Polymerase Chain Reaction Selects a Novel Disintegrin Proteinase from CD40-Activated Germinal Center Dendritic Cells

GCDCs were purified from human tonsils according to Grouard et al. (4). After collagenase IV and DNAse digestion of tonsils, cells were centrifuged through a 50% Percoll gradient for 20 min at 400 g. CD3⁺ T cells, CD19⁺ B cells, CD14⁺ monocytes (anti CD3[OKT3]), CD19 [4G7], and CD14 [MOP9] mAbs were purified from ascites, our laboratory), and CD16⁺CD56⁺ NK cells (ION16; Immunotech, Marseille, France; and NKH1; Ortho Diagnostic System, Raritan, NJ) were removed from the collected low density cell population by magnetic beads (anti–mouse Ig-coated Dynabeads M450; Dynal, Oslo, Norway). The remaining cells were stained with mouse anti– CD4-PE-Cy5 (Immunotech), anti–CD11c-PE (Becton Dickinson, Mountain View, CA), anti–CD3-FITC and CD34-FITC (Immunotech), anti–CD20-FITC and anti–CD16-FITC (Becton Dickinson) anti–CD1a-FITC (Ortho Diagnostic System). CD4⁺CD11c⁺ CD3⁻CD20⁻CD1a⁻ GCDCs were isolated by cell sorting using FACStar Plus^® (Becton Dickinson). In average, a tonsil pair yielded 3 × 10⁵ cells of 98% homogeneity, a total of 12 tonsil pairs was needed to collect 3 × 10⁶ cells.

Blood CD4⁺CD11c⁺ DCs were prepared from PBMCs essentially following the procedure for GCDCs. After cell sorting, the purity was >95%. GCDCs and blood DCs were CD40-stimulated for 24 h in complete medium in the presence of 10 μg/ml anti-CD40 antibody G28-5 (provided by Dr. E. Clark). Complete medium is RPMI 1640 (GIBCO BRL, Gaithersburg, MD) supplemented with 10% (vol/vol) FCS (Flow Laboratories, Irvine, UK), 10 mM Hepes, 2 mM l-glutamine, 5 × 10⁻² M 2-mercaptoethanol, and 0.08 μg/ml gentamycine (Schering-Plough, Levallois Perret, France).

Stem cell-derived DCs were obtained after 6 and 12 d of culture of CD34 progenitor cells in the presence of TNF-α and GM-CSF (5, 6). Monocyte-derived DCs were produced by incubating purified human monocytes for 5 d in the presence of GM-CSF and IL-4 (7). For CD40 activation, the in vitro–generated DCs were cultured over irradiated murine CD40L-transfected L cells established in this laboratory (8).

Blood mononuclear cells were obtained from human peripheral blood by Ficoll-Hypaque centrifugation. Cells were collected at the interphase and washed twice in complete medium.

T lymphocytes were purified from PBMCs by immunomagnetic depletion using a cocktail of mAbs (IOM2 [CD14], ION16 [CD16], IOT17 [CD35], and ION2 [HLA-DR] from Immunotech; and [CD19] [ascite]). The purity of CD3⁺ T cells (CD4⁺ and CD8⁺) is >95%. Naive T cells were positive selected by anti-CD45RA (ascite, our laboratory).

B lymphocytes were obtained from human tonsils as described (9). T cells were first depleted by rosetting sheep red blood cells and then the residual non–B cells were depleted by T cell–specific antibodies (CD2, CD3, and CD4) followed by anti–mouse IgG-coated magnetic beads (Dynal). The resulting cell population is >98% CD19⁺ B cells.

Monocytes were purified by CD14⁺ FACS^® sorting after preparation of PBMCs followed by 50% Percoll gradient.

Cell lines (TF-1, MRC5, CHA, U937, JY, and JURKAT) were obtained from American Type Culture Collection (Rockville, MD). Where indicated, cells were activated for 1 and 6 h with 1 μg/ml phorbol 12-myristate 13-acetate (PMA; Sigma Chemical Co., St. Louis, MO) and 1 ng/ml ionomycin (Calbiochem Corp., La Jolla, CA) and the time points were pooled.

Cells were lysed, and total RNA made as described (10). The integrity of the RNA was confirmed by denaturing agarose gel electrophoresis. For GCDC, 3 × 10⁶ cells were lysed in 0.8 ml lysis buffer (10), and 1/35 volume chloroform/isoamyl alcohol (1:24) was used rather than a 1/20 volume. For tester and driver preparation, polyA⁺ messenger RNA (mRNA) was selected by oligo dT₍₂₅₎-coupled magnetic beads (Dynabeads Oligo[dT]₂₅; Dynal). For U937, three rounds of polyA selection were done before uncoupling polyA⁺ RNA from the beads. In the case of GCDCs, 100 μl beads (1 mg/ml) was added to the 15 μg total RNA obtained, and after the uncoupling, the remaining total RNA was again selected by three more rounds of beads and finally detached. The quality of U937 polyA⁺ RNA was verified on a denaturing agarose gel.

All of the GCDC mRNA (estimated at 140 ng) was used as tester, and 2 μg U937 mRNA as driver. Complimentary DNA synthesis and subtraction was done in essence following the PCR-select kit (Clontech, Palo Alto, CA; reference 11) using Advantage™ KlenTaq polymerase (Clontech). To compensate for the low amount of tester cDNA, verification of RsaI digestion (step V.B) and the dilution step (step F1) were omitted. The first PCR reaction was done for 28 cycles and the second (nested) PCR for 12 cycles on a thermal cycler (480; Perkin-Elmer Corp., Norwalk, CT). To clone subtracted GCDC cDNA, 10 nested reactions were pooled and resolved on 2% low melting agarose. Aiming for individual bands, 10 gel slices in the 0.7–1.4 kb size range were cut out; the DNA was eluted and cloned either directly or after reamplification into a T/A-vector (pCRII Invitrogen Corp., San Diego, CA). The inserts were sequenced in both directions by automatic sequencing. Comparisons against GenBank and dbest databases as well as protein homology prediction were obtained from the NCBI blast server (http://www.ncbi.nlm.nih.gov/BLAST/).

In the case of the human cell lines and stem cell-derived DCs (Fig. 2,A) 50 μg of total RNA was treated with 20 units DNAse (Boehringer Mannheim GmbH, Mannheim, Germany) in the presence of RNAse inhibitor (RNAsin; Promega, Madison, WI) in standard reaction conditions. After phenol/chloroform extraction, the RNA was reverse transcribed using Superscript™II (GIBCO BRL, Gaithersburg, MD) and a dT_(12-18) oligonucleotide, according to instructions (GIBCO BRL). The reaction was stopped, and nucleic acids were ethanol precipitated and resuspended in water. About 20 ng cDNA was used in one PCR reaction. For all other cells, total RNA was prepared from ∼100,000 cells, reverse transcribed using Superscript^TMII (GIBCO BRL) and a random hexaoligonucleotide (pd[N]₆; Pharmacia) in a 20 μl reaction. 1 μl was used in a PCR reaction. The 50 μl PCR reaction contained 100 ng primers, 200 μM dNTP (Pharmacia, Uppsala, Sweden), and 0.5 units AmpliTaq™ polymerase (Roche Molecular Systems Inc., Branchburg, NJ) in standard PCR buffer (Roche Molecular Systems) and was subjected to 28 or 35 cycles of denaturing (30 s, 94°C), annealing (30 s, 60°C), and extension (2 min, 72°C) on the Perkin-Elmer thermal cycler 480. Decysin primers U137 and L677 are shown in Fig. 3 A, β actin primers were purchased (Stratagene, La Jolla, CA). The Marathon™ kit (Clontech) was used for RACE PCR and performed as recommended by the supplier. The outward primers were 5′-CCCATCAGACCAGATTTCCATACCTACC (upstream) and 5′-CCCATCTTCGGTTGCTGTTATTGAGGCT (downstream), and were used with the Marathon™ recommended cycling program 1. The two distinct PCR products were cloned into the T/A-vector and sequenced.

PolyA⁺ RNA blots of human tissues were purchased (Clontech). The original cloned 744-bp fragment was labeled by random priming with [³²P]dCTP (3,000 Ci/mmol; Amersham Intl., Buckinghamshire, UK; and HiPrime; Boehringer Mannheim GmbH) and unincorporated nucleotides were removed by spin column chromatography (Chromaspin-100; Clontech). Membranes were prehybridized at 65°C in Church solution (0.5 M NaHPO₄, pH 7.2, 7% SDS, 0.5 mM EDTA), heat-denatured probe was added, and incubated overnight at 65°C. The membranes were washed under high stringency conditions (0.1 × SSC/0.1% SDS at 65°C) and exposed for 2 wk.

In situ hybridization was done as described (12). Sense and antisense ³⁵S probes were made by runoff transcription of the 744-bp fragment. 6-μm sections of tonsils were fixed in acetone and 4% paraformaldehyde followed by 0.1 M triethanol amine/0.25% acetic acid. The sections were hybridized overnight, RNase A treated, and exposed for 3 wk. After development, the cells were stained with hematoxyline.

Germinal center DCs were purified from human tonsils (4) and activated by anti-CD40 antibody G28-5 in complete medium for 24 h. From 12 tonsil pairs, a total of three million GCDCs could be obtained, which were estimated to be 98% pure but represented too little material for a conventional subtracted cDNA library without PCR amplification. By modification of the subtractive hybridization technique, termed PCR-select (11), the amount of tester cDNA necessary could be lowered to the 140 ng GCDC mRNA obtained. GCDC cDNA (tester) was cut with RsaI, adapters ligated, and after hybridization in the presence of competitor (driver) cDNA from human monocytic cell line U937, amplified. Thus, the resulting PCR products (Fig. 1,A, lane 8) are restriction fragments of GCDC cDNA absent, or at least rare, in U937. Indeed, individual bands can be seen in lane 8 that are more clearly resolved in B. This is obviously different from amplified GCDC cDNA in absence of competitor (lane 6). As a positive control, subtracting skeletal muscle cDNA containing a trace amount of molecular weight marker against skeletal muscle cDNA only, was able to distinctly expose the added marker DNA (compare lanes 2 and 4). Given these results, GCDC cDNA fragments from panel B in the 0.7–1.4-kb size range were cloned, and 250 clones were sequenced. 30% of the clones contained unknown sequences. Among the known, we found a number of cDNA fragments corresponding to genes which are highly expressed in dendritic cells (Table 1). This demonstrates the feasibility of making a high quality subtraction library from a low number of cells.

Among the unknown genes, we retained a clone containing a 744-bp insert because it was not expressed in U937, nor in the following tested human cell lines: TF1 (myeloid precursor cell), JURKAT (T cell), CHA, MRC5 (kidney epithelial and lung fibroblastic cells), and JY (B cell) (Fig. 2 A). Yet, it is expressed in CD34 progenitor cell–generated DCs. As control, the cDNA is present in the nonsubtracted GCDC library, and enriched after subtraction. Among freshly isolated cells (B), the cDNA could be detected in PBMCs, monocytes, and B cells, but at a lower level compared to CD40-activated GCDCs where 28 PCR cycles were sufficient to generate a clearly visible product. Expression was never observed in T cells. As enough mRNA could not be obtained from GCDC, the expression of this gene in CD34 progenitor cell-generated DCs allowed us to clone the full-length cDNA.

Several RACE PCR products were sequenced to determine the full-length 2,187-bp nucleotide sequence of the gene which we have named decysin (Fig. 3 A). It encodes an open reading frame of 470 amino acids with strong homology to disintegrin metalloproteinases. It has two putative start codons, although the latter is likely to be used more regularly since it more closely resembles a Kozak consensus (13). It is followed by a hydrophobic signal sequence. As a member of the metalloproteinase family, decysin contains a prodomain with a cysteine-switch activation domain at cysteine 187 (14) and a furin cleavage site between residues 200 and 203 (15). Two zinc-chelating histidine (H) residues of the zinc binding pocket (B) are found at positions 352 and 356 as well as the methionine turn at residue 176 (16, 17). The glutamic acid after the first histidine is strictly conserved in active enzymes. However, distinct from all known mammalian disintegrin or matrix metalloproteinases (18, 19), the third zinc-chelating histidine is replaced by an aspartic acid (residue 362). In the bacterial metalloproteinase ScNP (20), an aspartic acid found at exactly this position was recently identified as the third zinc-chelating amino acid (21; B). All mammalian disintegrin metalloproteinases share an ∼90 amino acid stretch with snake venom disintegrins (22; C). Decysin comprises many of these conserved residues, but its open reading frame terminates half way along the consensus. It lacks a transmembrane region that together with the signal peptide, suggests that it is secreted. In summary, by a number of common criteria, the gene codes for a novel member of the disintegrin metalloproteinases with unique features so far unobserved in any other mammalian metalloproteinases.

Since decysin was identified in CD40-stimulated GCDCs, we wondered whether the metalloproteinase might be expressed at high levels in CD40-activated DCs. Indeed, 28 cycle PCR coupled to reverse transcribed RNA (RT-PCR) on freshly isolated GCDCs failed to detect decysin (Fig. 4 A). Its expression is induced by spontaneous maturation in culture (4) and increases in response to CD40 activation. Similarly, in another ex vivo isolate, blood CD11c⁺ DCs do not contain detectable decysin mRNA, but maturation in culture (23) and more importantly CD40 activation result in decysin induction. In vitro generated DCs from CD34 progenitor cells or monocytes rapidly synthesize the message in response to CD40 ligation and a mixed lymphocyte reaction (B) results in decysin expression together with CD83. Thus, the novel metalloproteinase represents a DC maturation marker synthesised in response to T cell signals.

By Northern blot analysis on different human tissues (Fig. 5), decysin is expressed as a single 2.4-kb message and is highly abundant in the small intestine and appendix. Database searches produced a single partial expressed sequencing tag from pig small intestine (not shown). Expression is also seen in lymph node, mucosal lining of the colon, thymus, spleen, and very weakly in bone marrow. Peripheral blood, ovary, testis, prostate, and fetal liver are negative, as well as other blots containing tissues such as heart, lung, or liver (data not shown).

To localize decysin mRNA in human lymph nodes, in situ hybridization was performed (Fig. 6). A tonsil section probed with the antisense RNA strand shows distinct hybridization signals primarily within germinal centers (A and C) in contrast to the same follicles probed with the sense strand (B). The follicle marked by an arrow is shown in higher magnification (C). Silver grains are in focalized clusters evenly distributed with the germinal center. This profile is identical to that obtained by anti-CD11c immunostaining of GCDCs (compare C and D), and confirms that DCs of the germinal center express high levels of decysin.

The study of DCs is hampered by their scarcity in vivo. In this paper, a PCR-based method was used to clone a cell type–specific cDNA from three million CD40-activated germinal center dendritic cells isolated from human tonsils. Several lines of evidence indicate the power of this technique. (a) Among the 250 sequenced clones, <5% contained common housekeeping genes (β actin could not be amplified from subtracted GCDC cDNA after 35 cycles of PCR [Fig. 2 A]). (b) A third were unknown genes. (c) 107 clones represented genes whose expression is either specific to or highly expressed by mature DCs (CD83, DC tactin, DEC205, MHC class II, Rel-B, IAP-c). On average, we determined that 1 out of 10 clones corresponds to a gene expressed in tester GCDCs and not in driver U937. All these data suggest that the PCR-based subtraction method used here is well applicable to clone unique genes from low number of ex vivo cells, including different DC subsets.

In this subtracted library we identified a novel member of the disintegrin metalloproteinases. It is a large family of mostly membrane-anchored proteases with conserved disintegrin and cysteine-rich domains and a zinc-chelating pocket, although not all members are active enzymes (19). They encode an adhesive function to cell-surface proteins through the COOH-terminal disintegrin domain and a potential antiadhesive/cleavage function through the zinc- dependent metalloprotease domain. Members of this group encode diverse functions. Fertilin α and β have been implicated in sperm–egg binding and fusion (24), meltrin in muscle cell fusion (25), and Kuzbanian in neurogenesis (26). Snake venom disintegrins bind the β₃α_v integrin and prevent its interaction with fibrinogen (27). By the use of specific inhibitors, metalloproteases have been implicated in shedding of a number of molecules which play critical roles in the immune system: TGF-α, TNF receptors p60 and p80, FasL, CD30, IL-6 receptor, and L-selectin (28– 33). Recently, the enzymes that process TNF-α have been identified as disintegrin metalloproteinases TACE (34, 35) and ADAM-10 (36).

Decysin is a member of the disintegrin metalloproteinase family by the following criteria: it has a hydrophobic leader followed by a prodomain with a cysteine-switch consensus, a mechanism by which a prodomain cysteine ligates the active site zinc, and retains the zymogen in an inactive state. Decysin comprises a zinc-chelating catalytic site with a methionine turn and most of the disintegrin domain. Yet, it is unique in three points. First, the third zinc-chelating residue, a histidine in all other disintegrin and matrix metalloproteinases, is replaced by aspartic acid in analogy to a bacterial proteinase. Second, the disintegrin domain is truncated. Third, decysin lacks a transmembrane domain, and is therefore potentially secreted. It will be important to examine whether this gene might represent the first member of a new subclass of mammalian disintegrin metalloproteinases.

RT-PCR distribution analysis showed that decysin is moderately expressed by normal B cells and monocytes, but not by T cells and a wide range of human cell lines, even though they had been treated with PMA/ionomycin which is known to promote processing of TGF-α, TNF-α, TNFR, IL-6R, and L-selectin by metalloproteinases (28– 30). Instead, decysin is induced or upregulated after spontaneous or CD40 ligand–promoted maturation of different types of immature DCs in vitro. These include blood CD4⁺CD11c⁺ DCs, tonsillar CD4⁺CD11c⁺ GCDCs, monocyte-derived DCs with GM-CSF, IL-4, and CD34⁺ progenitor-derived DCs with GM-CSF and TNF. In addition, a reaction with allogenic T cells induces decysin synthesis in stem cell–derived DCs. CD40 ligand failed to upregulate decysin expression of human B cells (data not shown). In situ hybridization confirmed the expression of decysin in germinal centers in a pattern identical to CD11c staining of GCDCs. These data show that decysin is selectively expressed by DC and upregulated by signals of activated T cells. The cloning of decysin together with genes of mature DC markers CD83, DEC205, and DC tactin from CD40-activated GCDCs further support the idea of decysin as a novel DC maturation maker.

The impressive expression of decysin in the small intestine and appendix is likely due to the high abundance of DCs in gut epithelium, the lamina propria, and the Peyers patches which comprise a large population of lymphoid cells (37). The continuous and high antigenic load in these sites may induce chronic DC–T cell interactions. Since the members of disintegrin metalloproteinase such as TACE and ADAM-10 have been shown to process membrane-bound TNF-α precursors, and metalloproteinases have been implicated in processing of other members of the TNF family (Fas-L, CD30, and TNFR), decysin may represent a key molecule in regulating DC–T cell interaction.

We thank C. Massacrier, O. de Bouteiller, and P. Garrone for cells, I. Durand for help in FACS^® sorting, and E. Bates for critical review of the manuscript. We also thank J. Banchereau, J. Chiller, and members of the laboratory for support and discussion.