Generation of truncated brain AE3 isoforms by alternate mRNA processing

AE3 is a member of the AE gene family of anion exchang-ers (reviewed by Kopito (1990)) that is expressed primar-ily in brain neurons and heart (Kopito et al., 1989; Kudrycki et al., 1990). Like the other members of this family, which include the erythrocyte ‘band 3’ protein (AE1p) and a broadly distributed isoform, AE2p (reviewed by Kopito (1990)), AE3p catalyzes the transmembrane exchange of chloride and bicarbonate (Kopito et al., 1989; Lee et al., 1991). By virtue of this activity the AE proteins contribute to the maintenance of cellular ionic homeostasis and the regulation of intracellular volume and pH (Kopito, 1990). All known AE family members share a common structural organization consisting of an ∼600 amino acid COOH-ter-minal membrane-associated domain (which spans the mem-brane 12-14 times) and a NH₂-terminal domain of variable size (Kopito, 1990). The former domain is highly conserved among the AE family members in both primary sequence and predicted secondary structure. The 81% sequence iden-tity between the COOH-terminal domains of AE1p and AE3p is reflected in the functional similarity between the two proteins. Expression of the COOH-terminal 645 amino acids of AE3p in COS or human embryonic kidney (HEK) cells (Lee et al., 1991) leads to a measurable increase in Cl⁻/HCO₃⁻ exchange activity, demonstrating that this domain is sufficient to mediate anion exchange (Kopito et al., 1989).

By contrast, the function of the NH₂-terminal domain has been established only for AE1p (Bennett and Stenbuck, 1980). In the erythrocyte, the ∼43 kDa NH₂-terminal domain of AE1p faces the cytoplasm, where it interacts with several cytoplasmic proteins including hemoglobin, gly-colytic enzymes and ankyrin (reviewed by Low (1986)). This latter interaction is significant as it forms the primary linkage between the phospholipid bilayer and the subcorti-cal spectrin/actin cytoskeleton, which confers on erythro-cytes their characteristic morphological and rheological properties (reviewed by Bennett (1990)). The high-affinity (Bennett and Stenbuck, 1980) ankyrin-binding site has not been mapped on AE1p, although indirect evidence suggests that it is somewhat removed from the extreme NH₂ termi-nus (reviewed by Low (1986)). The cytoplasmic domain of AE3p shares significant (59%) sequence identity with the corresponding domain of AE1p, raising the possibility that AE3p may interact with components of the neuronal cytoskeleton or membrane skeleton. Such an association has yet to be established, although several ankyrin isoforms have been identified in neurons (Otto et al., 1991; reviewed by Bennett (1992)). We have recently demonstrated that, when co-expressed in transfected HEK cells, AE3p forms a stable complex with the repeat domain of ankyrin (Mor-gans and Kopito, 1993). The AE3p cytoplasmic domain is also distinguished by the presence of an ∼260 residue, pro-line-rich extension at the extreme NH₂ terminus. Thus, while the COOH-terminal domains of AE1p and AE3p are highly conserved in both sequence and function in anion exchange, the role of the more divergent NH₂-terminal, cytoplasmic domain of AE3p remains unknown.

The AE3 gene is expressed in brain and heart and, at much lower levels, in several other tissues including kidney, stomach, skeletal muscle and testes (Kopito et al., 1989; Linn et al., 1992). The predominant AE3 transcript in heart is ∼1 kb shorter than that in brain (Kopito et al., 1989;

Kudrycki et al., 1990) due to tissue-specific alternative pro-cessing of four exons at the 5′ end of the AE3 gene (Linn et al., 1992). Cardiac AE3p is therefore expected to differ from brain AE3p only at the extreme NH₂-terminal portion of the cytoplasmic domain.

Alternative mRNA processing is widely used by eukary-otic cells to increase the coding power of a gene, and can be used to generate proteins with quite different biological activities. We report here the identification of two novel AE3 transcripts, which arise from tissue-specific alternative mRNA processing. Both these sequences encode AE3 polypeptides that are truncated within the cytoplasmic domain. One of these truncated AE3 polypeptides is expressed in brain as a 74 kDa polypeptide that is present within a high molecular mass, detergent-insoluble complex, suggesting a possible association with the cytoskeleton.

A mouse B10A genomic library was screened with a HindIII/SmaI fragment from plasmid pRK128 (nucleotides 2128 to 2510) of the cDNA sequence (Kopito et al., 1989) for the 3′ end of the AE3 gene, and the entire insert of pRK133 (nucleotides 413 to 815) for the 5′ end of the gene. From 8.5×10⁵ plaques, 21 duplicate positive clones were found and of these, six were purified. A 2.1 kb XbaI restriction fragment of one of the clones was found by Southern blotting (Maniatis et al., 1982) to contain both alternately spliced sequences. This fragment was subcloned into a Bluescript pSK⁺ vector (Stratagene) and called pCM105. Nested deletions of the insert were generated using exonuclease III and S₁ nuclease and then sequenced with the dideoxy chain termination method as previously described (Kopito et al., 1989).

Total RNA and polyadenylated RNA were prepared as previously described (Kopito et al., 1989). S₁ nuclease mapping was per-formed as described by Kopito et al. (1987). Briefly, RNA sam-ples were hybridized with 10,000 to 50,000 cpm of ³²P-labeled cDNA probe and then digested with 75 units of S₁ nuclease per sample for 1 hour. Protected fragments were resolved by gel elec-trophoresis on a 6% acrylamide/7.5 M urea gel. The probe used to analyze splicing of the 14 base sequence (Fig. 2a) includes the last 24 bases of exon 4, the 14 alternate bases, and 220 bases of exon 5, plus an additional 100 bases of Bluescript vector at the 3′ end to give a total length of 358 bases. The probe was labeled at the 5′ end by treatment with alkaline phosphatase followed by kinasing in the presence of [γ-³²P]ATP (New England Nuclear). The probe used to analyze the 311 base fragment (Fig. 2b), includes 40 bases of intron b, all of exon B and 84 bases of the 311 base alternate sequence, plus an additional 156 bases of Blue-script sequence to give a total length of 460 nucleotides. This probe was uniformly labeled during synthesis by primer extension in the presence of [³²P]dCTP (New England Nuclear). The labeled probes were isolated by electrophoresis on a 6% acrylamide/7.5 M urea gel followed by electroelution from the gel. The Blue-script sequence attached to the end of each probe allowed for differentiation based on size between the fully protected probe and undigested probe.

Amplification using the polymerase chain reaction (PCR), was performed with the kit from Perkin Elmer Cetus Corp. The procedure enclosed with the kit was followed except that the con-centration of each oligonucleotide was 5 μm. Amplification was achieved with 30 cycles of 1 minute at 94°C, 1 minute at 65°C, and 1.5 minutes at 72°C. The PCR oligonucleotides span the 311 bp alternate intron. The sequence of oligo 1 is 5′-CCCAAGCTTGTGTTCGTGGAGCTGAATGA-3′ of which the first 9 nucleotides contain a HindIII restriction site, and the remaining 20 nucleotides match nucleotides 505 to 525 of the genomic sequence. The sequence of oligo 2 is 5′-CCCAAGCTT-TAGCCCTGTCCTCAGGCCGG-3′. The first 9 nucleotides of this sequence also contain a HindIII restriction site, and the remaining nucleotides are complementary to bases 1080 to 1100 of the genomic sequence. Amplified fragments were separated by electrophoresis on a 1.5% agarose gel and visualized by staining with ethidium bromide.

Antiserum 7210 was raised against an AE3/β-galactosidase fusion protein in New Zealand white rabbits as described by Kopito et al., 1989. It binds to all three forms of the AE3 protein. Antisera 7017 and 7015 react with the C termini of 311-AE3 and 14-AE3, respectively. Synthetic peptides corresponding to the last 11 amino acids of 311-AE3 or the last 12 amino acids of 14-AE3 were cou-pled to keyhole limpet hemocyanin (KLH), and to bovine serum albumin (BSA). New Zealand white rabbits were immunized with KLH-coupled peptides, then boosted once a month with alternat-ing injections of peptide-BSA and peptide-KLH. The sera were tested by ELISA and by immunofluorescent staining of COS-7 cells transfected with 311-AE3 or 14-AE3. Antibody 7210 was affinity-purified by incubation overnight at 4°C with nitrocellulose strips onto which FL-AE3 had been electrophoretically transferred fol-lowed by elution in 0.1 M glycine, pH 2.5, for 10 minutes at room temperature. The eluate was neutralized by the addition of 50 μl/ml 1 M Tris-HCl, pH 9.5, and then concentrated and dialyzed against PBS in a Centricon-30 (Amicon) that had been preabsorbed with 2% BSA. The antibody was stored in PBS/1% BSA/0.02% NaN₃ at 4°C. Antibody 7015 was affinity-purified on an ACA-22 Ultra-gel column (LKB) to which the 14-AE3 C-terminal peptide had been coupled. It was eluted with 0.1 M glycine, pH 2.5. The eluate was treated as described above for 7210.

COS-7 cells were grown and transfected in Dulbecco’s modified medium supplemented with 10% fetal calf serum in a humidified incubator at 37°C and in 5% CO₂. They were plated at a density of 5×10⁵ per 100 mm culture dish 18 hours prior to transfection, and were then transfected with 8 μg DNA per dish as previously described (Oprian et al., 1987). At 48 to 56 hours after transfec-tion, the cells were either fixed and stained for immunofluorescent microscopy, or processed for immunoblotting by dissolving in Laemmli sample buffer (Laemmli, 1970).

Brains from Sprague-Dawley male rats from SIM were homoge-nized in a Potter-Elvjener homogenizer in 10 ml H buffer (0.32 M sucrose, 10 mM Tris, pH 7.4, 1 mM EDTA, 0.1 mM EGTA) per brain with 10 strokes of a loose-fitting Teflon pestle at 500 rpm. Homogenate corresponding to 100 μg of protein was diluted to 100 μl in H buffer containing 1% Triton X-100. The mixture was incubated on ice for 30 minutes, then centrifuged for 15 min-utes at 10,000 g in a microfuge at 4°C. The supernatant and pellet were diluted to 200 μl in Laemmli sample buffer for western blot analysis.

Samples were electrophoresed on a 7.5% acrylamide/SDS gel (Laemmli, 1970), then electrophoretically transferred to a nitro-cellulose filter. The filter was incubated for 30 minutes in 150 mM NaCl, 10 mM Tris-HCl, pH 7.5, 0.05% Tween-20, 5% w/v nonfat dry milk (TBSTM) to block non-specific binding, then in TBSTM containing the primary antibody for either 2 hours at room tem-perature or 4°C overnight. Following one 30 second wash and three 15 minutes washes in TBST (TBSTM with no milk), the filter was incubated in donkey anti-rabbit antibody coupled to horseradish peroxidase (Amersham) diluted 1:2000 in TBSTM for 30 minutes at room temperature, and then washed as before. The filter was then incubated for 1 minute in equal volumes of the two ECL reagents (Amersham), pressed between two pieces of blot-ting paper to remove excess liquid, and exposed to Kodak XAR-5 film.

Screening of a mouse brain cDNA library for full-length AE3 clones (Kopito et al., 1989) resulted in the isolation of two additional classes of cDNA that are identical to full-length AE3, with the exception of insertions of 311 bp and 14 bp at positions 1325 and 1646, respectively. The novel clones were designated 311-AE3 and 14-AE3. To deter-mine whether these cDNAs were derived from the same AE3 gene, a 2.1 kb genomic subclone containing this region was isolated (Fig. 1A). Alignment of this sequence with that of the published FL-AE3 cDNA (Kopito et al., 1989) indicates that it contains 5 exons, designated A-E (Fig. 1A,B). These exons correspond to exons 7-11 of the rat AE3 genomic sequence (Linn et al., 1992). Each of these exons is flanked by consensus splice donor and acceptor sequences (Smith et al., 1989). Analysis of the genomic sequence revealed that the 311 bp insertion identified in several cDNA clones corresponds to an unspliced intron between exons ‘B’ and ‘C’. The 14 bp insertion arises from the use of an alternate splice acceptor site between exons ‘D’ and ‘E’ at the end of intron ‘e’ (Fig. 1B). We conclude that all three AE3 transcripts arise from a common AE3 gene by alternative mRNA processing.

Conceptual translation of the amino acid sequences encoded by 311-AE3 and 14-AE3 indicated that both inser-tions disrupt the AE3 reading frame (Fig. 1B). The 14 bp insertion causes a frame-shift resulting in termination at a TGA 16 codons downstream of the insertion site. The polypeptide encoded by 311-AE3 (311-AE3p) is predicted to terminate at a TAG 12 codons downstream of the unused splice donor site. Thus, by contrast to the 135 kDa polypep-tide predicted from FL-AE3 (FL-AE3p), 311-AE3 and 14-AE3 are predicted to encode polypeptides of 43 kDa and 56 kDa, respectively. These polypeptides, which correspond to the NH₂-terminal domain of AE3, are predicted to lack the highly conserved membrane-associated domain shared by all three members of the AE gene family. They are thus not predicted to be integral membrane proteins and are not likely to participate directly in ion transport.

The presence of both alternate sequences in AE3 mRNA was investigated by S₁ nuclease protection analysis (Fig. 2). A 358 bp probe containing the 14 bp alternate sequence was 5 ′ end-labeled and hybridized with RNA from mouse brain, heart and kidney prior to digestion with S₁ nucle-ase and resolution on a polyacrylamide gel (Fig. 2A). This probe contained 100 nt of vector sequence at the 3′ end in addition to the AE3 sequence (Fig. 2C), which permit-ted discrimination between the undigested probe (358 nt) and the fully protected probe (258 nt). Two protected frag-ments were detected in all three tissues: a 258 nt fragment corresponding to 14-AE3, and a 220 nt fragment corre-sponding to transcripts lacking the 14 bp insert. Protected fragments of identical size were detected in all AE3-expressing tissues including brain, heart and kidney (Fig. 2A), and testes and skeletal muscle (data not shown). In all cases the 220 bp fragment was the predominant species detected. Densitometric analysis of the bands, however, showed that while only 3% of AE3 mRNA in brain and heart contains the 14 alternate bases, this isoform accounts for 33% of transcripts detected in kidney mRNA. The dif-ferent ratios of the two bands between tissues suggest the involvement of tissue-specific regulation of the splicing of intron ‘e’.

The relative abundance of 311-AE3 transcript was assessed by S₁ nuclease protection using a uniformly labeled 460 bp cDNA probe spanning from the middle of exon ‘A’ to the middle of intron ‘c’ and including 156 bp of vector sequence (Fig. 2C). This probe was predicted to generate 304 bp and 218 bp bands corresponding to tran-scripts either containing or lacking the 311 bp intron, respectively. This analysis generated rather broad bands centered around the expected sizes (Fig. 2B). The signifi-cance of these diffuse bands is unclear, but was repro-ducible. Such heterogeneity of S₁ nuclease digestion may reflect a degree of ‘nibbling’ of heteroduplex DNA and has been observed by others (Green and Roeder, 1980). Den-sitometric analysis of the autoradiograms indicates that the 311 base sequence is present in approximately 19% of heart, 33% of brain and 68% of kidney AE3 transcripts. Taken together, these data suggest that both 311-AE3 and 14-AE3 transcripts are present at steady-state in polyadenylated RNA and that their relative abundance is regulated by tissue-specific factors.

To confirm the S₁ nuclease protection data on the presence of the 311 bases in AE3 transcripts, cDNA from brain, heart, and kidney was analyzed by the polymerase chain reaction (PCR) using two primers, corresponding to AE3 sequences flanking the 311 bp intron (see Materials and Methods). The major products of the PCR were 316 and 627 bp, which correspond to the FL-AE3 and 311-AE3 sequences, respectively (Fig. 3A). These sizes were con-firmed by performing the PCR reaction on plasmids containing cloned copies of FL-AE3 and 311-AE3 cDNA. The results show that RNA samples from all tissues sources contained both types of transcript. The relative intensities of the two bands, however, could not be taken as a quantitative index of the relative abundance of the two mRNA species since the ratios were not preserved upon re-amplification of the PCR products (data not shown). The PCR results do argue though that the ‘smear’ observed in the S₁ analysis of 311-AE3 is not likely to be due to heterogeneity in the location of the actual splice sites since the PCR-generated bands are homogeneous.

In order to assess the expression of individual AE3 polypep-tides isoform-specific polyclonal antibodies were generated for use in immunoblotting and immunohistochemistry. Antisera were raised in rabbits to synthetic peptides corresponding to the unique COOH termini predicted from the DNA sequences of the individual isoforms (Fig 4A). Additionally, antibody 7210 (Kopito et al., 1989) was raised against a fusion protein encoding the NH₂-terminal ∼23 kDa region of AE3 common to all three AE3 isoforms. All anti-sera were affinity-purified.

Immunoblots of rat brain homogenate probed with anti-body 7210 revealed the presence of two immunoreactive bands (Fig. 4B, lane 2). The strongest signal was at ∼160 kDa, which corresponds well with the predicted size of gly-cosylated FL-AE3p, and is in agreement with the size of the protein detected in cells transfected with the FL-AE3 cDNA (Kopito et al., 1989). The other band, at 74 kDa, is larger than predicted from either of the two alternate AE3 isoforms, but is identical in mobility to the band detected in COS (Fig. 4B, lane 1) or human embryonic kidney cells (HEK-293; data not shown) transfected with the 14-AE3 construct. Significantly, a band of indistinguishable mobil-ity was also observed on immunoblots of rat brain homogenate probed with antibody 7015 (Fig 4B, lane 3), which was raised against the unique COOH-terminal pep-tide predicted from the 14-AE3p sequence. These data con-firm that both FL-AE3p and 14-AE3p are expressed as stable polypeptides in rat brain.

By contrast to the results with 14-AE3p, we obtained no evidence supporting the existence of 311-AE3p in rat brain. Only two bands were detected in immunoblots probed with antibody 7210 (Fig. 4B, lane 2), against the common frag-ment. No bands were observed in immunoblots probed with the 7017 antibody (data not shown). Although it is possi-ble that the antibody, which was raised against the mouse sequence, does not cross-react with rat, these results suggest that in contrast to FL-AE3 and 14-AE3, 311-AE3 is not translated into a stable polypeptide in rat brain.

Since the 14-AE3p sequence shares significant homology to the cytoskeleton-binding domain of AE1, we used non-ionic detergent extraction to detect possible association of the brain AE3 isoforms with the cytoskeleton. Crude rat brain homogenate was extracted in 1% Triton X-100 in isotonic buffer, then separated into ‘soluble’ and ‘insoluble’ fractions by centrifugation at 10,000 g. These fractions were then immunoblotted with antibody 7210 (Fig. 5). Virtually all of the material migrating at 160 kDa, which corresponds to FL-AE3p, was soluble under these conditions. In contrast, all of the detectable 14-AE3p was present in the insoluble pellet. No 14-AE3p was detected on immunoblots of purified brain nuclei (data not shown), indicating that the 14-AE3p in the pellet was not introduced by contaminating nuclei.

We have identified two novel isoforms of AE3 generated by alternate RNA processing of the mouse AE3 gene. Both of these transcripts are predicted to encode prematurely truncated AE3 polypeptides that lack the membrane-asso-ciated domain which in the full-length AE3 polypeptide (FL-AE3p), forms the anion exchange channel. We demon-strate that one of these alternate AE3 isoforms, 14-AE3p, is translated into a 74 kDa polypeptide in rat brain that may interact with the cytoskeleton.

The mouse AE3 genomic sequence corresponds well with the genomic sequence of rat AE3 (Linn et al., 1992). The location of the intron/exon boundaries is identical, although the introns diverge more in both sequence and length. The AE3 genomic sequence confirms that the FL-AE3, 311-AE3 and 14-AE3 transcripts all derive from a common gene by alternate mRNA processing. The 311 bp sequence is an unspliced intron since it contains consensus splice donor and acceptor sequences at its ends and lies between two known exons. In the rat, the equivalent of the 311 intron is 358 bp in length, and is also present in a subset of AE3 transcripts (Kudrycki et al., 1990). The sequence of this intron is, how-ever, rather poorly conserved between rat and mouse; it is not predicted to encode a polypeptide similar to 311-AE3p. Because of this poor conservation of the 311-AE3 sequence and our inability to detect a corresponding protein, it is likely that the 311-AE3 transcript represents a splicing intermedi-ate. However, similar use of an unspliced intron to create a truncated protein has been recently observed in the pro-cessing of the mRNA encoding the α-1 subunit of the Na⁺,K⁺-ATPase, another ion transporting, integral mem-brane protein (Medford et al., 1991).

In contrast to 311-AE3, the sequence of 14-AE3 is highly conserved between rat and mouse. The sequences of both alternate splice acceptors at the junction between intron ‘e’ and exon ‘E’ are perfectly conserved in the rat (where exon ‘E’ corresponds to rat exon 10). Moreover, the nonsense codon in the alternate reading frame is conserved between rat and mouse, as is the sequence intervening between the splice acceptor and the premature termination of the 14-AE3p peptide. Use of this alternative splice acceptor site in the rat should give rise to a 14-AE3p polypeptide of nearly identical sequence to mouse. This prediction is supported by immunoblots of rat brain showing that the mouse 14-AE3p specific antibody reacts with the appropriate polypeptide from rat brain. The use of alternate acceptor and donor sequences is common in the processing of viral mRNAs and has also been observed in the processing of several Drosophila genes (Smith et al., 1989). In vertebrates, the use of alternate donor sequences occurs in the processing of the mRNA for the two prohormones for gastrin-releasing peptide (Spindel et al., 1986). The splicing of the 14 bp sequence belongs in this class of alternate splicing and involves a choice between two potential acceptor sequences. The 14-AE3p isoform corresponds to the NH₂-terminal 485 codons of FL-AE3p with a unique 20 codon COOH-ter-minus generated from the insertion of the extra 14 bp and the consequent shift in reading frame. The resulting 14-AE3p protein is predicted to have a molecular mass of 56 kDa and to lack the entire transmembrane domain, which-in FL-AE3p-forms the anion exchange channel (Kopito et al., 1989). Immunoblots probed with antibodies to both the NH₂- and COOH-termini of the predicted 14-AE3p sequence con-firm that this alternately spliced AE3 polypeptide is expressed in rat brain. The mobility (74 kDa) of the rat brain 14-AE3p band is slower than predicted from the 14-AE3 cDNA sequence (56 kDa), but is identical to the 14-AE3p expressed in transiently transfected COS cells, arguing that the 74 kDa band is the mobility of authentic 14-AE3p. Since 14-AE3p lacks both a consensus sequence for Asn-linked glycosylation and hydrophobic membrane-spanning seg-ments that serve as ER insertion signals for FL-AE3p, it is unlikely that the slower mobility is due to the covalent addition of oligosaccharide residues. The presence of a high proportion of charged amino acids in this fragment (36%) may account for its anomalous migration on SDS-PAGE.

These data confirm the presence in rat brain of at least two isoforms of the AE3 gene. Full length AE3 (FL-AE3p) is present as an ∼160 kDa polypeptide associated with a detergent-soluble membrane fraction. Based on its strong homology with AE1 and on its activity when expressed in heterologous cells (Kopito et al., 1989; Lee et al., 1991), FL-AE3p probably functions as a plasma membrane anion exchanger. By contrast, 14-AE3p, derived from the AE3 gene by alternate RNA splicing, is not predicted to be an integral membrane protein, as it lacks the entire membrane-spanning domain of the other AE family members. It is there-fore unlikely to participate directly in ion transport, although we cannot exclude a possible role in regulating the activity of other membrane proteins. The portion of FL-AE3p to which 14-AE3p corresponds is highly polar and predicted to be cytoplasmic. We have recently shown that the NH₂-ter-minal cytoplasmic domain of FL-AE3p can interact with the ‘repeat’ domain of ankyrin, suggesting a possible link between FL-AE3p and the spectrin-based membrane cytoskeleton (Morgans and Kopito, 1993). The difference in detergent solubility between 14-AE3p and FL-AE3p may reflect a difference in the affinity of such an interaction. A possible role for 14-AE3p could be to modulate the interac-tion between integral plasma membrane proteins like FL-AE3p and the underlying membrane cytoskeleton.