The Migration of Human Follicular Dendritic Cell-Like Cell Is Facilitated by Matrix Metalloproteinase 3 Expression That Is Mediated through TNFα-ERK1/2-AP1 Signaling

Follicular dendritic cells are important stromal components of the germinal center (GC) and have pivotal roles in maintaining the GC microenvironment for high-affinity antibody production. Tumor necrosis factor-α (TNFα) is essential for the development and functions of follicular dendritic cells. Despite the importance of follicular dendritic cells in humoral immunity, their molecular control mechanisms have yet to be fully elucidated due to the lack of an adequate investigation system. Here, we have used a unique human primary follicular dendritic cell-like cell (FDCLC) to demonstrate that the migration of these cells is enhanced by TNFα-mediated metalloproteinase 3 (MMP3) expression. MMP3 was found to be highly expressed in normal human GCs and markedly upregulated in human primary FDCLCs by TNFα. TNFα induced ERK1/2 phosphorylation and the transcription of MMP3 through AP1. TNFα treatment increased FDCLC migration, and a knockdown of MMP3 significantly reduced the TNFα-induced migration of FDCLCs. Overall, we have newly identified a control mechanism for the expression of MMP3 in FDCLCs that modulates their migration and may indicate an important role in GC biology. Since GCs are observed in the lesions of autoimmune diseases and lymphomas, targeting the MMP3/TNFα-mediated migration of stromal cells in the B cell follicle may have great potential as a future therapeutic modality against aberrant GC-associated disorders.


Introduction
The germinal center (GC) is critical for the production of high-affinity protective antibodies and plays an important role in various immunologic disorders including autoimmune diseases and lymphomas [1,2]. Follicular dendritic cells (FDCs) are the major stromal components of GCs and are essential for their formation and maintenance. These cells form network-like structures with their long dendrites and provide crucial survival and differentiation signals to GC-B cells [3,4]. FDCs are developed and maintained by lymphotoxin α1/β2 and tumor necrosis factor-α (TNFα), which are produced from the B cells within the GC [5]. The spreading of the GCs together with FDCs is one of the important pathologic features of lymphoid diseases such as rheumatoid arthritis and follicular lymphomas [6]. Many of the molecular mechanisms underlying the actions of FDCs remain to be identified because of the lack of adequate investigation systems for these cells, particularly in relation to their migration.
TNFα is a pleiotropic cytokine that has pivotal roles in prolonged inflammation and autoimmune diseases [7,8]. Blocking TNFα signaling has become the major treatment for those diseases [9]. Clinical treatments with the TNFα inhibitor etanercept have been reported to reduce the GC number in rheumatoid arthritis patients [10]. TNFα is also required for the development of the FDC network and the GC. Mice deficient in the TNF receptor fail to develop the GC and FDC clusters [11]. TNFα knockout mice have a B cell follicle deficiency, defects in FDC network organization, and impaired humoral immunity [12]. In addition, several studies have reported that TNFα signaling induces the expression of matrix metalloproteinases (MMPs). MMP1 is increased by TNFα in human fibroblasts [13]. TNFα also induces MMP3 expression in nucleus pulposus cells [14] and synovial fibroblasts [15]. MMP3 degrades ECM components during tissue remodeling and can thus be implicated in GC organization and FDC migration [16,17].
Here, we investigate the molecular mechanisms underlying FDC migration using a primary human FDC-like cell (FDCLC) system. We show from our present analyses that MMP3 expression is required for TNFα-invoked FDCLC migration and is induced through TNFα-extracellular signal-regulated kinase 1/2-(ERK1/2-) activator protein-1 (AP1) signaling. This result sheds new light on the control of GC-related autoimmune diseases and FDC-associated fatal disorders including AIDS, prion infection, and follicular lymphoma. To our knowledge, this is the first report to describe the molecular processes that drive the migration of human primary FDCLCs.

Isolation of Human FDCLCs.
To investigate the molecular mechanism of FDCLC migration, we recently established isolation and culture methods for these cells derived from normal pediatric tonsil tissues. Human tonsils were obtained from routine tonsillectomies. Primary FDCLCs were then extracted from these tonsils via mechanical disruption and enzymatic digestion as previously described [18]. Briefly, specimens were cut into small fragments, placed in HBSS containing 2 mg/mL of collagenase IV at 37°C for 15 min. The released cells were collected and subjected to Ficoll gradient centrifugation. The interface layer that contained the FDCLCs was then collected, and the cells were centrifuged at 200 rpm for 10 min at 4°C over a discontinuous gradient of 7.5% and 3% BSA. FDCLC-enriched fractions were then carefully collected from the interface. The cells were subsequently washed with HBSS and cultured on tissue culture dishes. Nonadherent cells were removed and adherent cells replenished with fresh medium every 3-4 days. Adherent cells were trypsinized when confluence was attained.

Ethical
Approval for the Use of Human Tissues. The use of human tonsils that were obtained as leftover material after tonsillectomies was approved by the institutional review board of Asan Medical Center, Seoul, Korea (approval number, 2013-0864). Informed consent was waived because there was no additional risk to the participants and their identities were anonymized and completely delinked from unique identifiers.

Microarray Analysis.
Total RNAs from TNFα-stimulated FDCLCs were purified using NucleoSpin RNA (740955, Macherey-Nagel, Düren, Germany). The purity and integrity of the isolated RNAs were evaluated by denaturing gel electrophoresis, via the OD 260/280 ratio, and by analysis on an Agilent 2100 Bioanalyzer (G2939BA, Agilent Technologies, Palo Alto, CA). cRNAs were generated from total RNA extracts using an Ambion Illumina RNA Amplification Kit (AMIL1791, Ambion, Austin, TX) and hybridized to Human HT12 Expression v.4 Bead Arrays in accordance with the manufacturer's instructions (Illumina, Inc., San Diego, CA). Array signals were detected using Amersham fluorolink streptavidin-Cy3 (GE Healthcare Bio-Sciences, Little Chalfont, UK) and an Illumina BeadArray Reader Confocal Scanner.
2.5. Secreted Protein Analysis by Luminex. FDCLCs were seeded in a 24-well plate at 2 × 10 4 cells/mL/well to which 20 ng/mL of TNFα was added 24 h later. The culture supernatants were harvested after another 24 h, and the detached cells were removed by centrifugation. The supernatants were stored at -70°C. Three independent samples were prepared and pooled for Luminex analyzer (Rules-Based Medicine, Austin, TX).

Western
Blotting. For western blotting analysis, protein extracts from FDCLCs were separated on 12% SDSpolyacrylamide gels and electrophoretically transferred to an Immun-Blot PVDF Membrane (162-0177; Bio-Rad Laboratories). The membrane was then blocked for 1 h with 5% BSA and incubated overnight with MMP3 Abs. Unbound primary Abs were removed by washing 3 times with TBS/0.1% Tween 20, and the membrane was then incubated with horseradish peroxidase-conjugated anti-mouse secondary Abs (diluted 1 : 3,000 in TBS/0.1% Tween 20). Signals were visualized using an ImageQuant LAS 4000 biomolecular imager (GE Healthcare Life Sciences). β-Actin was used as an internal control.

Luciferase Reporter Assays for Human Mmp3.
Reporter plasmids for the human Mmp3 promoter were constructed as follows: the Mmp3 proximal promoter region (-1370 to +202) was amplified by PCR from FDCLC genomic DNA and inserted into pGreenFire (TR010PA-1, System Biosciences, Palo Alto, CA). As FDCLCs exhibit very low transfection efficiency, and also low transduction by general lentiviral transduction methods, the previously described measles virus glycoprotein-displaying lentivirus transduction system was used [20]. All virus clones (mCMV, hMMP3p, AP1-3, and hMMP3p-mut) were produced at the same time, and the virus copy number was measured by p24 qPCR. FDCLCs were transduced using equivalent amounts of the viral clones and subjected to reporter assays using the Luciferase Assay System (Promega Corporation, E1501, Madison, WI

Human FDCLCs in the GC Express MMP3 via TNFα
Induction. TNFα secretion in the GC has been reported previously [21]; therefore, we first investigated the effects of TNFα on FDCLCs. Primary FDCLCs were isolated from fresh human tonsils and cultured in the presence of TNFα. Both the cells and the culture supernatants were harvested for mRNA expression and protein secretion analyses. We confirmed that TNFα induced the expression of various transcripts and proteins in FDCLCs (Figures 1(a) and 1(b)). Among these factors, the expression of MMPs was found to be markedly increased in the presence of TNFα. Because the MMPs are indispensable for cell migration and tissue remodeling, we focused on their specific roles in FDCLCs. Subsequent mRNA expression data revealed that MMP3,    Journal of Immunology Research MMP1, and MMP12 transcripts were upregulated in these cells upon exposure to TNFα, whereas protein secretion results revealed an increase in the MMP3 and MMP9 levels ( Figure 1(c)). Only MMP3 was revealed to be upregulated by TNFα at both the mRNA and secreted protein levels (Figure 1(d)). We confirmed that MMP3 is expressed in the    Journal of Immunology Research human GCs in vivo by examining tissue slides of human tonsils (Figure 1(e), the borderline is assigned by the yellow dotted line).
To confirm that MMP3 is produced by FDCLCs and not by GC-B cells, primary FDCLCs and GC-B cells isolated from different donors were analyzed. All of the examined FDCLCs but none of the GC-B cells from different donors showed high MMP3 mRNA expression in the presence of TNFα (Figure 1(f)). MMP3 protein secretion from FDCLCs was also increased in the presence of TNFα (Figure 1(g)). Taken together, these results indicate that FDCLCs in the GC express MMP3 via TNFα signaling.

MMP3 Expression
Induced by TNFα Is Mediated through ERK1/2 and AP1 Activation in FDCLC. MMP3 expression has been shown to be regulated at the transcriptional level through the mitogen-activated protein kinase-(MAPK-) AP1 pathway in human chondrocytes and in non-small-cell lung carcinoma cells [22,23]. To examine if this same signaling pathway also operates in FDCLCs, we assayed the phosphorylation of ERK1/2, JNK, and p38 MAPK following TNFα treatment of these cells. TNFα significantly increased the phosphorylation of ERK1/2 but had no noticeable effect on the JNK and p38 MAPK phosphorylation status (Figure 2(a)). To confirm that MMP3 expression in FDCLCs is dependent on the ERK1/2 signaling pathway, we examined whether the inhibition of ERK1/2 activation would abolish TNFα-mediated MMP3 induction. Treatment of the cells with the ERK1/2 inhibitor U0126 markedly reduced MMP3 expression by TNFα to a level comparable with the vehicle-treated (DMSO only) negative control samples (Figure 2(b)). The inhibition of ERK1/2 activation in FDCLCs by U0126 was confirmed by western blotting. These results demonstrated that the MMP3 expression induction by TNFα in FDCLCs requires ERK1/2 activation.
The human MMP3 promoter harbors 2 AP1 binding sites [23]. To analyze whether TNFα-induced MMP3 production is dependent on these sites, we generated reporter vectors containing the MMP3 proximal promoter with 2 AP1 binding sites (hMMP3p), 3 repeat AP1 sites (cloned from the human MMP3 promoter) without the MMP3 promoter region (AP1 ×3), and an MMP3 promotercontaining mutations in the 2 AP1 sites (mut hMMP3p). A minimal CMV promoter-containing vector (mCMV) was used as the negative control in these analyses. TNFα exposure significantly increased the luciferase activity (by 5.3-fold) when using the hMMP3p construct, whereas no noticeable activity was seen with the mut hMMP3p vector (Figure 2(c)). FDCLCs containing the AP1 ×3 construct show even more stronger luciferase signal induction by TNFα stimulation. These results indicated that TNFα-induced MMP3 promoter activity in FDCLCs is strongly dependent on the activity of AP1 sites in the MMP3 promoter region. Collectively, our data demonstrate that TNFα induces MMP3 production in FDCLCs through ERK1/2 and AP1 activation.

FDCLC Migration Is Induced by TNFα, which Is
Dependent on MMP3 Expression. The molecular mechanisms underlying FDCLC migration had previously been uncharac-terized. Since MMP3 is known to be involved in tumor cell invasion, we investigated whether TNFα induces FDCLC migration and whether MMP3 expression is required for this. We conducted a migration assay with MMP3 siRNAtransfected (MMP3i) or nontargeting scrambled-RNAtransfected (Scr) FDCLCs in the presence or absence of TNFα in Matrigel-coated Transwells (Figures 3(a) and  3(b)). TNFα stimulation remarkably enhanced FDCLC migration (by 7.5-fold) compared to the controls. The MMP3i-expressing cells, however, showed a robust suppression of TNFα-induced migration by 98%. These results strongly indicated that the migration of FDCLCs in vitro is induced by TNFα but that MMP3 expression is essential for this process.
To verify the effectiveness of our MMP3 knockdown in the migration experiments, MMP3 mRNA expression and protein secretion were measured in Scr-or MMP3itransfected FDCLCs with or without TNFα treatment (Figures 3(c) and 3(d)). TNFα exposure increased MMP3 mRNA expression and protein secretion by 22.5-fold and 2.7-fold, respectively, in the Scr cells. The MMP3i cells, however, showed a reduction in these levels by 94% and 70%, respectively. The MMP3i knockdown in the FDCLCs also successfully blocked the upregulation of MMP3 mRNA expression, and almost completely abrogated the induction of MMP3 secretion, following exposure of the cells to TNFα. We additionally found that MMP3i FDCLC had reduced the length of dendrites by 34% on collagen-1-coated plates compared to Scr cells (Figures 3(e) and 3(f)). Overall, these results revealed that TNFα-induced FDCLC migration is enhanced by MMP3 expression, and the TNFα-mediated MMP3 expression is dependent on ERK1/2 and AP1 activation.

Discussion
A prolonged exposure to TNFα and GC spreading are characteristic features of many lymphoid lesions, including autoimmune diseases and some lymphomas. The molecular mechanisms that control the migration of FDCs, which are unique stromal cells in the GC, have not been actively investigated to date due in part to the lack of an adequate FDC study model. In our current study, however, we have now successfully explored the molecular pathways underlying the TNFα-induced migration processes of stromal cells in the B cell follicle by utilizing unique human primary FDCLCs that have been reported to retain some of the physiologic properties of normal FDCs [18,24].
We found from our present analyses that MMP3 is expressed in the GCs in vivo and that FDCLC migration is dependent on this expression in vitro. The expression of MMP3 in FDCs is mediated through the ERK1/2-AP1 signaling axis by TNFα, which is known to be produced by B cells in the GCs [25]. TNFα can activate multiple signaling pathways, including three MAPKs, i.e., ERK1/2, JNK, and p38 MAPK [26,27]. These kinases mainly induce the transcription factors NF-κB or AP1 [28,29]. AP1 is closely associated with TNF-mediated inflammatory responses, including chemokine and cytokine production [30]. In our current study, 7 Journal of Immunology Research we found that TNFα increases the activity of AP1 through ERK1/2 phosphorylation in FDCLCs. AP1 is bound to the MMP3 promoter region and induces MMP3 expression. These results are consistent with those of previous studies in other various cell types, which showed that ERK1/2 and AP1 control the expression of diverse MMPs. MMP3 expres-sion has been shown to be induced through the TNFα-MAPK-AP1 pathway in nucleus pulposus cells and synovial fibroblasts [14,15]. In addition, MMP9 is increased by TNFα-ERK1/2-AP1 in human trophoblastic cells [31]. TNFα also induces MMP1 expression in human fibroblasts through MAPK-AP1 signaling [32]. Similarly, the expression of  MMP13 is mediated by TNFα in human articular chondrocytes and is dependent on ERK1/2 and AP1 [33]. Further investigations are needed, however, to look into the expression changes of other MMPs and TIMPs in FDCLCs to uncover more details of the molecular mechanisms of FDCLC migration.
Another interesting finding from our present experiments was that FDCLCs with an MMP3 knockdown appear to have shorter dendrites than normal cells on collagen-1coated plates (Figures 3(e) and 3(f)). Similar findings have been described in a previous study which reported that FDCs isolated from the mouse GCs attached to a collagen-1 matrix, spread out, and generated FDC networks in vitro [34]. Collagen-1 is the major stromal collagen subtype of human GCs, and MMP3 can degrade collagen-1 through the conversion of MMP1 from its proform to its active form [35][36][37]. Our current results have revealed that an MMP3 knockdown suppresses TNFα-induced FDC migration. Taken together, our present data and previous evidence now indicate that FDC dendrite elongation in the GC and migration likely require the degradation of collagen-1 through a complex interaction of MMPs mediated via TNFα signaling from GC-B cells.
In conclusion, this is the first study to our knowledge to reveal the core molecular mechanisms underlying FDCLC migration. Our findings can be a basis for further delineation of FDC-GC biology for specific Ab production, and for the development of targeted treatments for GC-related disorders including follicular lymphoma and autoimmune diseases.

Data Availability
All data are available on request from the authors (H-KP: ledmilage@gmail.com; C-SP: csikpark@amc.seoul.kr).

Disclosure
These funding bodies had no role in the study design, data collection and analysis, decision to publish, or manuscript preparation.