo DLBCLs showed a trend toward higher MVDs compared with the transformed cases and FL (38.55 ± 17.04 versus 30.42 ± 10.74 versus 29.52 ± 3.19). Compared with the other indolent subtypes, CLL/SLL had a higher mean MVD (CLL/SLL versus DLBCL versus FL: 59.18 ± 17.30 versus 37.10 ± 16.24 versus 29.52 ± 3.19; Table 2). To examine the possibility that some of the CD34⁺ vessels may be of lymphatic origin, we did CD34 and podoplanin costaining (Fig. 1D). Podoplanin is specifically expressed on lymphatic capillaries but not in the blood vasculature. Up to 30% of CD34⁺ vessels in CLL/SLL tumors expressed podoplanin. In contrast, the benign control and the other NHL subtypes, including BL, DLBCL, and FL, had very few podoplanin/CD34 double-positive vessels. When adjusted to exclude the lymphatic vessels, the difference of the MVD of the CD34⁺ blood-specific vessels was less pronounced between the aggressive subtype of DLBCL and the indolent subtypes of CLL and FL. Collectively, the means of the CD34⁺ MVD were comparable between the aggressive and indolent subtypes of lymphoma [aggressive (n = 30) versus indolent (n = 12) versus benign hyperplasia (n = 5): 39.29 ± 18.71 versus 41.88 ± 18.65 versus 47.92 ± 13.05; P = 0.61; Table 2]. These data suggest that CD34⁺ MVD is not a sensitive biomarker for assessing neoangiogenesis in lymphoma and does not correlate with the clinical aggressiveness and histologic subtype of human lymphoma.

Marked stromal and perivascular infiltration of CD68⁺ hematopoietic cells is present in aggressive subtypes of B-cell NHLs but not in indolent subtypes. To delineate the contribution of the bone marrow–derived hemangiogenic cells present in aggressive or indolent lymphomas, we quantified the number of hematopoietic cells expressing CD68 by immunohistochemistry, which marks subsets of the hematopoietic progenitors and mature cells of myeloid and monocytic origin that may contribute to lymphoma neoangiogenesis. In benign controls, such as tonsils and reactive lymph nodes, the CD68⁺ cells were limited to the parafollicular sinus area, where they were closely associated with paracortical vasculature. Double staining with monoclonal antibodies against CD68 (hematopoietic cells) and CD34 (vasculature) showed a dichotomous pattern of CD68⁺ hematopoietic cells in indolent versus aggressive NHL (Fig. 1A). In aggressive subtypes, including BL and DLBCL, the tumor stroma was marked by intense CD68⁺ myelomonocytic infiltration within the tumor stroma and perivascular regions. In some tumor-involved areas, the CD68⁺ cells were aligned side by side with the CD34⁺ neovessels. In contrast, there was a profound decrease in the number of infiltrating CD68⁺ cells within the stroma of indolent subtypes. Additionally, the morphology of the myelomonocytic infiltrate was different. In aggressive lymphomas, the intrastromal and perivascular CD68⁺ cells were larger in size with ample cytoplasm, whereas in indolent lymphomas the CD68⁺ cells were small and often not associated with the vasculature.

The numbers of incorporated CD68⁺ cells were significantly higher in aggressive subtypes than those in indolent subtypes [aggressive (n = 30) versus indolent (n = 12): 235.48 ± 92.63 versus 35.98 ± 15.52 in 200× field; P < 0.0001; Fig. 1C; Table 2]. When the DLBCL cases were further segregated into de novo versus transformed cases, the average CD68⁺ counts were comparable between the two groups (241.77 ± 87.06 versus 193.97 ± 129.91; P = 0.38; Table 2). Within the DLBCL subtype, no significant differences of the CD68⁺ counts were detected according to the following variables: GBC (n = 3) versus ABC (n = 22; 299.63 ± 82.09 versus 227.80 ± 94.74); IPI 0 to 2 (n = 12) versus IPI >2 (n = 16; 258.46 ± 74.27 versus 216.10 ± 103.24); and alive (n = 18) versus deceased (n = 10; 227.83 ± 101.78 versus 245.82 ± 77.94). These data implicate a fundamental difference in the hemangiogenic switch between indolent and aggressive subtypes of lymphoma. The aggressive subtypes are substantially more capable of recruitment and stromal incorporation of the marrow-derived CD68⁺ hematopoietic cells. The differences of hemangiogenic activity within the DLBCL subtype may be less pronounced compared with the differences between the subtypes and could potentially be obscured by the small sample size.

Stroma of aggressive NHL is characterized by increased proliferation of VEGFR-1⁺ neovasculature and perivascular incorporation of VEGFR-1⁺CD68⁺ hematopoietic cells expressing VEGF-A. VEGFR-1 expression on the hematopoietic cells identifies a population of proangiogenic hematopoietic cells that contribute to neoangiogenesis in the primary tumors (21) and initiating the premetastatic niche (36, 37). Costaining with CD68 and VEGFR-1 on frozen sections was used to identify cells that could potentially participate in blood vessel assembly within the lymphoma stroma. In DLBCL, there were increased numbers of VEGFR-1⁺ vascular structures that were decorated and infiltrated by CD68⁺ hematopoietic cells. Importantly, the majority of the CD68⁺ hematopoietic cells coexpressed VEGFR-1 and seemed to be in close association with one or another VEGFR-1⁺ neovessels (Fig. 2F-I). The CD68⁺VEGFR-1⁺ double-positive cells were localized either within intravascular space, transiting across the vascular walls, juxtapositioning endothelial cells, or scattered in a short distance away from the vessels. To compare the hemangiogenic activity among different subtypes, we developed the term hemangiogenic index, which measures the percentage of the tumor-involved stromal area covered by the VEGFR-1⁺ vessels and VEGFR-1⁺CD68⁺ hematopoietic cells. In contrast to the aggressive subtypes, the hemangiogenic index of indolent subtypes of NHL, including CLL/SLL and FL, was significantly lower [DLBCL (n = 5) versus CLL (n = 4) versus FL (n = 5): 18.14 ± 2.83% versus 3.39 ± 1.95% versus 6.54 ± 2.82%; P < 0.0001; Fig. 2A-E]. No CD68⁺VEGFR-1⁺ double-positive cells were observed in indolent subtypes. To investigate whether the CD68⁺ cells also coexpress other vascular markers, we did double staining of CD68 with either CD34 or VEGFR-2 on frozen sections of DLBCL and BL (Fig. 3). Both CD34 and VEGFR-2 marked vasculature only and were not expressed in the CD68⁺ hematopoietic cells. The majority of the perivascular CD68⁺ hematopoietic cells expressed VEGF-A. These data suggest that hemangiogenic index as determined by the number of infiltrating VEGFR1⁺ and CD68⁺ cells provides a reliable means to assess the contribution of the proangiogenic hematopoietic cells to neoangiogenesis in different histologic subtypes of lymphoma. In addition to providing structural support, perivascular positioning of proangiogenic VEGF-A⁺VEGFR-1⁺CD68⁺ hematopoietic cells in aggressive lymphoma subtypes argues for a paracrine role in neovessel formation, where neoangiogenesis is essential to sustain rapid tumor growth.

Stroma of the indolent subtype of CLL/SLL is marked by incorporation of a large numbers of α-SMA⁺ cells. Newly formed neovessels within regenerating organs are stabilized by incorporation of α-SMA⁺ cells to the perivascular zone. In benign nodal hyperplasia, the α-SMA⁺ stromal cells were limited to the parafollicular sinus in association with paracortical vasculature. Remarkably, despite the abundance of the CD68⁺ hematopoietic infiltrates, the stroma of aggressive subtypes, such as DLBCL, was largely devoid of α-SMA⁺ cells. Most of the α-SMA⁺ cells detected in DLBCL were vascular smooth muscle cells, which demarcated the CD34⁺ vasculature. By contrast, the stroma of indolent subtype of CLL/SLL, which had scant incorporation of CD68⁺ cells, was marked by diffuse infiltration of α-SMA⁺ cells, which formed an elaborate meshwork bridging CD34⁺ vessels within the parenchyma of lymphomas. The magnitude of incorporation of α-SMA⁺ cells in CLL, which included both perivascular and intrastromal areas, was significantly higher than the other subtypes and benign nodal hyperplasia (CLL/SLL versus DLBCL versus FL versus FH: 11.80% versus 2.22% versus 1.42% versus 2.70%; P < 0.0001; Fig. 4). These findings indicate that stromal composition of lymphomas is dictated by differential recruitment of either mesenchymal α-SMA⁺ or hemangiogenic CD68⁺VEGFR1⁺ cells. CLL/SLL stroma has scant infiltration of CD68⁺VEGFR1⁺ cells but is populated by a large number of α-SMA⁺ mesenchymal cells. In contrast, aggressive subtypes recruit CD68⁺VEGFR1⁺ cells but lack incorporation of α-SMA⁺ stromal cells.

Non-Hodgkin' lymphomas are a heterogenous group of malignancies. The contribution of neoangiogenesis to the pathogenesis of various human lymphoma subtypes remains poorly defined. In preclicnical mouse models, it is well established that neoangiogenesis and growth of murine lymphomas is partially dependent on the recruitment of bone marrow–derived proangiogenic hematopoietic population contributing to neovessel assembly either instructively by releasing growth factors, such as VEGF-A, or constructively by localizing perivascularly to stabilize newly formed vessels (21, 24, 38). In this study, we show that there is a statistically significant correlation between hemangiogenic index, as measured by the differential stromal recruitment of either hemangiogenic CD68⁺VEGFR-1⁺ cells or mesenchymal α-SMA⁺ cells, and histologic subtype of human NHL (Fig. 5). In aggressive subtypes of lymphoma, there is increased proliferation of VEGFR-1⁺ neovasculature and enhanced recruitment and perivascular incorporation of VEGF-A-producing CD68⁺VEGFR-1⁺ hematopoietic cells, which provide paracrine support to neovasculature. Remarkably, the aggressive subtypes were depleted of α-SMA⁺ stromal cells. In contrast, the stroma of indolent subtype of CLL/SLL has a profound increase in the number of α-SMA⁺ mesenchymal cells both perivascularly and diffusely, which presumably enhance vascular stability and may contribute to chemotherapy resistance. Importantly, there is no correlation between the CD34⁺ microvessel densities in aggressive versus indolent lymphomas. These data suggest that the stromal hemangiogenic index, determined by the magnitude of the incorporation of VEGFR-1⁺, CD68⁺, and α-SMA⁺ cells, provides a novel surrogate marker for assessing bone marrow–derived hematopoietic contribution to neoangiogenesis in human NHL. As CD68⁺ hematopoietic and α-SMA⁺ stromal cells can directly be targeted by antiangiogenic agents, stromal hemangiogenic index may be used to predict response to antiangiogenic therapy.

The evidence highlighting the importance of stromal microenvironment during lymphomagenesis is emerging, most recently from large-scale gene expression array studies looking at differential genetic signatures among different subtypes or within one specific subtype of NHL (7–9). For instance, in DLBCL, genetic signatures expressed by the stromal and infiltrating immune cells, including lymphocytes, macrophages, and dendritic cells, have been shown to form a distinct prognostic group, which identifies the tumor microenvironment for risk stratification (8, 39). In FL, clinical survival was correlated to the nontumor immune infiltrating cells alone (40). Indeed, the content of the lymphoma-associated macrophages is an independent predictor of survival in FL (31). However, the mechanisms underscoring the prognostic contribution of the hematopoietic immune cells remain poorly characterized.

To characterize the tumor hematopoietic compartment, we have used the hematopoietic marker CD68, which is expressed by hematopoietic cells, including proangiogenic VEGFR-1⁺ myelomonocytic progenitors, monocytes, macrophages, and dendritic cells. Our present study clearly shows a statistically significant progressive increase in CD68 hematopoietic infiltrates within lymphoma stroma across different histologies from indolent to aggressive subtypes. Within DLBCL subtype, the de novo DLBCL cases exhibit a wide range of CD68 infiltrating cells, perhaps reflecting the underlying heterogeneity of the pathogenic tumor cells. No significant correlation was noted between CD68 cell counts and the clinical outcomes in the de novo DLBCL cohorts, although this may be masked by our sample size and the clustering of poor-risk features. Questions remain about the functional roles of the CD68⁺ hematopoietic infiltrates in the lymphoma stroma. CD68⁺ cells comprise a heterogeneous population, including monocytes, tissue macrophages, and dendritic cells. Our costaining studies on the CD68⁺ cells have shown that majority of the CD68⁺ cells coexpress VEGF-A and remain in close contact with the vascular endothelium, suggestive of paracrine support these cells may have on the developing neovasculature in aggressive lymphoma. Indeed, emerging evidence has shown that bone marrow–derived hematopoietic cells critically contribute to neovessel formation by promoting vessel stability (38) and enhancing in situ endothelial proliferation (26).

To characterize the tumor vascular compartment, we have used markers, including CD34 and VEGFR-1. Our study has shown that CD34 alone is not a sensitive biomarker for lymphoma neoangiogenesis, as it stains stable preexisting vessels and lymphatic structures as well. In contrast, VEGFR-1 expression showed a clear difference of vascular distribution with prominent microvascular staining in the aggressive subtype of DLBCL. Both the VEGFR-1 staining intensity and the VEGFR-1⁺ microvessel counts were significantly decreased in the indolent subtypes of FL and CLL/SLL. As such, the magnitude of VEGFR-1 expression either on the endothelial or hematopoietic cells provides a more sensitive means to assess the degree of neoangiogenesis in various subtypes of lymphomas. Recently, VEGFR-1⁺ hematopoietic progenitors have been implicated in initiating the premetastatic niche, a tumor-directed process proceeding the infiltration of actual tumor cells or tumor-associated neoangiogenesis (37). This raises a provocative notion that VEGFR-1⁺CD68⁺ hematopoietic cells may also direct the metastatic spread of lymphoma. Clearly, future in vivo studies in lymphoma mouse models are needed to further delineate the functional contribution of VEGFR-1⁺CD68⁺ hematopoietic population in lymphoma angiogenesis and progression.

Remarkably, in indolent subtype of CLL/SLL, the tumor stroma is marked by α-SMA⁺ stromal cells diffusely as well as perivascularly. α-SMA expression is up-regulated in fibroblasts in response to tumor stromal-specific growth factors, such as platelet-derived growth factor and transforming growth factor-β, causing a unique phenotype of tumor-associated myofibroblasts (32). α-SMA⁺ myofibroblasts have recently been shown to express high level of SDF-1α, a chemokine critical for mobilization and recruitment of CXCR4-expressing vascular progenitors (33). Alternatively, α-SMA⁺ pericytes/vascular smooth muscle cells can be recruited and proliferate into the interstitial stroma along with vascular endothelial cells under proangiogenic stimuli and regulate vessel stability (34, 41). Blood vessels with few pericytes coverage were more sensitive to VEGF-A inhibition (42). Thus, targeting α-SMA⁺ perivascular cells may compliment endothelial cell inhibition for effective antiangiogenic therapy in indolent subtypes of CLL/SLL (43).

Our study has provided a novel perspective into the complex interplay between several stromal and vascular components within the lymphoma microenvironment and the potentially divergent pathogenic processes whereby the bone marrow–derived hemangiogenesis contributes to lymphomagenesis. In aggressive subtypes, prominent incorporation of proangiogenic VEGFR-1⁺CD68⁺ cells and diminished localization of α-SMA⁺ stromal cells to the perivascular zone may contribute to enhanced neoangiogenesis and chemosensitivity, whereas decrease in CD68⁺ cells and increase in the α-SMA⁺ incorporation may promote neovessel stability and chemoresistance in CLL/SLL. Identification of critical hemangiogenic components specific to lymphoma subtypes will lay the foundation where efficacy in blocking these factors alone or in combination can be tested in clinical trials. In pilot studies, bevacizumab has shown clinical activity both as a single agent in relapsed aggressive NHL setting (44) and in combination with rituximab and cyclophosphamide-Adriamycin-vincristine-prednisone (CHOP) in upfront setting (45), lending support to antiangiogenic development in NHL. It is tempting to speculate that anti-VEGF-A therapy may be an effective addition to the standard chemotherapy for aggressive subtypes of lymphoma, which have a significant VEGFR-1⁺ hemangiogenic stromal component, whereas disrupting mesenchymal-tumor interaction may afford durable remission in the indolent subtypes.

We thank Dr. Sherry Ikalowych for assistance in patient data collection, Dr. Loic Vincent for helpful discussion and critical review of the article, and ImClone scientists Drs. Yan Wu, Zhenping Zhu, and Dan Hicklin for kindly providing anti-human VEGFR-1 antibody FB5.