Differential Regulation of Hypoxia-Induced CXCR4 Triggering during B-Cell Development and Lymphomagenesis Free

Lymphocytes are exposed to low oxygen tensions as they develop and acquire effector functions (1). The degree of hypoxia has been positively correlated with hypoxia-inducible factor 1 (HIF-1; ref. 2). HIF-1 is composed of the oxygen- and growth factor–regulated subunit HIF-1α, and the constitutively expressed HIF-1β subunit (3, 4). HIF-1α has been reported to be expressed in germinal centers (GC) and in the majority of non–Hodgkin's lymphomas (5). Through the use of chimeric mice, it has been reported that HIF-1α deficiency results in lineage-specific defects of B lymphocytes, including defects in development of B lymphocytes and autoimmunity (6). The mechanisms by which normal and malignant B lymphocytes adapt and respond to hypoxia are poorly understood.

The presence of hypoxic areas is a common feature to malignant tumors, including lymphomas (7). Clinical studies have shown that the low oxygen tension within a neoplastic lesion is an independent indicator of poor outcome and correlates with increased risk of metastasis (8–10). More than 60 target genes that are activated by HIF-1 have been identified; these include the chemokine receptor CXCR4 (11) and its ligand stromal cell-derived factor-1 (SDF-1/CXCL12; refs. 12, 13). Because B-cell lymphomas often express CXCR4, which plays a crucial role in numerous processes, among which tumor metastasis (14), and arise in lymphoid organs which constitutively produce CXCL12 (15, 16), an understanding of the regulatory role of hypoxia on CXCR4 expression and function can provide insight into its role in normal B-cell development as well as into lymphomagenesis.

Here, we report that the function of CXCR4 is differentially regulated under hypoxic conditions in lymphomas of different origin. We observed CXCR4 desensitization in GC B cells and GC-derived lymphomas and showed that this phenomenon is due to HIF-1α–dependent up-regulation of signal transduction molecules such as mitogen-activated protein kinase (MAPK) phosphatase 1 (MKP-1) and regulator of G protein signaling protein 1 (RGS1).

Cells and culture conditions. Hu/severe combined immunodeficient (SCID) lymphomas were generated as reported (17, 18). The following lymphoma cell lines were used: Burkitt's lymphoma cell lines (BRG and Ramos; from Dr. M. Ferrarini, National Cancer Research Institute, Genoa, Italy) and diffuse large B-cell lymphoma (DLBCL) cell lines OCI-Ly1, OCI-Ly7, OCI-Ly8, OCI-Ly10, VAL, and SUDLH4. Hypoxic treatment of cells was achieved by incubating cells in a PROOX model 110 chamber (Biospherix). During incubation, 0.1% to 0.5% O₂ was maintained. Cobalt chloride (CoCl₂; Sigma) was used at a concentration of 100 μmol/L.

Details concerning cytofluorimetric analysis, RNA extraction, and reverse transcription-PCR (RT-PCR), Northern and Western blot, confocal microscopy, real-time PCR, CXCR4 compartmentalization, chemotaxis, gene expression data, ELISA for extracellular signal-regulated kinase 1/2 (ERK1/2), statistical analysis, and B lymphocyte purification are provided as Supplementary Materials and Methods.

Lentiviral small interfering RNA cell lines. The lentiviral plasmids containing HIF-1α small interfering RNA (siRNA) and LUC siRNA target sequences (19) were a kind gift of Dr. O.V. Razorenova (Lerner Research Institute, Cleveland, OH). The lentiviral vector was produced as described (20). Cells expressing the siRNA were selected in puromycin [siRNA (P)] or sorted for enhanced green fluorescent protein [siRNA (E)].

Studies in lymphoma-bearing mice. SCID mice were purchased from Charles River. Procedures involving animals and their care conformed with institutional guidelines that comply with national and international laws and policies (EEC Council Directive 86/609, OJ L 358, December 12, 1987). Groups of 7- to 9-week-old SCID mice were injected i.p. with (a) 70 × 10⁶ to 100 × 10⁶ peripheral blood mononuclear cells (PBMC) to generate hu/SCID lymphomas or (b) 10 × 10⁶ BRG lymphoma cells. Tumors and other tissues of interest were processed as previously described (18). To identify hypoxic tumor cells, we used pimonidazole hydrochloride (Hypoxyprobe-1, Chemicon International; see Supplementary Materials and Methods).

Immunohistochemistry. Immunohistochemical analysis was done on frozen or formalin-fixed, paraffin-embedded sections, with the following primary antibodies: mouse anti–HIF-1α (ESEE 122, Novus Biologicals), rabbit anti–HIF-1α, rabbit anti–MKP-1, rabbit anti-RGS1 (Santa Cruz Biotechnology), mouse anti-CD34 (Novocastra), mouse anti-CD3, mouse anti-CD10, and mouse anti-IgD (Dakocytomation). Controls included isotype-matched murine or rabbit antibodies of irrelevant specificity.

Hypoxia-induced HIF-1α stabilization is associated with high CXCR4 expression in lymphoma cells in vivo. Little is known on the expression of hypoxia in hematologic malignancies (5). To better understand the role of hypoxia in a B-cell lymphoma setting, B lymphomas of different origin, hu/SCID lymphoma cells (originating from circulating B lymphocytes; ref. 21), and cells of Burkitt's-like lymphoma (BRG cell line; originated from GC B lymphocytes) were injected i.p. into SCID mice, and the tumor masses were studied for the expression of HIF-1α by immunohistochemistry. Hu/SCID lymphomas showed diffuse areas of necrosis in five of seven tumors analyzed, with viable perinecrotic cells staining for HIF-1α (Fig. 1A,, left), indicating the presence of intense hypoxia. Occasionally, HIF-1α staining was diffusely present within the tumor, possibly indicating increased HIF-1α levels driven by oxygen-independent mechanisms (Fig. 1A,, middle). On the other hand, BRG-derived lymphomas showed areas of necrosis in only one of six tumors (not shown), with the bulk of the tumor mass showing scattered cells staining for HIF-1α, except for lymphoma cells infiltrating adjacent tissues, where intense HIF-1α staining was found (Fig. 1A , right).

To better define the extent of hypoxia in our lymphoma models, we determined the presence of pimonidazole-labeled cells by flow cytometry as a surrogate marker of hypoxia. In 10 tumors examined, pimonidazole labeling was present in over 60% of hu/SCID lymphoma cells (Fig. 1B), whereas BRG lymphomas showed lower expression of pimonidazole adducts (<30% of the cells).

We recently showed that hu/SCID lymphomas are composed of two phenotypically distinct B-cell subsets (22): A CD23P^low population that expresses high surface CXCR4 (CXCR4P^int) and a CD23P^int subset that barely expresses CXCR4 (CXCR4P^low; ref. 23). Double staining of tumor cells with anti-CXCR4 and antipimonidazole antibodies revealed that the two hu/SCID lymphoma cell subpopulations showed a substantially different staining for pimonidazole (Fig. 1C), with the CXCR4P^int subpopulation staining much more intensely than cells not expressing CXCR4 (left). Similar results were obtained in BRG lymphomas, where cells expressing higher levels of surface CXCR4 stained more intensely for pimonidazole (Fig. 1C , right).

Hypoxia increases surface CXCR4 expression on B lymphoma cells through transcriptional and posttranscriptional mechanisms. The above data indicated that hypoxia correlates with preferential expression of CXCR4 in our tumors. To gain further insight on the mechanisms underlying this association, hu/SCID lymphoma cells and cells of Burkitt's-like lymphoma (BRG and Ramos cell lines) were exposed in vitro to severe hypoxia (<0.5%) or CoCl₂, which mimics the effects of hypoxia by inducing HIF-1α stabilization (24), for 16 to 18 h. As shown in Fig. 2A, both hypoxia and CoCl₂ significantly increased surface expression of CXCR4 in all the cell lines tested. Conversely, there was no significant increase in surface expression of other chemokine receptors, such as CCR6, CCR7, CXCR3, and CXCR5 (not shown). This phenomenon was reversible upon reoxygenation, supporting the causal role of hypoxia in regulating CXCR4 expression (Fig. 2B).

We also studied the effects of hypoxia on CXCR4 surface expression in a panel of DLBCL cell lines (also originating from GC B cells) representative of the three different gene expression subgroups recently described (25): GC B cell–like (GCB), activated B cell–like (ABC), and unclassified or type III. Most DLBCL cell lines showed a significant increase in surface expression of CXCR4 under hypoxic conditions or after CoCl₂ treatment, with GCB DLBCL cell lines showing a higher modulation compared with cell lines of the other two subgroups (Supplementary Table S1).

It has been shown that in some cell types, CXCR4 expression can be regulated by hypoxia through HIF-1 activation (11, 26). Thus, we investigated whether a transcriptional mechanism was also responsible for hypoxia-induced CXCR4 up-regulation in our lymphoma cells. As shown in Fig. 2C, CXCR4 mRNA, but not the mRNA of other receptors evaluated, was significantly up-regulated in response to hypoxia or CoCl₂ in hu/SCID lymphomas and BRG lymphoma cells. Northern blot analysis showed a strong increase in CXCR4 mRNA as early as 4 h, and persisting after 18 h, after incubation with CoCl₂ (Fig. 2C , right) or hypoxia (not shown) in BRG lymphoma cells. These results, however, do not exclude that increased CXCR4 transcript stability during hypoxia may also contribute to higher CXCR4 gene expression.

In view of the possibility that surface CXCR4 expression could also be regulated by posttranscriptional mechanisms (16), such as receptor redistribution, we compared the expression of CXCR4 in the extracellular and intracellular compartments by flow cytometry (27). In normoxic conditions, CXCR4 was predominantly distributed in the intracellular pool in both hu/SCID and BRG lymphoma cells (Fig. 2D); conversely, exposure to hypoxia or CoCl₂ determined a significant shift of the distribution of CXCR4 from the intracellular to the extracellular compartment. These results were also confirmed by confocal microscopy (not shown).

Differential response to CXCR4 triggering under hypoxia in B-cell lymphomas of different cellular origin may reflect a developmental regulation. Hu/SCID lymphomas resemble immunoblastic lymphomas arising from the expansion of the very few EBV-infected B lymphocytes present in the injected PBMC (18, 28), representing post-GC B cells. To assess whether the hypoxia-associated changes observed in CXCR4 expression were also accompanied by an effect on CXCR4 function and signaling, hu/SCID lymphoma cells were incubated in hypoxic conditions, and the number of lymphoma cells migrating in response to CXCL12 was recorded. As shown in Fig. 3A, the hypoxia-dependent increase in CXCR4 expression was accompanied by a significant rise in the number of hu/SCID lymphoma cells migrating in the presence of a CXCL12 gradient. We next evaluated the effects of hypoxia on CXCR4-mediated chemotaxis in BRG and Ramos Burkitt's lymphoma cells, which are derived from GC centroblasts. Notably, despite their higher CXCR4 surface expression under hypoxic conditions, both cell lines displayed a significant inhibition of the migratory response to CXCL12 (Fig. 3A).

A similar evaluation of the effects of hypoxia on CXCR4 function in DLBCL cell lines disclosed a similar dichotomic functional response (Fig. 3B), with GCB DLBCL cell lines (Ly7; left) displaying a significant inhibition of their chemotactic response to CXCL12 under hypoxic conditions, while the non-GCB cell line Ly8 showing an increased response (Fig. 3B , right) under the same conditions.

To determine whether this different response could be developmentally regulated, we investigated the functional response to CXCR4 triggering under hypoxic conditions in the normal counterpart of the lymphomatous cells; that is, GC B cells for Burkitt's lymphomas and DLBCL and peripheral blood B cells for hu/SCID tumors. As shown in Fig. 3C, (left), in analogy to hu/SCID lymphoma cells, hypoxia significantly increased the number of peripheral B cells migrating in the presence of a CXCL12 gradient. GC B cells (CD38⁺CD44⁻) instead were unresponsive to CXCL12 independent of hypoxia, confirming previous results (Fig. 3C , right; refs. 15, 29).

Using hu/SCID and BRG Burkitt's lymphoma cells as a model, we evaluated whether distinct alterations of the CXCR4 signaling pathway could be involved in determining the differential functional responses of naïve/memory versus GC B cells. Thus, we examined the phosphorylation status of p42/p44 MAPK/ERK 1 and 2, which are implicated in signal transduction (30, 31). As shown in Fig. 3D, under normoxic conditions, phosphorylation of ERK1/2 slightly increased in hu/SCID lymphoma cells after CXCL12 stimulation, whereas it underwent a dramatic increase under hypoxic conditions (Fig. 3D). Conversely, BRG cells displayed a marked increase of phosphorylated ERK1/2 after CXCL12 stimulation under normoxic conditions, whereas no significant increase was detectable under hypoxic conditions (Fig. 3D , right). These results indicate that the functional chemotactic response to CXCR4 triggering under hypoxic conditions in B cells correlates with distinct alterations of the CXCR4 signaling pathway and that this response may be developmentally regulated, with peripheral B cells exhibiting increased CXCR4 function and GC B cells undergoing CXCR4 desensitization.

Hypoxia-induced CXCR4 uncoupling in GC-derived lymphoma cells is associated with up-regulation of MKP-1 and RGS proteins. RGS proteins increase the intrinsic GTPase activity of Gα subunits (32) and can thus modulate chemokine responses. In addition, MKP-1 attenuates p44/p42 MAPK activity, which is required for chemokine-mediated activation and migration (33). Both mechanisms could be at play in hypoxia-mediated CXCR4 desensitization in GC-derived lymphoma cells. Of the RGS proteins evaluated by RT-PCR, after incubation under hypoxic conditions, a consistent increase was observed only for RGS1 transcript in both BRG and Ramos cell lines (Fig. 4A), whereas no significant up-modulation of any RGS transcript was found in hu/SCID lymphomas (not shown). Increased expression of RGS1 was confirmed by Western blotting in both cell lines (Fig. 4B), whereas no increased protein levels were recorded for hu/SCID lymphomas (not shown). Similar results were obtained for MKP-1, where a substantial increase was found by both quantitative RT-PCR and Western blotting in the Burkitt's cell lines BRG and Ramos (Fig. 4C and D, respectively). A similar evaluation of the effects of hypoxia or CoCl₂ on RGS1 and MKP-1 expression in DLBCL cell lines disclosed how only GCB DLBCL cell lines (Ly1 and Ly7) but not ABC (Ly10) or type III (unclassified; Ly8) DLBCL cell lines displayed a significant increase in their expression levels (Supplementary Fig. S2).

Numerous other mechanisms of receptor desensitization have been described (34). We found that hypoxia or CoCl₂ treatment did not modulate the expression of CD26 and CD45 (not shown), transforming growth factor-β1 and its receptor, or members of the β-arrestin family (ARRB1, ARRB2, or ARRB3) shown to be involved in receptor desensitization in other models (Supplementary Fig. S1). Also, phosphorylation of G protein–coupled receptors is thought to be essential for their ligand-induced endocytosis. We found that CXCR4 exhibited a low level of tyrosine phosphorylation in Burkitt's lymphoma cells (BRG) cultured under normoxic conditions, which did not appreciably change after hypoxic exposure (not shown). Thus, MKP-1 and RGS1 proteins are the main determinants of CXCR4 desensitization in GC-derived lymphoma cells, although we cannot exclude that additional mechanisms may be responsible for the observed receptor desensitization in GC-derived lymphoma cells.

Silencing of HIF-1α reverses hypoxia-induced CXCR4 up-regulation and prevents receptor uncoupling in Burkitt's-like lymphoma cells. Having established that hypoxia is associated with CXCR4 desensitization through complex alterations of the signaling pathways downstream of the CXCR4 receptor, we addressed whether HIF-1 plays a major role in hypoxia-associated modulation of CXCR4 expression and function by investigating the effects of its functional knockdown. To this end, we used lentiviral vectors expressing a siRNA specific for HIF-1α (>70% knockdown as determined by quantitative RT-PCR and Western blot; Fig. 5A), and investigated the effects of HIF-1α siRNA on CXCR4 expression and function under hypoxic conditions in BRG lymphoma cells. As shown in Fig. 5B, BRG cells stably expressing HIF-1α siRNA showed an important inhibition of CXCR4 up-regulation under hypoxic conditions as evaluated by quantitative flow cytometry compared with control cells. Similarly, by quantitative RT-PCR, there was a significant inhibition of CXCR4 transcript induction by hypoxia (not shown) or CoCl₂ in HIF-1α knockout BRG cells (Fig. 5B,, right). This inhibition of CXCR4 modulation under hypoxic conditions was associated with a substantial protection from hypoxia-induced CXCR4 uncoupling (Fig. 5C), with HIF-1α knockout BRG cells maintaining their responsiveness to CXCL12 under hypoxic conditions. To evaluate whether the induction of RGS1 and MKP-1 could play a role in hypoxia-induced CXCR4 desensitization, we also determined by quantitative RT-PCR the effect of HIF-1α knockdown on RGS1 and MKP-1 expression under hypoxic conditions. BRG cells stably expressing HIF-1α siRNA, in contrast to control cells, showed no significant increase in RGS1 and MKP-1 expression after CoCl₂ treatment (Fig. 5D) or hypoxic exposure (not shown). These findings were also confirmed at protein level by confocal microscopy of MKP-1 expression (Fig. 5D) and RGS1 (not shown). These data suggest that RGS1 and MKP-1 are downstream targets of HIF-1α and that this latter is required for hypoxia-induced CXCR4 up-regulation and receptor uncoupling in Burkitt's-like lymphoma cells.

Expression of HIF-1α and downstream molecules in the normal counterpart of Burkitt's-like lymphomas and primary GC-derived B-cell lymphomas. As previously mentioned, human GC B cells have been shown to have high CXCR4 expression and yet respond poorly to chemokines (15, 29); because Burkitt's lymphomas and DLBCL are presumed to arise from GC B cells and we have shown that under hypoxic conditions these cells undergo CXCR4 desensitization, we were interested in studying the expression of HIF-1α and the downstream molecules MKP-1 and RGS1 in these samples. As shown in Fig. 6A, HIF-1α seemed to be predominantly expressed in tonsil GCs, given the good overlap of HIF-1α staining with the GC B-cell marker CD10. This result was confirmed at higher magnification (Fig. 6B,, inset) where the staining for HIF-1α was found to be predominantly within GCs like CD10 staining and unlike IgD staining, which outlines GCs (not shown). The largest portion of HIF-1α staining was found in the cytoplasmic compartment of GC B cells, with occasional nuclear signal being detected. Interestingly, GCs were also found to be avascular, as shown by the lack of CD34 reactivity (Fig. 6A), whereas a rich vascular network was present outside of GCs, further suggesting that hypoxia is probably present within GCs. On the other hand, CXCR4 expression was found to be diffusely expressed both in GC and parafollicular areas. We next evaluated the expression of MKP-1 and RGS1 proteins in tonsil sections by immunohistochemistry and immunofluorescence. As shown in Fig. 6C, RGS1 and MKP-1 proteins were expressed predominantly within GCs, much like HIF-1α, but unlike IgD that stained mantle zone B cells. Confocal microscopic analysis of tonsil sections confirmed these data, as the fluorescence distribution of RGS1 and MKP-1 (Fig. 6C , right) was found to be predominantly within GCs. Thus, the expression of MKP-1 and RGS1 in GCs, possibly induced by HIF-1α, may indeed contribute to GC B-cell desensitization of CXCR4.

Similar evaluation of the expression of CXCR4 and HIF-1α in human B-cell lymphomas of GC origin (DLBCL and Burkitt's lymphomas) and non-GC origin (marginal zone and mantle cell lymphomas) disclosed how these two proteins were more highly expressed in GC-derived B-cell lymphomas (Fig. 6D; Supplementary Fig. S3), suggesting that in these tumors HIF-1α expression may modulate CXCR4 expression and function in vivo. In addition, the evaluation of the expression of the downstream molecules MKP-1 and RGS1 in primary Burkitt's lymphoma disclosed that the expression of these molecules is rather diffuse. As shown in Fig. 6D, (left), in most cases (9 of 10) HIF-1α staining was predominantly cytoplasmic with few cells showing nuclear staining, similarly to GC B cells; only rarely (1 of 10) was nuclear HIF-1α staining diffuse throughout the tumor mass, in the absence of evident necrosis (Fig. 6D,, bottom left). The expression of MKP-1 was also rather diffuse (10 of 10; Fig. 6D,, top middle), with a predominance in the cytoplasmic compartment, much like has been recently reported in other tumor types (35). On the other hand, RGS1 expression was diffuse and found to be prevalently nuclear in all cases analyzed (10 of 10; Fig. 6D,, bottom middle). CXCR4 expression was also found to be diffuse, with predominant cytoplasmic staining (10 of 10; Fig. 6D , right).

These results show that the expression of the hypoxia-associated marker HIF-1α and CXCR4 (together with RGS1 and MKP-1) are diffusely present in GC-derived B-cell lymphomas and suggest that the conditions for hypoxia-induced CXCR4 desensitization may be present in vivo.

CXCR4 is one of the most common chemokine receptors expressed by tumor cells. Previous reports have shown that low oxygen concentrations can induce high expression of CXCR4 in monocytes and epithelial-derived tumors (11) through the activation and stabilization of HIF-1α.

Here, we showed that hypoxia mediates selective up-regulation of CXCR4 through stabilization of HIF-1α in normal B cells and malignant B cells of different non–Hodgkin's B-cell lymphomas such as PBMC-derived hu/SCID lymphomas, GC B cell–derived Burkitt's, and DLBCL lymphoma cells. This receptor up-regulation persisted under chronic hypoxia, a situation that may be closer to what cancer cells are subjected to in the growing tumor mass, and more importantly it was reversible under normoxic conditions. Further, we expand the knowledge on the mechanisms that may be responsible for hypoxia-induced CXCR4 up-regulation. Indeed, we showed that hypoxia is not only able to act at the transcriptional level but also at the posttranslational level through a process of receptor redistribution.

Responsiveness to CXCL12 correlates with positioning of B cells within a peripheral lymphoid organ and is regulated by the differentiation state and by B-cell receptor engagement (29). It has been described that two rescue signals, such as antigen and CD40 ligand, raise RGS1 and RGS13 expression levels, respectively, in B cells (36, 37). Hypoxia, through modulation of CXCR4 expression and of the regulatory molecules RGS1 and MKP-1, may contribute to help GC B cells to negotiate a complex chemokine milieu such as that present in GC. Indeed, HIF-1α may participate in chemokine receptor desensitization, thereby allowing the cell to desensitize to a primary chemokine gradient and respond to a second one. A likely site of conflicting chemoattractant gradients is the dark and light zones of GC, where B cells have several alternative fates. In fact, CXCL12 amounts in the dark zone are higher than in the adjacent light zone (16). On the other hand, CXCL13 has a distribution in the GC that is opposite to that of CXCL12, being more abundant in the light zone than in the dark zone (16). HIF-1α may assist B cells in their navigation through the GC, modulating their responsiveness to CXCL12.

Concerning the functional importance of hypoxia-induced CXCR4 up-regulation in B cells, we found that lymphomas of different maturational stage (hu/SCID lymphomas originating from mature circulating B cells and Burkitt's/GCB-DLBCL lymphomas originating from GC B cells) showed an opposite response to CXCR4 triggering under hypoxic conditions. Indeed, CXCR4 function was enhanced by hypoxia in hu/SCID lymphomas, as shown by increased chemotactic responsiveness to CXCL12 and increased ERK1/2 phosphorylation after CXCR4 triggering, whereas this receptor underwent desensitization in hypoxic GC-derived lymphoma cells.

Several studies have shown that GC B cells respond poorly to chemokines, despite the expression of CXCR4. CXCR4 expression and response of B cells within lymphoid organs is heterogeneous. Indeed, within human tonsils, IgD⁺CD38-negative B cells residing in the mantle zone (considered as naive B cells) and IgD⁻CD38–negative B cells (considered as memory B cells) express moderate levels of CXCR4 but undergo efficient migration to CXCL12; meanwhile, GC B cells (CD38⁺ and IgD⁺ or IgD⁻), which express high levels of CXCR4, are essentially refractory to CXCL12-induced migration (36). This refractoriness has been attributed, at least in part, to high-level expression of RGS1 (36) and RGS13 (37). Also, B cells from RGS1−/− mice migrate better to CXCL12 than do those from wild-type mice (38). On the other hand, a recent study that used GC B cells from Bcl-2 transgenic mice (which had extended GC B cell survival) showed robust migratory responses to CXCL12 and CXCL13 (16). The authors argued that the lack of responsiveness of GC B cells to B-cell chemoattractants was caused by their poor in vitro viability, impeded by Bcl-2 overexpression. An alternative explanation (39) could be that Bcl-2 overexpression alters GC B-cell selection, thus allowing B cells to survive in the absence of normal rescue signals (such as antigen and CD40 ligand), which normally lead to increased expression of RGS1 and RGS13, thus leading to low levels of these RGS proteins and probably other factors in the Bcl-2 transgenic GC B cells.

We find here that hypoxia determines induction/up-regulation of RGS1 in GC-derived lymphoma cells. We also showed in situ expression of RGS1 predominantly in GC and that GC are poorly vascularized given the lack of CD34 staining; thus, it could be tempting to speculate that hypoxia may be involved with high expression of RGS1 in GC B cells, and thus contribute to their poor chemotactic response to CXCL12. The expression of RGS1 in GC B cells and primary Burkitt's lymphoma was unexpectedly found to be predominantly nuclear. It has previously been reported that RGS proteins can localize in the nucleus, in the cytoplasm, or shuttle between the nucleus and cytoplasm (40).

Recently, it has been reported that hypoxia is able to inhibit macrophage migration to numerous chemokines and N-formyl-Met-Leu-Phe, possibly through up-regulation of MKP-1 (41). In agreement with this report, we found that hypoxia up-regulated MKP-1 in GC-derived lymphoma cells. As phosphorylation and eventual activation of MAPKs are required for CXCL12-mediated migration (33), these results suggest that dephosphorylation of MAPK by MPK-1 may be required for hypoxia-induced inhibition of B lymphoma cell migration in response to CXCL12. We showed in situ expression of MKP-1 in GC, with a distribution similar to RGS1 and particularly to HIF-1α, thus providing evidence that hypoxia may contribute to CXCR4 desensitization in GC B cells, through expression of MKP-1.

Using immunochemical hypoxia marker techniques, we were able to show a rather diffuse presence of hypoxia in experimental hu/SCID lymphomas and BRG Burkitt's lymphomas. Nuclear HIF-1α expression in experimental GC B cell–derived BRG Burkitt's lymphomas was found predominantly in lymphoma cells infiltrating surrounding tissues, implicating that the phenomenon of CXCR4 desensitization may be localized and relevant when lymphoma cells acquire a more invasive phenotype. Primary GC-derived B-cell lymphoma samples and GC B cells had a more heterogeneous and variable expression of HIF-1α. In most cases, the expression was predominantly cytoplasmic. The function of cytoplasmic HIF-1α expression is not known, but a similar pattern of expression has been described in T lymphocytes infiltrating the synovial tissue of patients with rheumatoid arthritis (42). The unique aspects of HIF-1α expression in GC B cells and GC-derived lymphomas (presence of HIF-1α both in the cytoplasm and nucleus) indicate the possible involvement of oxygen-independent pathways of HIF-1α regulation.

Available gene expression data (Supplementary Fig. S4) from Lymphochip arrays done on lymphomas (25) showed that the expression levels of HIF-1α, RGS1, and MKP-1 were higher in GC-derived lymphomas (DLBCL and follicular lymphomas) compared with chronic lymphocytic leukemia samples, and that the expression levels of these three transcripts are related.

Lymphomas arise in lymphoid organs that constitutively produce the CXCR4 ligand, CXCL12 (15, 16). In addition, lymphoma cells are able to produce CXCL12 themselves (14). The significance of CXCL12 expression at the primary site is not fully understood. By analogy with its role in the bone marrow, high levels of CXCL12 should retain the tumor cells in situ rather than encourage their dissemination. Thus, CXCR4 up-regulation by hypoxia in primary tumors expressing CXCL12 could be expected to retain tumor cells at this site if functionality of CXCR4 is maintained, which could encourage their growth and survival, but discourage invasion and metastasis. Our results provide evidence that hypoxia is an important modulator of CXCR4 responsiveness in normal and malignant B cells and that hypoxia-induced CXCR4 receptor uncoupling may represent a novel mechanism for guiding normal and malignant cell migration/dissemination in sites of conflicting chemoattractant gradients; this phenomenon may be particularly relevant in early stages of nodal lymphomas where the integrity of the lymphoid tissue is almost intact.

Grant support: Ministero dell'Istruzione, dell'Università e della Ricerca (MIUR)-Fondo per gli Investimenti della Ricerca di Base and MIUR-Programmi di ricerca di Rilevante Interesse Nazionale; Istituto Superiore di Sanità (AIDS Project); Italian Association for Research on Cancer (AIRC), and Italian Foundation for Research on Cancer; Ministero dell'Universita e Ricerca Scientifica e Tecnologica 60%; Padua University grants; Ministero della Salute, Ricerca Oncologica Project RF-IOV-2006-408212. Azione Biotec 2005-Regione Veneto. V. Tosello was a recipient of an AIRC fellowship.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Drs. P. Marson and F. Comacchio (Padua, Italy) for precious help in lymphapheresis and tonsil supply, respectively; Dr. O.V. Razorenova (Lerner Research Institute, Cleveland, OH) for providing the pLSLG and pLSLP-HIF-1α/LUC-siRNA lentiviral plasmids; and Prof. N. Fujii (Kyoto University, Kyoto, Japan) for comments on the manuscript.