CXCL13 and CXCL12 in Central Nervous System Lymphoma Patients

Primary central nervous system (CNS) lymphoma is rare subtype of extranodal non-Hodgkin's lymphoma, mostly diffuse large B-cell lymphoma, confined to the CNS in the absence of systemic disease (1). Mechanisms leading to lymphoma development in the CNS normally devoid of a lymphatic system are still unclear. Lymphocyte homing to the CNS regulated by chemokines may play an important role in the pathogenesis of this disease.

CXCL12 and CXCL13 are important chemokines mediating germinal center organization in lymphoid tissue (2). CXCR5 is the prime receptor for CXCL13, which causes homing of B-helper cells and a subset of T-helper cells to the lymphoid follicle. Follicular dendritic cells located in the germinal centers of lymphoid follicles are considered the main source of CXCL13. CXCL13 expression was shown in several B-cell lymphomas, some with extranodal location such as gastric or ocular adnexal lymphomas (3–6).

CXCR4, the receptor for CXCL12, is physiologically expressed on B and T cells, bone marrow stroma cells, and other hematopoietic cells. Expression of CXCR4 on malignant cells has been reported for a variety of tumors such as glioblastoma, breast cancer, and B- and T-cell lymphoma. CXCL12 is expressed by a wide variety of tissues and plays a key role in the trafficking of hematopoietic cells but is also involved in tumor growth and metastasis (7–10).

Recently, the expression of CXCR4 and CXCR5, as well as their chemokine ligands, CXCL12 and CXCL13, has been shown on malignant lymphocytes in primary CNS lymphoma (11–14). These findings suggest a role for chemokines in the pathogenesis of primary CNS lymphoma.

We studied the levels of both chemokines in the cerebrospinal fluid (CSF) and serum of patients with CNS lymphoma before and during therapy and compared the results to controls with other CNS malignancies and nonmalignant CNS diseases.

CSF samples obtained by lumbar puncture and serum samples were collected from patients with newly diagnosed or relapsed primary or secondary CNS lymphoma. All had B-cell non-Hodgkin's lymphoma, confirmed either by brain biopsy or CSF immunocytology. Pretherapeutic sampling was done before chemotherapy was started. Corticosteroid pretreatment was allowed.

For the control group, patients undergoing diagnostic lumbar punctures at our hemato-oncology and neurology departments were enrolled. All patients gave written informed consent for the analyses.

CSF drawn by lumbar puncture was processed immediately at room temperature. CSF protein and albumin were measured by nephelometry, and total CSF cells were enumerated in a counting chamber. For chemokine measurements, CSF cells were pelleted by centrifugation (4 min; 4,000 rpm), and supernatants were stored at −80°C. Paired serum samples were also collected and stored at −80°C. For determination of blood-brain barrier disruption, the ratio of albumin in CSF and serum (Q_alb) was calculated (15). The extent of blood-brain barrier disruption was graded according to the Q_alb value as none (≤7.4 × 10⁻³), slight (>7.4 and ≤10 × 10⁻³), moderate (>10 and ≤20 × 10⁻³), and severe (>20 × 10⁻³).

For enzyme-linked immunosorbent assays of CSF supernatants and serum samples, the Quantikine Human CXCL13/BLC/BCA-1 and the Quantikine Human CXCL12/SDF-1α immunoassay kits were purchased from R and D Systems and were used according to the manufacturer's instructions. The minimum detectable chemokine doses were 0.4 pg/mL CXCL13 and 1.0 pg/mL CXCL12.

Descriptive and explorative data analyses of all parameters were done. Quantitative parameters were tested for normal distribution. To detect significant differences, the Kruskal-Wallis and Mann-Whitney U tests were used for quantitative parameters and the χ² square test for nonquantitative parameters.

For bivariate correlations, the Spearman's r was calculated. All tests were two-sided, with a P value of <0.05 indicating a significant difference. For all statistical analyses, the SPSS software package for Windows, release 15.0.1, was used.

Included were 23 patients with primary CNS lymphoma and 7 patients with secondary CNS lymphoma. CNS histology was diffuse large B-cell lymphoma in 28 patients, including three patients with Richter's transformation of a systemic indolent B-cell lymphoma. Two patients had an indolent CNS B-cell lymphoma, one of which was secondary. Four patients had leptomeningeal lymphoma spread in addition to their brain parenchyma manifestations, and none had intraocular involvement as controlled by slit lamp examination. All lymphoma patients were HIV negative.

There were 2 control groups: 10 patients with CNS involvement of cancer (7 with meningeal or brain metastases of solid tumors and 3 with glioblastoma) and 30 patients without CNS malignancy (6 lymphoma and 3 cancer patients without CNS involvement, 12 with normal pressure hydrocephalus, 4 with dementia, and 4 who had exclusion of inflammatory CNS disease).

CNS lymphoma patients were more often male compared with the control groups, CSF cell count and Q_alb were the highest in CNS lymphoma patients and lowest in no–CNS cancer controls, and total protein was higher in CNS lymphoma and cancer patients compared with no–CNS cancer controls. Moderate or severe blood-brain barrier disruption was significantly more frequent in the CNS lymphoma group as compared with the no–CNS cancer group (76% versus 26% of patients, respectively; Table 1).

CXCL12 serum levels were high both in CNS lymphoma patients and controls, without a significant difference between the groups (Table 2). There was a trend toward higher levels in CNS lymphoma compared with all controls (median values 2,088 and 1,500 pg/mL; P = 0.059). CXCL12 CSF levels were lower than serum levels in all groups and did not differ between CNS lymphoma patients and controls (Fig. 1).

CXCL13 serum levels were, unlike CXCL12, low in all patients (median, <100 pg/mL). In patients without CNS lymphoma also, CXCL13 CSF levels were very low. Only CNS lymphoma patients had high CXCL13 CSF levels; the difference to CNS cancer and noncancer controls was highly significant (Fig. 1).

Serum and CSF chemokine levels were correlated in CNS lymphoma patients to exclude a possible impact of high systemic chemokine levels on CSF levels in the presence of a disrupted blood-brain barrier. No correlation of CXCL12 and CXCL13 levels in CSF and serum was found with r's of 0.14 (P = 0.62) for CXCL13 and 0.13 (P = 0.67) for CXCL12, respectively. The lack of correlation persisted when all control patients were included into the analysis.

CSF CXCL13 levels in CNS lymphoma patients were significantly associated with Q_alb (R = 0.66; P = 0.003), whereas no such association was found for CSF CXCL12 levels. There was neither a correlation of both chemokines with the number of CSF cells per microliter nor with the presence of lymphoma cells in the CSF (detected either by cytomorphology or immunocytology).

Data on corticosteroid pretreatment were recorded in 26 of 30 CNS lymphoma patients: 20 patients received 3 to 24 mg of systemic dexamethasone per day, and 6 patients had not been pretreated with corticosteroids before sampling.

Serum and CSF chemokine levels were not significantly different in patients with and without corticosteroid pretreatment, although there was a trend toward lower CXCL12 serum levels in patients on corticosteroids (Table 3).

CXCL12 and CXCL13 serum levels in 9 patients with systemic B-cell lymphoma (5 with diffuse large B-cell lymphoma and 4 with indolent B-cell lymphoma, including 5 with secondary CNS involvement) and 19 patients without lymphoma were compared. Median CXCL12 serum levels were 2,059 pg/mL in systemic lymphoma patients and 1,357 pg/mL in patients without lymphoma (P = 0.069); median CXCL13 levels were 105 and 65 pg/mL (P = 0.039), respectively (Fig. 2).

Serial CSF measurements before, during, or after chemotherapy were done for CXCL12 in five and for CXCL13 in six patients, including one patient each with a progressive lymphoma. All other patients responded to therapy (partial or complete remission). Chemotherapy consisted of high-dose methotrexate (4 g/m² i.v. over 4 hours; day 1) and ifosfamide (1,5 g/m² i.v. per day; days 3-5) in all patients.

During chemotherapy CSF chemokine levels decreased 1.1 to 10.1 times (mean, 3.8) in all responding patients and increased in both patients with progressive disease (Fig. 3). Very low CSF CXCL12 and CXCL13 levels were detectable in one patient in complete remission, in whom relevant pretreatment specimens were not available.

In patients responding to therapy, CXCL13 levels dropped to values comparable with those measured in controls (<100 pg/mL) in all but one patient with a complete resolution of an intracerebral lesion and a persistence of the concomitant meningeal involvement on therapy. In one patient who had complete surgical removal of primary CNS lymphoma, a low CXCL13 level was found before chemotherapy.

In our study, we detected high concentrations of CXCL12 and CXCL13 in the CSF of CNS lymphoma patients, with a significant difference for CXCL13 as compared with control patients, whereas no difference was found for CSF CXCL12. It seems likely that these findings reflect the role of the chemokines in the pathogenesis, homing, and survival of malignant lymphoma cells in the CNS (16). Both chemokines are known as homing factors for B cells, but it remains unclear whether B cells undergo a malignant transformation before or after homing to the CNS. CXCL13 expression has not been detected in healthy CNS but in malignant B cells in primary CNS lymphoma, as shown by in situ hybridization for CXCL13 mRNA (13). Follicular dendritic cells are the only proven physiologic source of CXCL13; however, normal CNS lacks the formation of germinal centers. However, the formation of ectopic germinal centers within the meninges and intra-CNS production of CXCL13 has been shown in inflammatory CNS diseases such as neuroborreliosis and multiple sclerosis (17–22). An inflammatory CNS lesion initiating B-cell homing to the CNS may represent a first step in the pathogenesis of CNS lymphoma. CXCL12 expression on vascular endothelia facilitating the crossing of lymphocytes through the blood-brain barrier (11, 14) and the prevention of apoptosis (as shown for CXCL13 in B-cell chronic lymphatic leukemia and for CXCL12 in malignant glioma) may additionally contribute to the malignant lymphoma pathogenesis (10, 23).

Because primary CNS lymphoma tumor cells express both chemokines and their corresponding receptors, their role in supporting tumor growth and survival by autocrine mechanisms seems possible. Because follicular dendritic cells are not normally present in CNS lymphoma, one can assume that CXCL13 measured in the CSF directly represents its production by lymphoma cells. This is also indicated by higher serum CXCL13 levels in patients with systemic B-cell lymphoma. The sources of CXCL12 in the CSF are probably manifold. CXCL12 is considerably smaller than albumin (89 versus 584 amino acids). Because serum CXCL12 levels were higher than CSF levels, it cannot be ruled out that CSF CXCL12 was derived from serum, particularly in the presence of a disrupted blood-brain barrier. However, we found neither a correlation of serum and CSF levels nor a correlation of CXCL12 CSF levels and the degree of blood-brain barrier disruption so that a production of CXCL12 within the CNS can also be assumed. Apart from lymphoma cells, endothelial cells of tumor vessels and reactive glia are other potential sources of this chemokine within the CNS (24).

The decrease of CSF chemokine levels parallel to lymphoma response to therapy in CNS lymphoma patients suggests a correlation with tumor burden. This assumption is supported by the significant correlation of CXCL13 with the grade of blood-brain barrier disruption because large tumors have a stronger impact on the blood-brain barrier, and a disrupted blood-brain barrier reconstitutes during successful chemotherapy. Thus, in particular, CXCL13 CSF levels might have a prognostic value in primary CNS lymphoma similar to CSF protein, which also correlates with blood-brain barrier disruption and has been shown to predict outcome in this disease (25).

Corticosteroid application in primary CNS lymphoma patients might have biased our results because of the down-regulation of chemokine expression (26, 27), However, we did not find an impact of corticosteroid pretreatment on CSF chemokines levels. Serum CXCL12 levels were somewhat lower in corticosteroid pretreated patients, which possibly reflects the anti-inflammatory properties of corticosteroids.

According to our results, CXCL13 rather than CXCL12 seems to reflect the tumor burden and thus may represent a promising diagnostic and prognostic marker in CNS lymphoma warranting further studies. The potential role of both chemokines in the pathogenesis of CNS lymphoma should be evaluated.

No potential conflicts of interest were disclosed.