Primary cell cultures from an aggressive fibromatosis tumor harbouring a truncating APC mutation were established. These were transfected with either a wild-type APC gene or a control vector, and successful transfection verified using immunohistochemistry. These cell cultures were established for a previous study to determine the role of the interaction between APC and β-catenin in the regulation of cell proliferation in aggressive fibromatosis. The transfection conditions, mutational analysis, and results of proliferation studies have been previously reported 1011. RNA from the cells derived from an aggressive fibromatosis tumor with an APC mutation and the same cells after transfection of wild-type APC were compared using suppression subtractive hybridization. Both forward (cells transfected with a control vector compared to cells transfected with wild-type APC) and reverse (cells transfected with wild-type APC compared to cells transfected with a control vector) subtractions were conducted. Clones were identified, and verified using Northern blot. The most differentially expressed clones were sequenced and their identity searched using GenBank.

In this comparison, the most differentially regulated gene in the reverse subtraction was a novel gene that was initially identified on human chromosome 7 using GenBank sequence data. Both 5'- and 3'- rapid amplification of cDNA ends (RACE) using the sequence of the 200 bp clone isolated from suppression subtractive hybridization, as well as analysis of EST in the region were undertaken to identify the full length of this gene. Subsequent to the start of this work, Strausberg et al identified a gene in this region using high through put analysis of EST's and designed the gene sterile alpha motif domain 9 (SAMD9) 12. During this analysis, we also found a nearby gene, which was also subsequently identified by Strausberg et al as sterile alpha motif domain 9 like (SAMD9L) 12. Using a search of the DNA sequence database in GenBank, SAMD9 and SAMD9L were found to match human chromosome 7q21.2. SAMD9L is located 5' upstream from the SAMD9 gene (Fig. 1). Both genes are coded by the reverse strand with a head to tail orientation.

Compiling all of the sequence data, we located the most 5'- end sequence for SAMD9 in a 5'- RACE clone, and the most 3'- end sequence for SAMD9 in one of the EST clones [GenBank: AA628487] [see Additional file 1]. Using primers corresponding to the sequences of the most 5'- and 3'- ends of the SAMD9 gene, a predicted 7 kb transcript was amplified from an aggressive fibromatosis tumor cDNA. The full-length of this transcript, the longest transcript that we found, was sequenced [GenBank: AF445355]. The cDNA sequence of SAMD9 gene has 6847 bp with a poly-A signal (AATAAA) at 6831 bp downstream from the transcriptional start point (Fig. 1A). Mapping this cDNA sequence to the genomic DNA sequence in the GenBank, the SAMD9 gene is 18505 bp in length and contains three exons (161 bp, 101 bp and 6585 bp in length, respectively). There is a TATA signal (CCTATATTCT) at -75 bp and five predicted Lef/Tcf binding elements [(C,G) (A,T) (A,T)CAAA(G,C)] at -11440 bp, -10247 bp, -8014 bp, -4788 bp and -3321 bp. The clones isolated from 5'- and 3'- RACE of the SAMD9 gene were sequenced. Comparing the 5'- and 3'- RACE sequences to the genomic DNA sequence from GenBank, different transcription initiation sites and polyadenylation sites of SAMD9 are noted, and there are at least two alternative splices.

During the analysis of EST sequences near the initial clone, we identified a second gene,SAMD9L, which is located 12740 bp distal to SAMD9. Using the sequences from the most 5'- and 3'- EST in the region [Celera Database: THC 511215 and THC 513290, respectively] [see Additional file 1], the full-length sequence of the SAMD9L was amplified from an aggressive fibromatosis tumor cDNA, and sequenced [GenBank: AF474973]. The cDNA sequence of SAMD9L gene has 5821 bp with a poly-A signal (AATAAA) at 5659 bp downstream from the transcriptional start point (Fig. 1B). Mapping this cDNA sequence to the genomic DNA sequence in GenBank, the SAMD9L gene is 17583 bp in length, including six exons (152 bp, 59 bp, 132 bp, 148 bp, 102 bp and 5228 bp in length, respectively). There is a TATA signal (TCTATACTTC) at -73 bp and two predicted Lef/Tcf binding elements [(C,G) (A,T) (A,T)CAAA(G,C)] at -49139 bp and -3661 bp.

There are alternatively spliced transcripts in both SAMD9 [GenBank: AF453311] and SAMD9L [GenBank: AY195582-195587, DQ068177] [see Additional file 1]. These alternative splices lead to either inclusion or exclusion of an exon due to the usage of a different 5' donor splice site or a different 3' acceptor splice site. This alternative splicing also changes the internal coding region due to an in-frame insertion or deletion in both SAMD9 and SAMD9L. In the case of SAMD9, non-canonical dinucleotides AT and TC are used as donor and acceptor splice sites for intron 3, and there is an in-frame canonical pair of the donor and acceptor (GT and AG) splice sites near by. This alternative splicing changes the internal coding region due to an in-frame deletion, leading to the exclusion of sequence coding for a putative protein structure domain – sterile alpha motif (SAM) domain near the N-terminus.

The putative protein structures of both genes were determined using the open reading frame finder program on the National Center for Biotechnology Information (NCBI) website 13, and an open reading frame was identified in the last exon of both genes. The two genes share 78% homology at the DNA sequence level in this exon. The putative protein sequences were analyzed using the Conserved Domain search program in NCBI website 14 for SAMD9, the Motif Scan program from Swiss Institute of Bioinformatics 151617 and the Predict Protein server from European Molecular Biology Laboratory 18 websites for SAMD9L. Both protein products contain a sterile alpha motif (SAM) domain near the N-terminal region. Otherwise, there is no close homology to other well-characterized proteins or protein motifs. Based on a comparison of the SAM domain sequence of SAMD9 to the homology modeling of the RNA-binding Smg SAM domain 19, the SAM domain of SAMD9 lacks the residues that are essential for binding RNA, while it has 98% sequence homology with the Ephrin-B2-receptor SAM domain, which forms homo-oligomerization and provides a platform for the formation of larger protein complexes 20.

The expected 7 kb transcript of SAMD9 was identified by Northern analysis using RNA from different age placentas and a 1729 bp PCR amplified probe encompassing the 3'UTR (Fig. 2A). The expression pattern of SAMD9 in multiple tissues was investigated using a human multiple tissue cDNA panel (Clontech, USA) and RT-PCR. The primers were designed to recognize sequences at 3'UTR of SAMD9. Expression was detected in all human adult, fetal and tumor tissues that were tested, except for fetal brain. Only a very low signal was detected in skeletal muscle (Fig. 2B). Using the sequences at 3'UTR of SAMD9L as primers, SAMD9L was also found to be expressed in all tissues, except for the tumor types, colon cancer (CX-1), breast cancer (GI-101) and pancreatic cancer (GI-103) (Fig. 2B).

To determine the level of expression of SAMD9 in neoplasia, semi-quantitative RT-PCR and real-time quantitative PCR were performed using the sequence at 5'UTR of SAMD9 as a sense primer and a sequence at the beginning of the open reading frame of SAMD9 as an anti-sense primer. Expression was compared between tumor and normal control tissues in cases of aggressive fibromatosis, breast cancer and colon cancer. SAMD9 was expressed at a lower level (about 33% the level for normal control tissues) in aggressive fibromatosis (32.22% ± 29.18% vs 100%, p < 0.05 for real-time PCR, Fig. 3A). Expression of SAMD9 was lower in 20% cases of breast cancer and 35% cases of colon cancer than in the normal control tissues (Fig. 3C). In three cases of colon cancer, there was no RT-PCR product detected. However, there was no significant difference in expression between colon cancers and normal control tissues or breast cancers and normal control tissues when analysed as a group using real-time PCR. Expression studies of SAMD9L were also conducted using a sequence at 5'UTR which is common for all of the splice variants except one, as a sense primer and sequence at the beginning of the open reading frame as an anti-sense primer. There was no statistically significant difference in SAMD9L expression between aggressive fibromatosis or colon cancer and normal control tissues. However, the expression of SAMD9L was lower in breast cancers than in healthy breast epithelial tissues from the same patients (2.81% ± 2.40% vs 100%, n = 10, p < 0.001, Fig. 3B). There was no significant difference in expression of SAMD9L between aggressive fibromatosis samples and normal controls tissues (Fig. 3D). While two genes are expressed at a lower level in tumors, there is a discrepancy between the levels of expression of the two genes in different tumor types. For instance, SAMD9 is expressed at lower levels in aggressive fibromatosis, while SAMD9L is not.

In order to investigate the cellular localization of the putative SAMD9 protein, the open reading frame of the SAMD9 gene was cloned, an N-terminal EGFP tag was added and subcloned into a CMV driven vector pLP-EGFP-C1 (Clontech, USA). This vector was transfected into Cos-1 cells, and using an antibody to EGFP, the fusion protein was detected as a 200 kDa band (Fig. 4A). To determine its localization in a variety of cell types, it was transfected into a human fetal fibroblast cell line, MRC-5, and a colon cancer cell line, SW480. The EGFP fusion protein was detected diffusely in the cell cytoplasm (Fig. 4B–F).

RNA interference was utilized to study the cellular function of SAMD9. RNA interference was designed by searching the public available sequence database to find a 19 bp oligonucleotide sequence specific for SAMD9 and inserting into a pSUPER-RNAi-EGFP expression vector to generate the SAMD9 RNA interference vector (pSUPER-RNAi-SAMD9). One of the sequences is 846 bp after the start of translation, and other is 2371 bp after the start of translation. The pSUPER-RNAi-EGFP expression vector was generated based on pEGFP-C1 (Clontech, USA) and pSUPER 21. A nucleotide substitution from G to

A

SAMD9 expression reduced by RNA interference vector for SAMD9 was tested by co-transfection of an N-terminal EGFP tagged SAMD9 expression vector with either one of the two SAMD9 RNA interference vectors, RNA interference vector with mutated SAMD9 sequence to null its function as negative control, or its control empty vector into Cos-1 cells. The reduction of SAMD9 expression was confirmed using Western blot with anti-GFP (Santa Cruz, USA) for the fusion putative SAMD9 protein. A 90% reduction in expression of SAMD9 fusion protein by RNA interference was detected as measured using densitometry (Fig. 5A). The RNA interference vector that express EGFP was also expressed in SW480 cells along with an N-terminal HA tagged SAMD9 expression vector or its control vector. A 90% reduction in expression of SAMD9 fusion protein by RNA interference was detected (data not shown). Activity of the RNA interference vector for SAMD9 was further verified by the reduction of SAMD9 RNA expression in MRC-5 cells (Fig. 5B).

Because β-catenin stabilization is known to regulate fibroblast proliferation, motility and invasiveness 11, we studied the effect of down-regulating SAMD9 on these parameters. Proliferation rate, as measured by BrdU incorporation, was increased after RNA interference of SAMD9 in MRC-5 cells (12.5% ± 2.66% with the 1^st SAMD9 RNA interference vector, 12.2% ± 1.63% with the 2^nd SAMD9 RNA interference vector, and 7.5% ± 1.29% with the control vector, p < 0.01, Fig. 5C). There was not a statistical difference in cell motility as measured by the number of cells passing through the membrane of the culture insert (1.7 ± 0.65 or 2.3 ± 1.31 with RNA interference of SAMD9 vector vs 0.67 ± 0.65 with control vector). Cell invasion was increased, as measured by the number of cells passing through the membrane of the Matrigel invasion chambers, after SAMD9 expression was decreased in MRC-5 cells by RNA interference (16 ± 5.19 with RNA interference of SAMD9 vector vs 5.7 ± 1.73 with control vector, p < 0.05). These studies were also undertaken in Cos-1 cells, and showed a similar reduction in proliferation rate after over-expression of the putative SAMD9 protein in Cos-1 cells (20.7% ± 21.9% vs 76.8% ± 20.2%, p < 0.005). Taken together, this shows that SAMD9 plays a role regulating cell proliferation.

To explore the effect in neoplasia, cells from the SW480 colon cancer cell line were transfected with a SAMD9 expression vector to determine if increasing expression of this gene would suppress the neoplastic behavior in these cells. Upon transfection of the construct, cell proliferation was reduced (25.6% ± 3.3% vs 46.2% ± 9.4%, p < 0.01, Fig. 6A). Cell invasion index calculated as the ratio of the total number of nuclei counted in the membrane for invasion assay to that in the membrane for motility assay was reduced (0.0335 ± 0.0203 vs 0.0779 ± 0.0183, p < 0.05, Fig. 6B), while cell apoptosis, as indirectly measured by relative caspase-3 activity was increased (253.11% ± 45.90% of baseline, p < 0.005, Fig. 6C). The relative caspase-3 activity was also reduced in Cos-1 cells after expression of the SAMD9 siRNA construct (53.76% ± 0.99% of the baseline, p < 0.01). Thus expressing SAMD9 at higher levels causes cell effects which would be expected to suppress the neoplastic phenotype.

To investigate the role of SAMD9 in tumor growth, the colon cancer cell line, SW480, was transfected with SAMD9 expression vector or its control empty vector. The stably transfected SW480 cells were injected into the immune deficient mice. The tumor volume of the xenograft from SW480 cells over-expressing SAMD9 was reduced in both nude mice (62.5 ± 73.5 mm³ vs 162.5 ± 61.7 mm³, n = 4, p < 0.05, Fig. 6D) and NOD-SCID mice (62.4 ± 20.3 mm³ vs 98.6 ± 17.1 mm³, n = 21, p < 0.01, Fig. 6E). To verify the role of SAMD9 on neoplasia, we performed experiments when the expression of SAMD9 was lowered using RNA interference. The tumor volume of the xenograft from SW480 cells expressing the RNA interference of SAMD9 was increased in nude mice (310 ± 19.6 mm³ vs 270 ± 24 mm³, n = 5, p < 0.05).

The orthologous genes were identified for SAMD9 and SAMD9L by searching the NR and swissprot databases for best reciprocal matches using human SAMD9 and SAMD9L protein sequences. Genome sequences for all species available were searched by BLAT at the University of California at Santa Cruz genome browser 22. The results were summarized in Table 1 and the phylogram [see Additional File 1] was constructed using multiple sequence alignment of orthologous genes identified with ClustalW program from European Bioinformatics Institute website 23. The default parameters from website documentation were used 24. Based on their sequence similarity, SAMD9 and SAMD9L appear to have originated from a common ancestor by an ancient gene duplication. Since there are orthologous genes of both SAMD9 and SAMD9L in chimpanzees, dogs and rats, but not in chicken, frog or fish species, this duplication event likely occurred after the mammalian radiation. The genomic structures of both SAMD9 and SAMD9L, including the order and orientation of genes within the genomic region and the relative size of the intergenic region, are conserved in all available mammalian genome sequences including opossum, indicating the gene duplication event occurred between 175 to 200 MYA 25. No evidence for either gene was found in lower eukaryotes such as Drosophila, C. elegans or yeast.

There were several overlapping ESTs matching the open reading frame for SAMD9L, and no ESTs corresponding to the open reading frame for SAMD9 in the mouse genome. This was confirmed by bioinformatic searches of all genomic sequences and partial genome sequences (EST libraries). The 168 kb region in human chromosome 7, in which SAMD9 is located, has no match in the mouse genome. A syntenic map of the region was constructed for human, mouse and rat. A break of synteny at the SAMD9 locus was determined in mouse, with distal genes mapping to mouse chromosome 6, and the proximal genes mapping to mouse chromosome 5. The mouse specific rearrangement resulted in new centromere formation at the site of breakage and the subsequent loss of the SAMD9 gene. The SAMD9L region contains numerous gaps in the mouse genome assembly, and an enrichment of mouse specific segmental duplications (Fig. 7, also see Additional file 1). The relatively poor draft quality of the rat genome does not allow for high resolution mapping of this locus, but there is evidence to support the existence of both SAMD9 and SAMD9L genes in rats with no large genomic rearrangements, segmental duplications or breaks in synteny although numerous gaps do exist (Fig. 7, also see Additional file 1). There are no gaps or segmental duplications at this locus in the human genome, determined by searching the Human Genome Segmental Duplication Database.

SW480, MRC-5, Cos-1 and NIH3T3 cell lines were obtained from the American Tissue and Cell Collection. For each experiment in which cell lines were used for a quantitative assay, the experiment was conducted nine independent times, each time on a different day. The RNA populations used for subtractive hybridization were prepared from cells derived from an aggressive fibromatosis tumor with an APC mutation after transfection of wild-type APC or a control vector as previously reported 11. Cells, tumor and normal control tissues were collected from patients with sporadic aggressive fibromatosis, colon cancer, and breast cancer while the patients were under surgical excision of the tumors. All of the samples were cryopreserved as soon as possible after resection and stored in liquid nitrogen vapor for later nucleic acids extraction. The local ethical approvals for this research were obtained.

Suppression subtractive hybridization was used to compare the gene expression differences in a cell culture derived from an aggressive fibromatosis tumor with an APC mutation after transfection of a wild type APC to the same cell culture after transfection of a control vector using Clontech PCR-select cDNA subtraction kit (K1804-1, Clontech, CA, USA). Both forward (cDNA from cells transfected with control vector was subtracted by cDNA from cells transfected with wild-type APC) and reverse (cells transfected with wild-type APC was subtracted by cells transfected with control vector) subtractions were conducted. There were eight genes identified in the forward subtraction and 16 genes in the reverse subtraction that were found to be differentially expressed, based on the screening of dot blots containing 196 clones in the forward and reverse subtracted libraries, respectively. SAMD9 was confirmed as being differentially expressed in the reverse subtraction and selected for further analysis in this paper.

15 μg of total RNA from 8 weeks, 12 weeks and full term human placenta were electrophoresed on 1% RNase-free agarose gel and Northern transferred to a nylon membrane according to according to the method from Sambrook et al 34. A probe for SAMD9 was generated using PCR with 89729F and 88001R as primers for a 1729 bp fragment from the 3'UTR. The Northern membrane was first probed with the SAMD9, then stripped and probed with the human actin probe as a loading control. The probes were labeled with isotope ³²P.

Expression of SAMD9 and its paralogous gene SAMD9L were investigated using PCR of human multiple tissue cDNA panels including adult, fetal and tumor tissues (Clontech, USA). The sequences at the 3' UTR of the SAMD9 and SAMD9L genes were used to design the target primers according to the instruction from the manufacture since the DNase pretreated RNA was used for synthesis of those cDNA panels. 88347F and 88001R [see Additional file 1] were primers for SAMD9, while 119622F and 119170R [see Additional file 1] for SAMD9L. GAPDH was used as a control.

The total RNA was extracted from tumor tissue and normal control tissue from aggressive fibromatosis, breast and colon cancers using Trizol RNA reagents (Invitrogen, USA). The anti-sense and sense primers for SAMD9 and human beta-2 microglobulin (beta 2M) were designed using the sequence across the introns, and beta 2M was used as control. For beta 2M, beta 2M-F was used as sense primer and beta 2M-R as anti-sense primer to produce a 112 bp PCR product [see Additional file 1]. For SAMD9, 103270F was used as sense primer and 93871R as anti-sense primer to produce a 711 bp PCR product for SAMD9 with SAM domain coding sequence and a 222 bp PCR product for SAMD9 lacking SAM domain coding sequence [see Additional file 1]. The optimal amplification cycles were set within the linear range, 30 cycles for beta 2M and 37 cycles for SAMD9, after applying series of diluting cDNA samples and subjecting for series of PCR amplification cycles.

The polymerase cycling reactions of paired cDNA samples were conducted at the same time for both target and control genes. The amplified products were electropheresed on a 2% agarose gel and photographed under UV light. The expression of beta 2M was set as control to normalize the level of target expression. The expression level of the target gene SAMD9 was analyzed using computer software ImageQuant and represented as the ratio of its intensity verse the intensity of the beta 2M gene under the condition that the same amount of the cDNA was applied as template for the target gene and the control gene.

Real-time quantitative PCR was undertaken using 28S rRNA (28S) as the control gene. PCR primer pairs for human 28S rRNA were taken from Simpson et al, 2000 35, for SAMD9, 103270F was used as sense primer and 94424R as anti-sense primer, and for SAMD9L, 133857F was used as sense primer and 124170R as anti-sense primer [see Additional file 1]. Validation curves were carried out for the primer sets using RNA from SW480 cells diluted to 1:5, 1:10, 1:50, 1:100, and 1:1000. Delta delta Ct method was used for setting up the experiment and analysis of the data. An arbitrarily designed threshold was set at 0.2 for all analysis, while the baseline cycles were set for all analysis from 3 to 10 cycles for 28S and from 3 to 30 cycles for SAMD9 or SAMD9L. The threshold cycle, C_t was determined using the analysis software (SDS 2.1, Applied Biosystems). The result was analyzed using the relative quantitative 2^{-ΔΔ C} _t method 36. The expression level of the target gene SAMD9 or SAMD9L in tumor tissues were represented as the fold difference from normal control tissues.

The total protein from Cos-1 cells transfected with either pLP-EGFP-SAMD9 or its control vector pLP-EGFP were harvested using Nonidet P-40 lysis buffer (1% Nonidet P-40, 50 mM Tris-HCL [pH 7.5], 150 mM NaCl, 2 mM EDTA, 10% glycerol, proteinase inhibitor tablet [1 tablet per 10 mL buffer, Roche, USA]). The protein was electropheresed on 8% sodium dodecyl sulphacrylamide gel and transferred to membrane according to the method from Sambrook et al, 1989. The membrane was then probed with anti-GFP (1:200, Santa Cruz, USA) and secondary antibody (HRP-conjugated monoclonal anti-mouse IgG, 1:2000, Transduction Laboratory, USA). Finally, the membrane was exposed to X-ray film for 20 minutes. The total protein from Cos-1 cells transiently co-transfected with pLP-EGFP-SAMD9 and either one of the two SAMD9 RNA interference vectors (pSUPER-RNAi-SAMD9, pSUPER-RNAi-SAMD9-2^nd), RNA interference vector with mutated SAMD9 sequence (pSUPER-RNAi-SAMD9-mut9) to null its function as negative control, or its control empty vector (pSUPER-RNAi-EGFP) were also harvested for Western blot analysis to confirm the knock-down of SAMD9 expression by RNA interference.

After searching of the public available sequence database, a 19 bp oligonucleotide sequence (846 bp after the start of translation) specific for the RNA inference of the SAMD9 expression was found (5'-GTGCATTCGAGAGCCAAGA-3') and subcloned into a pSUPER-RNAi-EGFP expression vector to generate the SAMD9 RNA interference vector (pSUPER-RNAi-SAMD9). A nucleotide substitution from G to

A

SW480, MRC-5, or Cos-1 cells (ATCC, USA) were cultured in the DMEM medium with 10% fetal calf serum (Invitrogen, USA) in a CO₂ incubator at 37°C. Gene transfection was conducted using Fugene 6 reagent (Roche, USA). The transfected cells were either harvested for transient transfection or selected with the addition of 200 ng of G418 (Sigma, USA) for 2 weeks, and sorted by flow cytometry using GFP as marker. The cellular location of SAMD9 was analyzed under a confocal microscope in Cos-1, MRC-5 and SW480 cells transfected with pLP-EGFP-SAMD9. The effectiveness of SAMD9 RNA interference was documented by using confocal microscopy in SW480 cells co-transfected with pLP-CMV-HA-SAMD9 and either the SAMD9 RNA interference vector (pSUPER-RNAi-SAMD9) or its empty control vector (pSUPER-RNAi-EGFP), which were fixed, stained with anti-HA (1:200, Santa Cruz, USA) and Tex-red conjugated secondary antibody (1:100, Molecular Probes, USA).

The counting process was conducted by a blinded observer. Means and 95% confidence intervals were calculated. The t-test was applied to compare differences.