CCL21 Overexpressed on Lymphatic VesselsDrives Thymic Hyperplasia in Myasthenia

Sonia Berrih-Aknin, PhD,1,2 Nathalie Ruhlmann, MSc,1,2 Jacky Bismuth, MSc,1,2

Geraldine Cizeron-Clairac, PhD,1,2 Einat Zelman,3 Idit Shachar, PhD,3 Philippe Dartevelle, MD,4

Nicole Kerlero de Rosbo, PhD,1,2 and Rozen Le Panse, PhD1,2

Objective: Myasthenia gravis (MG), a neuromuscular disease mediated by anti-acetylcholine receptor (AChR) autoantibodies,is associated with thymic hyperplasia characterized by ectopic germinal centers that contain pathogenic antibody-producing Bcells. Our thymic transcriptome study demonstrated increased expression of CCL21, a recruiter of immune cells. Accordingly,we are investigating its implication in MG pathogenesis.Methods: The expression of CCL21 and its CCR7 receptor was analyzed by enzyme-linked immunosorbent assay andfluorescence-activated cell sorting, respectively. Chemotaxis of T and B cells to CCL21 was measured by transwell assay. Thenature of the thymic cells overexpressing CCL21 was investigated by immunochemistry and laser-capture microdissection com-bined with real-time PCR.Results: We demonstrate that CCL21 is overexpressed specifically in hyperplastic MG thymuses, whereas there is no variationin CCR7 levels on blood cells. We show that although CCL21 attracts both human T and B cells, it acts more strongly on naiveB cells. CCL21 overexpression is normalized in corticoid-treated MG patients, suggesting that targeting this chemokine couldrepresent a new selective treatment, decreasing the abnormal peripheral lymphocyte recruitment. Moreover, we locate proteinand messenger RNA overexpression of CCL21 to specific endothelial vessels. Investigation of the nature of these vessels dem-onstrated different angiogenic processes in MG thymuses: high endothelial venule angiogenesis and lymphangiogenesis. Unex-pectedly, CCL21 overexpression originates from afferent lymphatic endothelial vessels.Interpretation: We postulate that thymic overexpression of CCL21 on specialized lymphatic vessels results in abnormal pe-ripheral lymphocyte recruitment, bringing naive B cells in contact with the inflammatory environment characteristic of MGthymuses, where they can be sensitized against AChR.

Ann Neurol 2009;66:521–531

Acquired myasthenia gravis (MG) is a neurological au-toimmune disease caused by autoantibodies againstcomponents of the neuromuscular junction and leadingto disabling fatigability. Seropositive MG is caused byanti–acetylcholine receptor (AChR) autoantibodies andrepresents 85% of patients.1 Functional and morpho-logical abnormalities of the thymus occur frequently,and 50 to 60% of the seropositive patients exhibit thy-mic hyperplasia of lymphoproliferative origin with ec-topic germinal center (GC) development.2 These thy-mic abnormalities are correlated with the anti-AChRantibody titer,2 which decreases after thymectomy.3

The hyperplastic thymus includes all the componentsof the anti-AChR response: the AChR,4 B cells pro-ducing anti-AChR antibodies,5 and anti-AChR autore-

active T cells.6 The thymus thus plays a pivotal role inMG. An understanding of the mechanisms leading toectopic GC formation should shed light on the patho-genesis of this disease and help us to identify new ther-apeutic targets to avoid thymectomy or the nonselec-tive use of glucocorticoids.

Our analyses of the thymic transcriptome demon-strated an important chemotactic activity with an in-creased expression of CXCL13 probably involved inthe generalized B-cell infiltration observed for all MGpatients. Moreover, we also observed a specific overex-pression of CCL21 in thymuses of MG patients withthymic hyperplasia.7

CCL21 and CCL19 are two CCR7-binding chemo-kines that play an important role in thymopoiesis.

From the 1Centre National de la Recherche Scientifique-UniteMixte de Recherche 8162, Institut Paris-Sud Cytokines, HopitalMarie Lannelongue, Le Plessis-Robinson; 2Universite Paris-Sud,91405 Orsay, France; 3Department of Immunology, the WeizmannInstitute of Science, Rehovot, Israel; and 4Service de Chirurgie Tho-racique et Vasculaire et de Transplantation Cardiopulmonaire, Ho-pital Marie Lannelongue, Le Plessis Robinson, France.

Address correspondence to Dr Le Panse, CNRS UMR 8162, Ho-pital Marie Lannelongue, 133, Avenue de la Resistance, 92350 LePlessis Robinson, France. E-mail: rozen.lepanse@u-psud.fr

Potential conflict of interest: Nothing to report.

Received Jun 5, 2008, and in revised form Nov 27. Accepted forpublication Dec 5, 2008. Published online in Wiley InterScience(www.interscience.wiley.com). DOI: 10.1002/ana.21628

© 2009 American Neurological Association 521

They are mainly secreted by medullary thymic epithe-lial cells and promote the migration of CD4� orCD8� thymocytes toward the medullary zone.8 Theyalso play distinct roles because of their different pre-sentation patterns on distinct stromal components:CCL21 participates in prothymocyte recruitment inthe thymus,9 whereas CCL19 is involved in the exporttoward the periphery of mature thymocytes.10

CCL21 is also known to play a central role in im-mune surveillance and defense by controlling the cir-culation of T cells and dendritic cells within lymphoidand peripheral organs. In secondary lymphoid organs(SLOs), CCL21 enables naive T cells to encounter sen-sitized dendritic cells by mediating their recruitmentthrough its expression on high endothelial venules(HEVs) and afferent lymphatic vessels, respectively.11

Indeed, CCR7-deficient mice and paucity of lymphnode T-cell (plt) mice that do not express CCL19 andCCL21 show defective migration of lymphocytes anddendritic cells into T-cell zones.12

In view of the physiological role of CCL21 withinthe thymus, we decided to investigate the pathologicalrelevance of CCL21 overexpression in hyperplastic MGthymuses. We demonstrate that CCL21 is a potent re-cruiter of human naive B cells. We also show that dif-ferent angiogenic processes, HEV angiogenesis andlymphangiogenesis, occur in hyperplastic thymuses,and that CCL21 overexpression is concentrated onlymphatic endothelial vessels, suggesting a unique rolefor these vessels in an abnormal recruitment of periph-eral cells leading to thymic MG pathogenesis.

Subjects and MethodsSamplesThymic fragments (50–100mg) and sera were obtained fromMG female patients (16–42 years old) after thymectomy orfrom babies (1 week to 1 year old) and women (16–34 yearsold) undergoing cardiovascular surgery at the Marie Lan-nelongue Chirurgical Center (Le Plessis Robinson, France).Thymuses were thus classified as follows: babies (BB; n �13), non-MG adults (Ad; n � 12), MG patients with lowthymic hyperplasia (ML; with 2 or fewer GCs per section;n � 10) or high thymic hyperplasia (MH; with 3 or moreGCs per section; n � 21), or corticosteroid-treated MG pa-tients (Cortico, n � 9). Only seropositive MG caucasian fe-male individuals, without thymoma or other known diseases,were included. All MG patients included in this study hadreceived only anticholinesterase drugs.

Chemotactic assays and flow cytometry analyses were con-ducted with peripheral blood lymphocytes (PBL) fromhealthy adult donors. All studies on thymuses and bloodsamples were approved by the local ethics committee (CCP-PRB, Kremlin-Bicetre, France; agreement 06-018).

Immunohistochemistry on Thymic SectionsFrozen thymic sections (7�m) were fixed in ice-cold acetonefor 20 minutes. At least three different non-MG or MG do-

nors were used for each immunostaining. In immunohisto-logical analyses performed with a Leica DMRB microscope(Rueil-Malmaison, France), the relevant molecules were de-tected using the following combinations of antibodies: (1)goat anti–human CCL21 (AF366; R&D Systems, Lille,France) and Alexa-Fluor 488 rabbit anti–goat IgG (A21222;Invitrogen, Cergy-Pontoise, France) or biotinylated horse an-ti–goat IgG followed by Alexa-Fluor 594 streptavidin(S11227; Invitrogen, Eugene, OR), (2) mouse anti–humankeratin (A575; Dako, Trappes, France) and Alexa-Fluor 488goat anti–mouse (GAM) IgG (A11029; Invitrogen), (3) ratIgM anti–mouse (cross-reacting with human) peripheralnode addressin (PNAd) carbohydrate epitopes (MECA 79,553863; BD Biosciences, Le Pont-De-Claix, France) and bi-otinylated mouse anti–rat IgM (550330; BD Biosciences)followed by Alexa-Fluor 594 streptavidin, and (4) mouse an-ti–vascular endothelial growth factor receptor 3 (VEGFR3;MAB3757, Abcys, Paris) and Alexa-Fluor 488-GAM IgG.

In immunohistological analyses using the microarray scan-ner, serial thymic sections were stained with mouse anti–human CD21 antibody (555421; BD Biosciences) andR-phycoerythrin (RPE)-GAM-IgG (R0480; Dako), or anti-PNAd and biotinylated mouse anti-rat IgM followed by PEstreptavidin (349023; BD Biosciences). Sections were ana-lyzed on the 428 Affymetrix scanner (Aviso, Croissy-Beaubourg, France).

Enzyme-Linked Immunosorbent AssayAll antibodies and recombinant chemokines were from R&DSystems. CCL21 (AF366) and CCL19 (MAB361) antibodieswere diluted at 1�g/ml in coating buffer (50mM NaHCO3

and 50mM Na2CO3, pH 9.4) and incubated overnight at4°C. Thymic extracts (1.5mg/ml proteins) or standards(CCL21 or CCL19) were incubated for 3 hours, and subse-quently, 0.25�g/ml biotinylated anti-CCL21 (BAF366) oranti-CCL19 (BAF361) and streptavidin-horseradish peroxi-dase (Beckman Coulter, Villepinte, France) were added. Tet-ramethylbenzidine was used for color development, andplates were read at 450nm on an MRX-microplate reader(DYNEX Technologies, ThermoLabsystems, Cergy-Pontoise,France).

FACScan Flow CytometryThe expression of CCR7 was analyzed with anti-CCR7(MAB197; R&D Systems) and RPE-GAM (R0480; Dako)or anti-CCR7-fluorescein isothiocyanate (FITC) (FAB197F;R&D Systems) on T and B cells labeled with RPE-Cy5 anti-CD4 (C7069; Dako), RPE anti-CD8 (R0806; Dako), andRPE anti-CD19 (R0808; Dako).

Chemotaxis AssayHuman peripheral blood mononuclear cells from healthyadults were recovered on Ficoll gradient and incubated over-night in culture flasks in a quiescent medium: RPMI-1640-Glutamax-I supplemented with 0.5% fetal calf serum (In-vitrogen). Chemotaxis assays were performed using transwellchambers (5�m pore size; Corning/Fisher Scientific-Bioblock, Illkirch, France). Transwell upper and lower cham-bers were loaded with 1 � 106 cells in 100�l, and CCL21or CXCL13 (R&D Systems) in 600�l of quiescent medium,

522 Annals of Neurology Vol 66 No 4 October 2009

respectively. After a 3-hour migration step at 37°C, cells inthe lower chambers were labeled with antibodies for T- orB-lymphocyte subpopulations from BD Bioscience forCD69-FITC (555530) and CD25-PE (555431), fromAmersham Bioscience (Orsay, France) for CD27-FITC(M1764), or from Dako for CD3-FITC (F0818), CD19-RPE, CD45RO-RPE (R0843), CD25-RPE (R0811), andIgD (M0703). Cells were analyzed by fluorescence-activatedcell sorting (FACS) after 1-minute acquisition.

Chemotaxis assays on mouse lymphocytes were conductedas described previously.13 In brief, spleen cells were separatedas CD45R (B220)-positive (B cells) or CD45-negative (Tcells) with antimagnetic beads using the magnetic-activatedcell sorting system (Miltenyi Biotec, Auburn, CA). Upperand lower chambers of transwells were loaded with 5 � 106

cells in 100�l, and ccl21 or cxcl13 (400ng/ml; PeproTech,Rocky Hill, NJ) in 600�l of quiescent medium, respectively.After a 3-hour migration step at 37°C, cells in the lowerchambers were analyzed by FACS as described earlier.

Actin Polymerization AssayPolymerization of actin filaments was analyzed as describedpreviously.13 In brief, mouse spleen cells collected as for che-motactic assays were incubated for 15 seconds at 37°C with100ng/ml ccl21 or cxcl13, and fixed for 10 minutes in 3.7%formaldehyde. Cells were then permeabilized, stained withFITC-phalloidin, and cytoskeleton rearrangements were an-alyzed by FACS.

Laser-Capture MicrodissectionMicrodissection was conducted as described previously.14

Here sections were stained with anti-CCL21 (AF366) andbiotinylated rabbit anti–goat IgG (E0466; Dako), followedby streptavidin-horseradish peroxidase and counterstainedwith hematoxylin. CCL21-positive blood vessels and GCswere microdissected separately before RNA extraction.

Reverse Transcription and Real-Time PolymeraseChain ReactionTotal RNA was extracted as described previously,7 and 1� gwas reverse transcribed for 1 hour at 42°C using AMV (Eu-robio, Courtaboeuf, France) with oligo-dT (Invitrogen).RNA extracted after laser-capture microdissection was reversetranscribed using SuperScript II with oligo-dT and randomprimers (all from Invitrogen).

Real-time PCR reactions were performed on the LightCyclerapparatus as described previously.15 The primers used were asfollows: (1) for CCL21 (F: 5�-CCTTGCCACACTCTTT-CTCCC-3�, R: 5�-CAAGGAAGAGGTGGGGTGTA-3�),(2) for PROX1 (F: 5�-GGCTGAGACCTTGAAACAGG-3�,R: 5�-ATTGTGGTTTTCCCCATTGA-3�), (3) for VEGFR3(F: 5�-CCCACGCAGACATCAAGACG-3�, R: 5�-TGCAG-AACTCCACGATCACC-3�), (4) for 28S (F: 5�-CGGGT-AAACGGCGGGAGTAA-3�, R: 5�-GGTAGGGACAGT-GGGAATCT-3�), and (5) for D6 (PPH01335A) from Super-Array Bioscience Corporation (TEBU, Le Perray en Yvelines,France). All samples were normalized according to GAPDH or28S amplification.

Statistical AnalysesIn bar graphs, results are expressed as means of different ex-periments or duplicate values. Error bars represent standarderror of the mean. For two-by-two comparisons, nonpara-metric Mann–Whitney U or Wilcoxon tests were applied.For multiparameter analyses (dose effect or subpopulationanalyses), analyses of variance followed by post hoc tests wereused.

ResultsThymic Overexpression of CCL21 in HyperplasticMyasthenia Gravis ThymusesWe previously demonstrated a specific increased ex-pression of CCL21 messenger RNA (mRNA), by mi-croarray and real-time PCR, in hyperplastic thymusesof MG patients.7 To investigate in more detail the reg-ulation of CCL21 expression in the thymus of MGpatients, we have analyzed by enzyme-linked immu-nosorbent assay its expression and that of CCL19, thesecond ligand to CCR7 receptor. The data obtainedconfirmed the high overexpression of CCL21 in allMH thymuses (Fig 1A). An increased expression ofCCL19 was also observed in MH as compared withthymuses from non-MG adults and ML patients butnot to those from babies (Fig 1B). Glucocorticoids arewidely used to treat MG patients, decreasing the num-ber of ectopic GCs in the thymus,14 and for these pa-tients, we did not observe an increased expression ofCCL21 and CCL19 compared with non-MG adults(Figs 1A, B).

In the normal thymus, CCL21 and CCL19 play animportant role in prothymocyte recruitment and thy-mocyte export, respectively.9,10 Comparison of the thy-mic CCL21/CCL19 ratios indicated a major represen-tation of CCL19 in babies, which might be related tothe high thymic cell mass and the need to export ma-ture thymocytes to the periphery (Fig 1C). Accord-ingly, in involuted adult thymuses, the level of CCL19decreased (Fig 1C). In MH thymuses, the high level ofCCL21 compared with CCL19 could lead to changesin the patterns of thymocyte recruitment and export.In this context, these changes could explain the in-creased number of CD4� observed in MG thymuses.16

CCL21 overexpression in hyperplastic thymuses couldalso trigger an abnormal recruitment of peripheral lym-phocytes.

CCR7-deficient mice are prone to develop general-ized multiorgan autoimmunity with lymphocytic infil-tration in peripheral organs and has an altered thymicarchitecture.17 Moreover, variations in CXCR3 andCXCR5 expression have been detected on T lympho-cytes in MG patients.18,19 Accordingly, we measuredCCR7 expression on peripheral B, CD4�, or CD8� Tlymphocytes by FACS. Both fluorescence means (datanot shown) and percentages of CCR7-positive cellswere similar between non-MG adults and MH patients

Berrih-Aknin et al: CCL21 in MG Thymic Hyperplasia 523

in all populations of peripheral lymphocytes tested (Fig1D). We also did not observe variations in thymic ex-pression of CCR11, a decoy receptor for CCL21 thatcould have scavenged and controlled CCL21 overex-pression in MG thymuses (data not shown).

Altogether, these observations show that thymic hy-perplasia in MG patients is specifically associated withthe thymic overexpression of CCR7 ligands (CCL19and CCL21) rather than with variations of CCR7 orCCR11 expression. Because CCL21 is largely overrep-resented in MH patients compared with CCL19, wehave focused our investigations on the role of CCL21in MG patients.

Potent Chemotactic Effect of CCL21 on B CellsTo address the consequences of CCL21 overexpressionin the thymus of MG patients, we investigated the na-ture of peripheral lymphocytes preferentially chemoat-tracted by CCL21. Surprisingly, we observed that hu-man CCL21 was significantly more potent inrecruiting B cells than T cells (Fig 2A). For a represen-tative experiment, in terms of absolute number, the re-cruited cells correspond to 7,704 � 119 and 195,000� 420 T cells or 1,820 � 3 and 16,400 � 69 B cellsin controls and with 500ng/ml of CCL21, respectively.

To determine whether these observations were specificto human CCL21, we compared its effects with thoseof mouse ccl21. In a representative experiment, the re-cruited mouse T cells corresponded to 2,512 � 376 incontrol and 10,644 � 816 with 400ng/ml of ccl21.The recruited mouse B cells corresponded to 1,785 �223 in control, 3,422 � 186 with 400ng/ml of ccl21or 3,645 � 186 with 400ng/ml of cxcl13. Conse-quently, mouse B cells are also attracted by ccl21, al-though this chemokine is more efficient in recruitingmouse T cells as shown by chemotactic and actin po-lymerization assays (Figs 2C, D). We next comparedthe chemotactic effect of human CCL21 to that ofCXCL13, the well-known B-cell chemoattractant thatis also overexpressed in MG thymuses.7,14 CXCL13triggered B-cell migration from concentrations greaterthan 0.05ng/ml with an optimum effect around 0.5 to5ng/ml (average migration indexes around 200%; Fig2B). The effect of CCL21 on B-cell migration was ob-served from concentrations greater than 5ng/ml withaverage migration indexes of 323 and 875% at 50 and500ng/ml, respectively (Fig 2B). Thus, high concentra-tions of CCL21 can trigger B-cell migration more ef-ficiently than any concentrations of CXCL13 tested. Incontrast, the mouse chemokines, ccl21 and cxcl13,

Fig 1. Specific thymic overexpression of CCL21 in hyperplastic myasthenia gravis (MG) thymuses. Determination by enzyme-linkedimmunosorbent assay (ELISA) of (A) CCL21 and (B) CCL19 concentrations in thymic extracts from babies (BB), non-MG adults(Ad), MG patients with low (ML) or high (MH) thymic hyperplasia, or corticosteroid-treated MG patients (Cortico). (C) Distribu-tion of log(10) ratios of the concentrations of CCL21 versus CCL19 determined by ELISA. (D) Fluorescence-activated cell sorting(FACS) analysis of the percentage of CCR7-positive cells in peripheral CD4�, CD8�, or CD19� cells from non-MG adults (Ad;n � 6) and MH patients (n � 6). p values are assessed by the Mann–Whitney U test and indicated when less than 0.05.

524 Annals of Neurology Vol 66 No 4 October 2009

triggered mouse B-cell migration similarly (Figs 2E, F).These results demonstrate that, in the human system,CCL21 recruits B cells preferentially over T cells and iseven more efficient than CXCL13.

Potent Chemotactic Effect of CCL21 onNaive B CellsWe also investigated the chemotactic properties ofCCL21 on different human B- and T-cell subpopula-tions. Using CD69 and CD25 as makers of T-cell ac-tivation and CD45RO to distinguish between naive(CD45RO�) and memory (CD45RO�) T cells, weanalyzed the percentages of T cells recruited withCCL21 (500ng/ml). CCL21 recruited all T-cell sub-populations studied, albeit preferentially naive T cells(Fig 3A). In addition, we observed that CCR7 waspreferentially expressed on naive T cells as comparedwith memory T cells (Fig 3B), as Sallusto and col-leagues20 described previously.

Using CD27 to distinguish between naive or GC Bcells (CD27�, about 37–59% of B cells) and memory/plasma B cells (CD27�), we observed that CCL21 re-cruited preferentially CD27� B cells (Fig 3C). To de-termine whether within the CD27� B-cell populationCCL21 had comparable chemotactic effects on naiveand GC B cells, we analyzed the migration ofCD27�IgD� or CD27�IgD� B cells corresponding,respectively, to naive and GC B cells.21 As shown inFigure 3D, CCL21 triggered especially the recruitmentof naive B cells. By analyzing expression levels ofCCR7 on CD19�IgD� (naive B cells) andCD19�IgD� (differentiated B cells), we also observedthat CCR7 expression was higher on naive B cells (Fig3E). Altogether, these data confirm that, althoughCCL21 can recruit all T- and B-cell subtypes analyzed,its chemotactic activity is more pronounced on naive Bcells.

Fig 2. Preferential migration of human B cells toward CCL21. (A, B) Migration of T (CD3�) or B (CD19�) cells from humanPBL toward human CCL21 or CXCL13. The percentage of migrating cells analyzed by fluorescence-activated cell sorting (FACS) iscalculated from the number of cells in lower wells as the ratio of (chemokine/control) � 100. Control subjects corresponding tobasal migration are thus equal to 100. (A) One representative experiment out of three with bars corresponding to means of dupli-cate values for T-cell (black bars) and B-cell (gray bars) migration. A two-way analyses of variance followed by a Bonferroni posthoc test shows that B- and T-cell migration is significantly different from concentrations more than 5ng/ml with p values � 0.05.(B) Each point corresponds to the mean of three different experiments analyzing B-cell migration with CCL21 (solid line) andCXCL13 (dashed line). p values are assessed by the Mann–Whitney U test for each concentration, and p values are � 0.05 from50ng/ml. (C, E) Migration of purified mouse T or B cells toward mouse ccl21 or cxcl13 at 400ng/ml (means of duplicate val-ues � standard error of the mean). Cell migration was analyzed as described earlier. (D, F) Measure of actin-filament polymeriza-tion in purified mouse T or B cells treated 15 seconds with 100ng/ml of ccl21 or cxcl13. Actin polymerization was analyzed byFACS on cells stained with FITC-phalloidin. The percentage of increase in actin polymerization is calculated from the FITC-phal-loidin–positive cells as follows: (treated cells � control cells/control cells) � 100. (C–F) Bar graphs correspond to the mean of twodifferent experiments. p values are assessed by the Mann–Whitney U test, and only p values � 0.05 are indicated.

Berrih-Aknin et al: CCL21 in MG Thymic Hyperplasia 525

CCL21 Overexpression by Endothelial VesselsThe next step was to determine the nature of the cellsoverexpressing CCL21 in MG thymic hyperplasia. On-odera22 reported a strong CCL21 immunoreactivity inthymic stromal cells in hyperplastic MG thymuses. Al-though thymic epithelial cells express CCL21, we didnot detect any increased expression of CCL21 by thesecells in MG patients (data not shown). Immunohisto-chemical analysis indicated that CCL21 is overex-pressed on endothelial vessels in hyperplastic thymuses(Fig 4). To determine whether the CCL21 increase inhyperplastic MG thymuses was due only to CCL21-positive vessels, we performed laser-capture microdis-section to isolate CCL21-positive vessels from the thy-mic stroma and also from GCs. Analysis of CCL21mRNA expression in the microdissected fragments andin the remaining tissue section confirmed that CCL21overexpression in hyperplastic thymuses originates fromendothelial vessels (Fig 4D), as well as partly fromGCs, rather than from the thymic stroma. Altogether,these observations demonstrate that CCL21 is overex-pressed at the mRNA and protein level by endothelialvessels in hyperplastic MG thymuses.

Overrepresentation of High Endothelial Venules inHyperplastic Myasthenia Gravis ThymusesIn SLOs and in chronically inflamed tissues, lympho-cyte homing is directed through HEVs, a specializedendothelium bearing on its luminal surface diverse che-mokines and expressing high levels of PNAd carbohy-drate ligands.23 Using anti-PNAd antibody, we ob-served only a few HEVs in thymuses of babies andnon-MG adults (Figs 5A–C). In contrast, in the thy-mus of MG patients, we observed increased numbersof HEVs, located around GCs, which correlated withthe degree of thymic hyperplasia (Figs 5D, F).

We thus wondered whether HEVs were the vesselsoverexpressing CCL21. Double labeling for CCL21and PNAd antibodies (Figs 5G–I) did not show anycolabeling, indicating that the endothelial vessels over-expressing CCL21 in hyperplastic MG thymuses werenot HEVs.

Characterization of the Endothelial VesselsOverexpressing CCL21We further investigated the phenotype of CCL21-positive endothelial vessels in the thymus of MH pa-

Fig 3. Preferential migration of human naive B cells toward CCL21. (A) Migration of different T-cell subpopulations towardCCL21 (500ng/ml). (B) Analysis of the percentage of CCR7-positive cells in naive (CD3�CD45RO�) and memory(CD3�CD45RO�) T cells. (C) Analysis of the migration of B-cell subpopulations toward CCL21 (500ng/ml). CD19� B cellswere labeled with CD27 to distinguish naive- and germinal center (GC)–selected B cells (CD27�) from memory and plasma Bcells (CD27�). (D) To determine whether, within the CD27� B-cell population, CCL21 had comparable chemotactic effects onnaive- (IgD�) and GC-selected B cells (IgD�), we labeled B cells with both CD27 and IgD. (E) Analysis of the percentage ofCCR7-positive cells in naive- (CD19�IgD�) and memory (CD19�IgD�) B cells. (A, C, D) Percentage of migrating cells is calcu-lated from the number of cells in lower wells as the ratio of (chemokine/control) � 100. Each bar corresponds to the mean of dif-ferent experiments with PBL from five to seven different donors. p values are assessed by Kruskal–Wallis test followed by Dunn’smultiple-comparison test, showing a significant recruitment of naive T cells with CCL21 compared with control. (B, E) Each barcorresponds to the mean of eight different donors � standard error of the mean. p values are assessed by the Wilcoxon test.Fluorescence-activated cell sorting histograms correspond to representative experiments. FITC � fluorescein isothiocyanate.

526 Annals of Neurology Vol 66 No 4 October 2009

tients by double-labeling analysis of CCL21 andVEGFR3, a lymphatic endothelial cell (LEC) marker.We then observed that all CCL21-positive vessels werealso VEGFR3-positive (Figs 6A–G) and correspondedto small vessels. We also detected large VEGFR3-positive vessels in the thymus of babies, non-MG

adults, and MG patients, but these were not positivefor CCL21 (data not shown). Interestingly, these ob-servations made in humans differ from what has beenobserved in mice. Indeed, Odaka and coauthors24 havedemonstrated that both thymic lymphatic and bloodvessels, together with thymic epithelial cells, expressCCL21.

To confirm the LEC origin of CCL21 expression inhyperplastic thymuses, we analyzed the expression ofPROX1, a specific LEC transcription factor involved inendothelial cell differentiation into LECs,25 in the mi-crodissection extracts described earlier. We show thatthe microdissected CCL21-positive vessels expressedPROX1, whereas we could not detect its expression onthe remaining part of the sections (Fig 6H). The levelof PROX1 expression in CCL21-positive vessels ap-peared to be similar to the one observed in primarycultures of human dermal LECs used as a positive con-trol. In hyperplastic MG thymuses, the CCL21-positive lymphatic vessels could correspond to afferentor efferent vessels, allowing the import or export of pe-ripheral cells, respectively. We then analyzed the ex-pression of D6, a decoy receptor for most inflamma-tory chemokines known to be expressed on afferentlymphatic vessels of peripheral organs.26 We observedthat CCL21-positive vessels strongly expressed D6compared with the thymic stroma (Fig 6I).

To determine whether this overexpression of CCL21in hyperplastic thymuses is related to a greater numberof lymphatic vessels, we analyzed thymic expression of

Fig 4. CCL21 overexpression by endothelial vessels. Immunohis-tochemical analyses on thymic sections from (A) babies, (B)non–myasthenia gravis (non-MG) adults (Ad), and (C) highthymic hyperplasia (MH) patients. Thymic epithelial cells werelabeled in red with an anti-keratin antibody and CCL21 ingreen. Sections were viewed with a Leica DMRB microscope,and images were captured with a Sony DXC-930P color CCDcamera (Tokyo, Japan) connected to a Leica MSP 930 imageanalysis system (Leica Microsystems, Rueil-Malmaison, France)(bars � 20�m). (D) Real-time polymerase chain reaction(PCR) on laser-capture microdissection. Cryostat sections of hu-man thymic tissues were peroxidase-labeled to detect CCL21and counterstained with hematoxylin. CCL21-positive bloodvessels identified by a dark brown color and germinal centers(GCs) with mantle zones were isolated separately by laser-capture microdissection. The microdissected regions (CCL21�

vessels and GCs), the rest of the sections (whole sections withoutCCL21� vessels and GCs), and also whole independent sections(WS) were collected, and CCL21 RNA was quantified by real-time PCR. For all samples, the CCL21 RNA level was reportedto 28S. Each bar corresponds to the mean � standard error ofthe mean from three different non-MG (white bars) or MH(black bars) samples. One-tailed p values are assessed byMann–Whitney U test in each microdissected region to deter-mine whether CCL21 is significantly overexpressed in MG pa-tients; only p values � 0.05 are indicated.

Berrih-Aknin et al: CCL21 in MG Thymic Hyperplasia 527

VEGFR3 and PROX1 in non-MG and MG patients.Using real-time PCR analyses, we observed a high in-terindividual variation for these markers. Nevertheless,significant increased expressions of VEGFR3 (3.1times) and PROX1 (3.6 times; Figs 6J, K) were dem-onstrated in the thymus of MH patients comparedwith non-MG adults. Altogether, these data indicatethat CCL21-positive vessels correspond to afferentlymphatic vessels and show for the first time lym-phangiogenesis in hyperplastic thymuses.

DiscussionRole of CCL21 in Ectopic Germinal CenterDevelopment in the Thymus of MyastheniaGravis PatientsIn MG, thymic hyperplasia is characterized by thepresence of numerous GCs,2 and here we demonstratethat it is specifically associated with overexpression ofCCL21. The overexpression of this chemokine has alsobeen observed in nonlymphoid organs, especially in au-toimmune diseases characterized by ectopic GC devel-opment.27,28 Transgenic mouse models with a specificexpression of CCL21 in islet cells or in thyroid cellsshowed that ectopic expression of CCL21 is sufficientto trigger lymphoid neogenesis in pancreas or thyroid,respectively.29,30

To sustain the role of CCL21 in GC develop-ment within the thymus of MG patients, we observedthat CCL21 expression was strikingly reduced inglucocorticoid-treated compared with untreated pa-tients, and glucocorticoids are known to decrease thenumber of ectopic GCs in MG thymus.14 We suggestthat the efficiency of glucocorticoids in MG is at leastpartly due to their effects on CCL21 expression.

By analyzing in detail the chemotactic properties ofCCL21 on different T- and B-cell subpopulations, wedemonstrate that although CCL21 recruits T cells ef-ficiently, it is also a potent chemoattractant for B cells.Moreover, in humans, CCL21 could even be more ef-ficient than CXCL13 in recruiting B cells, especiallynaive B cells. We hypothesize that CCL21 overexpres-sion is probably decisive in allowing an additional re-cruitment of peripheral cells, in particular, naive Bcells. Once in contact with the inflammatory environ-ment characteristic of MG thymuses, these naive Bcells lead to ectopic GC development characteristic ofthymic hyperplasia. CCL21 may also play a final rolein the segregation of lymphoid compartments necessaryfor the complex cell organization of GCs, as Hopkenand coworkers31 suggested.

Altogether, these observations suggest that the spe-cific thymic overexpression of CCL21 in hyperplasticthymuses plays a central role in ectopic GC develop-ment. Consequently, we can envisage that therapies in-hibiting CCL21 overexpression selectively could lead toa decrease in thymic hyperplasia.

Neoangiogenic Processes with Specific Overexpressionof CCL21 by Lymphatic Endothelial Vessels inHyperplastic ThymusesOur investigation of the nature of the cells involved inthymic overexpression of CCL21 in MG demonstrateda specific overexpression of CCL21 by endothelial ves-sels in hyperplastic thymuses. In parallel, we observedthe development of numerous HEVs around GCs inhyperplastic thymuses. The protein CCL21 is known

Fig 5. Overrepresentation of high endothelial venules (HEVs)in hyperplastic myasthenia gravis (MG) thymuses. Immunohis-tochemical analyses of germinal center (CG) and HEV local-ization in thymuses. Serial thymic sections from (A) babies,(B, C) non-MG adults, (D) seropositive MG patients withlow thymic hyperplasia (ML), and (E, F) seropositive MGpatients with high thymic hyperplasia (MH) were stained withantibodies against CD21 (red), anti-PNAd (green), and anti-keratin (blue) to localize CG, HEVs, and the epithelial cells,respectively. The whole sections were then scanned using the428 Affymetrix scanner with a resolution of 10�m, and im-ages were superimposed. Immunohistochemical analyses ofHEV and CCL21 localization in thymuses. Thymic sectionsfrom (G) babies, (H) non-MG adults, and (I) MH patientswere labeled in red with an anti-PNAd antibody for HEVsand in green for CCL21. Scale bars � 200�m (A, B, D, E);10�m (G–I).

528 Annals of Neurology Vol 66 No 4 October 2009

to be present on HEVs, even if Carlsen and col-leagues32 demonstrated species-specific differences inits expression: mouse HEVs produce and express ccl21,whereas human HEVs lack detectable CCL21 mRNA.However, we did not detect any colabeling betweenCCL21 and thymic HEVs, suggesting that, unlike inSLOs, CCL21 in the thymus is not involved in therecruitment of peripheral cells through HEVs; rather,and similarly to what was recently described in otherhuman diseases with lymphoid neogenesis, such asrheumatoid arthritis, Sjogren syndrome, and ulcerativecolitis,33 CCL21-positive thymic vessels correspondedto lymphatic endothelial vessels. In addition, we ob-served that CCL21-positive vessels expressed D6, a de-

coy receptor for inflammatory chemokines consideredas a marker for afferent lymphatic vessels.26

The increased expression of lymphatic markers(VEGFR3 and PROX1) in hyperplastic thymuses alsosuggests an expansion of the lymphatic system. Lym-phangiogenesis occurs throughout life in homeostasisand diseases. It has been described in lymph nodesafter immunization, where it was shown to be depen-dent on the entry of B cells.34 In a model of trans-genic mice overexpressing CCL21 within the thyroid,lymphangiogenesis has also been observed and was re-lated to the influx of CD4� T cells.35 Altogether,these results demonstrate active angiogenic processeswithin the thymus of MG patients with HEV develop-

Fig 6. CCL21-positive vessels correspond to lymphatic endothelial vessels. Immunohistochemical analyses on thymic sections from seropos-itive MG patients with high thymic hyperplasia (MH): staining with (A, D) green-labeled anti–vascular endothelial growth factor re-ceptor 3 (VEGFR3) antibody, (B, E) red-labeled anti-CCL21 antibody, and (C, F) merge and (G) representative merge of isotype con-trol (G). Real-time polymerase chain reaction (PCR) on laser-capture microdissection as detailed in Figure 2. (H) PROX1 and (I) D6messenger RNA expression in CCL21�-microdissected vessels and in the rest of the sections (whole sections without CCL21� vessels andgerminal centers [GCs]; WS�) were quantified by real-time PCR and reported to 28S expression. Each bar corresponds to the mean ofexperiments with samples from different non-MG adults (n � 3) or MH patients (n � 3). Messenger RNA (mRNA) expression forthese genes was also analyzed using lymphatic endothelial cell and human umbilical vein endothelial cell (HUVEC) cultures as positiveand negative controls, respectively. p values are assessed by the Mann–Whitney U test for non-MG, MG, or cell culture samples; pvalues � 0.1 are indicated. (J, K) Real-time PCR for mRNA expression level of (J) vascular endothelial growth factor receptor 3(VEGFR3) and (K) PROX1 in the thymus of non-MG adults (n � 5) and MH patients (n � 7). mRNA expression level was re-ported to GAPDH. p values were obtained by the Mann–Whitney U test. Scale bars � 10�m.

Berrih-Aknin et al: CCL21 in MG Thymic Hyperplasia 529

ment and lymphangiogenesis associated with the overex-pression of CCL21.

Role of Lymphatic Overexpression of CCL21 inLymphocyte TraffickingHyperplastic MG thymuses present the major charac-teristics of SLOs and may behave similarly. First, thelarge number of HEVs observed in hyperplastic thy-muses suggests an entry of peripheral cells throughthese specialized vessels, and we recently demonstrateda significant increase in migration of PBL toward hy-perplastic thymic extracts in chemotactic assays.7 Sec-ond, the overexpression of CCL21,7 and alsoCXCL13,14 could account for an abnormal recruit-ment of peripheral B and T cells; however, the maindifference between hyperplastic thymus and SLOs isthat CCL21 and CXCL13 (unpublished data) were notdetected on thymic HEVs, suggesting that B- andT-cell recruitment in the thymus is different from thatin SLOs. According to Pearse,36 there are no afferentlymphatic vessels in the normal thymus; however, wecan hypothesize that, in hyperplastic thymuses,CCL21-positive vessels represent a pathological devel-opment of afferent lymphatic vessels, which, in addi-tion to thymic HEVs, could correspond to an entranceof circulating peripheral cells, in particular, naive Bcells, indispensable for triggering ectopic GC develop-ment. These thymic afferent lymphatic vessels couldalso mediate the recruitment of sensitized dendriticcells as Nagane and colleagues37 suggested.

In the human system, CCL21 is also strongly ex-pressed by afferent lymphatic vessels in peripheral tis-sues, where it mediates dendritic and T-cell exit, andtheir entry into afferent lymph, allowing their migra-tion to SLOs.38 Consequently, another hypothesis isthat hyperplastic MG thymuses behave similarly to aninflamed peripheral organ where CCL21-positive affer-ent lymphatic vessels may behave as an exit signal forCCR7-positive cells.

ConclusionWe show for the first time that the extent of thymichyperplasia in MG correlates with pathological angio-genic processes, ectopic HEV angiogenesis and lym-phangiogenesis, which suggest abnormal peripheral lym-phocyte recruitment. In vivo, overexpression of CCL21on lymphatic endothelial vessels could trigger a greaterrecruitment of naive B cells to the inflamed MG thy-mus, where they could be sensitized against AChR. Be-cause CCL21 overexpression is normalized upon corti-coid treatment, we postulate that molecules inhibitingthe effects of CCL21 could represent a new way to treatMG patients by decreasing the abnormal recruitment ofperipheral lymphocytes, especially B cells, and limitingthe development of thymic hyperplasia.

Grants from the European Community LSHM-CT-2006-037833and the “Association Francaise contre les Myopathies” (AFM, CGR37500711) obtained by Dr. S. Berrih-Aknin, and from the “AgenceNationale de la Recherche” ANR-06-MRAR-001-01 obtained byDr. R. LePanse.

We thank R. Alon for helpful discussions, F. Truffaultfor her technical assistance, S. Mussot and A. Serraf forthe thymic samples, and E. Dulmet and V. de Mont-preville for histological analyses.

References1. Aharonov A, Abramsky O, Tarrab-Hazdai R, Fuchs S. Hu-

moral antibodies to acetylcholine receptor in patients with my-asthenia gravis. Lancet 1975;2:340–342.

2. Berrih-Aknin S, Morel E, Raimond F, et al. The role of thethymus in myasthenia gravis: immunohistological and immuno-logical studies in 115 cases. Ann N Y Acad Sci 1987;505:50–70.

3. Kuks JB, Oosterhuis HJ, Limburg PC, The TH. Anti-acetylcholine receptor antibodies decrease after thymectomy inpatients with myasthenia gravis: clinical correlations. J Autoim-mun 1991;4:197–211.

4. Wakkach A, Guyon T, Bruand C, et al. Expression of acetyl-choline receptor genes in human thymic epithelial cells:implications for myasthenia gravis. J Immunol 1996;157:3752–3760.

5. Leprince C, Cohen-Kaminsky S, Berrih-Aknin S, et al. ThymicB cells from myasthenia gravis patients are activated B cells:phenotypic and functional analysis. J Immunol 1990;145:2115–2122.

6. Melms A, Schalke BC, Kirchner T, et al. Thymus in myasthe-nia gravis: isolation of T-lymphocyte lines specific for the nic-otinic acetylcholine receptor from thymuses of myasthenic pa-tients. J Clin Invest 1988;81:902–908.

7. Le Panse R, Cizeron-Clairac G, Bismuth J, Berrih Aknin S.Microarrays reveal distinct gene signatures in the thymus of se-ropositive and seronegative Myasthenia Gravis patients and therole of CCL21 in thymic hyperplasia. J Immunol 2006;177:7868–7879.

8. Annunziato F, Romagnani P, Cosmi L, et al. Chemokines andlymphopoiesis in human thymus. Trends Immunol 2001;22:277–281.

9. Liu C, Ueno T, Kuse S, et al. The role of CCL21 in recruit-ment of T-precursor cells to fetal thymi. Blood 2005;105:31–39.

10. Ueno T, Hara K, Willis MS, et al. Role for CCR7 ligands inthe emigration of newly generated T lymphocytes from theneonatal thymus. Immunity 2002;16:205–218.

11. Gunn MD, Tangemann K, Tam C, et al. A chemokine ex-pressed in lymphoid high endothelial venules promotes the ad-hesion and chemotaxis of naive T lymphocytes. Proc Natl AcadSci U S A 1998;95:258–263.

12. Gunn MD, Kyuwa S, Tam C, et al. Mice lacking expression ofsecondary lymphoid organ chemokine have defects in lympho-cyte homing and dendritic cell localization. J Exp Med 1999;189:451–460.

13. Flaishon L, Lantner F, Hershkoviz R, et al. Low levels of IFN-gamma down-regulate the integrin-dependent adhesion of Bcells by activating a pathway that interferes with cytoskeletonrearrangement. J Biol Chem 2001;276:46701–46706.

14. Meraouna A, Cizeron-Clairac G, Le Panse R, et al. The che-mokine CXCL13 is a key molecule in autoimmune MyastheniaGravis. Blood 2006;108:432–440.

530 Annals of Neurology Vol 66 No 4 October 2009

15. Poea-Guyon S, Christadoss P, Le Panse R, et al. Effects of cy-tokines on acetylcholine receptor expression: implications formyasthenia gravis. J Immunol 2005;174:5941–5949.

16. Berrih-Aknin S, Cohen-Kaminsky S, Truffault F. T-cell recep-tor expression in the thymus from patients with myastheniagravis. Ann N Y Acad Sci 1995;756:438–440.

17. Davalos-Misslitz AC, Rieckenberg J, Willenzon S, et al. Gener-alized multi-organ autoimmunity in CCR7-deficient mice. EurJ Immunol 2007;37:613–622.

18. Feferman T, Maiti PK, Berrih-Aknin S, et al. Overexpression ofIFN-induced protein 10 and its receptor CXCR3 in myastheniagravis. J Immunol 2005;174:5324–5331.

19. Saito R, Onodera H, Tago H, et al. Altered expression of che-mokine receptor CXCR5 on T cells of myasthenia gravis pa-tients. J Neuroimmunol 2005;170:172–178.

20. Sallusto F, Lenig D, Forster R, et al. Two subsets of memory Tlymphocytes with distinct homing potentials and effector func-tions. Nature 1999;401:708–712.

21. Klein U, Rajewsky K, Kuppers R. Human immunoglobulin(Ig)M�IgD� peripheral blood B cells expressing the CD27cell surface antigen carry somatically mutated variable regiongenes: CD27 as a general marker for somatically mutated(memory) B cells. J Exp Med 1998;188:1679–1689.

22. Onodera H. The role of the thymus in the pathogenesis of my-asthenia gravis. Tohoku J Exp Med 2005;207:87–98.

23. Girard JP, Springer TA. High endothelial venules (HEVs): spe-cialized endothelium for lymphocyte migration. Immunol To-day 1995;16:449–457.

24. Odaka C, Morisada T, Oike Y, Suda T. Distribution of lym-phatic vessels in mouse thymus: immunofluorescence analysis.Cell Tissue Res 2006;325:13–22.

25. Wigle JT, Harvey N, Detmar M, et al. An essential role forProx1 in the induction of the lymphatic endothelial cell phe-notype. Embo J 2002;21:1505–1513.

26. Nibbs RJ, Kriehuber E, Ponath PD, et al. The beta-chemokinereceptor D6 is expressed by lymphatic endothelium and a sub-set of vascular tumors. Am J Pathol 2001;158:867–877.

27. Barone F, Bombardieri M, Manzo A, et al. Association ofCXCL13 and CCL21 expression with the progressive organiza-tion of lymphoid-like structures in Sjogren’s syndrome. Arthri-tis Rheum 2005;52:1773–1784.

28. Takemura S, Braun A, Crowson C, et al. Lymphoid neogenesisin rheumatoid synovitis. J Immunol 2001;167:1072–1080.

29. Fan L, Reilly CR, Luo Y, et al. Cutting edge: ectopic expressionof the chemokine TCA4/SLC is sufficient to trigger lymphoidneogenesis. J Immunol 2000;164:3955–3959.

30. Martin AP, Coronel EC, Sano G, et al. A novel model forlymphocytic infiltration of the thyroid gland generated bytransgenic expression of the CC chemokine CCL21. J Immunol2004;173:4791–4798.

31. Hopken UE, Wengner AM, Loddenkemper C, et al. CCR7deficiency causes ectopic lymphoid neogenesis and disturbedmucosal tissue integrity. Blood 2007;109:886–895.

32. Carlsen HS, Haraldsen G, Brandtzaeg P, Baekkevold ES. Dis-parate lymphoid chemokine expression in mice and men: noevidence of CCL21 synthesis by human high endothelialvenules. Blood 2005;106:444–446.

33. Manzo A, Bugatti S, Caporali R, et al. CCL21 expression pat-tern of human secondary lymphoid organ stroma is conservedin inflammatory lesions with lymphoid neogenesis. Am J Pathol2007;171:1549–1562.

34. Angeli V, Ginhoux F, Llodra J, et al. B cell-driven lymphangio-genesis in inflamed lymph nodes enhances dendritic cell mobi-lization. Immunity 2006;24:203–215.

35. Furtado GC, Marinkovic T, Martin AP, et al. Lymphotoxinbeta receptor signaling is required for inflammatory lym-phangiogenesis in the thyroid. Proc Natl Acad Sci U S A 2007;104:5026–5031.

36. Pearse G. Normal structure, function and histology of the thy-mus. Toxicol Pathol 2006;34:504–514.

37. Nagane Y, Utsugisawa K, Obara D, et al. Dendritic cells inhyperplastic thymuses from patients with myasthenia gravis.Muscle Nerve 2003;27:582–589.

38. Debes GF, Arnold CN, Young AJ, et al. Chemokine receptorCCR7 required for T lymphocyte exit from peripheral tissues.Nat Immunol 2005;6:889–894.

Berrih-Aknin et al: CCL21 in MG Thymic Hyperplasia 531