Structure and Organization of the Human Ankyrin-1 Gene

Ankyrin-1 (ANK-1) is an erythrocyte membrane protein that is defective in many patients with hereditary spherocytosis, a common hemolytic anemia. In the red cell, ankyrin-1 provides the primary linkage between the membrane skeleton and the plasma membrane. To gain additional insight into the structure and function of this protein and to provide the necessary tools for further genetic studies of hereditary spherocytosis patients, we cloned the human ANK-1 chromosomal gene. Characterization of theANK-1 gene genomic structure revealed that the erythroid transcript is composed of 42 exons distributed over ∼160 kilobase pairs of DNA. Comparison of the genomic structure with the protein domains reveals a near-absolute correlation between the tandem repeats encoding the membrane-binding domain of ankyrin with the location of the intron/exon boundaries in the corresponding part of the gene. Erythroid stage-specific, complex patterns of alternative splicing were identified in the region encoding the regulatory domain of ankyrin-1. Novel brain-specific transcripts were also identified in this region, as well as in the “hinge” region between the membrane-binding and spectrin-binding domains. Utilization of alternative polyadenylation signals was found to be the basis for the previously described, stage-specific 9.0- and 7.2-kilobase pair transcripts of theANK-1 gene.

Erythrocyte ankyrin, ankyrin-1, is the prototype of a family of homologous proteins that are involved in the local segregation of integral membrane proteins within function domains on the plasma membrane, linking the cytoplasmic domains of integral membrane proteins to the spectrin-actin based membrane skeleton (1)(2)(3)(4)(5). This important cellular localization of membrane proteins may be provided by the relative affinities of different isoforms of ankyrin for target proteins. This specialization appears to have evolved through the tissue-specific, developmentally regulated expression of multiple protein isoforms.
The molecular mechanisms by which ankyrin has acquired distinct isoforms with specialized functions are beginning to be revealed. The isoform diversity of ankyrin arises from different gene products and from differential, alternate splicing of the same gene product (6 -19). The cDNAs for three human ankyrin proteins, ankyrin-1 (erythrocyte ankyrin, ANK R ) (6, 7), ankyrin-2 (brain ankyrin, ANK B ) (8), and ankyrin-3 (general ankyrin, ANK G ) (11), have been cloned and their gene products studied. Ankyrins share a common protein structure consisting of an NH 2 -terminal membrane-binding domain, a spectrin/fodrin-binding domain, and a COOH-terminal regulatory domain (2)(3)(4). Ankyrin binding has been described for a variety of proteins including membrane skeleton proteins, ion transport proteins, and cell adhesion molecules (reviewed in Refs. [1][2][3][4]. The membrane-binding domain of ankyrin-1 is composed of 24 tandem repeats of approximately 33 amino acids folded into a nearly spherical structure (6,7,20). Homologous ankyrin repeats are found in many proteins with diverse functions and in virtually all cellular compartments (4,21,22). They have been found in a variety of organisms including yeast, viruses, bacteria, worms, plants, insects, and vertebrates. Ankyrin repeats interact with an assorted array of proteins suggesting that ankyrin repeats function as a general protein-binding motif. In ankyrin-1, the 24 tandem repeats associate into four independently folded subdomains, each comprised of six ankyrin repeats (23). Different combinations of ankyrin repeat subdomains and varying determinants of the same combination of repeat subdomains lead to diversity in mediating ankyrin-membrane protein binding (24,25).
The region encoding the regulatory domain of ankyrin-1 is subject to alternate splicing (6,7,(15)(16)(17)(18). One of these alternate splices creates the "2.2" isoform of ankyrin-1 due to the deletion of 162 amino acids (6,7). This 2.2 isoform exhibits greater affinity for spectrin and band 3 binding (26). These 162 amino acids may act as a repressor by binding back on the rest of the ankyrin-1 molecule leading to allosteric changes (27). Alternate splicing of the region encoding the very COOH terminus of ankyrin-1 produces several isoforms with varying carboxyl termini (6,7,(15)(16)(17)(18). The functional significance of these different isoforms is unknown. The alternately spliced, extreme COOH-terminal regions are among the few regions of the regulatory domain highly conserved in ankyrin-1 between man and mouse (12). In the erythrocyte, ankyrin-1 provides the primary linkage between the spectrin-actin based membrane skeleton and the plasma membrane by attaching tetramers of spectrin to the cytoplasmic domain of band 3, the anion exchanger (20,28,29). Ankyrin-1 has been implicated in many cases of hereditary spherocytosis (HS), 1 a common inherited hemolytic anemia characterized by the presence of spherical red cells on peripheral blood smear with increased red cell osmotic fragility. The role of ankyrin in the pathogenesis of HS comes from a variety of sources including biochemical and genetic studies of HS patients (reviewed in Ref. 30). In some cases of HS, mutations in the ankyrin-1 gene (ANK-1) have been identified (31)(32)(33)(34)(35)(36)(37). An ankyrin-linked, murine model of HS, the nb/nb mouse, has also been described (38 -43).
To gain additional insight into the structure and function of ankyrin-1 and to provide the necessary tools for further genetic studies of HS patients, we cloned the chromosomal gene encoding ankyrin-1 and characterized its genomic structure. We constructed composite human erythroid ANK-1 cDNAs, including previously unpublished 3Ј-untranslated sequences. In the region encoding the regulatory domain, we identified novel erythroid-and brain-specific transcripts of the human ANK-1 gene created by alternate splicing. The molecular basis of the previously identified erythroid 9-and 7.2-kb ANK-1 mRNA transcripts was identified to be mediated by the developmental stage-specific use of alternate polyadenylation signals.

EXPERIMENTAL PROCEDURES
Genomic Cloning and Characterization of the Human Erythrocyte Ankyrin Gene-Overlapping human ANK-1 cDNA fragments isolated and sequenced by Lux et al. (7) that correspond to the entire coding region were used as hybridization probes to screen a human genomic DNA library. The library is a Charon 4A bacteriophage library containing fragments of human genomic DNA partially digested with AluI and HaeIII with EcoRI linkers added (46). A second genomic DNA library, DuPont Merck Pharmaceutical Co. human foreskin fibroblast P1 library (number 1), was screened with two oligonucleotide primers, A and B (Table I) as described (47). These intronic primers flank exon 2 of the erythroid ANK-1 gene (see below) and amplify an ϳ450-bp fragment from genomic DNA. Selected recombinants or DNA fragments that hybridized to the screening probes were purified and subcloned into pGEM-7Z plasmid vectors (Promega Corp., Madison, WI). Subcloned fragments were analyzed by restriction endonuclease digestion, Southern blotting, and nucleotide sequencing. Prior to sequencing, some plasmid vectors were manipulated using the ExoIII unilateral direction technique (48).
Nucleotide Sequencing-Nucleotide sequencing was performed using the dideoxy chain termination method of Sanger et al. (49) with T7 DNA polymerase (Sequenase, U.S. Biochemical Corp.). The sequencing primers were the Sp6 or T7 vectors of the pGEM-7Z plasmid vector or, for some reactions, synthetic oligonucleotides corresponding to known cDNA sequences. Deoxyinosine triphosphate was substituted for deoxyguanosine triphosphate to resolve band compressions and ambiguities (50).
RNA Isolation-Total RNA was prepared from human fetal liver tissue and human bone marrow using the guanidinium-thiocyanatechloroform method as described (51). Human reticulocyte RNA was prepared as described using acid precipitation (52).
3Ј RACE (Rapid Amplification of cDNA Ends)-Total human fetal liver RNA (1 mg) was reverse-transcribed using avian myeloblastosis virus reverse transcriptase using an oligo(dT) adapter primer as described (53). One-tenth of the reversed-transcribed cDNA was amplified by PCR using an adapter primer and one of two gene-specific primers (E or F, Table I). Amplification products were subcloned and nucleotide sequences determined.
Preparation and Amplification of cDNA and cDNA Libraries-cDNA was prepared by reverse transcription of total fetal liver, bone marrow, or reticulocyte RNA using reverse transcriptase of avian myeloblastosis virus. Primers C or D (Table I) were used for reverse transcription. These cDNAs were used as templates in PCR amplification using an automated DNA thermal cycler (Perkin-Elmer) as described previously (52). In some reactions, human cDNA libraries were used as templates in the polymerase chain reaction. These included an oligo(dT)-primed human fetal liver cDNA library in gt11 (54), an oligo(dt)-primed human bone marrow cDNA library in gt11 (CLONTECH, Palo Alto, CA), an oligo(dT)-primed human reticulocyte cDNA library in gt11 (a kind gift of Dr. John Conboy), an oligo(dT)-primed human brain cDNA library in gt11 (CLONTECH), and an oligo(dT)-primed human cerebellar cDNA library in gt11 (CLONTECH). Oligonucleotide primers used in PCR reactions are listed in Table I.
Northern Blot Analyses-Multiple tissue Northern blots containing 2 g of poly (A) ϩ mRNA per tissue were obtained from a commercial source (CLONTECH). Three probes were used in Northern blotting. Probe 1 is pAnk15, an ϳ2.3-kb ankyrin cDNA fragment of Lux et al. (7). Probe 2 is a 1.1-kb ApaI genomic DNA fragment that contains ϳ0.75 kb of the very 3Ј end of the ANK-1 cDNA. Probe 3 is a 2.0-kb human ␤-actin cDNA fragment used as a control for loading in Northern blot analyses (55).
Computer Analyses-Computer-assisted analyses of derived nucleotide and predicted amino acid sequences were performed utilizing the sequence analysis software package of the University of Wisconsin Genetics Computer Group (UW GCG; Madison, WI) (56) and the BLAST algorithm, National Center for Biotechnology Information (Bethesda, MD) (57).

TABLE I Oligonucleotide primers
yielded a large number of hybridization-positive plaques. Selected recombinants were analyzed. Twelve overlapping clones were identified that spanned over 190 kb of genomic DNA containing the ANK-1 gene (Fig. 1). An EcoRI restriction map of this 190-kb region is shown in Fig. 2. Two additional clones, 217 and 261, were isolated that contained ankyrin cDNA fragments but that did not overlap with the other ankyrin clones or each other (Fig. 1). The PCR-based screening of the human genomic DNA P1 library yielded four PCR-positive clones. Of these clones, clone 2032 (DMPC-HFF#1-795 A9) was further analyzed (Fig. 1). This clone contained genomic DNA spanning Ͼ80 kb of DNA and covered all the previously nonoverlapping areas. An Xbal restriction map of the 5Ј end of this clone that contains intron 1 (42 kb) and intron 2 (24 kb) is shown in Fig. 3.
Mapping the Exon/Intron Junctions of the ANK-1 Gene-The erythroid transcript of the human ANK-1 gene is encoded by 42 exons (Table II). Twenty-seven of the 42 exons are relatively short, Յ132 bp in length. The first and forty-second exons contain untranslated sequences. Comparison of the exon/ intron boundaries with reported consensus sequences reveals that the ag:gt rule was not violated in any junction (58,59). There are AG dinucleotides within the 15 bp upstream of the 3Ј (acceptor) splice junctions in five exons, 6, 16, 26, 33, and 40.
Correlation of Functional Domains with Genomic Structure-Comparison of the genomic structure with the protein domains reveals a near absolute correlation between 20 of the 24 tandem repeats encoding the membrane-binding domain of ankyrin with the location of the intron/exon boundaries in the corresponding part of the gene (Fig. 4). These tandem repeats of ankyrin are thought to fold into four subdomains of six repeats each (23). There is no correlation between the genomic structure and the four subdomains; however, fused exons do not cross any of the four subdomain boundaries. There is no correlation between genomic structure and the subdomains of the spectrin-binding domain except for the beginning of exon 26 and the beginning of the neutral subdomain (codon 913) as delineated by Platt et al. (29).
Polymorphisms of the ANK-1 Gene-A number of polymorphisms are present in the ANK-1 gene. A highly polymorphic AC dinucleotide is present in the 3Ј-untranslated region beginning at nucleotide 6589 ( Fig. 1) (7,60,61). To date, we have identified eight different alleles at this site by PCR typing according to the method of Weber and May (62). An informative NcoI polymorphism, initially described by Costa et al. (63,64), has been widely used in linkage analysis. We have mapped this polymorphism to a location in intron 38 (Fig. 1). Differences observed in the nucleotide sequences of the coding region derived from cDNA and genomic clones are shown in Table III. Only one of these polymorphisms, G to C at nucleotide 3049, changes an amino acid, Val 989 -Leu. The locations of a number of other polymorphisms present in the ankyrin cDNA have recently been reported (31).
Alternately Spliced Isoforms of ANK-1 mRNA Are Erythroid Developmental Stage-specific-Multiple isoforms of the ankyrin-1 protein, designated 2.1, 2.2, 2.3, 2.4, 2.6, and 2.9, have been previously identified (4,(65)(66)(67). Band 2.1 is the predominant isoform, and band 2.2 is the most prominent minor species. cDNA cloning of the ANK-1 gene identified several mRNA isoforms that may encode different protein isoforms (6, 7). Two cDNA clones encoding the regulatory domain of ankyrin differed by a 486-bp in-frame deletion resulting in the deletion of 162 highly acidic amino acids. It has been shown that band 2.1 is encoded by the cDNA clone without the deletion, and band 2.2 is encoded by the clone with the deletion. We have previously shown that this "activated" band 2.2 isoform of ankyrin is created by the use of an alternate acceptor site in exon 38 (17).
Cloning from a human reticulocyte cDNA library also identified three isoforms with differing sequences at the COOH terminus of the protein (6, 7). Isoform 1 encodes an acidic COOH terminus 33 amino acids in length, isoform 2 encodes a basic COOH terminus 32 amino acids long, and the third isoform encodes a neutral COOH terminus ending in the last 8 amino acids of acidic isoform one. Knowledge of the exon/intron organization of the ANK-1 gene allows us to determine the precise molecular basis of these alternate splicing events. The pattern of splicing involves splicing of entire exons, partial splicing of exons, and piece meal splicing of individual exons. Isoform 1 is the full-length cDNA that encodes the COOH terminus of the major ankyrin 2.1 isoform, isoform 2 lacks 25 amino acids due to the use of an alternative acceptor splice site in exon 41, and isoform 3 is created by the same alternate acceptor splice site in exon 41 used in isoform 2 and the use of additional new alternate splice donor and acceptor sites in exon 41.
Using human fetal liver, bone marrow, and reticulocyte cDNA or cDNA libraries as a template, we amplified this region of the ANK-1 cDNA encoding the regulatory region with primers G and H (Table I). These primers are placed 5Ј of the 2.1/2.2 splice (G) and in the 3Ј-untranslated region (H) and are designed to amplify almost the entire regulatory region including the sites of all previously described splices. Southern blot analysis of amplification products using genomic DNA clones cor-    responding to the entire regulatory region as hybridization probes is shown in Fig. 5A. This pattern of amplification products is highly reproducible. Nucleotide sequencing of shotgun subcloned amplification products identified 12 separate erythroid isoforms (Isoforms 1-12, Fig. 5B), and 4 of these (1-3 and 5) are the isoforms cloned from reticulocyte DNA. Isoforms 1-12 were found in amplified fetal liver and bone marrow cDNA; isoforms 9 and 10 were not found in amplified reticulocyte cDNA. Novel Alternately Spliced Isoforms of ANK-1 mRNA Are Specific to Brain-Additional ANK-1 gene transcripts in the region encoding the regulatory domain have previously been amplified from human, murine, and rat brain cDNA (12)(13)(14)(15). We looked for these brain-specific cDNA transcripts with PCR using human brain and cerebellar cDNA libraries as a template and, as above, primers G and H (Fig. 5A). In brain, isoforms 5-8 and 13-15 were found (Fig. 5B). Isoforms 13 and 14 contain a motif (motif 3, shown shaded black in Fig. 5B) present in exon 41 (motif 3, Fig. 6A) that is expressed in neural but not erythroid tissue. This previously unidentified motif encodes a peptide with a predicted molecular mass of 8.6 kDa and a pI of 8.3. This motif has a hydrophilic NH 2 terminus and a hydrophobic COOH terminus. Data base searching with this motif did not reveal any significant homologies to known genes. At the 5Ј end of the motif, 66 of 68 nucleotides matched a sequence deposited in the Non-redundant Data base of GenBank EST Division (dbEST T48090).
An Alternately Spliced Exon of the ANK-1 mRNA at the "Hinge" between the Membrane-binding and Spectrin-binding Domains Expressed in Brain-We amplified the "hinge" region between the membrane-binding and spectrin-binding domain using primers I ϩ J (Table I). A 24-bp insertion encoding an in-frame, 8-amino acid sequence was identified. This sequence encodes a neutral peptide with a predicted molecular mass of 843 Da and a pI of 7.5. Analysis of genomic DNA clones shows that this sequence (Fig. 7) is encoded by its own exon, desig-nated 22a in Table II. The sequence is nearly identical to that found in mouse and rat at the same location (Fig. 7) (12, 15). There is no homology of this sequence with either of the two short sequences that are inserted at the hinge between membrane-binding and spectrin-binding domains of murine ank-3 cDNA (13).
The 3Ј-Untranslated Region of the Human ANK-1 Gene-To identify the 3Ј end of the human ANK-1 cDNA, 3Ј RACE experiments were performed. The last 2 nucleotides of the most 3Ј end of the previously reported sequence, GG (7), are actually AC, and are followed by 1250 bp of additional sequence (Fig.  6B). Thus exon 42 is 2503 bp in length. Two polyadenylation signals are present in this sequence (see below). Remarkably, there was 76% similarity between the human and murine ANK-1 3Ј-untranslated regions over a 2561-bp region. There was no similarity between the human ANK-1 gene 3Ј-untranslated region and the 3Ј-untranslated regions of the human ANK-2 gene, the human ANK-3 gene, or the murine ank-3 gene.
Alternate Polyadenylation of the ANK-1 mRNA-Polyadenylation signals are located at positions 7014 -7019 and 8484 -8489 bp in the 3Ј-untranslated region of the cDNA. To determine if both polyadenylation signals are utilized, Northern blots were hybridized to cDNA sequences upstream and downstream of the 5Ј polyadenylation signal. While Northern blots clearly show the presence of two transcripts of ϳ9 and ϳ7.2 kb in length in fetal liver RNA when the upstream probe 1 (Fig.  8A) is used (Fig. 8B), only a transcript of ϳ9 kb is detected when the downstream probe 2 (Fig. 8A) is hybridized to the same Northern blots (Fig. 8B). Thus, these transcripts are most likely the result of alternate polyadenylation.
To provide additional evidence that these alternate polyadenylation signals are utilized in mRNA, we performed 3Ј RACE with an oligo(dT) primer (primer D, Table I) for reverse transcription using total human fetal liver RNA as template. Gene-specific primers, E or F, and a linker primer were used in PCR amplification of the cDNA. Poly(A) tails were discovered 17 and 16 nucleotides, respectively, downstream of the 5Ј and 3Ј polyadenylation signals (Fig. 9). Thus the molecular basis of the developmental stage-specific 9-and 7.2-kb transcripts is mediated via the use of alternate polyadenylation signals in the 3Ј-untranslated region of the ANK-1 cDNA.

DISCUSSION
Genetic analyses of patients with hereditary spherocytosis have been previously hampered by the lack of knowledge of the sequences of the ANK-1 chromosomal gene. A variety of mutations causing human disease have been described that affect RNA processing and translation (68). Many of these mutations are associated with dramatic decreases in steady state mRNA levels. This observation has important implications for the methodologies employed in mutation detection. Reverse-tran- FIG. 4. Correlation of the genomic organization of the human ANK-1 gene and ankyrin-1 protein domains. Top, a map of SmaI (S) and BamHI (B) restriction endonuclease sites in the ANK-1 cDNA. Middle, the exons encoding segments of the ANK-1 cDNA are shown with the exon numbers. The corresponding ankyrin repeats are shown below. There is exact correlation between 20 of the 24 tandem ankyrin repeats with individual exons. Bottom, the protein domains of human ankyrin-1. scriptase PCR-based techniques are unlikely to detect the mutations with decreased mRNA levels, necessitating study of these mutations at the genomic DNA level (69). Characterization of the genomic structure of the erythroid transcript of ANK-1 allows structural studies of the ANK-1 gene in patients with hereditary spherocytosis using genomic DNA.
Ankyrin repeats (also referred to as cdc 10 repeats, cdc10/ SWI6 repeats, and SWI6/ANK repeats) are found in varying numbers in a large number of functionally distinct proteins involved in assorted molecular associations including proteinprotein, intramolecular, and DNA interactions (20,21). Determination of the crystal structure of 53BP2, an ankyrin repeatcontaining protein identified to bind the p53 tumor suppressor in vitro, revealed that individual ankyrin repeats have an L-shaped structure consisting of a ␤-hairpin followed by 2 ␣-helices that pack in an antiparallel fashion (70). Adjacent repeats pack via both their ␤-hairpins, forming a continuous ␤-sheet, and via their helix pairs, forming helix bundles. Because the folding of an ankyrin repeat appears to depend on the presence of adjacent repeats, it seems unlikely that individual repeats are able to fold normally. This is compatible with the observation that almost all ankyrin repeat-containing proteins have four or more repeats (21). As revealed by the crystal structure, the nonglobular structure of ankyrin repeats offers many possibilities for diverse macromolecular interactions (70). This diversity, as well as specificity, is enhanced by variations in amino acid sequence of individual ankyrin repeats and their flanking sequences.
Prior to the availability of the crystal structure of an ankyrin repeat, attempts to identify the phasing of ankyrin repeats were complicated by the fact that there was no evidence of single repeats encoded by discrete exons. Intron/exon boundaries were found at varying locations within ankyrin repeats of different genes (71)(72)(73)(74)(75)(76). The crystal structure suggests that ankyrin repeats begin with the consensus sequence -(N/D)-----G-TPLH-AA (dashes indicate nonconserved amino acids) (70). Several ankyrin repeat-containing proteins, such as forked and plutonium of Drosophila, begin their repeat domains exactly with this phasing (71,72). All three human ankyrins begin and end their repeat domains (repeats 1 and 24) with "partial" repeats (6 -8, 11). Closer inspection suggests that these sequences are likely to be the initial ␤-strand of the ␤-hairpin of the first ankyrin repeat. Interestingly, comparison of the repeats in ankyrin-1, ankyrin-2, and ankyrin-3 shows that 10 residues of the repeat, TPLH-AA--G, are highly conserved. These conserved residues are predicted to be located in the first ␣-helix with individual residues playing critical roles; e.g. Thr in initiating ␣-helices, Gly in terminating ␣-helices, and His in supporting inter-repeat stabilization.
The sequence of ankyrin repeats in members of similar gene families are highly conserved, suggesting a common ancestor prior to divergence of individual members of a gene family. The ankyrin repeats of IKBA, BCL3, and NFKB2, members of the IB␣ family, are highly conserved, as are the genomic structures of these three genes (74 -76). The repeats of human ankyrin-1, ankyrin-2, and ankyrin-3 are also highly conserved. Although the genomic structures of the human ANK-2 and ANK-3 genes have not been reported, the intron/exon boundand 13-15 were amplified from human brain or cerebellar cDNA. The predicted molecular masses (kDa) and isoelectric points of isoforms 1-15, respectively, starting with serine 1387, the beginning of the regulatory domain (7) 5. Human ANK-1 cDNA isoforms. A, PCR amplification of the region encoding the regulatory domain of ankyrin. Erythroid or neural cDNAs were amplified, electrophoresed in an agarose gel, blotted, and hybridized to a [ 32 P]dCTP-labeled ankyrin genomic DNA probe encompassing the entire regulatory domain of ankyrin 1. A highly reproducible pattern of bands of varying intensity is observed. B, diagrammatic representation of the ANK-1 cDNA isoforms identified after PCR amplification of erythroid or neural cDNA or cDNA libraries followed by shotgun subcloning and nucleotide sequencing. Isoforms 1-12 were amplified from human fetal liver and bone marrow cDNA; isoforms 1-8 and 11 and 12 were amplified from reticulocyte cDNA. Isoforms 5-8 aries of a single ANK-2 gene exon (77) correspond exactly to the locations of intron/exon boundaries in the ANK-1 gene. Conservation of ankyrin repeats, as well as conservation of regions critical for spectrin binding, together provide strong evidence for a common origin for the different ankyrin genes, diverging after the formation of the membrane-binding and spectrinbinding domains.
The data presented here suggest that there is developmental stage-and tissue-specific diversity in ANK-1 cDNA transcripts. It is likely that additional transcripts have yet to be discovered.
These transcripts, if they are translated into functional proteins, may well play important roles in ankyrin-1 function, either as linker/adapter molecules or in other yet undiscovered capacities. It will be interesting to determine if these isoforms encode the previously described 2.3, 2.4, 2.6, and 2.9 ankyrin isoforms and the other isoforms detected on Western blots (4,66). The pattern of multiple ankyrin isoform expression observed due to alternate splicing is similar to that observed for other red cell membrane proteins such as ␤-spectrin, protein 4.1, and tropomyosin (78 -81).
Cleavage of primary mRNA transcripts and the addition of poly(A) to the newly formed 3Ј end of the transcript downstream of the highly conserved polyadenylation signal AAUAAA are features of almost all eukaryotic mRNAs (82). 3Ј-Untranslated regions can have an important influence on mRNA function including translation, localization, stability, and gene transcription (82,83). Alternate polyadenylation of 3Ј-untranslated regions may play a role in determining developmental or tissue-specific preferences for various mRNA transcripts (83,84). For example, alternative polyadenylation of the murine ␤-tropomyosin mRNA determines differentiation-dependent transcripts in BC 3 Hl muscle cells (85). The role(s) of the two ANK-1 gene transcripts that vary in their 3Ј-untranslated regions is (are) unknown, but considering the remarkable FIG. 6. Human ANK-1 gene cDNA sequence. A, nucleotide sequence and deduced amino acid sequence of exon 41 of the human ANK-1 gene. The sequences of the four alternatively spliced motifs shown in Fig. 5A are marked by brackets above the sequence. Motif 3 is a novel, brain-specific motif not previously identified. The termination codon present in motif 4 is double underlined. B, nucleotide sequence and deduced amino acid sequence of exon 42 of the human ANK-1 gene. The 3Ј-untranslated region of the human ANK-1 cDNA is also shown. The termination codon present in exon 42 is double underlined. Polyadenylation signals, AATAAA, are also double underlined.

FIG.
7. An alternately spliced exon of the ANK-1 cDNA at the hinge region between the membrane-binding and spectrinbinding domains expressed in brain. The nucleotide sequence and deduced amino acid sequence of an in-frame, 24-bp exon, exon 22a, identified by PCR amplification of a human brain cDNA library are shown. The deduced amino acid sequences identified in the same region of murine and rat ANK-1 cDNA are shown for comparison. homology between the 3Ј-untranslated regions of human and murine erythrocyte ankyrin gene, it seems likely that this region has an important function.  (7) upstream of both potential polyadenylation sites. Probe 2 is a 1.1-kb ApaI genomic DNA fragment that contains ϳ0.75 kb of the very 3Ј end of the ANK-1 cDNA and is 3Ј to the upstream polyadenylation site. B, Northern blot. Lane 1, samples of 2 g of poly(A) ϩ RNA from human fetal liver were hybridized to [ 32 P]dCTP-labeled probe 1. Abundant messages were detected at 9.0 and 7.2 kb. Lane 2, the same blot was stripped and rehybridized to probe 2. Only a 9.0-kb message is detected. Lane 3, the same blot was stripped and hybridized with a control 2.0-kb human ␤-actin cDNA probe.