| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
* Department of Surgery, McGill University, Montreal, Quebec, Canada;
McGill University and Genome Quebec Innovation Center, Montreal, Quebec, Canada; and
Département de Stomatologie, Université de Montréal, Quebec, Canada
4Correspondence: McGill University, Montreal General Hospital, 1650 Cedar Ave., Rm. C9–177, Montreal, Quebec H3G 1A4, Canada. E-mail: anie.philip{at}mcgill.ca
ABSTRACT
We have previously reported that keratinocytes defective in glycosylphosphatidylinositol (GPI)-anchor biosynthesis display enhanced TGF-ß responses. These studies implicated the involvement of a 150 kDa GPI-anchored TGF-ß1 binding protein, r150, in modulating TGF-ß signaling. Here, we sought to determine the molecular identity of r150 by affinity purification and microsequencing. Our results identify r150 as CD109, a novel member of the 
2-macroglobulin (2M)/complement superfamily, whose function has remained obscure. In addition, we have identified a novel CD109 isoform that occurs in the human placenta but not keratinocytes. Biochemical studies show that r150 contains an internal thioester bond, a defining feature of the 2M/complement family. Loss and gain of function studies demonstrate that CD109 is a component of the TGF-ß receptor system, and a negative modulator of TGF-ß responses in keratinocytes, as implicated for r150. Our data suggest that CD109 can inhibit TGF-ß signaling independently of ligand sequestration and may exert its effect on TGF-ß signaling by direct modulation of receptor activity. Together, our results linking CD109 function to regulation of TGF-ß signaling suggest that CD109 plays a unique role in the regulation of isoform-specific TGF-ß signaling in keratinocytes.—Finnson, K. W., Tam, B. Y. Y., Liu, K., Marcoux, A., Lepage, P., Roy, S., Bizet, A. A., Philip, A. Identification of CD109 as a TGF-ß1 accessory receptor in human keratinocytes.
Key Words: TGF-ß • receptor • CD109 • signaling • keratinocyte
TRANSFORMING GROWTH FACTOR-ß (TGF-ß) plays a critical role in skin development, homeostasis, and wound healing (1
2
3)
. Perturbations in the action of TGF-ß has been implicated in a variety of skin disorders, including impaired wound healing (4
, 5)
, hypertrophic scarring (6)
, psoriasis (7)
, and cancer (8)
. However, the regulation of TGF-ß signaling and action in skin cells is poorly characterized. TGF-ß superfamily members signal through type I and II transmembrane serine/threonine kinase receptors (9)
. Ligand binding induces the heteromerization of the type I–type II receptor complex, resulting in the transphosphorylation of the type I receptor by the type II receptor. The activated type I receptor then propagates the signal by phosphorylating its intracellular substrates, Smad2 and Smad3. The phosphorylated Smads then form heteromeric complexes with Smad4 and translocate into the nucleus where they control gene expression in a cell type-specific manner through interactions with transcription factors, coactivators, and corepressors (10)
.
Mechanisms regulating TGF-ß signal transduction include modulation of TGF-ß signaling receptor activity by accessory receptors, such as betaglycan (type III TGF-ß receptor) and endoglin (10)
. Betaglycan, a proteoglycan, binds all three TGF-ß isoforms with high affinity and facilitates TGF-ß binding to the TGF-ß signaling receptors (11
, 12)
. Furthermore, it can be phosphorylated by the type II receptor and can interact with ß-arrestin resulting in the endocytosis of both itself and the type II receptor (13)
. Endoglin, which shows 70% homology to betaglycan, does not bind TGF-ß on its own but can bind the TGF-ß1 and TGF-ß3 isoforms through its association with the type II receptor (14)
. Also, endoglin expression has been reported to alter the phosphorylation status of the type I and II receptors (15)
. Interestingly, ectopic expression of endoglin attenuates TGF-ß responses in monocytes and myoblasts, whereas overexpression of betaglycan enhances TGF-ß responses in these cells (11
, 14)
. The precise functional mechanisms by which endoglin and betaglycan modulate TGF-ß signaling are incompletely understood. Cell surface interactions of such accessory receptors with the TGF-ß signaling receptors, however, may represent modes of specifying signaling pathways and regulating diverse TGF-ß responses.
Several groups including ours have reported the occurrence of cell surface glycosylphosphatidylinositol (GPI)-anchored proteins of different molecular weights that bind TGF-ß in an isoform specific manner (16
17
18
19
20)
. However, the identities of these proteins remain unknown (10)
. We have recently reported that human keratinocytes defective in GPI-anchor biosynthesis display enhanced TGF-ß responses, indicating that GPI-anchored proteins regulate TGF-ß signaling in those cells (21)
. Our results suggested that an isoform-specific 150 kDa cell surface GPI-anchored TGF-ß1 binding protein, r150, which forms a heteromeric complex with the type I and II TGF-ß receptors, may play a role in this regard (21)
. However, since the molecular identity of r150 was unknown, the hypothesis that its loss is responsible for the enhanced TGF-ß signaling in these mutant cells could not be tested at the time. This, together with other known properties of r150, such as its high affinity for TGF-ß1, its ability to bind TGF-ß after release from the cell surface by phosphatidylinositol phospholipase C (PIPLC), and its capacity to form a heteromeric complex with the TGF-ß signaling receptors (22)
, provide compelling reasons to explore its molecular identity and significance as a regulator of TGF-ß signaling.
CD109 is a GPI-anchored protein, first identified using monoclonal antibodies raised against the human leukemia cell line KG1a, on activated T cells and platelets and on a subset of hematopoetic progenitor cells (23
, 24)
, with a wider distribution reported more recently (25
, 26)
. Although CD109 has been found to represent the Gov alloantigen on platelets (27)
, and its molecular cloning as a novel member of the 2-macroglobulin(2M)/complement gene family has recently been reported (28
, 29)
, its function has remained unknown. In the present study, we provide molecular and biochemical evidence that r150 represents CD109 and demonstrate that CD109 is a novel TGF-ß accessory receptor in keratinocytes, linking CD109 function to inhibition of TGF-ß signaling. Importantly, we show that CD109 interacts directly with the type I TGF-ß signaling receptor and that CD109 negatively modulates TGF-ß1 signaling, possibly through a mechanism that is independent of sequestration of TGF-ß1 ligand. The present study is the first report on the molecular identification of a GPI-anchored TGF-ß binding protein.
MATERIALS AND METHODS
Cell lines
The human keratinocyte cell line HaCaT (30)
was kindly provided by P. Boukamp (Heidelberg, Germany). It was cultured using standard techniques as described previously (21
, 22)
. The preparation and culture of HaCaT cells mutated in GPI anchor biosynthesis have been described previously (21)
. Stable transfectants were generated by transfecting HaCaT cells with CD109S or its empty vector (pCMVsport6) together with pCDNA3 (for neomycin resistance), followed by selection with 500 µg/ml geneticin (G418, Invitrogen). G418 resistant colonies were then pooled and expanded. Overexpression of CD109 was confirmed by Western blot. Mouse embryonic fibroblasts (MEFs) were kindly provided by A. Roberts (National Cancer Institute, Bethesda, MD). They were cultured in DMEM (low glucose) supplemented with 10% FBS, 100 U/ml penicillin, and 50 µg/ml streptomycin as described previously (31)
. Human embryonic kidney (HEK) 293 cells were purchased from the American Type Culture Collection and cultured in D-MEM supplemented with 10% FBS, 100 U/ml penicillin and 50 µg/ml streptomycin.
Antibodies
Mouse monoclonal anti-human CD109 (TEA 2/16), anti-fibronectin and anti-plasminogen activator inhibitor-1 (PAI-1) antibodies was purchased from BD Biosciences (Mississauga, ON, Canada). Other anti-human CD109 antibodies used include mouse monoclonal antibody 8A3 (kindly provided by R. Sutherland, Oncology Research Lab, Toronto, ON, Canada), and 1B3 (kindly provided by I. Bernstein, Fred Hutchinson Cancer Research Center, Seattle, WA). The anti-TGF-ß neutralizing antibody (Ab) (1D11) was kindly provided by Genzyme Inc. (Framingham, MA). Normal rabbit IgG, anti-type I TGF-ß receptor (V-22) anti-type II TGF-ß receptor (C-20), and anti-actin (H-300), anti-Smad2 (S-20) and anti-Smad3 (FL-425) antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Rabbit polyclonal anti-TGF-ß type III receptor (Get-1) and anti-phosphoSmad2 antibodies were gifts from S. Souchelnytskyi, Uppsala, Sweden. Rabbit polyclonal anti-phosphoSmad3 Ab was purchased from Cell Signaling Technologies (Beverly, MA). HRP-conjugated antimouse and anti-rabbit antibodies were obtained from BD Biosciences.
Purification, microsequencing and sequence analysis
To determine the identity of the 150 kDa GPI-anchored TGF-ß1 binding protein, r150 (17
, 21
, 22)
, HaCaT cells (834 confluent T75 cm2 flasks) were harvested in nonenzymatic cell dissociation medium (Sigma Aldrich, Oakville, ON, Canada) and incubated with 0.5 U/ml of PIPLC (Roche Diagnostics, Laval, QC, Canada) at 37°C for 1 h to release GPI-anchored proteins from the cell surface. The pooled supernatants were purified on a TGF-ß1-affinity column prepared by immobilization of TGF-ß1 (Genzyme Inc., Framingham, MA) using the AminoLink Immobilization kit (Pierce, Rockford, IL). Individual fractions were collected, resolved by SDS-PAGE and visualized by silver staining (32)
. Protein bands corresponding to 150 kDa were excised, and sequence analysis was performed at the Harvard Microchemistry Facility by microcapillary reverse-phase HPLC nanoelectrospray tandem mass spectrometry.
Sequence analysis
A 19 amino acid microsequence obtained from mass spectrometry was used to query the NCBI database. It matched to an IMAGE clone, CS0DI081YB18 (derived from a human placental library, Research Genetics Inc., Huntsville, AL) for which the 5'end and 3'end sequences were known. The full sequence of the Image clone was obtained using six primers, r150–1 to r150–6 (Table 1
), designed with information based on overlapping ESTs. It was found to be identical to that of the human CD109 (cloned from KG1a cells) reported by Lin et al. (29)
, except that the IMAGE clone is missing 51 bp corresponding to position 3652–3702 of the coding sequence. We designated the (human placental) IMAGE clone sequence as CD109S, and in the present study we refer to the sequence reported by Lin et al. (29)
as CD109. We constructed the CD109 form, which contains the 51 bp at positions 3652–3702 by reverse transcriptase-polymerase chain reaction (RT-PCR) of HaCaT cell RNA using primers r150–7 and r150–8 (Table 1)
. The PCR product was digested with Pae I and BglII and inserted into the Pae I/BglII site of CD109S, which was in a pCMVSport6 vector. The identity of CD109 was confirmed by sequencing.
|
Determination of the expression of CD109S and CD109 in human keratinocytes and placenta
We performed RT-PCR using total RNA extracted from HaCaT cells and term placenta. Equal amounts (1.0 µg) of total RNA were reversed-transcribed using oligo-dT and MMLV-reverse transcriptase at 37°C. PCR was performed to amplify the region containing the 51 bp region (3652–3702) deleted in CD109S using primers, r150–7 and r150–8 (Table 1)
, which amplify the sequence corresponding to positions 3341–3986 of CD109 or 3341–3935 of CD109S, to generate PCR products of 646 or 595 bp, respectively.
Affinity labeling and immunoprecipitation studies
Affinity labeling of cells with [125I]TGF-ß1 was performed as described previously (21
, 22)
. Membrane extracts were prepared and protein concentrations determined using a Bradford protein assay (BioRad, Mississauga, ON). The extracts were then not immunoprecipitated (NIP) or immunoprecipitated (IP) with the indicated antibodies or IgG (control), and analyzed by SDS-PAGE/autoradiography (21
, 22)
.
Methylamine treatment
HaCaT cells were affinity-labeled with [125I]TGF-ß1 and membrane extracts boiled for 30 min before or after incubation with 400 mM methylamine (Sigma) for 30 min at 37°C. Alternatively, extracts were immunoprecipitated with the anti-CD109 Ab, 8A3, and the immune complexes were left untreated or treated with 400 mM methylamine for 30 min at 37°C followed by boiling for 30 min. Samples were analyzed by SDS-PAGE/autoradiography.
Transient transfections and antisense delivery
HaCaT cells were transfected with CD109, CD109S or empty vector (pCMVSport6) using Superfect (Qiagen Inc., Mississauga, ON, Canada), as described previously (21)
. 293 cells were transfected with TGF-ß type I and II receptor cDNAs, with either CD109 or its empty vector, using Superfect. MEFs were transfected with the indicated cDNAs using Fugene 6 Transfection Reagent (Roche Diagnostics). Alternatively, HaCaT cells were treated with CD109-specific antisense morpholino oligos, GCC-CTG-CAT-CTC-GAC-GGC-GTC-TGC-C or inverse control oligos (Gene Tools, Philomath, OR), according to manufacturer’s instructions.
Luciferase reporter assays
At transient transfection, cells were cotransfected with 3TP-lux (33)
(provided by Dr. M. O’Connor-McCourt, NRC, Montreal) or (CAGA)12-lux (34)
(a gift from Dr. S. Huet, Laboratoire Glaxo Wellcome, France) reporter constructs, and with ß-galactosidase to monitor transfection efficiency. In one series of experiments, 293 cells were cotransfected with a constitutively active form of RI (T204D, obtained from J. Massague, Memorial Sloan-Kettering Cancer Center, NY) or its empty vector (pCMV5). Cells were allowed to recover for 24 h, and then incubated for 16 h in serum-free media containing 0–100 pM of TGF-ß1 (Genzyme Corp.). Cell lysates were prepared, analyzed for luciferase activity, and the values were normalized to ß-galactosidase activity (21)
.
Western blot and immunoprecipitation
For immunoprecipitation/Western blot studies, HaCaT cells were washed with a mild acid (0.1% glacial acetic acid) to ensure complete removal of endogenous TGF-ß (35)
, membrane extracts were prepared, and protein concentrations determined. The extracts were then left not immunoprecipitated or immunoprecipitated with the indicated antibodies and analyzed by Western blot using an anti-CD109 Ab (TEA 2/16, BD Biosciences). Alternatively, 24 h after transient transfection, or 48 h after antisense delivery, HaCaT or 293 cells were incubated for 15 or 60 min (for the determination of phosphoSmad2/3 levels) or 18 h (for the determination of fibronectin and PAI-1 levels) with the indicated amounts of TGF-ß1 under serum-free conditions. Cell lysates were prepared and analyzed by Western blot using the indicated antibodies (21)
.
Glutathione (GSH) S-transferase (GST) fusion protein interaction assay
The GST protein interaction assay was performed as described previously (36)
. Recombinant GST (gift from J. J. Lebrun, McGill University, Montreal, QC, Canada) and GST-RI, generated from the pGEX-RI plasmid (provided by P. ten Dijke, The Netherlands Cancer Institute, Amsterdam) were expressed in Escherichia coli and purified by adsorption to GSH-sepharose 4B (GE Healthcare Bio-Sciences Inc., Baie d’Urfé, QC, Canada). The concentrations of proteins immobilized on beads were semiquantified against BSA (Sigma) by SDS-PAGE and staining with Coomassie blue. 35S-labeled CD109 was generated by in vitro transcription and translation using the TNT kit (Promega). The radiolabeled mixture was incubated with 5 µg of either GST or GST-RI bound to the beads in binding buffer (50 mM Tris-HCl at pH 7.5, 200 mM NaCl and 0.5% Triton X-100) at 4°C for 2 h. Beads were collected by centrifugation and washed with binding buffer, and the adsorbed proteins were resolved by SDS-PAGE and visualized by autoradiography.
Thymidine incorporation
1) HaCaT cells stably transfected with CD109 or its corresponding empty vector, 2) HaCaT cells treated with CD109-specific antisense morpholino oligos or inverse oligos, or 3) GPI mutant or its parental HaCaT cells, were incubated without or with 50 pM TGF-ß1 in serum-free media for 48 h. Cells were pulsed with 1.0 µCi of [3H]thymidine for the final 6 h of the incubation, and the incorporated radioactivity was determined (17)
.
In vitro wound closure assay
Confluent monolayers of HaCaT cells stably overexpressing CD109 or its empty vector were wounded by manually scraping the cells with a pipette tip to create wounds of similar size. The cell culture medium was then replaced with fresh medium containing 0 or 25 pM of TGF-ß1 and incubated for 24 h. Wound area was selected by Adobe Illustrator and measured using Band Leader (M. Aharoni, TechKnowledge, Tel-Aviv, Israel). Wound size is expressed as a percentage of its size at time zero.
Cancer profiling array
The expression of CD109 in skin and other human tissues was determined using a Cancer Profiling Array (BD Biosciences-Clontech, Mississauga, ON, Canada). A CD109 cDNA probe was generated by PCR using oligonucleotide primers r150–9 and r150–10 (Table 1)
to amplify a sequence corresponding to nucleotide (nt) positions 1112–1528 of the coding sequence. PCR was performed as described above and the PCR product resolved by agarose gel electrophoresis. A single band corresponding to the expected size of 417 bp was extracted and labeled with [32P]-dCTP using the Megaprime Labeling System (GE Healthcare Bio-Sciences Inc.). The expression of CD109 in tissues (cDNA) on the Cancer Profiling Array membrane was determined using the probe according to manufacturer’s instructions.
RESULTS
Microsequencing and cloning of r150
We have recently reported that human keratinocytes (HaCaT cells) defective in GPI anchor biosynthesis display enhanced TGF-ß responses and that an isoform-specific 150 kDa cell surface GPI-anchored TGF-ß1 binding protein, r150, might be involved (21)
. This, together with other properties of r150 such as high affinity for TGF-ß1 and ability to associate with the TGF-ß signaling receptors (17
, 22)
, prompted us to seek its molecular identity. HaCaT cells were treated with PIPLC to release the GPI-anchored proteins and the supernatant containing these proteins was purified on a TGF-ß1 affinity column. Affinity labeling of keratinocytes with [125I]TGF-ß1 before PIPLC treatment shows the presence of membrane anchored r150 on the cell surface (Fig. 1
, left panel), while affinity labeling of fraction #6 collected from the TGF-ß1 affinity column shows the presence of the released r150 (Fig. 1
, right panel). The relatively high background in lane 3 is likely explained by the fact that affinity labeling was performed in solution. The high-affinity binding of TGF-ß1 by the membrane anchored and released (soluble) r150 is demonstrated by competition with unlabeled TGF-ß1 (lanes 1 vs. 2 and 3 vs. 4). The labeled band seen in fraction 6 at 
75 kDa appears to be nonspecific, as it was not displaced by unlabeled TGF-ß1. Fractions 6 and 7, but not other fractions collected from the affinity column, displayed a 150 kDa protein specifically bound to [125I]TGF-ß1 (data not shown).
|
Resolution of the affinity-purified fractions 6 and 7 containing the 150 kDa protein by SDS-PAGE, followed by analysis of the excised 150 kDa bands by mass spectrometry yielded a microsequence of 19 amino acids, SNLIQQWLSQQSDLGVISK, which matched to the 5'end of an IMAGE clone, CS0DI081YB18 (from a placental cDNA library) for which the 5' and 3'end sequences were known, and to the human CD109 sequence (29)
reported during the course of this study. When the IMAGE clone was fully sequenced and compared to the CD109 sequence reported by Lin et al. (29)
, it was found to have a deletion of 51 nucleotides in exon 28 corresponding to position 3652–3702 of the coding sequence (Fig. 1B
). This sequence codes for a 17 amino acid sequence, VKFLIDTHNRLLLQTAE. We designated the IMAGE clone sequence as CD109S and the reported CD109 sequence (29)
as CD109. The CD109S sequence was deposited in GenBank under the accession number AY788891. The protein encoded by CD109 has 1445 amino acids, with an N-terminal signal peptide of 21 amino acids and a C-terminal consensus GPI-anchor signal sequence with the cleavage predicted to occur after amino acid 1420 (1404 in CD109S) (Fig. 1B
). CD109 belongs to the 2M/complement superfamily and shares several structural motifs with members of that gene family, including a putative bait region (aa 651–683), a thioester signature motif (aa 918–924) and a thioester-reactivity defining hexapeptide (aa 1030–1035). A comparison of the nt and predicted amino acid sequences of the CD109 cDNAs from placenta (IMAGE clone), KG1a cells (29)
(GenBank Accession # AF410459), and U373 cells recently reported by Solomon et al. (28)
(GenBank Accession # AY149920) are shown in Supplemental Information Figs. 9 and 10, respectively. Other than the 17 amino acids missing in CD109S, the amino acid sequences of all three clones are identical except for the following amino acid substitutions (relative to the KG1a clone): I627M, Y703S, N797S, G803S, V845I, M1241T. Taken together, the sequencing results indicate that the 150 kDa TGF-ß1 binding protein purified from PIPLC-supernatants of HaCaT cells represents CD109.
CD109 but not CD109S is expressed in keratinocytes, while placenta expresses both isoforms
The cloning data above indicated the occurrence of two forms of CD109. To determine which isoform is expressed in keratinocytes, we performed RT-PCR on HaCaT and placental total RNA using primers r150–7 and r150–8 (Table 1)
to amplify the 51 bp region present in CD109, but missing in CD109S. As shown in Fig. 1C
, a single PCR product of 646 bp representing "CD109" is detected in HaCaT cells (lane 1), whereas 2 bands of 646 and 595 bp corresponding to "CD109" and "CD109S", respectively, are detected in the placenta (lane 2). As expected, the PCR products generated from CD109 and CD109S cDNAs (positive controls) yielded bands of 646 and 595 bp, respectively, (lanes 3 and 4). The 18S band indicated (lower panel) demonstrates that equal amounts of HaCaT and placental RNA were used. These data suggest that the keratinocytes express only CD109, while the placenta expresses both CD109S and CD109.
r150 contains an internal thioester bond and the 150 kDa form of r150 is derived from a 180 kDa form by autolytic cleavage
A unique characteristic shared by many 2M/complement superfamily members is the presence of an intact internal thioester bond between a cysteine and a glutamine in the sequence CGEQ (37)
. Under denaturing conditions, these proteins, including CD109, undergo an autolytic cleavage, which can be prevented if the thioester bond is first disrupted by treatment with proteases or nucleophilic amines such as methylamine (29)
. To determine whether r150 displays similar characteristics, HaCaT cells were affinity labeled with [125I]TGF-ß1, and membrane extracts, not immunoprecipitated (NIP) or immunoprecipitated (IP) with the anti-CD109 Ab, 8A3, were boiled before or after methylamine treatment (Fig. 1D
). SDS-PAGE/autoradiography analyses of extracts not immunoprecipitated reveal that boiling before the addition of methylamine results in the occurrence predominantly of a 150 kDa complex (lane 1). In contrast, the addition of methylamine before boiling led to the appearance of a 180 kDa complex with a concomitant decrease in the 150 kDa complex in membrane extracts (lane 2). The 150 and 180 kDa band are more readily detectable when the membrane extracts are immunoprecipitated with an anti-CD109 Ab (8A3) prior to methylamine/boiling treatment (lanes 3 and 4). No radiolabeled protein complexes were precipitated with IgG (lane 5), demonstrating that the immunoprecipitations are specific. These results indicate that r150 contains an intact thioester bond, and that the 150 kDa form is derived from the 180 kDa form with the involvement of a thioester mediated autolytic cleavage.
Altering CD109 levels changes TGF-ß1 binding to r150
To determine whether varying CD109 levels alters ligand binding on the cell surface, we performed affinity labeling of keratinocytes with [125I]TGF-ß1 following CD109 transient overexpression or antisense morpholino oligo treatment. As shown in Fig. 2
A, overexpression of CD109 leads to an approximate 2-fold increase in the amount of [125I]TGF-ß1 bound to r150 in extracts not immunoprecipitated, as determined by densitometric analysis (NIP, lane 2 vs. 1). Furthermore, membrane extracts immunoprecipitated with the 1B3 anti-CD109 Ab showed approximately a 2-fold increase in [125I]TGF-ß1 binding to r150 in HaCaT cells overexpressing CD109, relative to empty vector (EV) transfectants (lane 4 vs. 3), as determined by densitometric analysis. Similar results were obtained with overexpression of CD109S instead of CD109 (data not shown). No labeled proteins were observed when precipitations were performed with control IgG (lanes 5 and 6), indicating that the immunoprecipitations were specific. Western blotting of cell lysates with an anti-CD109 Ab (TEA 2/16) demonstrates that CD109 levels are moderately higher in CD109 transfected cells as compared to the EV transfectants (lane 8 vs. 7). Lower panel shows Ponceau staining of a 75 kDa band, demonstrating that equal amounts of protein were loaded in each lane. No appreciable difference in ligand binding to the type I and II receptors was observed with overexpression of CD109 under the conditions used (lane 1 vs. 2).
|
In contrast to overexpression studies, when CD109 expression was blocked in HaCaT cells using CD109-specific antisense oligonucleotides (as-oligo), the amount of [125I]TGF-ß1 bound to r150 was decreased in membrane extracts immunoprecipitated with 1B3, as compared with control inverse oligonucleotide (inv-oligo) treated cells (Fig. 2B
, lane 2 vs. 1). Densitometric analysis of the data indicated a 40% decrease in [125I]TGF-ß1 binding to r150. No labeled proteins were observed with control IgG (lanes 3 and 4). Western blotting of cell lysates with an anti-CD109 Ab (TEA 2/16) demonstrated that CD109 levels are reduced in cells treated with CD109-specific as-oligo as compared to inv-oligo treated control cells (lane 6 vs. 5). The lower panel shows Ponceau staining of a 75 kDa band demonstrating that equal amounts of protein were loaded in each lane. Together, these results support the conclusion that r150 corresponds to CD109. No appreciable difference in ligand binding to the type I and II receptors was observed with blocking CD109 expression levels under the conditions used (data not shown).
CD109 forms a heteromeric complex with the TGF-ß signaling receptors in the presence of TGF-ß1 ligand
We have previously shown that r150 associates with the TGF-ß signaling receptors in the presence of TGF-ß (17
, 22)
. To confirm that CD109 displays this property, we performed affinity labeling/immunoprecipitation studies (Fig. 3
A). HaCaT cells affinity labeled with [125I]TGF-ß1 (not immunoprecipitated, NIP) showed the characteristic TGF-ß receptor profile reported previously (17
, 22)
, demonstrating the presence of types I (RI), II (RII), and III (RIII/betaglycan) TGF-ß receptors and r150 (lane 1). The band seen between RI and RII may represent RI cross-linked to a TGF-ß1 dimer in which the two monomers of TGF-ß1 became crosslinked inadvertently. The anti-CD109 Ab, 1B3, immunoprecipitated endogenous r150, and coimmunoprecipitated small amounts of RI and RII, demonstrating that CD109 associates with the TGF-ß signaling receptors in the presence of TGF-ß (lane 2). Possible explanations for the relatively low amounts of RI and RII coimmunoprecipitated as compared with the amounts present in the NIP lane may include: 1) not all of RI and RII being associated with CD109, and 2) masking of the CD109 epitope because of its complex formation with RI/RII, hindering Ab recognition. The identity of the 45 kDa band observed is not known but may represent soluble RI (38)
or a proteolytic product of betaglycan (39)
. Similar results were obtained using another anti-CD109 Ab, 8A3 (data not shown). These results confirm that CD109 binds TGF-ß1 and forms a heteromeric complex with the TGF-ß signaling receptors in the presence of ligand, as previously demonstrated for r150 (17
, 22)
. The association of CD109 with the signaling receptors is also supported by the observation that the anti-RI Ab (lane 3) and anti-RII Ab (data not shown) (17)
, coimmunoprecipitates CD109, as previously shown for r150 (17
, 22)
. The coimmunoprecipitation of RII with the anti-RI Ab (lane 3), reflects their heteromeric complex formation on the cell surface (40
, 41)
. No labeled proteins were immunoprecipitated with the control IgG (lane 4), indicating that the immunoprecipitations were specific.
|
CD109 forms a heteromeric complex with the TGF-ß signaling receptors and betaglycan in the absence of TGF-ß1 ligand
Whether r150 (or CD109) associates with the TGF-ß signaling receptors in the absence of TGF-ß has not been examined previously. This was determined using immunoprecipitation in tandem with Western blot analysis (Fig. 3B
). HaCaT cells were washed with mild acid to remove endogenous TGF-ß (35)
, and membrane extracts not immunoprecipitated or immunoprecipitated with various antibodies were immunoblotted with an anti-CD109 Ab, TEA2/16. The anti-CD109 Ab detected the CD109 protein at 150 kDa in the nonimmunoprecipitated lysate (NIP, lane 1). Appreciable amounts of CD109 were coimmunoprecipitated with the anti-RI (lane 2), and detectable amounts with anti-RII (lane 3) and anti-RIII (lane 4) antibodies, while control IgG (lane 5) failed to immunoprecipitate any CD109. These results demonstrate that CD109 associates with the TGF-ß signaling receptors and betaglycan in a ligand-independent fashion, and at physiological receptor concentrations. The IgG fraction shown in the bottom panel demonstrates that equal amounts of immune complexes were loaded.
CD109 interacts directly with the type I TGF-ß (RI) receptor in vitro
The above results indicate that the anti-RI Ab coimmunoprecipitates substantial amounts of CD109 protein (Fig. 3B
, lane 2 vs. lane 3 and 4) raising the possibility that RI and CD109 might interact directly. To explore this possibility, we in vitro translated CD109 and examined its ability to bind recombinant GST-RI, as compared to GST, in vitro. Figure 3C
(upper panel) demonstrates that in vitro translated [35S]CD109 is specifically retained by GST-RI, but not GST, immobilized to GSH-sepharose beads. Coomassie blue staining of the gel confirms that approximately equal amounts of GST and GST-RI were used in the assay. These data suggest that CD109 interacts directly with the type I TGF-ß receptor.
CD109 inhibits TGF-ß1-induced phosphorylation of Smad2 and Smad3
We then determined the effect of gain and loss of function of CD109 on TGF-ß signaling by examining TGF-ß1-induced phosphorylation of Smad2 and Smad3. Figure 4
A (upper panel) shows that overexpression of CD109 in HaCaT cells led to a marked reduction in TGF-ß-induced Smad2 phosphorylation as compared with empty vector transfected controls, at all concentrations of TGF-ß1 tested. Total Smad2 levels were not altered under these conditions (Fig. 4A
, middle panel). The relative overexpression of CD109 in the transfected cells is also shown (Fig. 4A
, bottom panel). Similar results were obtained with the overexpression of CD109S instead of CD109 (data not shown). In 293 cells, overexpression of CD109 led to a decrease in TGF-ß1-induced phosphorylation of Smad3 compared with the empty vector controls (Fig. 4B
, upper panel). Phosphorylated Smad3 protein is detected as a double band, which is in agreement with previous studies using this Ab (42)
. Reprobing the membrane with an anti-Smad3 Ab (FL-425 from Santa Cruz Biotechnology, which detects both Smad2 and Smad3) indicates that neither total Smad2 nor Smad3 levels were altered under these conditions (Fig. 4B
, middle panel). The relative levels of CD109 protein expression are also shown (Fig. 4B
, bottom panel).
|
In complementary studies, blocking the expression of CD109 with antisense oligonucleotides was examined in HaCaT cells. Figure 4C
(top panel) shows that TGF-ß-induced phosphorylation of Smad2 is enhanced in antisense oligo (as-oligo)-treated cells as compared to the inverse oligo (inv-oligo)-treated controls. The levels of total Smad2 remained unchanged under these conditions (Fig 4C
, middle panel). The relative decrease in CD109 levels in the as-oligo treated compared with the inv-oligo treated cells is also shown (Fig. 4C
, bottom panel). Together, these results support the notion that CD109 is a negative modulator of TGF-ß signaling in keratinocytes and other cell types.
CD109 inhibits TGF-ß1-induced transcriptional activity
Next we explored the effect of CD109 on TGF-ß1-induced transcriptional activity by using two different TGF-ß-responsive luciferase reporter constructs, 3TP-lux, which contains both PAI-1 and collagenase promoter elements (40)
, and (CAGA)12-lux, which is a Smad3-specific synthetic construct containing 12 repeats of CAGA (34
,43)
. Overexpression of CD109 in HaCaT cells leads to a significant (P<0.05) reduction in TGF-ß1-induced 3TP-lux activity when compared to empty vector (EV) transfectants (Fig. 5
A). A significant decrease (P<0.05) in the basal level of 3TP-lux activity detected in the absence of exogenous TGF-ß could be due to the autocrine effect of TGF-ß. These results confirm that CD109 is a negative modulator of TGF-ß signaling in keratinocytes. Furthermore, overexpression of CD109 in MEFs resulted in a decrease in basal and TGF-ß1-induced 3TP-lux activity (Fig. 5A
), suggesting that CD109 inhibits TGF-ß responses in cell types other than keratinocytes.
|
We then determined the role of CD109 on Smad3-specific signaling in keratinocytes using the (CAGA)12-lux reporter construct (34
, 43)
. Figure 5B
demonstrates that TGF-ß1 causes a dose-dependent increase in (CAGA)12-lux activity in HaCaT cells, with an 80-fold increase at 50 pM TGF-ß1. Importantly, overexpression of CD109 in HaCaT cells results in a significant reduction in TGF-ß1-induced stimulation of (CAGA)12-lux activity at all of the TGF-ß1 concentrations tested (P<0.05) when compared with empty vector transfected cells (Fig. 5B
). Similar results were obtained with overexpression of CD109S (data not shown). These results show that CD109 inhibits TGF-ß1-induced Smad3-driven signaling. The high concentration of activation with the (CAGA)12-lux reporter relative to 3TP-lux reporter detected in the present study is consistent with those reported previously by others (34
, 43
, 44)
and may be ascribed to the higher sensitivity due to multiple CAGA elements in the (CAGA)12-lux construct.
CD109 inhibits TGF-ß1-induced fibronectin and PAI-1 production
Since TGF-ß1 is a key regulator of extracellular matrix synthesis and breakdown, we next examined whether CD109 regulates TGF-ß1-induced plasminogen activator inhibitor-1 (PAI-1) or fibronectin expression. Figure 5C
demonstrates that empty vector (EV) transfected HaCaT cells show a robust dose-dependent increase in PAI-1 and fibronectin protein expression in response to TGF-ß1 treatment (upper and middle panels, respectively). Importantly, overexpression of CD109 led to an approximate 2-fold and 7-fold decrease in PAI-1 levels at 5 pM and 50 pM TGF-ß1, respectively (Fig. 5C
, upper panel), and 2- and 3-fold decreases in fibronectin levels at 5 pM and 50 pM TGF-ß1, respectively (Fig. 5C
, middle panel) as compared to empty vector (EV) transfectants (Fig. 5C
). Reprobing the membrane with an anti-actin Ab demonstrates that equivalent amounts of protein were loaded in each lane (Fig. 5C
, bottom panel).
We have previously shown that HaCaT cells deficient in GPI-anchor mutant synthesis, and hence functional CD109 protein, display enhanced responses to TGF-ß1 (21)
. Therefore, we examined whether these cells exhibit enhanced TGF-ß1-induced PAI-1 production. As indicated in Fig. 5D
, GPI M cells show a marked increase in PAI-1 expression in response to 50 pM TGF-ß1 compared with parental HaCaT cells (upper panel). Reprobing the membrane with an anti-actin Ab demonstrates that equivalent amounts of protein were loaded in each lane (Fig. 5D
, lower panel). Taken together, these data suggest that CD109 is a negative modulator of TGF-ß1-induced PAI-1 and fibronectin protein expression in human keratinocytes.
CD109 inhibits TGF-ß1-induced antiproliferative effect
Since TGF-ß1 has a potent growth inhibitory effect on most epithelial cell types, including keratinocytes (45)
, we examined whether CD109 could modulate the antiproliferative effect of TGF-ß1 on keratinocytes. HaCaT cells stably transfected with CD109S show a diminished (P < 0.05) TGF-ß1-induced inhibition of thymidine incorporation as compared with empty vector transfectants (Fig. 6
A). On the other hand, blocking the expression of CD109 with antisense oligonucleotides results in an enhanced (P<0.05) TGF-ß1-induced inhibition of thymidine incorporation when compared to cells treated with control inverse oligonucleotides (Fig. 6A
). Furthermore, HaCaT cells defective in GPI anchor biosynthesis display more potent TGF-ß1-induced growth inhibition as detected by markedly reduced (P<0.05) thymidine incorporation when compared to parental HaCaT cells. Together, these results indicate that CD109 negatively regulates the antiproliferative effect of TGF-ß1 on human keratinocytes.
|
CD109 inhibits TGF-ß1-induced wound closure in vitro
TGF-ß is an enhancer of keratinocyte migration which is of critical importance for re-epithelialization during wound healing (46)
. We examined the effect of CD109 on TGF-ß1-induced wound closure in monolayers of HaCaT cells stably transfected with CD109S or empty vector, by measuring wound size at 24 h postwounding and expressing it as a percentage of its size at time zero. HaCaT cells stably overexpressing CD109S exhibit diminished (P<0.05) TGF-ß1-induced wound closure at 24 h postwounding when compared with empty vector (EV) transfected cells (Fig. 6B
). A significant decrease (P<0.05) in wound closure at 24 h postwounding is also observed in CD109 overexpressing cells compared to EV transfectants, in the absence of treatment with exogenous TGF-ß.
Inhibition of TGF-ß signaling by CD109 does not require ligand sequestration
Our results thus far showed that CD109 associates with the TGF-ß signaling receptors in the presence or absence of TGF-ß1 ligand, raising the possibility that the mechanism by which CD109 inhibits TGF-ß responses may involve cell surface interactions between CD109 and the signaling receptors. However, since our previous studies and the present study demonstrate that CD109 binds TGF-ß1, it was of interest to test whether CD109 exerts its effect by sequestering TGF-ß1 ligand away from the signaling receptors. Figure 7
A shows that increasing the concentration of CD109 on the cell surface increases TGF-ß1 binding to CD109, but does not alter TGF-ß1 binding to the type I and II receptors, suggesting that CD109 does not sequester ligand away from the receptors. This is consistent with the results shown in HaCaT cells (Fig. 2A
). Furthermore, results shown in Fig. 7B
demonstrate that CD109 inhibits transcriptional activity induced by a constitutively active type I receptor (T204D) in the absence of ligand as compared to empty vector transfectants (P<0.05). The contribution of autocrine TGF-ß to transcriptional activation in 293 cells is negligible as demonstrated by virtually identical results obtained in the presence of a TGF-ß neutralizing Ab (1D11, 10 µg/ml). These results confirm that CD109 exerts its inhibitory effect on TGF-ß signaling independently of ligand sequestration.
|
CD109 expression is down-regulated in human malignant melanoma
Since our results show that CD109 is an inhibitor of TGF-ß signaling and responses in skin cells, we examined whether its expression is altered in malignant melanoma, a form of skin cancer with high metastatic potential, using a cancer profiling array. As shown in Fig. 8
A, CD109 expression is decreased in 5 out of 7 patients with malignant melanoma as compared to the corresponding adjacent normal skin tissue. The negative control RNA and DNA did not yield CD109-specific signals, whereas a CD109-specific signal was detected in the human genomic DNA positive control (Fig. 8B
).
|
DISCUSSION
The multifunctional nature of TGF-ß implicates a requirement for strict regulation of its cellular signaling. This regulation is achieved at multiple levels of the signaling pathway and may include modulation of signaling receptor activity by accessory receptors or other cell surface TGF-ß-binding proteins. Several groups including ours (16
17
18
19
20
21
22)
have reported the occurrence of cell surface GPI-anchored proteins that bind TGF-ß in an isoform-specific manner. However, the identities and functional significance of these proteins remain unknown. We have previously reported that human keratinocytes defective in GPI-anchor synthesis show enhanced TGF-ß responses, suggesting that GPI-anchored proteins regulate TGF-ß signaling in those cells (21)
. Our results suggested that a specific cell surface GPI-anchored TGF-ß1 binding protein of 150 kDa, r150, may play a role in this regard. Here we report the molecular cloning of r150 and show that it represents CD109, a novel member of the 2M/complement superfamily (28
, 29)
. Our results demonstrating that CD109 is a component of the TGF-ß receptor system in keratinocytes and a negative modulator of TGF-ß responses link CD109 function to regulation of TGF-ß signaling. Furthermore, we show that CD109 interacts directly with the type I TGF-ß receptor and regulates TGF-ß signaling independently of ligand binding. The present study represents the first report on the molecular characterization of a GPI-anchored TGF-ß binding protein.
TGF-ß accessory receptors that have been shown to regulate TGF-ß signaling include betaglycan and endoglin (10)
. Results from the present study together with our previous findings (17
, 21
, 22)
show that CD109 exhibits several properties that are similar to those of endoglin and betaglycan. Like these two proteins, CD109 can bind TGF-ß with high affinity, form a heteromeric complex with the TGF-ß signaling receptors on the cell surface, and regulate TGF-ß responses, indicating that it is a component of the TGF-ß receptor system. However, unlike betaglycan, which binds all three TGF-ß isoforms with similar affinity (11
, 12)
, CD109 binds ligand in an isoform specific manner with high affinity for TGF-ß1, a moderate affinity for TGF-ß3, and virtually no affinity for TGF-ß2 (22)
. Also, in contrast to endoglin, which binds TGF-ß only in the presence of the type II receptor (14)
, CD109 can bind the ligand in the absence of receptors or an intact membrane [(22)
; Present Study, Fig. 1A
]. Furthermore, while betaglycan enhances TGF-ß signaling (11
, 14)
, endoglin (14
, 47)
and CD109 negatively modulate TGF-ß signaling. The distinct properties of CD109 suggest that it may play a unique role in the regulation of isoform-specific TGF-ß signaling in keratinocytes.
Recent molecular cloning of human CD109 from KG1a cells (29)
and U373 cells (28)
has shown that CD109 comprises an open reading frame of 4338 nucleotides encoding a protein of 1445 amino acids. Available information on CD109 sequences reveals at least five amino acid substitutions (Supplemental Information, Figs. 9 and 10), the most notable being the Tyrosine703Serine (Y703S) polymorphism, which defines the Gov alloantigen on platelets and is implicated in refractoriness to platelet transfusion, neonatal alloimmune thrombocytopenia and posttransfusion purpura (27
, 48)
. Whether these amino acid substitutions alter CD109 conformation or function (interaction with TGF-ß or TGF-ß receptors) warrants further study. In the present study, we have identified a novel CD109 isoform, designated as CD109S, (deposited in GenBank under accession number AY788891), that is expressed in human placenta and contains a 51 bp deletion (Fig. 1B
; Supplemental Figs. 9 and 10). Interestingly, the 51 bp sequence (position 3652–3702) in CD109 is flanked by GU(GT) and AG dinucleotides, which represent consensus sites for RNA splicing in higher eukaryotes (49)
. Our results suggest that the placenta expresses both the short (CD109S) and long (CD109) forms, but that HaCaT cells express only the CD109 form, raising the interesting possibility that alternative splicing of CD109 mRNA may occur in a tissue-specific manner. We observed no detectable difference between the two isoforms in any of the parameters of TGF-ß signaling and responses measured. Thus, the functional significance of the 17 amino acid deletion in CD109S, if any, remains to be determined.
One of the defining characteristics of CD109 and other members of the 2M/complement superfamily is the presence of an internal thioester bond (37)
. During heat denaturation, these proteins may undergo nucleophilic attack on thioester resulting in autolytic cleavage of the protein. This autolytic cleavage can be prevented if the thioester bond is first disrupted with small nucleophiles such as methylamine (37)
. Our data showing that methylamine prevents the heat-induced cleavage of endogenous r150 provide evidence for the presence of a thioester bond in r150 and its identity as CD109. Furthermore, our results demonstrate that the 150 kDa form of r150 is derived from a 180 kDa form by thioester-mediated autolytic cleavage. Although it has been argued that the 150 kDa and 180 kDa forms of CD109 represent alternative splice variants (28)
, our data are consistent with the report of Lin et al. (29)
, supporting the thesis that the 150 kDa form is generated from the 180 kDa form by thioester-mediated autolytic cleavage. The functional significance of the thioester sequence and other domains of CD109 in TGF-ß binding warrants further studies.
Our results showing that CD109 binds TGF-ß1 and forms a heteromeric complex with the TGF-ß signaling receptors in the presence of ligand are consistent with our previous work on r150 (22)
. Our data further suggest that the interaction of CD109 with the signaling receptors occurs at endogenous receptor concentrations, since the experiments were performed using untransfected cells. In addition, we show for the first time that CD109 associates with the signaling receptors on the cell surface in the absence of ligand and that CD109 interacts directly with the type I TGF-ß receptor. These results suggest that CD109 and the signaling receptors can exist as a preexisting heteromeric complex and that ligand binding to CD109 might occur when CD109 is already a component of this complex. This is as opposed to the alternative situation where TGF-ß1 would bind CD109 and induce its association with signaling receptors. Whether the regulation of TGF-ß signaling by CD109 is dependent on its mode (preformed vs. ligand-induced) of heterooligomerization remains to be determined. Interestingly, differential signaling by preformed heteromeric receptor complexes (with ligand binding afterward) vs. ligand induced heteromeric receptor complexes has been demonstrated for BMP receptors (50)
.
Our data from gain and loss of function studies show that CD109 is an inhibitor of TGF-ß1-induced Smad2/3 signaling and transcriptional responses, substantiating our previous results obtained with keratinocytes defective in GPI anchor synthesis (21)
. In the present study, we further demonstrate that CD109 decreases TGF-ß1-induced PAI-1 and fibronectin production, cell growth inhibition, and in vitro wound closure of keratinocytes. Interestingly, data in the current report also suggest that CD109 may inhibit autocrine TGF-ß action since CD109 overexpression inhibits basal transcriptional activity in HaCaT cells and MEFs. Collectively, these data indicate that CD109 is a negative modulator of TGF-ß signaling and responses in keratinocytes. Furthermore, since similar results were obtained using MEFs and 293 cells (data not shown), CD109 may also negatively modulate TGF-ß signaling in other cell types.
Although CD109 has been shown to be an activation antigen for T cells and platelets (24)
, to carry the platelet Gov alloantigen system (27)
, and to be expressed widely in human tissues (29)
, the physiological function of this protein is not known. Results from the present study linking CD109 function to regulation of TGF-ß1 signaling in vitro suggest that it may play a similar role in vivo. Our results showing that CD109 expression is down-regulated in 5 out of 7 malignant melanomas suggest that CD109 may be involved in the initiation or progression of this cancer. Recently, CD109 expression has been shown to be high in other human tumors, including lung and esophageal squamous cell carcinomas, but low in other types of lung carcinomas such as adenocarcinoma and small and large cell carcinoma (26)
. Thus, CD109 might be a molecular marker for certain types of cancers. In addition, since CD109 is a negative modulator of TGF-ß signaling and responses in vitro, it is possible that gain or loss of CD109 may modulate the tumor suppressive or prometastatic action of TGF-ß (45)
during different stages of cancer.
Potential mechanisms by which CD109 may inhibit TGF-ß signaling include reducing ligand availability to the signaling receptors and/or modulating signaling receptor activity. Our data are consistent with a mechanism whereby CD109 modulates TGF-ß1 activity independently of ligand sequestration. This conclusion is based on the observations that: 1) increasing or decreasing the levels of CD109 expression on the cell surface is associated with corresponding changes in TGF-ß1 binding to CD109, but does not result in alterations in TGF-ß1 binding to the signaling receptors; and 2) CD109 can inhibit transcriptional activity induced by a constitutively active form of type I TGF-ß receptor in the absence of TGF-ß ligand. Furthermore, our results show that CD109 interacts with the signaling receptors both in the presence or absence of TGF-ß1 ligand, supporting the notion that CD109 exerts its effect on TGF-ß1 signaling by directly modulating receptor activity. Although our previous work has shown that CD109 can be released from the cell surface, and that the released (soluble) CD109 can bind TGF-ß1 and reduce TGF-ß1 binding to signaling receptors, the release occurs only in small amounts in the absence of exogenously added PIPLC (22)
. The relative contribution of ligand sequestration by soluble CD109, if any, remains to be determined.
In conclusion, we show that r150, a 150 kDa GPI-anchored TGF-ß1 binding protein that we have previously reported on human keratinocytes, represents CD109. Our results define CD109 as a novel TGF-ß accessory receptor and suggest that it may play a unique role in the regulation of isoform-specific TGF-ß signaling in these cells. Furthermore, our data demonstrate that CD109 negatively regulates TGF-ß signaling and responses by a mechanism that does not involve ligand sequestration but instead may involve direct modulation of receptor activity. Delineation of the precise molecular mechanism(s) by which CD109 acts as a negative regulator of TGF-ß signaling will be a promising area for future investigation.
ACKNOWLEDGMENTS
We thank Ying Wang for excellent technical assistance. We also thank D.R. Sutherland, I. Bernstein, S. Souchelnytskyi, S. Huet, P. Boukamp, J. Massague, P. ten Dijke, A. Roberts, J.J. Lebrun, M. O’Connor-McCourt and Genzyme Corporation for reagents, and Harvard Microchemistry for microsequencing. This work was supported by a CIHR operating grant to A.P. (FRN13732).
FOOTNOTES
1 These authors contributed equally to this work. 
2 Present address: TargeGen Inc., 9380 Judicial Dr., San Diego 92121, CA, USA.
3 Present address: Dept. of Colorectal Surgery, Tianjin Medical University Cancer Hospital, Tianjin, P.R. China.
Received for publication October 31, 2005. Accepted for publication March 14, 2006.
REFERENCES
