OUP user menu

Chemokines in multiple sclerosis: CXCL12 and CXCL13 up-regulation is differentially linked to CNS immune cell recruitment

Markus Krumbholz, Diethilde Theil, Sabine Cepok, Bernhard Hemmer, Pia Kivisäkk, Richard M. Ransohoff, Monika Hofbauer, Cinthia Farina, Tobias Derfuss, Caroline Hartle, Jia Newcombe, Reinhard Hohlfeld, Edgar Meinl
DOI: http://dx.doi.org/10.1093/brain/awh680 200-211 First published online: 9 November 2005


Understanding the mechanisms of immune cell migration to multiple sclerosis lesions offers significant therapeutic potential. This study focused on the chemokines CXCL12 (SDF-1) and CXCL13 (BCA-1), both of which regulate B cell migration in lymphoid tissues. We report that immunohistologically CXCL12 was constitutively expressed in CNS parenchyma on blood vessel walls. In both active and chronic inactive multiple sclerosis lesions CXCL12 protein was elevated and detected on astrocytes and blood vessels. Quantitative PCR demonstrated that CXCL13 was produced in actively demyelinating multiple sclerosis lesions, but not in chronic inactive lesions or in the CNS of subjects who had no neurological disease. CXCL13 protein was localized in perivascular infiltrates and scattered infiltrating cells in lesion parenchyma. In the CSF of relapsing–remitting multiple sclerosis patients, both CXCL12 and CXCL13 were elevated. CXCL13, but not CXCL12, levels correlated strongly with intrathecal immunoglobulin production as well as the presence of B cells, plasma blasts and T cells. About 20% of CSF CD4+ cells and almost all B cells expressed the CXCL13 receptor CXCR5. In vitro, CXCL13 was produced by monocytes and at much higher levels by macrophages. CXCL13 mRNA and protein expression was induced by TNFα and IL-1β but inhibited by IL-4 and IFNγ. Together, CXCL12 and CXCL13 are elevated in active multiple sclerosis lesions and CXCL12 also in inactive lesions. The consequences of CXCL12 up-regulation could be manifold. CXCL12 localization on blood vessels indicates a possible role in leucocyte extravasation, and CXCL12 may contribute to plasma cell persistence since its receptor CXCR4 is retained during plasma cell differentiation. CXCL12 may contribute to axonal damage as it can become a neurotoxic mediator of cleavage by metalloproteases, which are present in multiple sclerosis lesions. The strong linkage of CXCL13 to immune cells and immunoglobulin levels in CSF suggests that this is one of the factors that attract and maintain B and T cells in inflamed CNS lesions. Therefore, both CXCL13 and CXCR5 may be promising therapeutic targets in multiple sclerosis.

  • lymphocytes
  • chemokines
  • cerebrospinal fluid
  • inflammation
  • immune cell migration
  • BBB = blood–brain barrier
  • Ig = immunoglobulin
  • mAb = monoclonal antibody
  • NIND = non-inflammatory neurological diseases
  • OIND = other inflammatory neurological diseases (other than multiple sclerosis)
  • PBMC = peripheral blood mononuclear cells
  • PP = primary progressive
  • RR = relapsing–remitting
  • SP = secondary progressive


Multiple sclerosis is considered to be an immunopathological disease in which activated T cells enter the CNS and trigger an inflammatory cascade that leads to recruitment of other immune cells. Modulation of immune cell migration to the CNS offers great therapeutic potential (Kivisakk et al., 2001; Miller et al., 2003). Migration of immune cells is largely regulated by chemokines, which comprise a family of >50 molecules (Nelson and Krensky, 2001). The possibility of treating inflammatory diseases by targeting chemokines or their receptors has strongly stimulated research efforts in this field.

Chemokines are loosely divided into two functional classes: inflammatory chemokines that govern leucocyte trafficking to inflamed tissues ‘on demand’, and homeostatic chemokines that regulate constitutive trafficking, mainly, to lymphoid sites. Previous studies on chemokines in multiple sclerosis lesions focused largely on inflammatory chemokines. For example, the CCR1 and CCR5 ligand RANTES (CCL5), the CCR5 ligand MIP-1β (CCL4), the CXCR3 ligand IP-10 (CXCL10) and the CCR2 ligand MCP-1 (CCL2) have been associated with lesion development in multiple sclerosis (Simpson et al., 1998, 2000; Balashov et al., 1999; Sorensen et al., 1999; Trebst et al., 2001). These chemokines attract mainly monocytes and T cells.

However, some of the most prominent immunological abnormalities in multiple sclerosis are based on B cells: patients typically show intrathecal immunoglobulin (Ig) production, which presents as oligoclonal bands and a polyspecific Ig response. Oligoclonal Ig may persist in the same patient for many years (Walsh and Tourtellotte, 1986). Clonal expansion and somatic hypermutation of B cells (Qin et al., 1998; Colombo et al., 2000, 2003; Owens et al., 2003) and also plasma cells (Ritchie et al., 2004) have been observed in the CSF and in multiple sclerosis brain lesions (Owens et al., 1998; Baranzini et al., 1999). Lymphoid follicle-like structures have recently been described in the meninges of two out of three patients with secondary progressive (SP) multiple sclerosis (Serafini et al., 2004). Thus the currently available evidence indicates that B cells, plasma blasts or plasma cells migrate to the CNS of multiple sclerosis patients and persist there for many years, perhaps lifelong (reviewed, Uccelli et al., 2005). This long-term persistence of B cells might be promoted by local production of the B cell survival factor BAFF by astrocytes (Krumbholz et al., 2005). B-lymphocyte and plasma cell clonal expansion were observed in monosymptomatic optic neuritis CSF (Haubold et al., 2004), which supports the implication of B cell autoimmunity in multiple sclerosis, since it was observed at the first detectable phase of disease. The cellular composition of the CSF of patients shows considerable interindividual variation, but the pattern tends to be stable within the same patient and a high B cell/monocyte ratio may be associated with more rapid disease progression (Cepok et al., 2001). In vitro human B cells migrate across the brain endothelium more rapidly than autologous T cells (Alter et al., 2003), but almost nothing is known about chemokine regulation of B cell migration to the inflamed CNS (Ransohoff et al., 2003).

To address mechanisms of B cell trafficking to the CNS we analysed CXCL12 (SDF-1) and CXCL13 (BCA-1) in multiple sclerosis lesions. These two chemokines mediate germinal centre organization in lymphoid tissue (Allen et al., 2004).

The receptor for CXCL13, CXCR5, is expressed on virtually all B cells, a subset of T cells in blood and lymphatic tissue (Kim et al., 2001), and is transiently induced on T cells upon activation (Langenkamp et al., 2003). Recently, CXCL13 has been associated with inflammatory as well as homeostatic functions in the settings of neolymphoid development in the thyroid (Aust et al., 2004), in Helicobacter pylori-induced mucosa-associated lymphoid tissue and gastric lymphoma (Mazzucchelli et al., 1999), Sjoegren's syndrome (Salomonsson et al., 2002), rheumatoid arthritis (Shi et al., 2001) and ulcerative colitis (Carlsen et al., 2004).

CXCL12 is a potent chemoattractant for different immune cells including monocytes, T cells, B cells and plasma cells, and it regulates haematopoiesis. Beyond this, CXCL12 has complex functions in neurobiology. It is crucial for neuronal guidance in development (Lazarini et al., 2003; Klein and Rubin, 2004) and also expressed in mature brain (Stumm et al., 2002), but its function in the adult CNS is unclear.

In the present study we determined the presence of CXCL12 and CXCL13 in actively demyelinating and chronic inactive lesions. Because we found that CXCL13 is localized in infiltrating immune cells in active multiple sclerosis lesions, the regulation of the production of this chemokine by immune cells was analysed and found to be unusual. The CSF levels of these two chemokines were elevated and a linkage to intrathecal Ig production and immune cell subsets was analysed.

Materials and methods

Samples from patients and controls

A total of 14 snap-frozen CNS lesions from 6 multiple sclerosis patients were analysed. Patient 1 (two lesions analysed) had a rapidly progressive disease, which started at age 23 years and the patient died 3 years later. The patient did not receive immunomodulatory treatment within his last 2 years. Patient 2 (two lesions) had SP disease with a 20 year history and an onset at age 27 years. Starting 3 years before she died anti-CD4 therapy was given for 6 months followed by cyclophosphamide. Patient 3 (one lesion) had an SP course with a 17 year history with only short remissions and an onset age 39 years. Patient 4 (one lesion) had an SP course with a history of 8 years and an onset at age 21 years. Patients 3 and 4 did not have immunosuppressive therapy in the 2 years before death. Samples from Patients 1–4 were determined using a B cell monoclonal antibody (mAb) panel to contain B cells in perivenular cuffs and scattered in the parenchyma throughout these lesions. Patient 5 with a relapsing–remitting course has been described [Patient 2 in (Babbe et al., 2000)]. The two tissue blocks from this patient were active and probably in the early stage of lesion development as the lesions were removed 2 weeks after clinical manifestation of multiple sclerosis (Babbe et al., 2000). Scattered CD20+ B cells and CD138+ plasma cells/plasma blasts were present, but no lymph follicles were seen. Numerous CD14+ monocytes and CD68+ macrophages were present perivascularly throughout the whole lesion area and CD68+ cells were also abundant in the lesion parenchyma (Theil et al., 2005). This patient was untreated at the time of biopsy. The lesions from Patients 1–5 were classified as actively or recently demyelinating based on the presence of oil red O-positive macrophages and perivascular infiltrates. Patient 6 had an SP course with a history of 24 years and an onset at age 26. From this Patient, 6 blocks representing chronic inactive lesions were analysed. This patient had received mitoxantrone, the last time 11 months before he died. A total of 15 blocks from 3 adults without clinical or histological evidence of brain disease who died following accidents were also examined.

CSF samples were obtained from 30 patients with relapsing–remitting (RR) multiple sclerosis, 8 patients with primary progressive (PP) multiple sclerosis and 14 patients with SP multiple sclerosis. For comparison, CSF was also obtained from 28 patients with other inflammatory neurological diseases (OIND), including neuroborreliosis, viral inflammatory CNS disease (meningitis, encephalitis, zoster oticus and FSME), intracerebral abscess and mycoplasma meningitis. In addition, 14 patients with other non-inflammatory neurological disease (NIND) including headache, back pain, normal pressure hydrocephalus or seizure disorder, whose CSF did not show any evidence of inflammation such as intrathecal Ig production, presence of oligoclonal Ig, elevated cell numbers, or blood–brain barrier (BBB) disruption, were analysed. A disturbance of the BBB was diagnosed if the ratio of albumin in CSF and serum (Qalb) exceeded 7.4 × 10−3. Serum samples were obtained from 7 healthy blood donors, 8 of the NIND patients and 13 of the RR-multiple sclerosis patients. This study was approved by an Ethical Committee of the University of Munich.

Cell culture

Peripheral blood mononuclear cells (PBMC) were isolated from the blood of healthy donors by density gradient centrifugation. Monocytes were isolated by negative selection with immunomagnetic beads (Miltenyi, Bergisch Gladbach, Germany) and differentiated into macrophages for 9 days with M-CSF (20 ng/ml) as described (Nardelli et al., 2001). PBMC, monocytes and macrophages were cultured in RPMI supplemented with 5% FCS (PAN Biotech, Aidenbach, Germany). The following cytokines were added for 24 h: 25 ng/ml TNFα (Roche, Mannheim, Germany), 50 ng/ml IL-1β (R&D Systems, Wiesbaden, Germany), 100 U/ml IFNγ (Roche), 1000 U/ml IL-4 (PromoCell, Heidelberg, Germany), and subsequently supernatants were collected. RNA was isolated and cDNA was prepared using RNeasy Columns with a DNase digestion step (Qiagen, Hilden, Germany) and M-MLV reverse transcriptase (Promega, Mannheim, Germany).

Quantitative PCR

Frozen sections from tissue samples were collected in tubes, RNA was isolated and cDNA was prepared. A total of 14 lesions from the 6 patients described above and 15 samples from the three controls were analysed. Quantitative PCR was performed on the ABI 5700 (Applied Biosystems, Darmstadt, Germany) using the qPCR Core kit and uracyl-N-glycosylase for carry-over prevention (both Eurogentec, Seraing, Belgium). The reaction volume was 25 μl containing 25–50 ng RNA converted to cDNA. For all reactions the annealing temperature was 60°C. For detection of CXCL13 the following oligonucleotides were designed using Primer Express software (Applied Biosystems) and tested not to amplify genomic DNA up to 100 ng per reaction: forward (exon 1) CATCTCGACATCTCTGCTTCTCAT, reverse (exon 2) TCTCTTGGACACATCTACACCTCAA, probe FAM-AGCCTCTCTCCAGTCCAAGGTGTTCTGG-TAMRA.

Two housekeeping genes, cyclophilin A (peptidyl-prolyl isomerase A, PPIA) and GAPDH (both Applied Biosystems), were used.

The efficiency of these quantitative PCRs was determined as 2, meaning that the number of PCR products doubled at each cycle within the exponential phase of the PCR. The SD of the quantitative PCR experiments was calculated by the formula Math, which is based on Muller et al. (2002).


Enzyme-linked immunosorbent assay (ELISA) of CSF, serum and cell culture supernatant was performed on Maxisorp 96 well plates (Nunc, Wiesbaden, Germany) using DuoSet ELISA development kits for CXCL13 (R&D Systems). The sensitivity of this ELISA was ≤10 pg/ml. Samples from patients and controls were analysed blind. The SDF-1α Quantikine Immunoassay kit and SDF-1β ELISA antibody pair (monoclonal capture antibody MAB350, polyclonal biotinylated detection antibody BAF351) were used according to manufacturer's instructions (all from R&D Systems).

Analysis of CSF, blood and serum by standard diagnostic methods and flow cytometry

Paired CSF and serum samples were examined as described (Cepok et al., 2001, 2005). In brief, CSF and serum were analysed for protein, albumin and IgG by nephelometry and for oligoclonal bands by isoelectric focusing and IgG immunoblotting. Total CSF cells were enumerated in a counting chamber and further characterized by flow cytometry using mAbs recognizing CD3, CD14, CD19 and CD138. Paired CSF and blood samples were analysed as described (Kivisakk et al., 2003) using anti-CD4 PerCP, anti-CD45RO APC, anti-CD19 FITC (all from BD Biosciences), anti-CXCR5 PE (R&D Systems) and isotype matched control mAbs (BD Biosciences).


A total of eight lesions from Patients 5 and 6 were analysed. Cryostat sections were air-dried, fixed in acetone or formalin, and incubated with 1.5% methanolic hydrogen peroxide for 10 min and with 5% normal rabbit or goat serum in 1% BSA for 30 min. Monoclonal anti-human CXCL13 MAB801 (18 μg/ml), anti-CXCL12 MAB310, or anti-CXCL12 MAB350 (8 μg/ml; all R&D Systems) was applied overnight at 4°C. CXCL13 immunostaining was detected with biotinylated rabbit anti-mouse antibody (1 : 500) followed by peroxidase-conjugated streptavidin (1 : 300 for 30 min), or directly with peroxidase-conjugated goat anti-mouse antibodies (1 : 100). CXCL12 immunostaining was visualized using the PAP or SABC system according to manufacturer's instructions (all secondary reagents by DAKO, Hamburg, Germany). Sections were developed with diaminobenzidine as chromogenic substrate for up to 10 min and lightly counterstained with haematoxylin. Negative controls included omission of the first mAb and control Igs. Tonsils and adenoids were used as positive controls and to set up the staining protocol.


Localization of CXCL12 in healthy brain and in active and inactive multiple sclerosis lesions

CXCL12 was detected on blood vessels in each of the analysed brains from subjects without neurological disease (Fig. 1A). In multiple sclerosis lesions CXCL12 was also seen on astrocytes. Immunolabelling of blood vessel walls and astrocytes was seen in both active and chronic inactive lesions (Fig. 1B and C). Occasionally CXCL12 was also detected on a few cells in perivascular infiltrates. Both monoclonal antibodies against CXCL12 gave similar staining patterns, although MAB350 staining was more sensitive and clearer (Fig. 1).

Fig. 1

Localization of CXCL12 and CXCL13 in CNS tissue. Immunostaining with monoclonal antibodies against CXCL12 (A and B, MAB350; C, MAB310) and CXCL13 (D and E, MAB801) was performed on cryosections of normal brain (A), chronic inactive (B, Patient 6) and early active (CE, Patient 5) multiple sclerosis lesions. CXCL12 was detected on blood vessels in normal brain (A) and multiple sclerosis lesions (B and C, arrows). In addition, CXCL12 was detected on astrocytes in both chronic inactive (B) and early active (C) multiple sclerosis lesions (arrowheads). CXCL13 was only detected in active multiple sclerosis lesions, where it was localized within perivascular cuffs (D) and also in cells scattered within the inflamed parenchyma (E, arrows). Specific binding of the mAb was developed with diaminobenzidine as a substrate resulting in brown staining. Cell nuclei were stained with haematoxylin. Magnification bars indicate 20 μm.

CXCL13 is locally produced within actively demyelinating multiple sclerosis lesions

Immunohistological analysis localized CXCL13 within the perivascular infiltrates in actively demyelinating lesions (Fig. 1D). CXCL13 immunostaining was found also on extracellular matrix in these lesions (Fig. 1D and E). A subset of infiltrating cells in the parenchyma outside the perivascular area was labelled (Fig. 1E), but no CXCL13 was detected in any of the multiple sclerosis CNS samples outside the inflammatory lesions (data not shown). The six chronic inactive lesions contained scattered CD68+ macrophages. No cells displaying CXCL13 were detected in chronic inactive lesions or in brain from subjects without neurological disease.

This differential expression of CXCL13 in active versus inactive multiple sclerosis lesions was also detected by quantitative PCR. Eight actively demyelinating lesions from multiple sclerosis Patients 1–5, six inactive lesions from Patient 6, and 15 brain specimens from three control subjects were analysed by quantitative PCR to assess the presence of CXCL13. Expression of CXCL13 was elevated in actively demyelinating multiple sclerosis lesions (Fig. 2). Data were normalized to two housekeeping genes, GAPDH (Fig. 2) and PPIA (data not shown). Elevated levels of CXCL13 expression in actively demyelinating multiple sclerosis tissues were detected by both approaches. In contrast to the actively demyelinating multiple sclerosis lesions, CXCL13 transcripts were not elevated in chronic lesions. In six chronic lesions the transcript level of CXCL13 was noted with a median of 0.001% GAPDH, which is about the same level as we observed in normal CNS (Fig. 2).

Fig. 2

Local production of CXCL13 in active multiple sclerosis lesions. cDNA was obtained from eight actively demyelinating lesions of five multiple sclerosis patients, six chronic inactive lesions from one patient, and fifteen samples from three control donors. Within the actively demyelinating lesions, the values for the lesions from Patient 5 (early active lesions of a patient with RR-multiple sclerosis) are plotted as squares. Expression levels of individual tissue samples determined by quantitative PCR are shown as percentage of the housekeeping gene GAPDH. The median of each group is indicated. The P-value was calculated by the Mann–Whitney rank sum test.

CXCL12 and CXCL13 are elevated in CSF of multiple sclerosis patients

The two isoforms of CXCL12 (SDF-1α and SDF-1β) were analysed separately by ELISA. SDF-1β was not found in CSF samples from eight multiple sclerosis and five NIND patients with a detection limit of 40 pg/ml. In contrast, SDF-1α was found to be a constitutive component of CSF (median 223 pg/ml in NIND, Fig. 3A). In RR-multiple sclerosis patients CXCL12 levels were significantly higher than in NIND patients (Fig. 3A). OIND patients displayed even higher CXCL12 CSF levels than multiple sclerosis patients (Fig. 3A). Among these OIND patients those with a disturbed BBB showed a more extensive elevation of CXCL12 (Fig. 3A).

Fig. 3

Elevated levels of CXCL12/SDF-1α and CXCL13 in the CSF of multiple sclerosis and OIND patients. CSF samples were analysed for CXCL12/SDF-1α (A) and CXCL13 (B) by ELISA. The horizontal lines indicate the medians. P-values were calculated using the Mann–Whitney rank sum test. A significant difference was observed between NIND and multiple sclerosis patients (P = 0.002) for CXCL12 (A) and CXCL13 (B). If serial samples from one patient were available, only the first one was included in this figure.

CXCL13 was only detected in 1 out of 14 CSF samples from NIND patients, but was >30 pg/ml in the CSF of 17 out of 30 RR-multiple sclerosis patients. In contrast, the CXDL13 serum levels were in a similar range in RR-multiple sclerosis patients (median 51 pg/ml) and control subjects (56 pg/ml). In most of the multiple sclerosis patients with detectable CXCL13 in the CSF, this value was even higher than the corresponding serum level indicating the intrathecal production of this chemokine.

The CXCL12 and CXCL13 levels were also analysed in patients with PP- and SP-multiple sclerosis. Two out of eight patients with PP-multiple sclerosis and 4 out of 14 with SP-multiple sclerosis had elevated levels of CXCL13 in the CSF. Patients with PP- and SP-multiple sclerosis tended to have higher CXCL12 levels in the CSF than NIND patients (Fig. 3A).

The majority of the multiple sclerosis patients (RR-multiple sclerosis: 26/30; PP-multiple sclerosis: 6/8; SP-multiple sclerosis: 13/14) displayed no evidence in the CSF of a disturbed BBB, which is in accordance with the well-characterized features of CSF from multiple sclerosis patients. Only these multiple sclerosis patients without a disturbance of the BBB were included in this analysis (Fig. 3) to exclude a potential effect of pathological transfer of CXCL12 or CXCL13 from blood to CSF. The samples from multiple sclerosis patients with a disturbance of the BBB as judged by CSF analysis had the following chemokine levels for CXCL12: RR-multiple sclerosis 120, 995, 1207 and 1272 pg/ml; PP-multiple sclerosis 607 and 668 pg/ml; SP-multiple sclerosis 1229 pg/ml, and for CXCL13: RR-multiple sclerosis <10, 35, 93 and 487 pg/ml; PP-multiple sclerosis (two patients) <10 pg/ml; SP-multiple sclerosis 13 pg/ml.

Overall, CXCL13 but not CXCL12 levels showed a striking heterogeneity. In each group of multiple sclerosis patients there was a subset of patients with low or undetectable level of CXCL13 and another subset with high levels of CXCL13 in the CSF. The CXCL12 levels followed a Gaussian distribution in all subgroups, but not the CXCL13 levels. Therefore parametric tests also could be used to compare the CXCL12 levels. The t-test (P = 0.005 for NIND versus RR-multiple sclerosis patients) gave essentially the same result as shown by the Mann–Whitney rank sum test (Fig. 3A).

Among the OIND patients there were great variations in the levels of CXCL13. Both OIND patients without a disrupted BBB and high CXCL13 levels had neuroborreliosis. Not only neuroborreliosis, but also other inflammatory diseases of the CNS were associated with high levels of CXCL13 in the CSF. Eight OIND patients had a CXCL13 level in the CSF >1000 pg/ml. Five of them had acute neuroborreliosis, one a viral encephalitis, one a cerebral abscess and one was diagnosed with a cranial neuritis distinct from neuroborreliosis, since he was repeatedly negative by PCR and antibody testing. Patients with a BBB disturbance had higher CXCL13 levels in the CSF than those without (Fig. 3B). We assume that the higher level of CXCL13 in the CSF of the subgroup of OIND patients with a disturbed BBB due to the higher intensity of CNS inflammation, a more pronounced meningeal involvement, and possibly transfer from blood.

CXCL13 in the CSF of multiple sclerosis patients correlates with intrathecal Ig production

The QIgG (IgGCSF/IgGserum) reflecting intrathecal IgG production in the absence of BBB damage correlated strongly with the CXCL13 levels in RR-multiple sclerosis patients (P < 0.001, Fig. 4H). In contrast, neither the Qalb (R = 0.26, P = 0.17) nor the serum CXCL13 concentrations (R = 0.04, P = 0.9) correlated with the CSF CXCL13 levels. To exclude the possibility of a pathological transfer from blood, only multiple sclerosis patients without CSF signs of a disturbance of the BBB were included for this analysis. In contrast to CXCL13, the CXCL12 levels in the CSF did not correlate with the intrathecal Ig production (Fig. 4C).

Fig. 4

CXCL13, but not CXCL12 in CSF from RR-multiple sclerosis patients correlates strongly with the presence of B cells, plasma blasts, intrathecal Ig-production and T cells. CSF samples from RR-multiple sclerosis patients without an elevated Qalb as a sign of BBB disruption were analysed. Leucocyte subset counts per μl of all CD19+ B cells (A and F), CD19+CD138+ plasma blasts (B and G), CD3+ T cells (D and I), and CD14+ monocytes (E and J) were determined by flow cytometry and plotted against the CXCL12 and CXCL13 level in the CSF. IgG was determined in paired CSF and plasma samples and the QIgG (IgGCSF/IgGserum) was calculated and plotted against CXCL12 (C) and CXCL13 (H). Spearman coefficients of correlation and P-values were calculated.

CXCL13 in the CSF of multiple sclerosis patients correlates with the presence of B cells, plasma blasts and T cells

We then analysed the relationship between immune subsets and levels of CXCL12 and CXCL13 in the CSF of RR-multiple sclerosis patients. As previously, only CSF samples from multiple sclerosis patients without CSF signs of a BBB disturbance were included. Numbers of CD3+ T cells, CD14+ monocytes, CD19+ B cells and CD19+CD138+ plasma blasts were determined in the same CSF samples. Whereas the vast majority of CSF cells, ∼70–90%, consisted of T cells, the levels of the other immune subsets showed considerable differences. The number of B cells ranged from 11.2 to 3892 cells/ml, that of plasma blasts from 0 to 1657 cells/ml and that of monocytes from 55.7 to 1901 cells/ml.

The levels of CXCL13 correlated strongly and significantly with the number of all B cells, plasma blasts and T cells in the CSF of multiple sclerosis patients (Fig. 4). The number of monocytes also correlated with the CXCL13 level, but less strongly (Fig. 4J). Multiple sclerosis patients with a higher B cell/monocyte ratio had more elevated CXCL13 levels in the CSF. The B cell/monocyte ratio correlated significantly with the CXCL13 level (P = 0.002; R = 0.6). In contrast to CXCL13, the CXCL12 levels in the CSF did not show a significant correlation with numbers of T cells, B cells, plasma blasts or monocytes in the CSF (Fig. 4A–E).

We analysed CSF immune cell subsets, QIgG, CXCL12 and CXCL13 levels also in OIND patients. These patients had more pronounced inflammatory changes than the multiple sclerosis patients, as indicated by a higher percentage of OIND patients with a disturbed BBB, and higher cell numbers in the CSF (median RR-multiple sclerosis 11 cells/μl, OIND 60 cells/μl). The specific findings in the OIND group were as follows. Both CXCL12 and CXCL13 levels showed a strong positive correlation with B cells, plasma blasts and T cells: CD19+ cells (CXCL12: P = 0.02, R = 0.59/CXCL13: P < 0.001, R = 0.84), CD19+CD138+cells (P < 0.001, R = 0.75/P < 0.001, R = 0.79) and CD3+ cells (P < 0.001, R = 0.82/P < 0.001, R = 0.76). For CD14+ cells, the positive correlation was obvious for CXCL12 (P = 0.006, R = 0.67), but the association with CXCL13 was less pronounced (P = 0.047, R = 0.53) (CXCL12: 15 samples from 10 patients; 10 samples with a disturbed BBB; CXCL13: 14 samples from 9 patients; 9 samples with disturbed BBB). Further we noted also a correlation between CXCL12 and QIgG (P < 0.001, R = 0.64; 37 samples from 25 patients; 26 samples with disturbed BBB), and CXCL13 with QIgG (P < 0.001, R = 0.76; 38 samples from 26 patients; 26 samples with disturbed BBB).

Expression of CXCR5 by T cells and B cells in the CSF

CXCR5 expression was analysed by flow cytometry on parallel samples from CSF and blood. Virtually all CSF and blood B cells expressed the CXCL13 receptor CXCR5 and ∼20% of the CD4+ cells in blood or CSF expressed CXCR5 (Fig. 5). The CSF is known to be enriched for CD4+CD45RO+ cells. Also for this immune cell subset a similar percentage of CXCR5+ cells was observed in CSF and blood (Fig. 5). Patients with or without inflammation in the CNS did not show a striking difference in terms of percentage of CXCR5+ immune cell subsets. The CXCR5 expression in CSF compared with blood was slightly reduced, most obviously for the CD4+CD45RO+ and CD19+ subsets. Whether this reflects a ligand-induced receptor modulation remains to be determined.

Fig. 5

Expression of CXCR5 on CD19+ cells, CD4+ cells, and CD4+CD45RO+ cells in the CSF. CXCR5 expression was analysed on parallel samples from blood and CSF within the indicated immune cell populations. Samples from a total of 12 patients were examined (2 multiple sclerosis, 2 OIND, 8 NIND). Eleven samples were available for analysis of blood cells, while CSF cells could be analysed in twelve (CD4+ and CD4CD45+ cells) and nine (CD19+ cells) samples, respectively. Not all data points are easily distinguishable as some are very close together. All samples displayed a similar percentage of CXCR5+ cells in blood and CSF.

Production of CXCL13 by blood-derived cells and regulation by cytokines

Since CXCL13 protein was localized to infiltrating immune cells in actively demyelinating multiple sclerosis lesions, we asked whether the CXCL13 production by PBMC was regulated by inflammatory cytokines and during macrophage differentiation. Analysis of subsets of PBMC indicated that the CXCL13 transcribing cells were within the CD14+ monocyte fraction (Fig. 6 and data not shown). The regulation of CXCL13 production by immune cells was studied in a total of seven experiments, two with PBMC, two with purified monocytes and three with differentiated macrophages. Figure 6 shows representative experiments. Purified monocytes and macrophages were stimulated with the indicated cytokines. The amount of CXCL13 secreted was measured by ELISA. In parallel the level of CXCL13 transcripts was determined by quantitative PCR. The production of CXCL13 was enhanced by IL-1β and TNFα, whereas IL-4 and IFNγ inhibited the spontaneous and induced CXCL13 production (Fig. 6). Similar results as with purified monocytes were obtained with PBMC (data not shown). Differentiation of monocytes to macrophages with M-CSF greatly enhanced the spontaneous CXCL13 expression. These macrophages constitutively produced a high amount of CXCL13 that was not further enhanced by TNFα and IL-1β. Again, an inhibition with IL-4 and IFNγ was observed. The CXCL13 protein release measured by ELISA mirrored the transcriptional regulation analysed by quantitative PCR (Fig. 6).

Fig. 6

CXCL13 production is induced by TNFα and IL-1β, but inhibited by IL-4 and IFNγ. Differentiation into macrophages induces expression of CXCL13. Negatively isolated monocytes (A) or monocytes that had been differentiated to macrophages with M-CSF (B) were stimulated with the indicated cytokines for 24 h (PCR) or 48 h (ELISA). The level of CXCL13 transcripts in relation to the housekeeping gene GAPDH was determined by quantitative PCR. The amount of secreted CXCL13 was determined by ELISA. Note the different scales on the y-axis in monocyte and macrophage cultures. The error bars of the quantitative PCR represent a compound SD, considering the SD of both the housekeeping gene and the target gene calculated as described in Materials and methods. The error bars of the ELISAs represent SDs of replicates.


In order to gain further insight into mechanisms of B cell trafficking into the inflamed CNS, the chemokines CXCL12 and CXCL13 were analysed. Both chemokines mediate together the dark zone and light zone organization of germinal centres in lymphatic tissue (Allen et al., 2004). We report that CXCL12 is elevated in both active and inactive multiple sclerosis lesions, whereas CXCL13 was detectable only in active lesions. Levels of both CXCL12 and CXCL13 were elevated in the CSF. CXCL13, but not CXCL12, correlated strongly with intrathecal Ig synthesis, presence of T cells, B cells and plasma blasts.

We detected CXCL12 on blood vessel walls in the grey and white matter of subjects without neurological disease. Previous studies indicated that neurons and endothelial cells are the main producers of CXCL12 in the healthy adult CNS (Klein and Rubin, 2004). In multiple sclerosis lesions this chemokine was strongly displayed on astrocytes and blood vessels. The up-regulation of CXCL12 in multiple sclerosis lesions might be due to two different pathological alterations. First, the inflammation could induce enhanced production of CXCL12, since this chemokine is prominent in inflammatory skin (Pablos et al., 1999) and on the synovium in rheumatoid arthritis, where CXCL12 production was enhanced by CD40-stimlation (Nanki et al., 2000). IL-1 and TNF-α stimulate CXCL12 production by human astrocytes (Ambrosini et al., 2005). Neuroinflammation results in elevated levels of CXCL12 in the CSF [this study and (Pashenkov et al., 2003)]. The higher CSF level of CXCL12 in OIND patients with a disrupted BBB as compared with multiple sclerosis patients might be related to a more pronounced neuroinflammation in these cases. A second mechanism of CXCL12 induction in multiple sclerosis lesions could be based on a hypoxia-like metabolic state. Hypoxia and the hypoxia-induced transcription factor HIF-1 stimulate the production of CXCL12 (Hitchon et al., 2002; Ceradini et al., 2004). CXCL12 is up-regulated in the ischaemic penumbra following stroke and there localized on astrocytes (Hill et al., 2004). Interestingly, a hypoxia-like tissue injury was recently considered to be a component of multiple sclerosis lesions (Graumann et al., 2003; Lassmann et al., 2003). This hypoxia-like metabolic state in multiple sclerosis lesions can principally be induced via two ways, firstly by defective microcirculation due to inflammatory damage of the vascular wall or secondly by toxic metabolites that interfere with mitochondrial energy metabolism (Lassmann, 2003).

The consequences of up-regulated CXCL12 could be manifold. CXCL12 is a chemoattractant for T cells, monocytes, B cells, and also plasma cells and could contribute to the maintenance of immune cells in the CNS. CXCL12 is, in particular, important for the maintenance of plasma cells, since B cells lose most of their chemokine receptors during differentiation to plasma cells, but keep CXCR4, the receptor for CXCL12 (Hargreaves et al., 2001). The effects of CXCL12, however, exceed regulation of migration, e.g. it induces the production of TNF, IL-1 and CCL5/RANTES in astrocytes, which might promote inflammation and tissue destruction (Han et al., 2001). CXCL12 can induce neuronal apoptosis in certain conditions (Hesselgesser et al., 1998) and CXCL12 immunoreactivity has been linked to AIDS-related dementia (Langford et al., 2002). On the other hand, CXCL12 can promote survival of neurons (reviewed, Lazarini et al., 2003). Interestingly, cleavage by metalloproteinase (MMP)-2 converts CXCL12 into a neurotoxic form (Zhang et al., 2003). MMPs including MMP-2 were found to be present at elevated levels in post-mortem multiple sclerosis tissue (reviewed, Yong et al., 2001). Therefore, the elevated level of CXCL12 in multiple sclerosis lesions we describe here could contribute to the neuronal damage and axonal loss observed in multiple sclerosis (Kornek and Lassmann, 1999).

The cellular localization and the expression pattern of CXCL13 were very different from that of CXCL12. CXCL13 was undetectable in the normal CNS, up-regulated in active multiple sclerosis lesions, but not in chronic inactive lesions. This differential expression of CXCL13 in active versus inactive lesions was seen by both quantitative PCR and immunohistology of multiple sclerosis lesions. Also our CSF analysis supported this conclusion: CXCL13 levels were linked to inflammatory activity, in particular intrathecal Ig production as well as accumulation of B cells and T cells in the CSF of multiple sclerosis patients. This observation has implications for the still largely unknown mechanisms for recruitment of B cells into the inflamed CNS in multiple sclerosis, and extends the biology of CXCL13/CXCR5 beyond homeostatic trafficking to the regulation of chronic inflammation in the CNS. CXCL13 was also elevated in the CSF of OIND patients. Our results confirm a recent report about elevation of CXCL13 in neuroborreliosis (Rupprecht et al., 2005). In addition, we show that a high CXCL13 level can also be observed in non-neuroborreliosis patients, in particular in those with a disturbed BBB. Our PCR analysis of a total of 14 multiple sclerosis lesions and the CSF examinations indicated that intracerebral CXCL13 production is a feature of acute inflammation in the CNS.

Our immunohistological staining localized CXCL13 to infiltrating immune cells in early active multiple sclerosis lesions with highly abundant CD68+ macrophages in the parenchyma and around blood vessels (Theil et al., 2005). In the perivascular cuffs CXCL13+ staining could be seen between the infiltrating cells and infiltrating cells were labelled in the parenchyma. Additionally, CXCL13 staining was also seen on extracellular matrix within these lesions. This is in line with two other studies, one that co-localized CXCL13 in fibronectin+ fibrils in ulcerative colitis and rheumatoid arthritis (Carlsen et al., 2004) and another one that described CXCL13 in the wall of a blood vessel in a chronic active multiple sclerosis lesion, but not directly to any cells (Corcione et al., 2004). The localization of CXCL13 to infiltrating cells we have described here, probably monocytes/macrophages, in actively demyelinating multiple sclerosis lesions is in line with two other studies, one which localized CXCL13 to monocytes/macrophages in rheumatoid arthritis and ulcerative colitis (Carlsen et al., 2004) and another which found up-regulation of CXCL13 in an animal model of Lyme borreliosis and detected CXCL13 on myeloid cells (Narayan et al., 2005). This inspired us to analyse how this chemokine, classified as homeostatic, can be regulated by inflammatory cytokines and we identified four cytokines that affect its production in a complex way. The monocyte fraction of PBMC produced CXCL13 after stimulation with IL-1β and TNFα, whereas both IFNγ and IL-4 inhibited this. These findings were unexpected as IL-4 and IFNγ frequently have opposing effects on monocyte and macrophage activation. In harmony with a recent report (Carlsen et al., 2004) we observed in vitro that differentiation of monocytes to macrophages greatly enhanced their expression of CXCL13. It is presently unclear how the contrasting effects of IL-4 and IFN-γ on the one hand versus TNF-α, IL-1β, and macrophage differentiation on the other hand affect the CSF level of CXCL13.

A recent study reported immunostaining of CXCL13 in lymphoid follicle-like structures in the cerebral meninges in two out of three patients with SP-multiple sclerosis but not in patients with RR-multiple sclerosis and PP-multiple sclerosis (Serafini et al., 2004). CXCL13 was not detected by immunohistochemistry in the parenchyma of chronic active or chronic inactive multiple sclerosis lesions (Serafini et al., 2004), which is in line with our analysis of six chronic inactive lesions by histology and quantitative PCR. We observed CXCL13 immunostaining in a full blown early active lesion; no such lesion was included in a study that failed to detect CXCL13 on infiltrating cells within the lesions (Serafini et al., 2004). Our study detected a differential expression of CXCL13 in actively demyelinating and inactive lesions by combining quantitative PCR and immunostaining.

Overall, a picture is emerging of two different sites of CXCL13 production in the multiple sclerosis CNS. In lesions with strong inflammation the infiltrating macrophages could be a source of this chemokine, whereas in some patients with an SP-multiple sclerosis, CXCL13 might be produced in follicle-like structures in the meninges. Our study shows that the CXCL13 level in the CSF is linked to the degree of inflammation. Only a few SP-multiple sclerosis and PP-multiple sclerosis patients had considerable levels of CXCL13 in the CSF, whereas in the majority of CP patients very little or no detectable CXCL13 was found in the CSF. It remains to be established whether follicle-like structures, which were described in the meninges of two out of three patients with SP-multiple sclerosis (Serafini et al., 2004), are linked to a high level of CXCL13 in the CSF.

To analyse the potential functional relevance of the intracerebrally produced CXCL13, we examined expression of its receptor CXCR5 by CSF cells. In CSF virtually all B cells and ∼20% of the CD4+ T cells or of the CD4+CD45RO+ T cells expressed CXCR5, the receptor of CXCL13. The expression of CXCR5 was observed at similar levels in blood and CSF cells. This resembles previous observations on the chemokine receptors CCR1-6 (Ransohoff, 2002) and CCR7 (Kivisakk et al., 2004), whereas CXCR3 was present at higher levels on CSF cells than on the corresponding population in blood (Kivisakk et al., 2002).

We went on to analyse whether the intracerebral production of CXCL13 could be linked to recruitment or persistence of a particular immune cell subset. The CXCL13 level showed a strong correlation with the intrathecal Ig-production and also a significant correlation with the presence of B cells and plasma blasts, which are the main B cell effector subset during the course of multiple sclerosis (Cepok et al., 2005). This is the first report that correlates the number of B cells and plasma blasts in CSF to a particular chemokine and suggests that CXCL13 is important in directing B cells to the inflamed CNS.

Remarkably, the CXCL13 level in the CSF also correlated very strongly with the presence of T cells in the CSF. Whereas the role of CXCL13 and its receptor CXCR5 on B cells in the development of lymph follicles is well established (Ansel et al., 2000), far less is known about functions of CXCR5 on T cells. CXCR5-expressing T cells are functionally heterogeneous (Kim et al., 2001). In tonsils, a subset of CXCR5+ T cells are specifically localized in germinal centres and function as follicular B helper T cells (Moser and Ebert, 2003). CXCR5 expressing T cells in blood are, unlike the CXCR5+ T cells in the germinal centres, in a resting state and need to be activated to help B cells (Kim et al., 2001). In vitro CXCR5 is transiently induced on T cells after polyclonal activation (Langenkamp et al., 2003). The functional properties of CXCR5+ T cells in the CSF have yet to be defined.

In addition to its regulation of cell migration, CXCL13 is known to have several other effects, including a stimulatory cytokine loop: CXCL13 up-regulates membrane LTα1β2 on B cells, promoting the development of follicular dendritic cells and thus the production of CXCL13 itself (Ansel et al., 2000). This raises the possibility that CXCL13 in the multiple sclerosis lesion is not only involved in immune cell recruitment, but also in the regulation of the cytokine environment.

In summary, we have demonstrated elevated levels of CXCL12 in active as well as inactive lesions. CXCL12 might be involved in the maintenance of immune cells, in particular plasma cells, in the CNS. CXCL13 was up-regulated in active, but not inactive lesions. The intracerebral CXCL13 production strongly correlated with intrathecal Ig production, as well as the number of B cells, plasma blasts and T cells in CSF. Thus, CXCL13 is identified as a chemokine that is involved in recruitment of B cells and T cells to the inflamed CNS. Therefore, targeting CXCL13–CXCR5 interactions may be a promising novel therapeutic approach in multiple sclerosis.


We thank D. Zech for technical assistance, Dr S. Wagenpfeil for help with the statistical analyses, and Dr A. Flügel and Dr D. Jenne for comments on the manuscript. This work was supported by the DFG (SFB 571, GRK 688, He2386/5-1 and 5-2), the DMSG, the Verein zur Förderung der Therapieforschung für MS-Kranke and the USA National Institutes of Health (PO1 NS38667 to R.M.R.). The Institute for Clinical Neuroimmunology is supported by the Hermann and Lilly Schilling Foundation.


View Abstract