Wnt/β-catenin signaling plays a critical role during development of both normal and malignant colorectal cancer tissues. Phosphorylation of β-catenin protein alters its trafficking and function. Such conventional allosteric regulation usually involves a highly specialized set of molecular interactions, which may specifically turn on a particular cell phenotype. This study identifies a novel transcription modulator with an FLYWCH/Zn-finger DNA-binding domain, called “FLYWCH1.” Using a modified yeast-2-hybrid based Ras-Recruitment system, it is demonstrated that FLYWCH1 directly binds to unphosphorylated (nuclear) β-catenin efficiently suppressing the transcriptional activity of Wnt/β-catenin signaling that cannot be rescued by TCF4. FLYWCH1 rearranges the transcriptional activity of β-catenin/TCF4 to selectively block the expression of specific downstream genes associated with colorectal cancer cell migration and morphology, including ZEB1, EPHA4, and E-cadherin. Accordingly, overexpression of FLYWCH1 reduces cell motility and increases cell attachment. The expression of FLYWCH1 negatively correlates with the expression level of ZEB1 and EPHA4 in normal versus primary and metastatic colorectal cancer tissues in patients. Thus, FLYWCH1 antagonizes β-catenin/TCF4 signaling during cell polarity/migration in colorectal cancer.
This study uncovers a new molecular mechanism by which FLYWCH1 with a possible tumor suppressive role represses β-catenin-induced ZEB1 and increases cadherin-mediated cell attachment preventing colorectal cancer metastasis.
This article is featured in Highlights of This Issue, p. 1817
Activated Wnt-signaling has significant implications both in cellular homeostasis and carcinogenesis depending on the context of activation and the surrounding microenvironment. Accumulated evidence identified several proteins that interact with nuclear-β-catenin and antagonize its transcriptional activity leading to transcriptional repression of Wnt-target genes. Accordingly, any alterations of the Wnt/β-catenin pathway through β-catenin corepressors is often involved in crucial biological processes ranging from normal development to cancer. For instance, Sox9 regulates the progression of the chondrocyte differentiation pathway during endochondral bone formation through inhibition of β-catenin (1), whereas the Wnt-signaling attenuator RUNX3 was proposed to function as a tumor suppressor during intestinal tumorigenesis (2). Moreover, expression of Chibby, a nuclear-β-catenin associated antagonist of the Wnt/Wingless pathway, was also linked to embryonic development and cancer (3). These findings indicate that β-catenin co-repressors function as potential gate-keepers for the activity of β-catenin, ensuring that a proper threshold of β-catenin is achieved before its interaction with members of T-cell factor (TCF) and lymphoid enhancer factor (LEF) transcription factors (TCF/LEF).
Furthermore, the paradox of β-catenin/TCF interaction, however, is challenged by identification of new strategies for β-catenin-dependent but TCF-independent gene regulation. It has been reported that β-catenin interacts with Prop-1 rather than members of TCFs to regulate the cell-lineage determination during pituitary gland development (4). β-Catenin was also found to control the pluripotency of embryonic-stem cells through interaction and enhancement of Oct-4 activity independently of TCFs (5). Likewise, the involvement of several other DNA-binding transcription factors with β-catenin was reported in the literature. Some are enhancing the promoter binding ability of β-catenin and facilitate the chromatin remodeling at the promoter site of Wnt-target genes such as p300/CBP and Brg-1 (6, 7), whereas others are promoting nuclear localization of β-catenin such as Pygopus (8). Corepressors such as ICAT (inhibitor of β-catenin and TCF4) and p15RS antagonize Wnt-signaling through inhibiting the interaction and formation of β-catenin/TCF4 complex (9). Moreover, RUNX3 forms a ternary complex with β-catenin/TCF4 by which attenuates Wnt-signaling activity (2). Similar to this mechanism, RAR, Chibby, Sox-9, and HIF1α compete with members of TCF/LEF for binding to β-catenin (1, 3, 10, 11).
Notably, some of the β-catenin-binding transcription factors including KLF4, Glis2/3, HIC1, and Osx belong to the Cys2-His2 (C2H2) family of Zinc-finger proteins characterized by having multiple C2H2-type Zinc-finger DNA-binding domains (12–15). These proteins could potentially regulate the gene expression programs controlled by Wnt/β-catenin-signaling via their interaction with β-catenin and/or TCFs, disrupting the formation of β-catenin/TCF-complexes, and subsequently recruiting nuclear-β-catenin to the promoter of specific target genes independently of TCFs. However, this view of nuclear-β-catenin recruitment has not yet been fully explored.
To further delineate the nuclear events of the Wnt-signaling pathway, we used a modified yeast two-hybrid Ras-recruitment system (RRS) using mouse embryonic-cDNA library and identified several new proteins that bind β-catenin in a GSK-3β phosphorylation-dependent and/or independent manner. One of the proteins was FLYWCH1, a conserved nuclear protein containing multiple FLYWCH-type zinc-finger domains with unknown function. FLYWCH motif has first been described and annotated based on the presence of FLYWCH consensus sequence (F/Y–Xn–L–Xn–F/Y–Xn–WXCX6-12CX17-22HXH; where X indicates any amino acid) in Drosophila (16, 17). In addition, FLYWCH motifs were also identified and studied in two more proteins of C. elegans: PEB-1 (18) and FLYWCH transcription factors; FLH-1, FLH-2, and FLH-3 (19). Based on Drosophila and C. elegans studies, FLYWCH motifs may function in protein–protein interactions and serve as DNA-binding domains (16–19). Accordingly, it can be predicted that human FLYWCH1 may have similar biological activities. However, no evidence for the presence of a transactivation-domain within the coding region of this protein was detected. To further characterize FLYWCH1, we set out to functionally describe its interaction with β-catenin and explore its importance and role(s) in colorectal cancer.
Materials and Methods
A total of 49 colorectal cancers tissues were arrayed on a tissue microarray (TMA) block in triplicates by the Molecular Pathology Unit from the consenting patients through the Nottingham tissue bank, as described previously (20).
Antibodies were purchased as follow: β-catenin (610154, BD-Transduction and 9582; Cell Signaling Technology), ZEB2/SIP1 (H260; Santa Cruz Biotechnology), Vimentin (RV202; Santa Cruz Biotechnology), ZEB1 (H-102 and E-20X; Santa Cruz Biotechnology), E-cadherin (BD Biosciences), TCF4 (Upstate; 05-511), FLYWCH1 (Sigma; HPA040753; Santa Cruz Biotechnology; V18), GFP (3E6; Invitrogen), Flag-tag (F1804; Sigma), Myc-tag (9E10; Millipore), and β-actin (Abcam).
The IMAGE-clone of human FLYWCH1-cDNA purchased from Geneservice (ID: 4839062/AK34; GenBank: 84256). Human FLYWCH1 gene is composed of 2,151 nucleotides encoding a protein with 716aa residues called FLYWCH1. The IMAGE-clone plasmid DNA then digested with several restriction-enzymes and cloned into different vectors, while fused with the MYC-tag, His6, and eGFP. The FLAG-β-cateninWT, FLAG-β-cateninS35A, FLAG-β-cateninS33A, S37A, T41A, S45A, TCF4, human c-jun-promoter-Luc, TOP/FOP-FLASH-Luc reporters previously described (21, 22). The −1129/+55 of ZEB1-promoter-Luc was from Dr. M. Saito (Tokyo, Japan), FLAG-GSK-3β-plasmid was from Prof B.W. O'Malley (Houston, TX), and the packaging plasmids were kindly provided by Dr. D. Bonnet (London, UK). pEGFP-C2 was purchased from Clontech, pBluescript-SK (Stratagene), pcDNA3.1 (Invitrogen), pET-His6, pGEX-KT (Addgene), the scrambled piLenti-scrambled control shRNA (scs)-GFP, and piLenti-FLYWCH1-set-shRNA-GFP (set of 4) from ABM Inc. (https://www.abmgood.com/FLYWCH1-shRNA-Lentivectors-i008000.html).
Modified yeast-2-hybrid based RRS
The RRS uses the growth defective yeast strain cdc25-2, which is deficient in Ras activity and cannot grow at 37°C (23). Briefly, the yeast cells stably transformed with pMET3-GSK-3β then cotransformed with a bait composed of the FLAG-tagged construct of β-catenin lacking the 7-8 armadillo-domains (ΔARM) and fused to oncogenic RasV12 (RasV12-FLAG-β-catenin/ΔARM). Furthermore, the expression of GSK-3β and substrate-phosphorylation dependency is dependent on the presence (promoter-off) and absence (promoter-on) of methionine in yeast. Thus, the activated-GSK-3β (under the control of a methionine-regulated MET3-promoter) induces phosphorylation of FLAG-β-catenin−ΔARM. These cells then transformed with mouse embryonic-cDNA library fused to the Src-myristylation signal (Stratagene). Glucose plates were grown for 5 days at 25°C then plated onto minimal galactose plates (+/−methionine) at a restrictive temperature 36 to 37°C. Yeast colonies exhibiting efficient methionine-dependent growth isolated and further analyzed.
Both sense and antisense FLYWCH1 RNA-probes generated by cloning a short nucleotide sequence of human FLYWCH1-cDNA into a pcDNA3 (details can be provided upon request). Both probes transcribed with T7-RNA polymerase using DIG-RNA-labeling mix (Roche) and ISH-assay carried out through the core facility at the CRUK-LRI as described previously (24).
For immunofluorescence (IF), colorectal cancer cell lines were grown on coverslips, fixed for 30 minute s with 4% paraformaldehyde and permeabilized with 0.5% Triton-X100 in PBS. Samples were exposed to goat anti-rabbit antibodies conjugated to Alexa Fluor 594 (A11037; Invitrogen) and/or rabbit anti-mouse antibodies conjugated to Alexa Fluor 488 (A11059; Invitrogen). Tetramethylrhodamine-B isothiocyanate (TRITC) conjugated Phalloidin (P1951; Sigma) used to label actin filaments according to manufacturer's instruction. Finally, slides mounted in Vectashield-DAPI medium (Vector Labs) and analyzed by fluorescent microscopy.
Transient and stable cell transfections
Human colorectal cancers (HCT116, DLD-1, and SW480), TIG119 fibroblasts, and HEK293T cells maintained in RPMI supplemented with 10% FBS. Cells transiently transfected using Lipofectamine-2000 (Invitrogen). After 48 hours, cells processed for transcriptional analysis, mRNA, or protein extraction. SW480 cells stably knocked-down for FLYWCH1 by lentiviral transduction of piLenti-FLYWCH1-set-shRNA-GFP and control piLenti-scsRNA- (scrambled control shRNA) GFP. Both cells selected in a Puromycin-containing medium for 2 weeks then the resistant cells trypsinized, counted, and seeded into 96-well plates (1 cell/well). Finally, the resistant colonies derived from a single-seeded cell further amplified and expanded. The knocked-down status of FLYWCH1 validated through RT-PCR and IF analysis. R-spondin1 (293T-HA-Rspo1-Fc) and Wnt-3a (293T-HA-L) expressing cell lines were a gift from Prof Hans Clevers (Hubrecht, the Netherlands). R-spondin1/Wnt-3a conditioned media validated by a TOP/FOP-flash luciferase-reporter assay before using it.
Western blot, immunoprecipitation, and chromatin-immunoprecipitation assays
We performed Western blots (WB) and immunoprecipitations (IP) as previously described (25, 26). For in vitro binding assays, cDNAs encoding FLAG-β-catenin, and GFP-FLYWCH1 were translated by a T7-in vitro-coupled transcription/translation system (Promega). λ-Phosphatase (BioLab) treatment was carried out with 200 units for 60 minutes at 30°C. Immunocomplexes were captured by rotating for 2 to 3 hours with protein-A, Protein-G (Sigma), or secondary precoated Dynal-magnetic beads. In some cases, conjugated Agarose beads were used. Immunoprecipitates washed four times with lysis buffer, and SDS-PAGE separated sample proteins then transferred onto PVDF-membranes (Amersham) following standard procedures. Chromatin-immunoprecipitation (ChIP) analysis performed as described previously (20, 21).
Recombinant human β-catenin protein 1 μg/mL (Abcam) was used to coat a 96-well ELISA plate overnight at 4°C. The plate was washed three times with PBS-T and then blocked 2 hours at room temperature. For the TCF4-detection/calibration curve; after three washes, TCF4 recombined protein (Autogen Bioclear) added to the plate (range: 0.0–500 ng) and incubated at 37°C for 6 hours. The plate was washed three times PBS-T and incubated with 2 μg/mL of TCF4-FITC (Autogen Bioclear) for 1 hour at room temperature. Following washing with PBS-T (three times) 50 μL PBS was added. The level of fluorescent was estimated in a fluorescence microplate reader green setting excitation 494 nm and emission 518 nm. For the FLYWCH1-detection/calibration curve; after 3× washes, FLWYCH1 recombined protein (Sigma) added to the plate (range: 0.0–500 ng). Plate incubated at 37°C for 6 hours, and after three washes with PBS-T, 2 μg/mL of FLYWCH1-HRP (Sigma) added to the wells and incubated 1 hour at room temperature. The plate washed with PBS-T (3×). Tetramethylbenzidine (TMB) reagent A and B mixed and 100 μL of mixed solution transferred to each well and incubated at 37°C for 20 minutes. Fifty microliters of TMB stop solution was added to each well and absorbance read at OD (450 nm) in the microplate reader. The experiment for secondary antibodies interaction have been carried out, and there was no significant difference observed. Next, based on the calibration curve, TCF-4 100 ng/mL and FLWYCH1 80 ng/mL and mixer of both proteins added to wells already coated with human β-catenin protein (each triplicate). The plate incubated at 37°C for 6 hours and after three washes with PBS-T incubated with 2 μg/mL of TCF4-FITC. The plate incubated for 1 hour at room temperature and washed with PBS-T and added 50 μL PBS and read in fluorescence microplate reader green setting excitation 494 nm and emission 518 nm.
Electrophoretic mobility shift assays
For studying DNA–protein interactions, we used the LightShift chemiluminescent electrophoretic mobility shift assays (EMSA) Kit (Thermo Scientific) that uses a non-isotopic method to detect DNA–protein interactions. Biotin end-labeled DNA contains 3× of Tcf-DNA-binding site incubated with SW480 nuclear protein extract in the absent and/or present of recombinant TCF4 and FLYWCH1 proteins as outlined above. This reaction is then subjected to gel electrophoresis on a native polyacrylamide gel. The biotin end-labeled DNA is detected using the streptavidin–horseradish peroxidase conjugate and the chemiluminescent substrate according to the manufacturer's protocol (Thermo Scientific).
A firefly reporter gene (TOP-FLASH) driven by three copies of Tcf-binding elements (ATCAAAG) upstream to TK-minimal-promoter (which is specifically regulated by Wnt/β-catenin signaling) was used. Luciferase-activity measured using the Dual-Luciferase system (Promega) as described (20). For statistical analysis, the ratio of Firefly-Luc activity was normalized to Renilla-Luc. Data were expressed as fold-induction from three independent experiments.
Cell spreading, migration, and wound healing assays
Colorectal cancer cells synchronized in G0-phase of the cell cycle by serum-free starvation for 18 hours before cell motility assays. Electric cell-substrate impedance-sensing (ECIS) system was used for cell attachment and spreading assays in a confluent layer, according to manufacturer's guidelines. The system was run for 24 hours after which the quantification data of cell attachment and spreading was collected. Cell migration and wound healing assays assay were performed as previously described (24). All assays were performed at least three times in triplicate.
RT-PCR and quantitative RT-PCR assays
We extracted total RNA from homogenized primary tumors and cultured cells using Trizol (Sigma) or RNeasy Purification Kit (Qiagen). Superscript-III first-strand synthesis (Invitrogen) and oligo(dT) primers were used to convert the RNA to cDNA according to the manufacturer's instructions. qPCR was accomplished with SYBR-green (Roche), and the data were analyzed using LightCycler-480 software (Roche) and primers below.
5′-AGGTGACAGCATTGCTTCTG (forward: β-actin)
5′-AGGGAGACCAAAGCCTTCAT (reverse: β-actin)
5′-GTCTGTAGGAAGGCACAGCCTGTCG (forward: CDH1)
5′-AGGACCAGGACTTTGACTTGAGC (reverse: CDH1)
5′-GCTGGAGGTCTGCGAGGA (forward: CCND1)
5′-CATCTTAGAGGCCACGAACA (reverse: CCND1)
5′-AGAAGGTGTGGGCAGAAGAA (forward: CD44)
5′-AAATGCACCATTTCCTGAGA (reverse: CD44)
5′-ACCCCAAGATCCTGAAACAG (forward: c-Jun)
5′-ATCAGGCGCTCCAGCTCG (reverse: c-Jun)
5′-AAAACCAGCAGCCTCCCGC (forward: c-Myc)
5′-GGCTGCAGCTCGCTCTGC (reverse: c-Myc)
5′-GGCAAGCATGAGACTGTGAA (forward: ENFB1)
5′-ACTCCAAGGTGGCATTGTTC (reverse: ENFB1)
5′-CTGCTGGATCAACCAGGAAT (forward: ENFB2)
5′-TCTAGCACAGACGGCAACAG (reverse: ENFB2)
5′-AGGATTACCCTGTGGTGGTC (forward: EPHB2)
5′-TACAACGCCACAGCCATAAA (reverse: EPHB2)
5′-TCGTGGTCATCGCTATCGTCT (forward: EPHB3)
5′-AAACTCCCGAACAGCCTCATT (reverse: EPHB3)
5′-CGACAAAGAGCGTTTCATCA (forward: EPHA4)
5′-GCTTCACCCAAGTGGACATT (reverse: EPHA4)
5′-GCAAAGGTCGAAGACCAGGA (forward: FLYWCH1)
5′-TTCCTGGTGTACGAGTCCTT (reverse: FLYWCH1)
5′-GGAGACCGACCAGGAGAC (forward: FSCN1)
5′-CATTGGACGCCCTCAGTG (reverse: FSCN1)
5′-GACAACAGCAGTATGGACG (forward: LGR5)
5′-GCATTACAAGTAAGTGCCAG (reverse: LGR5)
5′-GGCATACACCTACTCAACTACGG (forward: ZEB1)
5′-TGGGCGGTGTAGAATCAGAGTC (reverse: ZEB1)
5′-AATGCACAGAGTGTGGCAAGGC (forward: ZEB2)
5′-CTGCTGATGTGCGAACTGTAGG (reverse: ZEB2)
The significance of differences between mean and median was determined using the Student t test and the Mann–Whitney U test, as appropriate. Significance testing was performed using SPSS version 15. Mean ± SD (*, P < 0.05; **, P < 0.01; ***, P < 0.001) values are shown.
FLYWCH1 and its physical interaction with β-catenin
FLYWCH1 is an uncharacterized protein product of the human FLYWCH1 gene. The sequence of this gene is publicly available (http://www.ncbi.nlm.nih.gov/gene/84256), yet its function remained undefined. The first evidence of FLYWCH1/β-catenin interaction comes from our large-scale screening using a modified yeast-2-hybrid based RRS (Fig. 1A–C; ref. 23). As the bait, we used the FLAG-tagged construct of β-catenin lacking the seven to eight armadillo-domains (ΔARM; FLAG-β-cateninΔARM) to avoid any interaction with TCF/LEF and other known nuclear proteins (Fig. 1B). The expression of GSK-3β and substrate-phosphorylation dependency is dependent on the presence (promoter-off) and absence (promoter-on) of methionine in yeast (Fig. 1C). This RRS utilizes an embryonic cDNA library (pMyr-cDNA library). Thus, cdc25-2 mutant yeasts can grow at 37°C, when a phosphorylation-dependent and -independent interaction between a protein target and RasV12-β-cateninΔARM takes place. Therefore, to determine GSK-3β phosphorylation-independent interactions between human β-catenin and myristoylated target proteins, yeast colonies exhibiting efficient methionine-dependent growth isolated and further analyzed. The β-cateninΔARM(bait)-dependent growth of these clones was further analyzed on galactose-containing medium at 37°C. This screening allowed us to identify several proteins that specifically bound to unphosphorylated-β-catenin (Supplementary Table S1). The FLYWCH1 protein contains a tandem array of five FLYWCH-type Zinc-finger motifs (Fig. 1D and E) and a putative nuclear localization signal (NLS) motif (KRAK; Supplementary Fig. S1A) closely resembling the classical NLS motif consensus sequences (K-R/K-X-K/R). We tested multiple antibodies from various commercial sources, unfortunately despite a lot of efforts no commercial antibody can detect the endogenous FLYWCH1 protein in WB and IP, while working on IF assays. Therefore, the full-length cDNA of human FLYWCH1 tagged with MYC-epitope (MYC-FLYWCH1) or eGFP (GFP-FLYWCH1) were used for biochemical/molecular analyses (WB, IP, Co-IP, etc.) in this study.
To confirm our initial screening data regarding FLYWCH1/β-catenin interaction, we first carried out a coupled in vitro transcription/translation (IVT) assay of both FLYWCH1 and β-catenin in the presence and absence of constitutively active GSK-3β. Treatment of the IVT product with λ-phosphatase resulted in a single unphosphorylated, faster-migrating form of FLAG-β-cateninWT resembling FLAG-β-cateninS33A (Supplementary Fig. S1B, bottom, lanes 2 and 3). Interestingly, our IP assay (using α-GFP antibody) confirmed the interaction of GFP-FLYWCH1 with both FLAG-β-cateninWT (treated with λ-phosphatase) and FLAG-β-cateninS33A (Supplementary Fig. S1B, top, lanes 2 and 3).
Next, to further address the association of unphosphorylated β-catenin with FLYWCH1, we expressed and purified GST-β-cateninWT, GST-β-catenin4A, and His-FLYWCH1 proteins from bacteria for an in vitro binding assay. Co-IP with His-antibody and WB with β-catenin-antibody revealed that FLYWCH1 could directly bind the unphosphorylated β-catenin4A (a mutant clone that lacks the phosphorylation sites S33, S37, and S45, which needed to prime β-catenin for subsequent phosphorylation at S41, S33, and S37 by GSK-3; Fig. 1F). We also confirmed the interaction of GFP-FLYWCH1 with ectopically expressed β-catenin (FLAG-β-cateninS33A and FLAG-β-catenin4A) in HEK293 cells using both α-GFP and α-FLAG antibodies, respectively (Fig. 1G; Supplementary Fig. S1C). Further Co-IP analyses showed that β-catenin4A interacts with FLYWCH1 regardless of GSK status (Supplementary Fig. S1D, lane 3 vs. lane 4). A single mutation of β-catenin (β-cateninS33A) showed stronger interaction with FLYWCH1 than the wild-type (β-cateninWT; Supplementary Fig. S1D, lane 5 vs. lane 1), whereas the GSK3 inhibitor-BIO (5micromolar) enhanced the interaction of both β-cateninS33A and β-cateninWT with FLYWCH1 (Supplementary Fig. S1D, lanes 6 and 2). Overall, these data indicate that phosphorylation of any of these sites may dampen the FLYWCH1 interaction with β-catenin.
The β-catenin has three main domains, the N-terminus, C-terminus, and the armadillo repeats, to which most of its nuclear partners bind (27–29). Thus, we generated a series of β-catenin internal deletions to investigate which region is involved in the FLYWCH1/β-catenin interaction (Supplementary Fig. S2A). Two β-catenin mutants (MD3 and ΔC) were weakly expressed, presumably due to their degradation in the cytoplasmic β-catenin-destruction complexes (Supplementary Fig. S2A, lanes 7 and 8). These two deletions were therefore excluded from further Co-IP analyses. Co-IP analysis showed that the N-terminal deletion of β-catenin (ΔN) was sufficient to abort FLYWCH1/β-catenin interaction (Fig. 1H). The β-catenin protein has no classical nuclear localization or nuclear export signals (30). Therefore, β-catenin deletions may not affect the subcellular distribution of the mutant proteins. Intriguingly, the armadillo repeats region of β-catenin (represented by deletions MD1 and MD2 in Supplementary Fig. S2A) containing the TCF/LEF binding domain was not essential for this interaction (Fig. 1H). Furthermore, we generated several deletion mutant clones of FLYWCH1 to map the interaction of FLYWCH1 with β-catenin (Supplementary Figs. S2B–S2D). Co-IP revealed that β-catenin lost its interaction with the C-terminal deletion mutant of FLYWCH1 (ΔC350) in which the last two FLYWCH motifs are removed (Supplementary Fig. S2E). Full-length FLYWCH1 is expressed in the nucleus in a punctate format (Supplementary Fig. S3A) presumably due to the presence of an NLS within the C-terminus domain (Supplementary Fig. S1A). However, the protein product of GFP-FLYWCH1-ΔC350 is diffused into both the cytoplasm and nucleus (Supplementary Fig. S3A), and a significant fraction of its protein remained in the nucleus (Supplementary Fig. S3B). Therefore, the cellular distribution of FLYWCH1-ΔC350 protein may not result in a complete loss of FLYWCH1/β-catenin interaction. Next, we investigated the interaction of FLYWCH1 with endogenous β-catenin in SW480 cells, which contain a high level of nuclear β-catenin due to mutations in the APC gene. Our Co-IP assay showed a direct interaction between endogenous β-catenin and FLYWCH1, whereas the C-terminal deletion of FLYWCH1 (ΔC350) has lost the ability to interact (Fig. 1I). Taken together, these data suggest that the FLYWCH1-C-terminus domain is essential for efficient β-catenin interaction and complete nuclear localization of FLYWCH1.
FLYWCH1 represses the transcriptional activity of β-catenin by preventing the association of TCF4/β-catenin to the Tcf-DNA-binding elements
To elucidate the consequences of FLYWCH1/β-catenin interaction on the transcriptional activity of β-catenin/TCF4, we first performed a dual-luciferase TOP/FOP-FLASH reporter assay as described previously (20, 21). Intriguingly, FLYWCH1 effectively suppressed the reporter activity induced by β-catenin in different cell lines (Figs 2A; Supplementary Fig. S3C). The luciferase-activity of all deletion mutants of β-catenin except the N-terminal deletion (β-catenin−ΔN) was significantly inhibited by FLYWCH1 (Supplementary Fig. S3D, lane 3). Similarly, the C-terminus removal of FLYWCH1 (ΔC350) has lost its suppressive activity against β-catenin signaling (Fig. 2B, lane 6). Thus, in line with our Co-IP results, the luciferase data also confirmed that the β-catenin-N-terminus and the FLYWCH1-C-terminus are required for both their physical and functional interaction.
To explore the downstream effect of the functional interaction of FLYWCH1/β-catenin, we show that the c-jun-promoter-Luc (a putative β-catenin-target gene) was effectively suppressed by FLYWCH1 (Fig. 2C). This suppression, however, was impaired by FLYWCH1-siRNA, whereas the TCF4 and β-catenin protein levels were unaffected (Fig. 2C). Intriguingly, our Co-IP assays showed that the interaction between β-catenin and TCF4 is mitigated by the ectopic expression of FLYWCH1 (Fig. 2D). Moreover, the luciferase assay showed that overexpression of TCF4 only slightly rescued the suppression effect of FLYWCH1 on β-catenin (Fig. 2E) giving the notion that FLYWCH1 may compete with TCF4 for binding to β-catenin. To further confirm this notion, we performed a competitive binding assay of human β-catenin protein to determine if the β-catenin and TCF4 interaction values influenced by FLYWCH1 using competition and an ELISA based binding assays in vitro (Supplementary Fig. S4A–S4C). The result shows that FLYWCH1 protein, indeed, competes with TCF4 protein for binding to β-catenin.
FLYWCH1 may compete with TCF4 via its binding to the Tcf-DNA binding sites. To explore this possibility, we first used the LightShift Chemiluminescent EMSA assay using SW480 nuclear protein extract incubated with a biotin end-labeled DNA contains 3× of Tcf-DNA-binding site in the absent and/or present of recombinant TCF4 and FLYWCH1 proteins. Interestingly, the amount of DNA–TCF4–protein interaction is substantially reduced by FLYWCH1 protein (Supplementary Fig. S4D). Our biotinylated-oligonucleotide-mediated ChIP assay (31) also showed that the biotinylated oligos (containing 3× Tcf-DNA consensus sites, TOP) could not equally pull-down the HA-TCF4 from streptavidin/biotinylated beads in the presence of MYC-FLYWCH1 (Fig. 2F). Furthermore, our endogenous TCF4-ChIP assay in HCT116 cells (contains high endogenous level of TCF4) show that binding of the TCF4 to the Tcf-sites of the c-jun-promoter was reduced by both endogenous and exogenous levels of FLYWCH1 (Fig. 2G). These data strongly indicate the competition of FLYWCH1 with TCF4 to bind to the Tcf-DNA binding sites.
Next, we assessed the role of β-catenin in this competition. Our data show that knockdown of β-catenin (20) remarkably rescued the suppression effect of FLYWCH1 (Fig. 2H), indicating the necessity of β-catenin for the FLYWCH1-mediated transcriptional repression. To further support this notion, we performed a β-catenin-ChIP assay based on standard and qPCR analysis (Fig. 3A–B) to the Tcf-DNA-binding sites within c-jun promoter. β-Catenin-ChIP experiments were performed in normal fibroblasts (TIG119) that lack nuclear-β-catenin (Supplementary Fig. S4E; ref. 21) and colorectal cancer cells (DLD-1 and SW480) that contain nuclear β-catenin. Notably, in contrast to the TIG119 cells, immunofluorescent assay demonstrated that the β-catenin differentially expressed in the normal colonic epithelial cell line CCD-CoN-841, including some nuclear expression (Supplementary Fig. S4E). Expectedly, the ChIP assays showed an efficient binding of both β-catenin/TCF4 and β-catenin/FLYWCH1 (to a different extent) to the Tcf-DNA-binding elements of the c-jun-promoter in colorectal cancer cells, but not fibroblasts (Fig. 3A). These data further confirm the β-catenin dependency of the FLYWCH1-mediated transcriptional repression. Moreover, FLYWCH1 overexpression also reduced the interaction of β-catenin/TCF4 to the c-jun-Tcf-sites (Fig. 3B). We also show that the competition of FLYWCH1 and TCF4 to the Tcf-DNA-binding elements is not through the direct interaction of TCF4 and FLYWCH1 proteins as no FLYWCH1 was co-IPed with TCF4 in TIG119-fibroblasts that lack nuclear β-catenin (Supplementary Fig. S4F).
To elucidate the physiologic relevance of FLYWCH1/β-catenin interaction and activation of Wnt-signaling, we examined the FLYWCH1 expression under the influence of canonical Wnt3a-ligand and R-spondin-1 that bind to and activate Wnt-receptors (32, 33). The WB assays showed that Wnt3a/R-Spondin treatment remarkably increased the accumulation of nuclear β-catenin and Wnt-target genes such as cyclin-D1 and c-Jun (Fig. 3C), whereas downregulated the FLYWCH1 gene expression (Fig. 3D and E). Further analyses showed that overexpression of FLYWCH1 in the Wnt3a/R-Spondin-treated SW480 cells leads to decreased expression of both cyclin-D1 and cyclin-D2 (Fig. 3F, lane 3 vs. lane 4), indicating that FLYWCH1 acts as a negative transcriptional modulator for the Wnt-target genes.
Altogether, our findings suggest that FLYWCH1 represses the transcriptional activity of β-catenin by preventing the association of TCF4/β-catenin to the Tcf-DNA-binding elements via competing with TCF4. Our data also indicate that FLYWCH1 may act as a negative regulator of Wnt/β-catenin signaling pathway through transcriptional repression of Wnt-target genes.
FLYWCH1 modulates cell morphology
To explore the biological activity of FLYWCH1, we first examined FLYWCH1 transfected cells under the microscope and observed changes in cell shape and morphology. It is noteworthy to mention that stable overexpression of full-length FLYWCH1 has been attempted in SW480 and HCT116 cells using GFP-FLYWCH1 lentiviral vector (pLVX-GFP-FLYWCH1-PuroR) based on Puro-selection with no success (Supplementary Fig. S5A). GFP-FLYWCH1 showed different cellular localization pattern (Supplementary Fig. S5A) and different truncated forms of the protein (Supplementary Fig. S5B), indicating possible spontaneous mutation(s) of FLYWCH1 in these cell lines. To explore this possibility, the N-terminal coding region of the genome-integrated FLYWCH1 flanked by eGFP was amplified by PCR and sequenced. Expectedly, the sequencing analysis revealed a point mutation, resulted in a stop-codon at aa277 (stable ≠1), and a relatively large internal in-frame deletion (537nt) mapped to nucleotides 1405–1942 within the C-terminal domain (Δ468-648aa, stable ≠2) of FLYWCH1 (Supplementary Fig. S5C and S5D). These deletions were corresponding to the truncated protein bands detected on WB (Supplementary Fig. S5B). Indeed, consistent with our previous data, the stably expressed truncated form of FLYWCH1 (Δ468-648aa) lost its physical interaction and suppression activity against TCF4/β-catenin (Supplementary Fig. S5E–S5F). Data from the The catalogue of Somatic Mutations in Cancer (COSMIC) database revealed that somatic mutation in FLYWCH1 is rare in human cancers. Out of 32,801 primary human cancer samples, only 154 samples (a mutation rate of 0.47%) have somatic mutations. Interestingly, over 23% of these mutations found in the large intestine and 3.9% of these were frameshift deletions (Supplementary Fig. S6A). Therefore, colorectal cancer cells transiently expressing GFP-FLYWCH1 were used for biological analysis.
With the possible frameshift mutation and loss of action associated with FLYWCH1 overexpressing cells, we first tested the effect of FLYWCH1-knockdown on cell-cycle progression in SW480 and HCT116 cells stained with propidium iodide (PI) using flow cytometric analyses. We observed that FLYWCH1-knockdown increased cell number at G2–M transition in both SW40 and HCT116 cells (Supplementary Fig. S6B). Next, we tested the morphologic changes associated to transient overexpression of FLYWCH1. Our microscopic study for Phalloidin stained SW480 cells showed that untransfected cells displayed branched, flat, and elongated shape with prominent actin fibers. Although FLYWCH1-expressing cells showed a different phenotype, characterized by shorter, circular epithelial-like morphology with actin fibers concentrated at the edges of the cells (Fig 4A, middle horizontal panel). Consistent with this observation, FLYWCH1-knockdown cells exhibited irregular branched shapes with numerous elongated projections (filopodia) in all directions compared with controls (Fig. 4A, lower; Supplementary Fig. S7C, S7D and S7F vs. S7G), suggesting that FLYWCH1 may modulate cellular biological activities through changes in cell morphology.
Previous studies have showed that β-catenin protein complexes redistribute from the membrane to the nucleus upon Wnt, irradiation, TGFβ1, and Rac1 activation or fibrosis and the nuclear substructures formed were readily identifiable (34–38). To explore this, we tested TIG119 cells, a normal embryonic fibroblast lacking nuclear β-catenin (Supplementary Fig. S4E), when treated with Wnt3A and stained for anti-β-catenin (Supplementary Fig. S7I). Thus, activation of β-catenin signaling may show a nuclear substructures foci-like pattern similar to the subnuclear localization of FLYWCH1.
Changes in cell morphology through actin cytoskeleton reorganization and filopodia formation have been involved in cell proliferation, migration, and invasion (39, 40). Therefore, we investigated whether alteration of cell morphology caused by FLYWCH1 correlates to cell motility. Interestingly, our in vitro scratch assay showed that the wound-closure of FLYWCH1 expressing cells versus FLYWCH1-knockdown and control cells is significantly reduced in SW480 cells (Fig. 4B; Supplementary Fig. S7E), indicating negative effect of FLYWCH1 on cell migration. To further delineate the effect of FLYWCH1 on cell attachment and motility, we evaluated the attachment and spreading capacity of FACS-sorted SW480-expressing GFP, GFP-FLYWCH1, and GFP-FLYWCH1shRNA using electric cell-substrate impedance sensing (ECIS; ref. 41). The ECIS analyses showed that cells overexpressing GFP-FLYWCH1 established a significantly higher resistance value comparing to FLYWCH1shRNA and control cells (∼6,000 ohms vs. 2,000–∼3,000 ohm). But the capacitance decreases in cells overexpressing GFP-FLYWCH1 versus FLYWCH1-knockdown and control cells (∼1.5F vs. ∼2.5–∼4 μF; Fig. 4C, right). These data indicate migration/invasion defects caused by FLYWCH1 possibly due to increased cell–cell attachment. Indeed, the ECIS analyses showed that cells-expressing GFP-FLYWCH1 were quickly and tightly attached within 1 to 2 hours period compared with cells expressing GFP (10–12 hours), whereas cells expressing FLYWCH1shRNA were delayed for over 24 hours (Fig. 4C, left). Also, cells expressing GFP-FLYWCH1 have restricted the current flow by slow spreading over the electrode leading linearly to decreased capacitance. This indicates increased cell attachment of GFP-FLYWCH1 overexpressing cells. To investigate if such activity may also modulate metastatic colorectal cancer cells, we tested the effects of FLYWCH1 on cell morphology and migration using a metastatic colorectal cancer cell line, SW620 (Supplementary Fig. S8A–S8C). These data indicate that FLYWCH1 inhibits cell migration and invasion in SW620 cells. Thus, these results suggest that FLYWCH1 may negatively regulate cell migration through alterations in cell–cell adhesion.
Molecular mechanisms of FLYWCH1-regulated cell migration
To explore the mechanisms underlying the effect of FLYWCH1 on cell migration, we first used a human WNT-signaling pathway cDNA-array to determine the expression profiles of 84 genes related to WNT-mediated signal transduction according to the manufacturer's instructions; http://www.sabiosciences.com/rt_pcr_product/HTML/PAHS-243Z.html (Supplementary Table S2). These data further highlighted differences in transcript expression between FACS-sorted cells overexpressing GFP-FLYWCH1 and GFP proteins. Notably, many of these changes associated with cell polarity and invasion (e.g., MMP7, DVL, FZD, JUN, and several WNT-target genes; Fig. 5A; Supplementary Table S2). Interestingly, our further qRT-PCR analysis showed that expression of a limited subset of genes (CD44, LGR5, c-JUN, EPHA4, CCND1, EFNB2, and FSCN1) were suppressed by the ectopic expression of FLYWCH1, whereas expression of others (c-MYC, EFNB1, EPHB2, and EPHB3) were unaffected in both SW480 and HCT116 cells (Fig. 5B, left, blue columns). Strikingly, the transcriptional repression of CD44, LGR5, c-JUN, EPHA4, CCND1, EFNB2, FSCN1, ZEB1, and c-MYC was significantly reversed by the FLYWCH1-knockdown, whereas the expression of others such as EFNB1 and EPHB3 were unchanged (Fig. 5B, left, red columns). These data suggest selective repression of Wnt-target genes via FLYWCH1/β-catenin interaction. Moreover, our WB analysis showed a sufficient correlation between transcript and protein levels of some of these target genes in FLYWCH1-knockdown cells but less correlation in FLYWCH1-overexpressing cells (Fig. 5B, right). Not surprising that mRNA levels cannot always correspond to protein levels due to posttranslational modifications (42). Furthermore, in a normal embryonic fibroblast lacking nuclear β-catenin and stably expressing the MYC-FLYWCH1, the cyclin-D1 expression is downregulated whereas FSCN1, SFRP1, c-MYC, and the c-JUN mRNA levels remained unchanged (Supplementary Fig. S8D and S8E) further supporting the notion of selective negative regulation of Wnt-target genes by FLYWCH1/β-catenin interaction. However, effects of FLYWCH1 on cyclin-D1 gene expression via β-catenin-independent TCF/LEF activity cannot be excluded.
Notably, the expression of the cell–cell adhesion receptor, E-cadherin (CDH1) was upregulated by FLYWCH1 overexpression in contrast to FLYWCH1-knockdown. Some of the prementioned responded genes such as ZEB1, CDH1, and EPHA4 are well known for their involvement in epithelial–mesenchymal transition (EMT) and cell migration. Thus, next, we asked if FLYWCH1 could regulate cell migration via EMT/MET process. Our WB showed no significant differences in the levels of a mesenchymal marker, Vimentin, or an EMT transcription factor, ZEB2 (Fig. 5C). However, significant differences between E-cadherin and ZEB1 mRNA and protein expression were detected in response to the presence and/or absence of FLYWCH1 (see above and Fig. 5B and C). Moreover, the level of Snail protein is accumulated in FLYWCH-knocked down cells (similar to ZEB1), whereas the level of Twist protein is downregulated (Supplementary Fig. S9A). Overall, these data suggest a synergistic action of these EMT-activating transcription factors may ultimately lead to a partial or steady state level of EMT process transition in colorectal cancer cells with altered FLYWCH1 expression. However, further investigation is required to explore the possible mechanism in more details.
Previous studies found that β-catenin/TCF4 activity modulates expression of ZEB1 in SW480 cells (43). Accordingly, it is expected that FLYWCH1 interaction with β-catenin may also regulate β-catenin–mediated transcription of ZEB1. Indeed, overexpression of FLYWCH1 massively decreased the transcription of human ZEB1-promoter in SW480 cells (Fig. 5D and E). This effect, however, was not found when a mutant form of FLYWCH1 (FLYWCH1ΔC), which lacks the β-catenin–binding domain was used (Fig. 5D). Notably, overexpression of ZEB1 rescued the inhibitory effect of FLYWCH1 overexpression on cell migration (Supplementary Fig. S9B and S9C). These data support the notion that FLYWCH1-mediated suppression of ZEB1 may depend on the formation of a transcriptional activation complex with nuclear β-catenin.
ZEB1 and EPHA4, which were among the most affected genes by the FLYWCH1-knockdown (Fig. 5B), are important determinants of tumor progression serving as an inducer of invasion and metastasis in colorectal cancers (44, 45). Therefore, we sought to examine whether the expression of FLYWCH1 correlates with the expressions of these genes in human pathologic samples. First, to examine the correlation between FLYWCH1 and ZEB1, CDH1 and EPHA4 gene expression, we obtained the expression data of colon cancer from the TCGA. Level 3 Illumina RNA-Seq data for colon adenocarcinoma (COAD) were downloaded from cBioPortal, which contained gene expression data for 469 primary tumor samples of COAD (Supplementary Fig. S9D–S9G). The data indicated a significant negative correlation between FLYWCH1 and ZEB1 and EPHA4 (Supplementary Fig. S9D and S9E), whereas the correlation between FLYWCH1 and CDH1 or ZEB1 and CDH1 is very weak and is not significant (Supplementary Fig. S9F and S9G). These data suggest that FLYWCH1 may possess some antimetastatic activity.
Next, the expression of FLYWCH1 in a series of early and advanced/metastatic patient's colorectal cancers was assessed using ISH assay on a TMA. Our analysis showed a negative correlation between FLYWCH1 expression and tumor progression (70.8% vs. 25.7%, P ≤ 0.017; Fig. 6A). Furthermore, the expression pattern of ZEB1 and EPHA4 versus FLYWCH1 in a cohort of 33 (normal, primary, and metastatic) tumor tissues were also investigated. Our data showed that ZEB1 expression is highly increased in advanced tumors, whereas FLYWCH1 expression is significantly repressed (Fig. 6B). Interestingly, a significant linear correlation between downregulation of FLYWCH1 and upregulation of ZEB1 and EPHA4 in normal versus advanced stages, but not in the primary tumors, was observed (Fig. 6B).
Despite the fact that substantial progress toward identification of full human proteome, not all the annotated protein-coding genes have been characterized. One such example is the protein product of FLYWCH1 gene, which contains five conserved FLYWCH-type Zinc-finger domains (Fig. 1). This work demonstrates identification, characterization, and functional analysis of FLYWCH1 protein for the first time. Based on bioinformatics analysis, FLYWCH1 may have potential DNA-binding abilities due to the presence of highly conserved NLS and FLYWCH-type Zinc-finger motifs, which appears to be evolutionarily conserved in mammals (Fig. 1; Supplementary Fig. S1). It should be noted that, although the endogenous FLYWCH1 can be detected by IF, despite lot of efforts no commercial antibody can detect the endogenous FLYWCH1 by WB or IHC analysis.
We demonstrated that FLYWCH1 inhibits β-catenin–mediated transcriptional activation by competing with TCF4 to bind to β-catenin. We also show direct evidence of cell migration and invasion defects caused by overexpression of FLYWCH1 that mimic a canonical Wnt loss-of-function phenotype. Moreover, our analyses revealed repression of a surprisingly limited subset of β-catenin-target genes by FLYWCH1 including those involved in regulation of cell migration and morphology. Alterations in cell motility and morphology are initial steps toward invasion and metastasis, one of the critical hallmark of cancer. Wnt/β-catenin signaling pathway, through different mechanisms, is often linked to regulation of various biological traits such as modulation of collective cell migration and cell branching (46–48). Indeed, signaling through this pathway modulates the expression of multiple genes that control filopodia formation and consequently, cell morphology, migration, and invasion. Examples of these genes include CDH1, Fascin1, c-Jun, and Eph/ephrin molecules (49–52). One of the most widely studied critical protein for the establishment of filopodia is the actin-binding protein, Fascin. Notably, our data show that Fascin1 gene expression is regulated by FLYWCH1 (Fig. 5B). During development, Fascin promotes cell migration through filopodia formation. High Fascin expression is correlated with invasion and metastasis of various tumors (53, 54).
In line with these findings, the reduced cell migration and morphologic changes observed in this study might be due to modulation of the genetic program downstream of β-catenin/TCF4 by FLYWCH1. Especially, as FLYWCH1 negatively regulates the β-catenin/TCF transactivation with no effect on the level of either β-catenin or TCF4 proteins (Fig. 2). The possible mechanism by which continuous Wnt3a/R-spondin1 treatment downregulates FLYWCH1 is currently unclear. It is possible that the Wnt/β-catenin-TCF4 via a negative feedback loop regulates the FLYWCH1 transcription product. In turn, FLYWCH1, together with Wnt via the β-catenin/TCF4, activates “on” or “off” decision to control the downstream genes fundamental to cell migration. Although this genetic modulation is likely to represent a transcriptional hierarchy, nevertheless, it is essential to understand why and how FLYWCH1 selectively represses specific β-catenin/TCF-target genes and uncover the molecular mechanisms behind this. This study, so far, focused on the assessment of the consequences of FLYWCH1 aberrations on the Wnt/β-catenin canonical pathway. However, a possible role of FLYWCH1 in combination with different ligands and receptors in regulating cell morphology and migration via alternative noncanonical Wnt/signaling responses (e.g., mediated by small GTPases, Rac1, and RhoA; ref. 55), remains unknown.
Defects in cell migration and actin cytoskeleton rearrangement could be related to EMT (56). However, the decreased cell migration found in our study is likely due to mechanisms other than EMT. No significant change in Vimentin expression indicates that its expression might not be regulated by FLYWCH1 and, thereby, ruling out the involvement of FLYWCH1 during EMT. Thus, it appears that FLYWCH1 may control cell migration via β-catenin-dependent E-cadherin upregulation and ZEB1 downregulation leading to increased cell–cell and cell–matrix adhesion.
The role of E-cadherin and the adherens junctions in mediating epithelial differentiation, by the establishment of cellular adhesion and polarity, is well recognized. Consistent with our observations, upregulation of E-cadherin has been widely associated with reduced cell motility and invasion in addition to morphologic changes from mesenchymal-like to epithelial-like shape in various types of cancer (57, 58). Another interesting observation of our study is the negative correlation between FLYWCH1 expression and tumor progression (Fig. 6A and B). One possible hypothesis is that tumor cells have developed mechanisms to suppress FLYWCH1 expression (Fig. 6C), facilitating changes in cell morphology and migration which promotes metastasis. FLYWCH1 is a newly characterized gene. Our findings demonstrate that FLYWCH1 could have potential tumor suppressor activity. In advanced colorectal cancer, FLYWCH1 similar to most tumor suppressor genes is significantly inhibited. A likely hypothesis is that tumors with high degrees of methylation are more likely to inactivate genes critical for tumor progression and metastasis. However, the underlying mechanisms associated with the possible epigenetic and posttranscriptional regulation of FLYWCH1 gene expression remained unknown. Furthermore, FLYWCH1 functions as a transcription modulator that can bind to DNA and/or RNA and in β-catenin-dependent or independent manners. Therefore, outside the scope of the current studies, a comprehensive genomic analysis to identify the direct target genes regulated by FLYWCH1, using both ChIP-sequencing and RNA-sequencing or RIP-sequencing analyses, will further clarify other functions of this protein.
In conclusion, our findings revealed a new molecular mechanism underlying negative signal cross-talk between FLYWCH1 and β-catenin. In this manner, the FLYWCH1/β-catenin complex represses selective Wnt-target genes, including Eph/ephrin and ZEB1, which mainly associated with cell polarity and migration during colorectal cancer development and metastasis. Prospectively, FLYWCH1 has potential to represent a biomarker and therapeutic target against metastatic colorectal cancer in the future.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: B.A. Muhammad, A. Saadeddin, A.S. Nateri
Development of methodology: B.A. Muhammad, R. Babaei-Jadidi, A. Saadeddin, M. Ilyas
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.A. Muhammad, S. Almozyan, R. Babaei-Jadidi, E.K. Onyido, S. Hossein Kashfi, M. Ilyas
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.A. Muhammad, S. Almozyan, R. Babaei-Jadidi, E.K. Onyido, S. Hossein Kashfi, M. Ilyas, A. Lourdusamy, A.S. Nateri
Writing, review, and/or revision of the manuscript: B.A. Muhammad, R. Babaei-Jadidi, E.K. Onyido, S. Hossein Kashfi, B. Spencer-Dene, M. Ilyas, A. Lourdusamy, A.S. Nateri
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.A. Muhammad, B. Spencer-Dene, A. Lourdusamy, A. Behrens, A.S. Nateri
Study supervision: A. Saadeddin, A.S. Nateri
We are grateful to B.W. O'Malley, M. Saito, and A. McIntyre for providing essential reagents. We thank R. Muraleedharan and W. Fadhil for technical help and advice on histology. We also thank the fantastic fundraising efforts of Alison Sims and her family in memory of Daz Sims to support the work in our laboratory. This work was supported by the Medical Research Council (grant number G0700763) to A.S. Nateri and University of Nottingham, Nottingham, UK.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.