Identification of Pactolus, an Integrin β Subunit-like Cell-surface Protein Preferentially Expressed by Cells of the Bone Marrow*

We have sought to develop methodologies to identify genes that are preferentially expressed during the differentiation of mast cells from their hematopoietic stem cell precursors. By using a modified differential display protocol, we compared a subset of transcripts expressed in bone marrow cells differentiated into immature mast cells with the exogenous addition of stem cell factor (SCF) or interleukin 3. One gene was identified that was preferentially expressed in the SCF-derived cells and encodes a novel murine integrin β subunit-like molecule, dubbed Pactolus-1 (Pactolus). Two distinct forms of Pactolus mRNA were detected which, via alternative splicing, are predicted to encode a membrane-bound form and truncated version of the protein. The full-length Pactolus gene product is very similar to a number of β subunit integrin chains, particularly β2, with the notable exceptions of the apparent deletion of the metal-binding site within the putative metal ion-dependent adhesion site-like domain of the Pactolus gene product and a cytoplasmic domain that shares no obvious homology to similar domains of the other β subunit integrin proteins. Although the Pactolus sequence was first identified in immature mast cell samples, screening of murine tissues indicated the highest level of Pactolus expression was found in the bone marrow, suggesting that the expression of Pactolus is confined to immature and maturing bone marrow-derived cells, and that the SCF-derived mast cells are more representative of this state than are the interleukin 3-derived mast cells. Immunoprecipitation of Pactolus revealed a cell-surface protein with an apparent molecular mass of about 95 kDa. Surprisingly, no associating α integrin subunit could be identified suggesting that either Pactolus does not associate with another integrin subunit or the association is too weak to be identified. These data suggest that Pactolus represents a gene and gene product related to those of the integrin β subunits but whose function(s) may be quite distinct from those of the integrin β subunits.

We have sought to develop methodologies to identify genes that are preferentially expressed during the differentiation of mast cells from their hematopoietic stem cell precursors. By using a modified differential display protocol, we compared a subset of transcripts expressed in bone marrow cells differentiated into immature mast cells with the exogenous addition of stem cell factor (SCF) or interleukin 3. One gene was identified that was preferentially expressed in the SCF-derived cells and encodes a novel murine integrin ␤ subunit-like molecule, dubbed Pactolus-1 (Pactolus). Two distinct forms of Pactolus mRNA were detected which, via alternative splicing, are predicted to encode a membrane-bound form and truncated version of the protein. The fulllength Pactolus gene product is very similar to a number of ␤ subunit integrin chains, particularly ␤2, with the notable exceptions of the apparent deletion of the metal-binding site within the putative metal iondependent adhesion site-like domain of the Pactolus gene product and a cytoplasmic domain that shares no obvious homology to similar domains of the other ␤ subunit integrin proteins. Although the Pactolus sequence was first identified in immature mast cell samples, screening of murine tissues indicated the highest level of Pactolus expression was found in the bone marrow, suggesting that the expression of Pactolus is confined to immature and maturing bone marrow-derived cells, and that the SCF-derived mast cells are more representative of this state than are the interleukin 3-derived mast cells. Immunoprecipitation of Pactolus revealed a cell-surface protein with an apparent molecular mass of about 95 kDa. Surprisingly, no associating ␣ integrin subunit could be identified suggesting that either Pactolus does not associate with another integrin subunit or the association is too weak to be identified. These data suggest that Pactolus represents a gene and gene product related to those of the integrin ␤ subunits but whose function(s) may be quite distinct from those of the integrin ␤ subunits.
It has been the goal of many investigations to isolate those genes whose expression is intimately tied to hematopoietic cell differentiation and maturation (reviewed in Refs. 1 and 2). This goal has been approached using a variety of different experimental approaches. One of the most successful techniques used to isolate cell-specific gene products has been cDNA subtraction (3). This protocol has been used to identify and isolate a number of significant gene products, notably the T cell receptor genes (4,5). There are, however, several drawbacks to the subtraction technique when it is applied to study hematopoiesis. Most conspicuously, it requires relatively large amounts of mRNA and is labor intensive. A successful subtraction normally requires several rounds of hybridization and physical dissociation. This protocol also leads to the preferential enrichment of abundant mRNA species over the rarer species.
More recently, an alternative approach to subtraction was introduced in which mRNA species are randomly amplified from total RNA preparations. This technique, known as differential display (DD) 1 (6,7), is based on the theory that every mRNA in the cell can be amplified, via a cDNA intermediate, with a specific combination of a poly(T) containing anchoring primer and a random decamer oligonucleotide. This protocol allows for the rapid comparison of transcript species between two closely related cell types such as normal and transformed or quiescent and activated cells. By using this protocol the total complexity of transcripts within a cell can be displayed, thus achieving transcript saturation. Any transcript product that is specific for the cell type in question can, with this protocol, be identified, isolated, and sequenced. The advantages of this protocol are many and include the requirement of much less RNA, the simplicity of PCR amplification, and the ease of product resolution (6 -8). The primary difficulties with DD have been the large number of false positives generated and the requirement of closely matched cell types with which to compare.
Mast cells arise from the multipotent bone marrow stem cells. There are two types of tissue mast cells in the mouse (mucosal mast cells and connective tissue mast cells) that are believed to be derived from the same precursor cells. A variety of researchers have demonstrated that immature mucosal-like and connective tissue-like mast cells can be derived in vitro by culturing mouse bone marrow with either IL-3 or SCF, respectively (9 -12). The molecular events that drive the differentiation of these two cell types have remained an enigma despite the recent characterization of mast cell progenitor cells (13). Since these two cell types are very similar to one another and can, given the correct circumstances, shift their phenotype back and forth, they are ideal candidates for the utilization of DD to identify those gene products specific for each cell phenotype.
In this work, we describe a modification of the original differential display protocol designed to identify transcripts implicated in bone marrow maturation. This protocol not only allows for a more rapid progression of the protocol but also appears to increase the specificity of the products such that the generation of false positives is greatly diminished. By using this protocol we identified a transcript preferentially expressed in SCF-derived mast cells but absent in those derived with IL-3. The gene fragment identified in this protocol was used to screen a mast cell cDNA library from which an apparent fulllength transcript was obtained. This cDNA predicts a novel murine integrin ␤ subunit-like molecule, Pactolus, that, via alternative splicing, would be expected to produce both a membrane-bound and truncated form. The expression of this gene is most pronounced in the murine bone marrow. The expression of Pactolus in cells derived in SCF but not the IL-3 cells suggests that the SCF-derived cells may represent a more immature mast cell type than those derived in IL-3 culture.

EXPERIMENTAL PROCEDURES
Mice and Tissue Culture-5-wk-old female NIH(s) mice were obtained from the National Institutes of Health. Mice were used at 5-10 weeks of age. Bone marrow-derived mast cells cultured in IL-3 or SCF were produced as described previously (14).
RNA Preparation and cDNA Synthesis-Total RNA from various cells and tissues was isolated using the CsCl/guanidine method (15). cDNA was synthesized by mixing 5 g of RNA, 10 l of 5ϫ first strand buffer, 5 l of 5 mM dNTP, 5 l of 0.1 M dithiothreitol, 2 l of perspective anchoring primer (6, 7), 2 l of Moloney murine leukemia virus reverse transcriptase, and water to a final volume of 50 l. The reaction mixture was incubated at 37°C for 1 h. 2 l of DNase-free RNase (1 mg/ml) was then added, and the reaction mixture was incubated for an additional 5 min followed by phenol/chloroform extraction and ethanol precipitation.
Differential Display PCR Amplification-PCR amplification was done by mixing 200 ng of cDNA (reversed transcribed with anchoring primers), 1 l of 500 M dNTP, 1 l of 10ϫ PCR buffer, 0.5 l of 1 g/l decamer, 0.15 l of Taq DNA polymerase (Life Technologies, Inc.), 0.25 l of [ 32 P]dCTP, and water to a total volume of 10 l per reaction. Each reaction was done in triplicate. The reaction mixture was then put into a capillary tube (Idaho Technology) and amplified under the following conditions. For the first five cycles, the amplification was carried out at 94°C for 1 s, 40°C for 1 s, and 72°C for 10 s. The following 35 cycles were carried out at 94°C for 1 s, 50°C for 1 s, and 72°C for 10 s. The reaction was then quenched by adding 10 l of stop buffer, and then 5 l of the reactions was loaded onto a 6% sequencing gel. After 2 h of electrophoresis, the gel was dried followed by autoradiography.
Primers used in this study have been described previously (6 -8) and are shown in the following: anchoring primer, 5Ј TTT TTT TTT TTT (ACG)G 3Ј and random decamer, 5Ј TGGATTGGTC 3Ј.
Bands containing candidate DNA fragments were eluted with 450 l of water and 25 l of 5 M NaCl overnight. The supernatant was recovered followed by ethanol precipitation. The DNA pellet was resuspended in 20 l of water. 2 l was then used for the second round amplification.
Cloning and Sequencing-Purified DNA fragment was ligated into pmm5 vector (a kind gift from Dr. Eric Kofoid, University of Utah) which has a single dT overhang on both ends. DNA sequence was generated by standard fluorescence DNA sequencing in the University of Utah core facility.
cDNA Library Preparation-Total cellular RNA was extracted from bone marrow-derived CTMC as described previously (15). Poly(A) ϩ RNA was isolated by using the Oligotex mRNA Kit (Qiagen). cDNA library was constructed according to the instruction manual of lambda ZAP II Cloning Vector Kit (Stratagene). Greater than 500,000 independent clones were screened for gene-specific inserts.
In Vitro Transcription and Translation-In vitro transcription was performed according to the manufacturer's manual of RNA Transcription Kit (Stratagene). mRNA was then translated in vitro using the rabbit reticulocyte lysate system (Amersham Pharmacia Biotech) in the presence of 35 S-labeled methionine.
Generation of Pactolus-specific Antisera-Two rabbits were injected with a peptide derived from the Pactolus cytoplasmic domain (CGTQ-KAAKLPRKG) using the keyhole limpet hemocyanin/bovine serum albumin conjugation protocol and reagents from Pierce. The first injection was done with an equal volume of Freund's complete adjuvant, whereas the subsequent injections were with Freund's incomplete adjuvant. Antisera from one of the rabbits was utilized.
Immunoprecipitation-Bone marrow cells were harvested from mice, and red blood cells were lysed with red blood cells lysis solution (0.15 M NH 4 Cl, 1.0 mM KHCO 3 , and 0.1 mM EDTA, pH 7.2) for 5 min at room temperature. EL-4 cells were maintained in RPMI media supplemented with 5% fetal bovine serum. Before labeling, cells were washed two times with phosphate-buffered saline. Cells were labeled by incubating with 0.5 mg/ml sulfo-NHS-LC-biotin (Pierce) for 30 min at room temperature. After washing with ice-cold phosphate-buffered saline three times, cells were resuspended in lysis buffer (0.5% Nonidet P-40, 0.5% sodium deoxycholate 50 mM NaCl, 2 mM CaCl 2 , 25 mM Tris, pH 7.5, 1% bovine serum albumin, 0.2 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, and 1 g/ml pepstatin A) at 5 ϫ 10 7 cells/ml, and the reactions were incubated on ice for 1 h. Lysates were immunoprecipitated with either polyclonal rabbit antisera against Pactolus cytoplasmic peptide or anti-mouse ␤2 monoclonal antibody (PharMingen). After incubating the lysates with respective antibody for 1 h at 4°C, the protein-antibody complex was absorbed with either protein A-Sepharose (for rabbit antibody) or protein G-Sepharose (for rat antibody). After the absorption, the Sepharose beads were washed 4 times as follows: once with lysis buffer, twice with lysis buffer with 150 mM NaCl, and once with 0.05 M Tris, pH 6.8. The samples were boiled in 1ϫ SDS loading buffer for 5 min before they were loaded on to 10% SDSpolyacrylamide gel. After the electrophoresis, proteins were then transferred to a polyvinylidene difluoride membrane (Millipore). The membrane was then incubated with 1:10,000 dilution of horseradish peroxidase-avidin D (Vector) for 1 h at room temperature. The biotinlabeled protein was visualized by chemiluminescence (MEN TM Life Science).

Isolation of Differentially Expressed Mast Cell Products-To
differentiate transcript products specific for mast cells derived in either IL-3 or SCF, we utilized the differential display protocol. This protocol was altered for use in the air thermocyler which differs from conventional PCR thermocyclers in reaction volumes (10 l versus 50 -100 l in conventional machines), cDNA requirements (100 ng versus 1 g or more for conventional machines), cycle times (23 s per cycle versus 2-6 min in conventional machines), and reagent vessels (glass capillary tubes versus plastic Eppendorf tubes in conventional machines). The specifics of the modified DD protocol are detailed under "Experimental Procedures." The RNA to be analyzed was isolated from four different mast cell sources. Bone marrow cells were cultured in the presence of either IL-3 or SCF for 21 days to develop into mast cells phenotypically similar to immature mucosal (MMC) or connective tissue mast cells (CTMC), respectively. RNA was isolated from such cells. Additionally, RNA was isolated from IL-3-derived cells that had been transferred into SCF without IL-3 for 24 h (MMC ϩ SCF) and SCF-derived cells transferred into IL-3 without SCF for 24 h (CTMC ϩ IL-3). The isolated total RNA was reverse-transcribed with a specific anchoring primer and expanded by PCR-based amplification with ran-domly designed decamers. Each sample was amplified independently in triplet (three 10-l reactions) and resolved by denaturing acrylamide gel electrophoresis (Fig. 1). As shown, this procedure provides consistent and reproducible band patterns comparable with those obtained from the conventional DD protocols. The two products found to be differentially expressed in this experiment with a single decamer/primer oligonucleotide set are marked by the arrows A and B.
To identify the gene products denoted by the DD products A and B, the candidate DNA fragments were eluted from the gel slice, re-amplified with the same decamer/primer oligonucleotides, and resolved within a second sequencing gel side by side with the original amplification reaction. This secondary amplification gave rise to several DNA fragments with varying mobility (Fig. 2). The correct fragment band (marked A or B) was identified by its alignment to the original amplification products and re-excised from the gel. The fragment was then reamplified and cloned into a T overhang plasmid for DNA sequencing (see "Experimental Procedures").
The DNA sequence obtained from the A and B fragments was used to search GenBank TM for homologous sequences. The sequences derived from these fragments did not match any sequence in GenBank TM . Additional analyses of the fragment A sequence are in progress and will not be reported here; however, they were clearly not derived from the same gene as fragment B. The sequence from fragment B was utilized to design nested primers to confirm the specificity of expression and to generate a probe with which to screen a murine mast cell cDNA library.
Identification of a Novel Integrin ␤ Subunit-like Sequence-When the nested primers derived from the fragment B sequence were used to re-analyze the primary mRNA samples (using the standard RT-RPCR protocol) the expression of the fragment B gene was primarily confined to the SCF-derived cells, not those initially derived in IL-3 (Fig. 3). Accordingly, a cDNA library constructed with RNA isolated from mast cells derived in SCF was screened with the fragment B probe. Twelve clones were isolated from this screening. The sequence of the largest of these cDNA clones possessed an insert with 2,585 nucleotides. A single large open reading frame was determined to start at an ATG at base 64 and terminate at a stop codon at base 1686, encoding a protein with 540 amino acids. This sequence thus predicted a long 3Ј-untranslated sequence of 901 nt. This predicted amino acid sequence suggested a secreted protein possessing a signal sequence but lacking a transmembrane domain for membrane anchoring (see below). A GenBank TM search with this cDNA sequence (and its derived amino acid sequence) indicated it was very similar to those within the ␤ integrin subunit family of genes. The highest degree of homology was found with the murine and human ␤2 integrin subunits. The gene and gene product was named Pactolus-1.
Alternative Pactolus Transcripts Predict a Membrane-bound and Truncated Form of the Protein-When the nucleotide sequence of Pactolus was plotted against that for murine ␤2 (18) using a homology matrix analysis, two features were striking. First, two distinct gaps were noted between the two sequences ( Fig. 4) that denoted apparent insertions in the ␤2 sequence FIG. 1. Differential display analysis of two distinct bone marrow-derived mast cell types: resolution of differential display products. An example of the DNA pattern generated by rapid capillary PCR-based mRNA differential display. PCR parameters are described under "Experimental Procedures." CTMC represents mast cells grown exclusively in SCF for 22 days; CTMCϩIL-3 represents mast cells grown exclusively in SCF for 21 days followed by 24 h in IL-3 alone; MMC represents mast cell grown exclusively in IL-3 for 22 days; MMC ϩ KL represents mast cells grown exclusively in IL-3 for 21 days followed by 24 h in SCF (KL) alone. Arrows A and B indicate two differentially expressed products.
FIG. 2. Secondary analysis of differential display products: isolation of differentially displayed products. Eluted DNA products A and B, shown in Fig. 1, were re-amplified with the same primers used in the original amplification reaction and compared with the original differential display products. Products corresponding to the size of the original fragment were then re-isolated for further characterization. Contaminant refers to inappropriate DNA bands contaminating the original fragment isolate.
FIG. 3. Confirmation of specificity of differential display products. Total cDNA samples derived from the same mast cell RNA samples described above were amplified using primers designed according to the fragment B sequence (5Ј TGG AGG AAG CAT GGT TTG CTG 3Ј; 5Ј ATA GGT CCT CAA AGT AAC GTC 3Ј). RT-RPCR reactions were performed as described under "Experimental Procedures." Shown at the bottom are ␤-actin transcripts utilized as amplification and loading controls. ␤-Actin products were amplified for 16 cycles while the fragment B products were amplified for 26 cycles. compared with that of Pactolus. And second, a significant level of homology remained between the two sequences after the putative stop codon of the Pactolus sequence (base 1692). This region corresponded to the 3Ј-untranslated sequence of the Pactolus transcript and extracellular coding sequences (aminoterminal of the transmembrane sequence) of the ␤2 integrin. DNA sequence analysis of the other Pactolus cDNA clones isolated from the CTMC cDNA library confirmed the sequence of the Pactolus-1 cDNA at these sites (data not shown).
The presence of these unexpected gaps within the Pactolus sequence suggested there may be different isoforms of the Pactolus gene product(s) created by alternative splicing events. We therefore established a RT-RPCR assay in which the region of the Pactolus coding sequence which included these gaps could be evaluated for alternative RNA products. Oligonucleotides were generated which flanked the two gaps and were used to analyze RNA isolated from murine bone marrow and spleen, two tissues we suspected would possess Pactolus transcripts. As shown in Fig. 5, the Pactolus-specific primers spanning gap 1 only produced a single product of the expected size. This product was excised from the gel, cloned, and sequenced and was found to be identical to that predicted from the Pactolus sequence.
The oligonucleotides spanning gap 2 generated two distinct products differing in size by about 40 bp (Fig. 5). The size of the smaller of the two was predicted from the Pactolus cDNA sequence. Both fragments were excised from the gel, cloned, and sequenced. As expected, the shorter of the two was identical to the Pactolus sequence. The larger fragment possessed the expected Pactolus sequences plus a 43-bp insertion (Fig. 6,  panel A). When this additional sequence was inserted within the Pactolus coding sequence, the reading frame was altered such that the predicted stop codon was shifted to base 2,286, predicting the inclusion of an additional 223 amino acids. These additional amino acids encode a 23-amino acid transmembrane domain and a cytoplasmic tail of 43 amino acids (Fig. 6, panel B). Thus the full-length Pactolus would predict a protein with transmembrane and cytoplasmic domains. The truncated form would predict a premature termination of the protein.
The insertion of the 43-bp alternative sequence into the Pactolus coding sequence would predict a protein similar to that of the integrin ␤ subunits. This prediction allowed for a more complete amino acid homology comparison between the mouse Pactolus gene product and that of the most similar integrin ␤ subunit, murine ␤2 (18). As shown (Fig. 6, panel C) there are regions of very high amino acid homology between these two proteins. This homology is lost toward the carboxyl terminus of the protein and is not evident within the transmembrane and cytoplasmic domains. The cytoplasmic domain of Pactolus does not demonstrate significant homology with any of the eight characterized ␤ subunit integrin chains (data not shown) (18 -25). All characterized ␤ subunit integrin chains have been proposed to possess a MIDAS motif (metal ion-dependent adhesion FIG. 5. RT-RPCR analysis with Pactolus-specific primers predicts two discrete gene products. Pactolus-specific PCR primers were designed to span gap 1 and gap 2 on the cDNA sequences. Standard RT-RPCR was performed as described under "Experimental Procedures." Shown in the left panel is the result from primers spanning gap 1, and to the right is the result from primers for gap 2. For each panel, ␤-actin was amplified for 16 cycles, and Pactolus transcripts were amplified for 26 cycles. Two distinct PCR bands are readily identifiable with primer set spanning gap 2. The lower band agrees with the predicted size from the Pactolus-1 cDNA sequence. The upper band is approximately 40 bp larger than the lower form. site) at about residue 130 (Fig. 6, panel D). This site, characterized in the CD11b protein (26), consists of a conserved DX-SXS sequence followed by a threonine (about 65 aa carboxyl to the DXSXS sequence) and an aspartate residue (about 25 aa carboxyl of the threonine residue). This site within the integrin ␤ subunit has been demonstrated to be critical for ligand binding (27), ␣ chain association (at least two mutations in the integrin ␤2 chain at this site are responsible for the lymphocyte adhesion deficiency syndrome and are underlined in the ␤2 sequence) (28), and Mg 2ϩ ion binding (conversion of the aminoterminal aspartate of the MIDAS motif to alanine abolishes metal binding at this site) (29 -31). All of the eight known ␤ integrin subunits demonstrate sequence conservation of this site, as demonstrated by the comparison between the mouse integrin subunit ␤1 and ␤2 sequences (Fig. 6, panel D). Interestingly Pactolus does not because it lacks the critical aspartate residue as part of the DXSXS sequence. Additionally the spacing between the SXS site to the threonine is altered (gap 1) such that this region is approximately 21 amino acids shorter than the consensus. Since the proposed MIDAS site is formed by the ternary association of these conserved amino acid residues (26), it is likely this site in the Pactolus chain will not FIG. 7. Different forms of Pactolus transcripts are generated via the use of alternative acceptor sites. Panel A, genomic clones possessing the Pactolus gene were obtained by screening a mouse genomic DNA library. The subfragment possessing the exons encoding the gap 2 sequences was cloned and analyzed. DNA sequence was obtained by standard fluorescent sequencing. The 5Ј exon is noted followed by an intron (shown in italics). Gaps were introduced between the intron and exon sequences to demarcate the sets of sequences. A well conserved splice donor site is found at the end of the 5Ј exon (CAG GTG). Two splice acceptor sites are found either within the intron (CAG AAA) or at the 3Ј end of the insert sequence (underlined) (CAG CTT). The insert of 43 bp found within the "larger" DNA sequence shown in Fig. 7 is underlined. Panel B, a schematic representation of the Pactolus alternative splicing event showing the use of the alternative splice acceptor sites to generate the two different isoforms of Pactolus. the murine integrin subunit ␤2 sequence (18). The vertical bar indicates identical amino acids, and the dotted line represents sequences missing in the Pactolus sequence (gap 1) to conserve alignment. These two proteins are 63% identical to each other for the region shown; there is no significant sequence homology between the two following residue 631 of the Pactolus sequence. Panel D, sequence alignment between the murine Pactolus, ␤2 and ␤1 integrin subunits in the putative MIDAS region (26) starting at residue 134 of Pactolus. The consensus sequences for the MIDAS motif is shown below the murine sequences. Spacer amino acids are shown either as X (DXSXS) or by sequence length (12 aa, etc). Residues demonstrated to be mutated in the human ␤2 sequence and thus causative for the lymphocyte adhesion deficiency syndrome (28) are underlined. The underlined residues are shown to demonstrate the break point between these two predicted sequences due to the alternative splicing event. Panel C, sequence alignment between the predicted Pactolus protein sequence (Pact) and function in an analogous fashion to that of the integrin ␤ subunits.
The inclusion of this 43-bp sequence via alternative splicing could occur either by the inclusion of a small independent exon or via the use of alternative acceptor/donor sites within a single exon. To determine which of these mechanisms was utilized for the generation of the different Pactolus isoforms, the Pactolus gene sequence was used to screen a 129/sv genomic library to isolate genomic sequences encoding Pactolus. The region of the gene encoding these alternatively spliced sequences was cloned and sequenced (Fig. 7, panel A). This sequence indicated the presence of two different acceptor sites that could be used to create the two transcript products (Fig. 7, panel B). Thus the utilization of the most 5Ј acceptor site would allow for the inclusion of the 43-bp sequence (termed insert) creating the membrane-bound form of the protein, whereas the use of the 3Ј acceptor site would generate the truncated form of the protein.
mRNA Expression of Pactolus-As shown in Figs. 3 and 5, Pactolus transcripts are evident in bone marrow-derived mast cells (cultured in SCF) and in bone marrow and spleen. Due to the extreme sensitivity of RT-PCR, the expression of Pactolus transcripts in the bone marrow was further confirmed by Northern blot analysis. As shown in Fig. 8, panel A, a 2.8 -3.0kilobase pair band was evident when 5 g of total RNA from mouse bone marrow was hybridized with the Pactolus cDNA sequence. The size differential between the two Pactolus mRNA isoforms is too small to detect two distinct bands with this type of assay.
The tissue distribution of Pactolus transcripts was further analyzed by using a semi-quantitative RT-RPCR assay. These assays were done with the oligo set which discriminates between the truncated and full-length Pactolus transcript (thus covering gap 2) and were done at low cycle number to inhibit saturation of the product. As shown in Fig. 8, panel B, the mouse bone marrow clearly possesses the highest level of Pactolus transcripts. The high amount of transcripts for the truncated form (compared with the full-length product) may be due to the competitive advantage of the smaller fragment over that of the larger fragment during PCR amplification. When such an analysis was done with higher cycle numbers, the presence of Pactolus transcripts in the spleen and lung samples was evident (data not shown).
To compare the expression level of Pactolus with the expression of other genes whose products are known to be influential in bone marrow development, RT-RPCR analysis was done using oligo sets specific for Pactolus, integrin subunits ␤1, ␤2, ␣L, ␣M, and L-selectin. The Pactolus-specific product is that which would only be found within the transcript encoding the full-length protein product. All of the reactions were done for the same number of cycles (and produce products of approximately the same size) for this comparison. As shown in Fig. 8, panel C, Pactolus is expressed at a lower level than ␤1, ␤2, and ␣M but in comparable quantities as those seen for ␣L and L-selectin. These data suggest that the full-length Pactolus FIG. 8. Pactolus transcript expression. Panel A, 5 g of total RNA from bone marrow was analyzed by denaturing Northern blot. The RNA was transferred onto a nitrocellulose membrane before it was hybridized with a Pactolus cDNA probe. The positions of the 28 S and 18 S RNA are shown. Panel B, pactolus tissue expression. Total RNA was prepared from a variety of mouse tissues. RT-RPCR was performed with a Pactolus-specific primer set (24 cycles) which spans the gap 2 on the cDNA sequence. With the exception of bone marrow samples, all the tissues have low level of expression of both forms of Pactolus transcripts, which was evident after prolonged exposure to x-ray film. Actin samples (16 cycles) were analyzed to ensure equivalent quantities of cDNA were amplified. The 1st lane (next to bone marrow) was a water PCR control (no cDNA added). Panel C, comparison of the full-length Pactolus expression with other adhesion molecules. Oligo sets were designed against cDNA sequences of a variety of adhesion molecules. The oligo set for Pactolus only recognizes the fulllength form (for details, see "Experimental Procedures"). Bone marrow cDNA sample was amplified for 26 cycles for all reactions. Panel D, Pactolus expression by maturing IL-3-derived mast cells. Bone marrow cells were cultured in the presence of IL-3 over a period of 34 days. Cell aliquots were taken from the culture at different time points (as shown in the figure) and RNA derived. Pactolus transcripts (24 cycles) were visualized using the primer set specific for gap 2; actin controls (16 cycles) were included to ensure equivalent quantities of cDNA were amplified. cDNA derived from total bone marrow was included as a positive control: water (H 2 O) was a negative control (no cDNA added). gene product is expressed in a quantity that is physiologically significant.
The high level of expression of the Pactolus transcript in the marrow sample compared with the spleen suggested its expression is highest in immature bone marrow-derived cells. Our previous screening of Pactolus expression by the IL-3 and SCFderived mast cells was done after a minimum of 21 days in culture. To determine if the maturation of the IL-3-derived cells led to the loss of Pactolus expression, we screened RNA samples derived from mast cells cultured in IL-3 for a variable length of time. As shown in Fig. 8, panel D, the expression of Pactolus is progressively lost as the mast cells mature with IL-3 until, after 24 -30 days in culture, the transcript is almost undetectable. Similar experiments analyzing variably aged SCF-derived cells show a constitutive level of expression of Pactolus at about 20% the level of total bone marrow (compared with a ␤-actin control) (data not shown). Thus Pactolus expression is maintained in the long term SCF-derived mast cell cultures but is down-regulated in those cells induced to differentiate into mast cells via IL-3.
Identification of the Pactolus Protein-To analyze the protein product of the Pactolus gene, a peptide corresponding to the Pactolus cytoplasmic tail was synthesized, and polyclonal rabbit antisera were generated. To determine whether the antisera recognized Pactolus protein, 35 S-labeled protein was generated by in vitro translation of Pactolus mRNA in the presence of [ 35 S]methionine and immunoprecipitated with either preimmune or post-immune serum. As shown in Fig. 9, panel A, only the post-immune serum specifically recognized the Pactolus gene product. To identify Pactolus protein in vivo, mouse bone marrow cells were surface-labeled with biotin, and the lysate was immunoprecipitated with either preimmune or post-immune serum. As shown in Fig. 9, panel B, a 95-kDa protein was readily identified by this approach. However, in identically labeled splenocytes, this protein was not evident (data not shown). The increase in apparent molecular weight in the Pactolus gene product between the two panels is presumably due to post-translational modification of the protein in the bone marrow sample (i.e. glycosylation) and charge differences via the addition of the biotin complexes.
All integrin ␤ subunits are known to form a heterodimeric complex with an integrin ␣ chain. Interestingly, no associating subunit was evident with Pactolus in Fig. 9, panel B, despite that, under identical labeling and lysis conditions, the ␣ subunits associated with the ␤2 subunit can be co-immunoprecipitated with anti-␤2 antibody from EL-4 cells. Bone marrow samples have been biotin-labeled and -treated using a variety of lysis and immunoprecipitation conditions utilizing a variety of stringencies. Under no conditions have we been able to detect another protein associated with Pactolus that would suggest an ␣ integrin subunit (data not shown). In addition, radioiodination analyses of similar samples have detected only the single Pactolus protein following low stringency immunoprecipitation of bone marrow cells (data not shown). These data thus sugggest that either Pactolus does not require a partner to be expressed on the cell surface or that the association of Pactolus with another subunit is so weak that the partner cannot be resolved using the standard approaches.

DISCUSSION
In this report, we describe the isolation of a novel murine gene (Pactolus) that shares a high degree of homology with the family of adhesion molecules known as the integrin ␤ subunits. The Pactolus sequence was obtained from a differential display (DD) analysis of murine bone marrow-derived mast cells in which transcripts derived from cells cultured in IL-3 were compared with those derived in SCF. Based upon our data and data from others (11,14,32), those cells derived solely in IL-3 possess a phenotype associated with mast cells found in the intestinal mucosa, whereas those derived in SCF possess characteristics of mast cells found in the skin and peritoneal cavity of the animal. Following this cell culture rationale and utilizing a modified DD protocol, we isolated a novel sequence from bone marrow cells cultured in SCF which was lacking, as an apparent transcript, from those cells cultured in IL-3. Subsequent cloning and sequence analysis indicated this novel gene transcript was very similar to the members of the integrin ␤ subunit gene family.
The protein predicted by the Pactolus sequence is related to the integrin ␤ subunits but is clearly divergent in two important domains. The first of these is within the proposed I (inserted) domain-like structure (also termed A for activation domain) which is also present within the majority of the ␣ integrin subunits (reviewed in Ref. 33). This region resides at the amino terminus of the protein (thus extracellular) and is implicated in ligand binding, heteroduplex formation, and metal ion binding. Single amino acid mutations in this site abrogate stable expression of the integrin heterodimers and block ligand binding. Structural analysis of the I domain of the CD11b chain of the CR3 integrin complex suggested the presence of a MIDAS motif (metal ion-dependent adhesion site) that was formed in a three-dimensional fold utilizing a conserved DXSXS-(65 amino acids)-T-(25 amino acids)-D sequence, where the X residues and spacer amino acids are not conserved (26). The analagous sequence present in Pactolus differs from that of the ␣ and ␤ integrin subunits. Not only does Pactolus lack the amino-terminal Asp residue, but the spacing between the Ser and Thr residues, which presumably is critical for the ternary folds of the protein, is 21 amino acids shorter than the consensus sequence. Of the more than 40 gene sequences that possess the proposed MIDAS motif, the shortest distance between the analogous Ser and Thr residues is 61 Bone marrow cells were surface-labeled with biotin and subsequently lysed in the presence of 0.5% Nonidet P-40 and 0.5% deoxycholate. Total cellular lysate was then immunoprecipitated with either preimmune or post-immune sera. EL-4 cells were treated identically, and the lysate was immunoprecipitated with monoclonal antibody against mouse ␤2 integrin subunit. The experiment using bone marrow cells is shown on the left, with Pactolus demonstrating an apparent molecular mass of about 95 kDa. The EL-4 sample is shown on the right; lines to its right indicate integrin ␤2 subunit and three different integrin ␣ subunits of the ␤2 integrin family. The thick band representing ␣L and ␣M could be differentiated into two bands with lighter exposure. amino acids, compared with 44 for Pactolus. This deletion would, with respect to the proposed MIDAS motif structure, apparently delete the ␣2,␣3 helices, thus placing the Thr residue in opposite orientation to the Mg 2ϩ ion within the MIDAS structure. These two alterations in this region of Pactolus suggest that this site may not be functionally similar to those of the integrin ␤ subunits.
The other site of high homology within the integrin ␤ subunit family is the cytoplasmic domain. Sequences have been defined in the cytoplasmic tail of these subunits that are implicated in the binding of the ␤ chain to cytosolic proteins including ␣-actinin, talin, paxillin, and others (reviewed in Ref. 34). Pactolus is lacking these conserved residues in its proposed cytoplasmic domain suggesting it may bind to a different set of cytoplasmic proteins than those described for the integrin ␤ subunits.
The Pactolus gene transcripts also differ from members of the integrin ␤ subunit family in predicting two distinct forms of the protein, plus or minus the transmembrane and cytoplasmic domains. Attempts to generate an antisera specific for the truncated form of the Pactulus protein have so far been unsuccessful; thus we cannot conclude whether the protein produced by the truncated transcript is stably expressed in mammalian cells. Some integrin ␤ subunits utilize alternative splicing to produce variant isoforms of the proteins. In particular, at least four distinct alternative cytoplasmic domains have been described for integrin subunit ␤1 which alter the functional characteristics of the protein (35)(36)(37)(38). However, we do not know of any ␤ integrin subunit that would, like Pactolus, predict a secreted form of the protein.
The expression of the Pactolus gene appears to be limited to immature cells of bone marrow derivation. The murine tissue demonstrating the highest level of expression is the mouse marrow. Based upon ␤-actin transcript levels, the quantity of Pactolus transcripts in the splenic sample was 10% or less than that of the bone marrow. Since the spleen is primarily populated by mature cells of bone marrow origin (B cells, T cells, and macrophages), the absence of appreciable Pactolus transcripts in the spleen may suggest the down-regulation of this gene during cellular maturation.
Two key findings were provided in the analysis of the Pactolus protein. First, the protein is expressed on the surface of the bone marrow cells with an apparent molecule mass of 95 kDa. Second, we cannot detect any significant association of Pactolus with any other cell-surface protein, as might be expected if Pactolus was to function as a ␤ integrin subunit. Based upon the sequence of the Pactolus gene product and the lack of an apparent heterodimer complex, it is likely that Pactolus does not function as the typical ␤ integrin subunit despite its sequence homology with ␤2. Therefore placing Pactolus within the integrin gene/protein family would imply a functionality of the protein that it may not possess.
The major question left open by this study is the function of Pactolus. Its expression pattern suggests it is expressed by immature and maturing cells of bone marrow derivation. Its similarity in structure to the integrin ␤ subunit gene family suggests it may be a receptor mediating adhesion of such cells within the marrow stroma. Previously the ␣4 integrin has been shown to be critical in marrow maintenance (39,40). Alternatively, Pactolus may be playing a signaling role for cells within the marrow. For example, one of the alternative cytoplasmic domains for the integrin ␤1 subunit (␤1c) acts to directly inhibit cell cycle progression (38). The marrow consists of many cell types, some of which are held in a state of low replicative activity. The two different forms of the Pactolus protein may, upon ligation with ligand, send two quite distinct signals into the cell. This model might be especially appropriate if a modulation of splicing between the two forms is evident during bone marrow cell maturation and if we can detect the stable expression of the truncated form of the Pactolus protein.