AC133-2: A Novel Isoform of Human AC133 Stem Cell Antigen

followed by Texas-Red anti-mouse IgG (red). Green fluorescence of β 1 intergrin subunits colocalize with red fluorescence of AC133-2 (not shown). Single optical xy plane sections at center of the cells are shown. The results are representative of three independent experiments


Introduction
Human AC133 antigen is a glycoprotein with a molecular weight of ~120 kDa. Based on its predicted amino acid sequence, AC133 contains an extracellular N-terminus, two large extracellular loops, five transmembrane domains, two small cysteine-rich cytoplasmic loops, and a cytoplasmic C-terminus (1). AC133 antigen was first detected on CD34 bright hematopoietic stem cells using a monoclonal antibody (mAb) named clone AC133 that was raised against human CD34 + cells (2). AC133 antigen has since been widely used to facilitate the analysis and isolation of hematopoietic primitive cells (3,4,5). Subsequently, Peichev et al. showed that endothelial progenitor cells co-express AC133 antigen and the endothelial cell specific receptor KDR in subpopulations of CD34 + cells derived from fetal liver, bone marrow, cord blood and peripheral blood (6,7). Recently, human central nervous system-stem cells were also reported to express AC133 antigen (8). A characteristic feature of this protein is its rapid down-regulation during cell differentiation (7,9), which makes it a unique cell surface marker for identification and isolation of stem cells and progenitor cells.
A structural and sequence related protein, prominin, was identified as the mouse orthologue of human AC133 antigen (9). The two polypeptides share ~65% amino acid sequence identity. It has been shown that prominin is phylogenetically conserved from mammals to zebrafish, and in fruit flies and nematodes (10,11). Prominin is selectively expressed in microvilli of various embryonic and adult epithelial cells, and in plasma membrane protrusions of nonepithelial cells including murine CD34 + bone marrow progenitor cells (9,12). Studies have shown that prominin specifically interacts with membrane cholesterol (11), suggesting prominin has a role in membrane organization and membrane-to-membrane interactions. Recently, Corbeil et al. reported that human AC133 antigen was also expressed in the plasma membrane of human epithelial cell line Caco-2 (9). Human retinal degeneration has been associated with a frameshift mutation in AC133 gene, which results in a truncated protein that fails to reach to the cell surface  Immunofluorescence and Confocal Microscopy-Near confluent 293 cells grown on glass coverslips were transfected with 2 µg of the AC133-2 cDNA and 10 µl of Lipofectamine 2000 6 (Life Technology Inc.). After 48 h, cells were carefully rinsed with PBS and fixed in 3% paraformaldehyde in PBS for 30 min at room temperature. WERI-RB-1 cells grown in suspension were directly plated on glass coverslips in 50 µl PBS/0.5% BSA for 20 min before fixation.
Incubation and washes were done directly on the coverslips. Indirect immunofluorescence staining was performed as previously described (9). Cells were observed with a BioRad uRadience2000/ Nikon Eclipse 800 (Bio-Rad Laboratories) confocal laser scanning scope using BioRad LaserSharp2000 imaging software. The images shown were prepared from confocal image files using Adobe Photoshop 5.5 software.
Cell Surface Biotinylation and Immunoprecipitation-Biotinylation was performed as reported (9), except that Sulfo-NHS-LC-LC-biotin (Pierce Chemical) was used. Protein concentrations of the cell lysates were quantified by Bio-Rad Protein Assay. 400 µg protein was incubated with mAb AC133/2 at 10 µg/ml for overnight. Immune complexes were collected with protein A/G-agarose (Santa Cruz Biotechnology). Immunoprecipitated proteins were analyzed by SDS-polyacrylamide gel electrophoresis on a 4-12% gradient gel (Invitrogen), and transferred to a PVDF membrane (Millipore). The blot was blocked with 5% low fat milk/ 0.3% Tween-20.
Biotinylated proteins were identified by horseradish-conjugated NeutrAvidin (Pierce Chemical) and detected by enhanced chemiluminesce western blotting detection reagent (KPL).

LX-1 Lung Carcinoma Implantation and Tumor
Cell Isolation-Frozen LX-1 tumor fragment was acquired from the National Cancer Institute. 2 x 2 x 2 mm 3 size tumor cubes were implanted in the axillary region of 6~7 week-old nude (nu/nu) mice (13). After 3 to 4 weeks, about 2 x 2 x 2 cm 3 size tumors were dissected. Tumor tissue was minced in PBS and gently homogenized in a 50-ml tube (Fisher Scientific) with a fitted pestle. Tumor cells were then 7 sequentially filtered though 100 µm and 40 µm mesh. Single tumor cells were subjected to flow cytometry or cell surface biotinylation followed by immunoblotting to detect AC133-2 protein.

AC133-2 Expression in Human Skin
Epidermis-Human neonatal foreskins were obtained from routine circumcisions performed at Brigham and Women's Hospital, Boston under an IRB approved protocol. Epidermal basal keratinocytes were isolated as reported (14). Isolated cells were labeled with PE-or FITC-conjugated isotype controls, PE-conjugated AC133/2 mAb, or FITC-conjugated CD29, a mAb against β1 integrin subunits (Beckman Coulter). Cells were analyzed by flow cytometry or indirect immunofluorescence staining. AC133-2 + /β1 + cells purified by fluorescence-activated cell sorting (FACS) were cultured on a feeder layer of 3T3 mouse embryo cells using the method of Rheinwald (15). To monitor the loss of AC133-2 expression, very confluent keratinocytes were further grown in DME/F12 containing 2% FBS for a week before FACS analysis. For indirect immunofluorescence staining, cultured keratinocytes were plated on glass cover slips pre-seeded with irradiated 3T3 cells, and grown in the same conditions as described above. Confluent cells were fixed and stained with anti-keratin (clone LP34) or anti-involucrin (NeoMarkers) as described (16).

Results
Cloning and Identification of AC133-2-To make a Northern blot cDNA probe for human AC133 antigen, we amplified a N-terminal fragment of 458 bp (GenBank Accession No. AF027208) by RT-PCR from the retinoblastoma cell line WERI-RB-1, from which the first AC133 cDNA was cloned (1). DNA sequencing data revealed that one out of four clones had a deletion of a 27bp segment corresponding to positions 314 to 340. This did not result in a frame shift in the coding sequence, but created a 9 amino acid deletion (Fig. 1A, 1B). Using RT-PCR with primers flanking the start and stop codons of AC133, a novel full-length cDNA clone was 8 isolated and sequenced. This novel sequence, which will be referred to as human AC133-2, differs from the previously published AC133, which will be referred to as human AC133-1, by the 27 bp deletion (Fig. 1A). A literature search revealed that a homologous mouse sequence to human AC133-1 had previously been reported by Maraglia et al. (17). It was named as mouse AC133, which will be referred to as mouse prominin-1 in this paper (Fig. 1B). Interestingly, this sequence is 27 bp longer than the previously characterized mouse prominin, which will be referred to here as prominin-2. The relationship between the gene structure of prominin-1 and -2 has not been reported. The corresponding nine amino acids in human AC133-1 are highly homologous to that in mouse prominin-1 (Fig. 1B). Human AC133-2 and mouse prominin-2 encode proteins that lack a 9 amino acid segment at same location in the N-terminal extracelluar region just proximal to the first transmembrane domain. The deletion produces no significant change in the predicated membrane spanning domains of AC133-2, compared to AC133-1, by hydrophobicity analysis (data not shown). The discovery of human AC133-2 and the existence of its mouse counterpart prominin-2 indicate that AC133-2 may be a result of alternative mRNA splicing.
Human AC133-2 is a Novel Splice Variant-Exons of AC133 gene were identified by aligning human genomic sequences with AC133-1 or -2 cDNA sequences using the database at the National Center for Biotechnology Information. We found two human chromosome 4 clones  Table 1, see supplemental data). Our analysis revealed that AC133-9 2 resulted from an deletion of exon 3 of 27 nucleotides, flanked by splice acceptor and donor consensus sequences (Fig. 1C). Using a forward primer in the intron prior to exon 3 and a reverse primer in exon 4, exon 3 and its surrounding intron sequences were amplified by PCR from human genomic DNA, and confirmed by DNA sequencing. These results indicate that AC133-2 is a novel alternatively-spliced isoform of AC133-1.

AC133-2 is a Membrane-associated Glycoprotein-
To determine if AC133-2 cDNA could be transported to the cell surface, as shown for the endogenous AC133 (1), we analyzed the protein product and its cellular localization in 293 cells transfected with AC133-2 cDNA. Cell Surface AC133-2 was detected by flow cytometry (Fig. 2A) using a commercially available PEconjugated mAb designated AC133/2 (1, 2). More than 95% transfected cells expressed AC133-2. In contrast, mock transfected 293 cells did not show any immunoreactivity to this antibody (data not shown). Localization of AC133-2 to plasma membrane of transfected 293 cells was observed by indirect immunofluorescence staining with the same antibody followed by confocal laser scanning microscopy ( Fig. 2B, a, b). The cellular localization of endogenous AC133 antigen in retinoblastoma cell line WERI-Rb-1 was similar (Fig. 2B, c, d). Cell surface distribution of AC133-2 was further confirmed by biotinylation of the intact cells, followed by immunoprecipitation with mAb AC133/2. Recombinant AC133-2 was detected by NeutrAvidin blotting as a single band with an apparent molecular mass of ~115 kDa (Fig. 2C).
Deglycosylation with PNGase F yielded a ~94 kDa band, which is the predicted size from AC133-2 amino acid sequence. The 9 amino acid deletion in AC133-2 should not affect the degree of N-glycosylation compared to AC133-1, because none of the consensus sequences for glycosylation sites are within the deleted region.
Tissue mRNA Expression of AC133 Isoforms-Similar to AC133-1, the transcript of AC133-2 is approximately 4 kb as determined by Northern blotting (data not shown). However, the size by guest on March 24, 2020 http://www.jbc.org/ Downloaded from difference between AC133-1 and -2 mRNA is too small to be resolved by this technique. Specific primers flanking the truncated sequences were used to amplify AC133-1 fragment of 180 bp and AC133-2 of 153 bp in the same PCR reaction. The two fragments were then separated electrophoretically on a 3% MetaPhor agarose gel. The identity of the cDNA fragments was confirmed by DNA sequencing. To determine the constitutive mRNA expression profiles of AC133 isoforms, PCR was performed on commercially available cDNA panels. Since the CLONTECH multiple tissue cDNA panels have been normalized against several housekeeping genes, mRNA expression level between samples can be directly compared.
AC133-2 mRNA was ubiquitously expressed in a variety of human fetal and adult tissues (Fig.   3A). It was predominant in fetal liver, skeletal muscle, kidney, and heart, as well as adult pancreas, kidney, liver, lung and placenta. AC133-1 mRNA was not detectable in fetal liver and kidney, adult kidney, pancreas and placenta. However, AC133-1 was abundantly expressed in fetal brain, but poorly expressed in adult brain. In skeletal muscle and heart, the relative abundance of the two isoforms was reversed between fetal and adult tissues. Interestingly, AC133-2, but not AC133-1, was high in fetal liver, low in bone marrow, and barely detectable in peripheral blood (Fig. 3B). These relative levels of AC133-2 mRNA in hematopoietic tissue are consistent with reported AC133 protein levels (3). Low levels of AC133-1 transcripts were observed only after AC133 + cells were purified from fetal liver, bone marrow, cord blood and G-CSF immobilized peripheral blood using the mAb AC133/2 conjugated magnetic beads. The ratios of AC133-1 and AC133-2 transcripts were identical among these samples (data not shown).
In vivo AC133-2 Protein Expression-Due to the lack of isoform-specific antibody, we have not investigated the protein tissue distribution profile of AC133 isoforms. We know AC133-2 11 protein is expressed on hematopoietic stem cells, in which AC133-2 is the predominant transcript ( Fig. 3B), because AC133 + cells have been isolated from hematopoetic tissues (2,3,4,5,6). To show that AC133-2 protein is expressed in vivo in other tissues, we analyzed cell surface expression of this protein in lung carcinoma LX-1. LX-1 tumors expressed only the AC133-2 mRNA (Fig. 3C). Mononuclear cells isolated from peripheral blood and WERI-RB-1 retinoblastoma cells were included as a negative and a positive control, respectively. Flow cytometric analysis showed that about 80% of LX-1 tumor cells were recognized by AC133 mAb (Fig. 4B). In addition, the endogenous AC133-2 protein was detected by immunoprecipitation from biotinylated LX-1 tumor cells (Fig. 4D). AC133-2 appeared as a single band with a similar molecular mass to the AC133 antigen expressed in WREI-RB-1 cells. The relative protein levels of AC133-2 in LX-1 and WERI-RB-1 determined by immunoprecipitation and flow cytometry were similar.

Expression of AC133-2 in a Human Stem Cell Niche-
The wide tissue distribution of AC133-2 (Fig. 3A, B) suggests that the utility of this antigen for stem cell analysis and purification is not restricted to hematopoietic tissue. To determine if AC133-2 was expressed in other stem cell niches, we chose to isolate basal keratinocytes from human neonatal foreskin epidermis due to its availability. Epidermal stem cells reside in the basal layer of epidermis, in which cells mainly express α2β1, α3β1, and α5β1 integrins (5). Only AC133-2 was detected in epidermal cells (Fig.   5A, lane 1). Flow cytometric analysis showed that over 90% of the isolated epidermal cells were positive for β1 integrin subunits, indicating they were basal cells (Fig. 5B). In the total isolated cell population, the percentage of AC133-2 + cells was the same as the AC133-2 + /β1 integrin + cells, suggesting AC33-2 + cells were a subpopulation of the epidermal basal cells (data not shown). About 10% of the basal cells were positive for PE-conjugated mAb AC133/2 and FITCconjugated mAb against β 1 integrin subunits. Previous studies have shown that about the same 12 numbers of the basal keratinocytes have stem cell properties (21,22). Expression of AC133-2 was then further confirmed by indirect immunofluorescence staining of FACS-sorted AC133 + cells with a different mAb (Fig. 5C). Over 95% cells were AC133-2 positive, and were also β1 integrin-positive (data not shown).
Down-regulation of AC133-2 in Culture-To test whether isolated AC133 + cells could proliferate and differentiate in culture, isolated epidermal basal cells were either labeled with PEconjugated AC133 mAb or doubled with PE-conjugated AC133 mAb and FITC-conjugated anti-β1 integrin mAb, and then AC133 + cells or AC133 + /β1 integrin + cells were isolated by two rounds of fluorescence-activated cell sorting (Fig. 6B, left panel). The positive cells formed visible colonies as early as 6 to 10 days after culture initiation on irradiated 3T3 embryonic cells.
Colonies formed faster among AC133 + cells than AC133 + /β1 integrin + cells, perhaps because antibody binding to β1 integrins decreased the efficiency of cell attachment through integrin receptors. Cells continued to grow and formed tightly adherent epithelioid colonies and underwent stratification, which was difficult to see under phase contrast (Fig. 6Aa) but was manifested by indirect immunofluorescence staining with a mAb to anti-keratin 5, 6 and 18 ( Fig.   6 Ac). Clearly, all cultured cells were keratinocytes. The flat stratified cells in layers also expressed involucrin (Fig. 6Ad), a terminal differentiation marker (22). Isotype-matched control monoclonal Ab did not show any immune reactivity (Fig. 6Ab). Flow cytometry showed that AC133-2 expression was completely lost in cultured basal cells after confluent keratinocytes were grown in basal medium containing low serum for 6 days (Fig. 6B, right panel). This was confirmed by indirect immunofluorescence staining with anti-AC133 mAb (data not shown).

Discussion
In this study we have identified and characterized a second isoform of AC133. As a result of mRNA alternative splicing, the novel isoform, AC133-2, differs from the previously identified isoform, AC133-1, by absence of exon 3. As a consequence, AC133-2 encodes 856 amino acid residues with a deletion of 9 amino acids in the N-terminal extracellular domain.
How might loss of the 9 amino acids affect the structure and function of AC133-2? The nine amino acid deletion does not interfere with the signal peptide, asparagine-linked glycosylation sites, or any of the five transmembrane domains. Similar to the endogenous human AC133-1 antigen (1) and mouse prominin-2 (12), human AC133-2 can localize to the plasma membrane, indicating a possible role in cell-cell interactions or ligand receptor interactions. Whether the two isoforms are functionally redundant or serve distinct functions remains unclear, as little is known about their biological functions. Nevertheless, since both AC133 isoforms contain 5 tyrosine residues in the cytoplasmic C-terminal domain, the biological functions may depend on an interaction with an unknown ligand(s) as well as intracellular proteins, and thereby relay AC133 isoform-specific signaling. We speculate that the deletion in the AC133-2 isoform may affect ligand binding affinity and specificity, and therefore delineate its biological function from AC133-1.
The mRNA distribution profiles of AC133-1 and -2 suggest distinct roles for the two isoforms in development and mature organ homeostasis. Whereas AC133-2 was expressed ubiquitously, AC133-1 was not detectable in fetal liver and kidney, adult pancreas, kidney and placenta.
AC133-1 was strongly expressed in fetal brain but not in adult brain tissue, implying a possible role for AC133-1 in fetal brain development. AC133-2 was more abundant than AC133-1 in fetal skeletal muscle and heart but much less than AC133-1 in adult counterparts, suggesting AC133-2 may be involved in fetal development in these tissues. In addition, we found that AC133-2 was strongly expressed in four poorly differentiated human tumors, including lung carcinoma LX-1, pancreatic adenocarcinoma GI-103, colon adenocarcinoma CX-1 and breast carcinoma GI-101, but not in prostatic adenocarcinoma PC3, lung carcinoma GI-117, ovarian carcinoma GI-102, and colon adenocarcinoma GI-112. Variable expression of AC133 isoforms in tumors may be associated with the degree of differentiation of the tumor cells.
In vivo, AC133 antigen was first detected in hematopoietic CD34 bright cells using a mAb raised against CD34 + cells from human blood (1). AC133-1 cDNA was originally cloned from a retinoblasoma cell line WERI-RB-1, and has been the only reported transcript encoding AC133 antigen (2). Here, for the first time, we demonstrate that AC133-2 transcript is predominant in hematopoietic tissue, while AC133-1 is the predominant transcript in WERI-RB-1 cells (Fig. 3).
Based on abundant evidence in the literature (2,3,4,5,6), it appears that the AC133-2 isoform has been the antigen recognized in many Ab-based selection protocols to isolate hematopoietic stem or progenitor cells.
In addition, we show that AC133-2 is expressed in human stem cell niches other than hematopoietic tissue. About 10% basal keratinocytes in foreskin epidermis co-express AC133-2 and β1 integrin subunits, consistent with previously reported numbers of stem cells in human skin epidermis (14). We showed that isolated AC133-2 + /β1 integrin + cells from the epidermal basal layer can self-renew, differentiate in culture, and that AC133-2 is lost as the cells differentiate into non-stratified and stratified involucrin-expressing keratinocytes. These results demonstrate that AC133-2 serves as a marker of undifferentiated cells. Previously, isolation of skin epidermal stem cells relied solely on β1 integrin as a marker, which cannot efficiently differentiate stem cells and other basal cells due to subtle difference in their integrin levels (20). Therefore, AC133- The expression pattern of AC133-2 suggests that stem cells/ progenitor cells are present in more human fetal and adult tissues than previously thought. Accumulating evidence supports the concept that the utility of AC133 antigen as a stem cell marker is not restricted to hematopoietic tissue (8,21). Recently, Uchida et al. reported the isolation of human central nervous system-stem cells (hCNS-SC) by using a monoclonal antibody to AC133 antigen and other specific markers for hCNS-SC (8). Our study supports that AC133 antigen is an excellent marker for stem and progenitor cells when used in combination with other tissue and cell specific markers. The AC133-2 isoform may be of significance in the analysis and isolation of stem and progenitor cells from specific tissues and thereby facilitate functional characterization and application to tissue engineering and regenerative medicine.
In conclusion, we have isolated and characterized a novel isoform of the AC133 antigen. Our findings clarify the sequence identities of the two isoforms on the surface of stem cells and progenitor cells. The discovery of AC133-2 isoform increases complexity of this novel family of 5-transmembrane cell surface glycoproteins. Further investigation is needed to unmask the biological significance of these two AC133 isoforms.