Complex Patterns of Sequence Variation and Multiple 5’ and 3’ Ends Are Found among Transcripts of the Erythroid Ankyrin Gene*

The structural protein ankyrin functions in red blood cells to link the spectrin-based membrane skeleton to the plasma membrane.

The spectrin-based membrane skeleton forms a supporting protein network underlying the plasma membrane in red blood cells (for review, see Refs. 1 and 2). Erythroid ankyrin functions in the membrane skeleton as a high affinity binding protein linking the p subunit of spectrin to the cytoplasmic domain of band 3, the anion channel (3-6). Once thought to be unique to red blood cells, a spectrin-based membrane skeleton and the linking protein ankyrin are now known to be present in many, if not all, cell types (for review, see Ref. 7). Immunological and biochemical analyses of ankyrin from different cell types have revealed a large family of related proteins. The best characterized members of this family are erythroid and brain ankyrins (ANK-1 and ANK-2, respectively) (%lo).' Erythroid and brain ankyrins are coded for by HL29305 and DK27726 (to J. E. B.) and HL15157, HL32262, and * This work was supported by National Institutes of Health Grants DK34083 (to S. E. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in thispaper has been submitted to the GenBankm/EMBL Data Bank with accession number(s) X69063, X69064, and X69065. ll These authors contributed equally to this work. Throughout this paper, we have adopted a nomenclature system based upon the conventions used to name mouse ankyrin genes (19). Briefly, the genes are referred to as Ank-1 (erythroid) and Ank-2 (brain). The proteins are ANK-1 and ANK-2. Note that gene names are italicized; protein names are not. A detailed discussion of ankyrin nomenclature is given in Ref. 20. separate genes (11)(12)(13)(14), and each gene is known to produce alternatively processed transcripts (15)(16)(17)(18). This allows for the production of multiple ankyrin isoforms from a single gene. It is likely that other ankyrin genes exist and will be found as more is learned about the ankyrin proteins in other tissues.
Three laboratories have recently isolated and sequenced overlapping cDNA clones for human and mouse erythrocyte ankyrins (15,16,18). The three structural domains of ankyrin as defined by chymotrypsin sensitivity (8) have been confirmed by sequence data to include an NH2-terminal89-kDa domain, a central 62-kDa domain, and a COOH-terminal55-kDa domain (16). The NH2-terminal domain consists primarily of 22 tandem 33-amino acid repeats and contains band 3 and tubulin binding activity (16). Ankyrin-type repeats have been identified in a variety of other proteins with diverse functions, including cell cycle proteins and transcription factors (see Ref. 16;for review,see Refs. 20 and 21). The central 62-kDa domain is responsible for P-spectrin and vimentin binding (16). The 89-and 62-kDa domains are relatively well conserved among the erythroid ankyrins sequenced thus far. The COOH-terminal55-kDa regulatory domain is more variable and appears to function as a modifier of the binding activities of the other two domains (22). Two regions of alternative sequence within the regulatory domain of human erythroid ankyrin are known. The first refers to a 486-base pair (bp)' segment missing in some clones and accounts for the 2.2 ankyrin isoform found in human red blood cells (16). The second occurs at the COOH terminus where three alternatives have been identified (15,16). Comparison of mouse Ank-1 sequence (18) to the human sequences shows >95% amino acid identity in the band 3-and spectrin-binding domains. The 55-kDa regulatory domain shows somewhat lower overall homology (79% amino acid identity); however, several segments are 100% conserved, possibly indicating regions essential for ankyrin function in erythroid cells.
The existence of a large family of ankyrin proteins and the observation that more than one type of ankyrin can be expressed in the same cell type suggest that in addition to binding the membrane skeleton to the plasma membrane, ankyrins may have other more specialized functions. For example, some ankyrins have binding affinities for integral membrane proteins other than band 3, notably Na+/K+-ATPase in kidney (23) and the voltage-dependent Na' channel in brain (24) and the neuromuscular junction (25). Some ankyrins are localized to specialized regions of cells such as Ranvier's nodes in myelinated nerves (26,27). Studies on mice with a mutation (nb) in the erythroid ankyrin gene have The abbreviations used are: bp, base pair(s); kb, kilobase(s); PCR, polymerase chain reaction.

9533
indicated that the presence of an ANK-1 isoform in Purkinje cells is critical for the survival of these cells in brain (14).
TO define the functional significance of the various ankyrin proteins, a clear description of their origin (distinct genes or alternative processing of single gene products), their structure, and their pattern of expression is needed. This study was designed to gain such information about transcripts of the mouse erythroid ankyrin gene Ank-1 on mouse chromosome 8 (11). Two transcripts (9.0 and 7.5 kilobases (kb) in size) are known from erythroid and cerebellar tissues (14). On RNA hybridization blots, we have detected two additional Ank-lpositive tissues (skeletal muscle and heart) and four additional transcripts (5.0,3.5,2.0, and 1.6 kb in size). We have prepared mouse reticulocyte and cerebellar cDNA libraries and screened them for erythroid ankyrin. Characterization of the ankyrin clones obtained has produced a full-length coding sequence for a cerebellar transcript and sequence for several partial erythroid and cerebellar transcripts that represent variations from previously described Ank-1 sequence. We have found two distinct NH2 termini and three functional polyadenylation sites and characterized a transcript from spleen that has deleted most of the sequence coding for the regulatory domain. RNA hybridization blot data using domain-specific probes indicate that the three small transcripts from skeletal muscle are missing large regions of coding sequence from the band 3-and spectrin-binding domains. Several new inserts and deletions involving small sequence segments are documented. Our data show that the expression of the Ank-I gene involves complex patterns of sequence variation and suggest the potential for an even greater diversity of ankyrin proteins than previously suspected.

MATERIALS AND METHODS
Animak-All mice used in this study were of the inbred strain C57BL/6J produced in our research colony at the Jackson Laboratory. (The Jackson Laboratory is fully accredited by the American Association of Laboratory Animal Care.) Hybridization Probes-Mouse erythroid ankyrin clones were detected using the cDNA clone mAnk-l, a 4.6-kb mouse anemic spleen clone isolated and sequenced by White et al. (18). This clone starts within the 15th repeat of the NH2-terminal domain of erythroid ankyrin (bp 1650 of White et al,), continues through the COOH terminus, and extends 539 bp into the 3'-untranslated region. For use as a probe, the mAnk-1 insert was released from the plasmid by EcoRI digestion and purified by phenol extraction after electrophoretic separation on low melting agarose (28). A second probe (Er4/ 272) was prepared by amplification of a 272-bp segment of the anemic spleen cDNA clone Er4, which was isolated and characterized in this study (refer to Fig. 6 for location and sequence of the primers used). This region represents previously undescribed coding and 3"untranslated sequences for Ank-I mRNA. Amplification was done by polymerase chain reaction (PCR) techniques (29) using the GeneAmp DNA amplification kit (Perkin-Elmer Cetus Instruments). Both probes were 32P-labeled by random oligonucleotide primer extension (30).
Preparation of RNA-When exposed to anemic stress, normal mice respond by production of erythrogenic spleens, bone marrow hyperplasia, and an increase of circulating reticulocytes (31). We exposed mice to anemic stress by injection of the hemolytic agent phenylhydrazine (Sigma) (32). These mice were the source of the reticulocytes and anemic spleens used in this study. For reticulocyte collection, animals were anesthetized with Avertin, and blood was collected from the heart cavity using heparin as the anticoagulant. Animals for tissue RNA preparation were anesthetized and perfused through the heart with cold (4 "C) phosphate-buffered saline to remove contaminating blood cells. Tissues were removed, immediately frozen in liquid nitrogen, and stored at -70 "C until use. Total RNA was prepared using the guanidinium/thiocyanate/chloroform method (RNAzol, Cinna/Biotecx, Houston, TX) (33). Poly(A)' RNA was isolated by oligo(dT) chromatography (34).
Preparation and Screening of cDNA Libraries-Reticulocyte and cerebellar libraries derived from the C57BL/6J strain of mice were prepared by random and oligo(dT) priming, respectively. cDNAs were rl. Gene Transcripts synthesized using the Riboclone cDNA synthesis system (Promega Biotec). To minimize the number of globin clones, reticulocyte cDNAs d . 0 kb were excluded by electrophoretic separation on agarose gels, followed by electroelution (Geneluter, Invitrogen, San Diego, CA). Cerebellar cDNA was size-selected by a spin column to exclude cDNA 6 0 0 bp (Size Select 400 Spun columns, Pharmacia, Uppsala). Reticulocyte and cerebellar cDNAs were cloned into the XZap I1 vector (Stratagene, La Jolla, CA) using an EcoRI adaptor ligation system (Promega Biotec). Gigapack I1 Gold (Stratagene) was used for X packaging. An anemic spleen cDNA library derived from the CD-1 strain of mice was provided for us by R. Kopito. This library contained size-selected cDNAs (>2.0 kb) cloned into Xgtll (35). The libraries were screened with the mAnk-1 cDNA clone described above using the plaque hybridization method (36) with Duralon-UV nylon membranes (Stratagene). Hybridizations were performed at 42 "C with 50% formamide, 0.75 M NaCI, and 1% SDS. 32P-Labeled probe was added to the bag without changing the prehybridization solution. Final filter wash was 0.1 X SSC, 0.1% SDS at 60 "C. Positive Xgtll clones were subcloned into the pGEM7ZF(-) plasmid vector (Promega Biotec), and positive XZap I1 clones were excised and recircularized to generate subclones in the pBluescript SK(-) phagemid vector.
DNA Sequencing-Plasmid DNA was prepared either by cesium chloride banding (36) or by Magic minipreps (Promega Biotec). Double-stranded plasmids were sequenced using the dideoxynucleotide chain termination method (37) and T7 DNA polymerase (Sequenase, U. S. Biochemical Corp.). Sequence was obtained by generating unidirectional nested deletions using controlled exonuclease III/Sl nuclease digestion (Erase-a-Base, Promega Biotec) and by the use of sequence-specific synthetic oligonucleotides as primers. Sequence data analysis was done using Microgenie DNA analysis software (Beckman-Spinco).
Hybridization Blots and PCR Analyses-Genomic DNA was prepared as previously described (38). DNA was restriction-digested and blotted according to the method of Southern (39). RNA was prepared as described above, run on formaldehyde-agarose gels (36), and blotted onto Zetabind nylon filters (AMF CUNO, Meriden, CT). The amounts and quality of the RNA on the filter after transfer were checked by UV shadowing (40). Labeling of probes, hybridization, and washing procedures were as described for cDNA library screening. Regions of the cDNA clones and reverse-transcribed mRNAs were amplified by PCR (29), electrophoresed on 1.5% agarose gels, and visualized by ethidium bromide staining.

RESULTS
Tissue Survey of Ank-1 Expression-The finding of an Ank-1 transcript expressed in cerebellum and the demonstration of its important role in Purkinje cell stability (14) suggested the possibility that the Ank-1 gene is expressed in other tissues as well. We did an RNA hybridization blot analysis of total cellular RNA from 13 different tissues of the mouse using mAnk-1 as the probe. In addition to the previously described Ank-1-positive tissues of bone marrow, spleen, reticulocytes, and cerebellum (14), we identified two other positive tissues, skeletal muscle and heart. No positive signal was detected in eye, lung, liver, kidney, pancreas, small intestine, or testes. The RNA hybridization blot pictured in Fig. 1 summarizes the transcripts detected in our experiments by the mAnk-l probe. Reticulocytes express the 7.5-kb transcript, and cerebellum expresses the 9.0-kb transcript almost exclusively (lanes 1 and 3 ) . In some preparations, small amounts of a 9.0-kb transcript are seen in reticulocytes. The 9.0-and 7.5-kb transcripts are expressed in nearly equal amounts in anemic (erythrogenic) spleen (lane 2 ) . In preparations in which degradation products from the larger bands do not obscure smaller bands, transcripts of -5.0, 3.5, and 2.0 kb can be resolved in anemic spleen (lane 2 ) . Lune 4 shows the pattern in skeletal muscle. Three bands -3.5, 2.0, and 1.6 kb in size are seen. The 9.0-and 7.5-kb bands were not detected in skeletal muscle. The pattern in heart was identical to that in skeletal muscle (data not shown).
Isolation and Characterization of cDNA Clones-As a basis for understanding the clones and sequences reported in this paper, Fig. 2A summarizes known features of erythroid ankyrin. The protein structural domain sizes and primary functions are given above the line, and the locations of the repeat region and known splice regions are indicated below the line.
Using the mAnk-1 cDNA clone as a probe to screen the libraries, we isolated 1 clone from the anemic spleen library, 15 clones from the reticulocyte library, and 11 clones from the cerebellar library. All clones were partially sequenced and positioned relative to each other by comparison to the mouse erythroid ankyrin sequence of White et al. (18). This sequence is referred to as erythroid 1 (Erl) throughout this paper. Five clones were chosen for detailed study: the anemic spleen clone (Er4), the reticulocyte clone (Er18), and the cerebellar clones (Cbll, Cb12, and Cb14) (Fig. 2B). Regions of these clones that were identical to the sequence of Erl were sequenced on one strand. Regions unique to Erl were sequenced on both strands. These regions are underlined in Fig. 2B. We report (i) sequence for a full-length cerebellar cDNA clone (Cb14), (ii) sequence for the alternative COOH-terminal segments of Er18 and Cb12, (iii) sequence for the 3"untranslated regions and alternative polyadenylation sites of Er18 and Cbll, and (iv) sequence for the alternative COOH-terminal domain of Er4.
Erythroid and Cerebellar cDNA Sequences-Two overlapping clones (Cb14 and Cbll) formed a complete coding and 3"untranslated region for a cerebellar transcript. The nucleic acid sequence and deduced amino acid sequence for these clones are designated Cb14/11 and are reported in Fig. 3. The total length of Cb14/11 is 8145 bp and includes 357 bp of 5'untranslated sequence, 5544 bp of coding sequence, and 2244 bp of 3"untranslated sequence. From RNA hybridization blots, the estimated size of the cerebellar transcript is 9.0 kb. Assuming a 200-bp poly(A) tail, this would predict an additional 655 bp of 5"untranslated region. The site of the start of translation is reasonably certain as there is a well conserved Kozak start site consensus sequence (41) surrounding the initiator codon (ATG) and an in-frame stop codon 39 bp upstream. The open reading frame codes for 1847 amino acids and predicts a protein with a molecular weight of 202,468 and a PI of 6.12. The 5"untranslated region has properties of a CpG island in that it is rich in GC nucleotides (71%) and does not show the CpG dinucleotide suppression seen in interisland regions (42). There is a polyadenylation signal (AATAAA) at base pair 8124, 16 base pairs upstream from the start of a poly(A) tail of >lo0 bases.
The primary sequence of Cb14/11 is essentially identical to that of Erl. Seven single base changes, resulting in only three amino acid changes, were found as follows: base 471, T for C; base 588, A for G; base 1017, C for T , base 2414, C for T (changed amino acid P for L); base 2817, G for C; base 3699, C for G (changed amino acid N for K); and base 4847, T for G (changed amino acid V for G). These changes probably reflect normal differences between the C57BL/6J and CD-1 strains of mice used to construct the libraries. However, it is possible that the leucine-to-proline and glycine-to-valine substitutions could have some functional significance. In addition to the single base changes among the new clones, several other differences were noted.
First, there appear to be at least two and possibly three first exons utilized by Ank-1 transcripts. This is illustrated by the complete divergence of the 5"untranslated and NH2terminal regions of Erl, Cb14/11, and the human erythroid ankyrin sequence. The three NHp-terminal amino acid sequences are aligned in Fig. 4A. The asterisk indicates the point where the sequences become homologous, and this site coincides with the position of a splice junction identified in the human erythroid ankyrin gene (43). A second difference is a 24-bp insert in Cb14/11, coding for 8 amino acids, just 6 residues 5' to the predicted 89/62-kDa junction (see Cb14 in Fig. 2B and Fig. 3). The amino acid sequence of the insert (TAHISIMG) predicts a neutral peptide with a molecular weight of 829 and a PI of 6.23 and an overall hydrophobic nature.
A new alternative was found in the 2.2 region of the 55-kDa regulatory domain. Unlike human erythroid ankyrin, no evidence has been found for the occurrence of the 2.2 isoform in mouse. However, we have detected a cDNA clone with a 90-bp deletion within the 2.2 splice region (see Cb14 in Fig.  2B and Fig. 3). The 3' splice junction of the 90-bp segment

GQQRV Cb14/11
FIG. 4. Protein sequence alignments. A, alignment of ANK-1 5' ends. The amount of 5'-untranslated sequence known for each clone is indicated as nucleotide number (nt#). The asterisk represents the point where the amino acid homology resumes among the three clones (residue 13 of Cb14/11). The human sequence comes from Lux et al. (16). Retic, reticulocytes; Cb, cerebellum. E , protein sequence alignment of segments of Erl and Cb14/11 showing the 90-bp deletion, The asterisk marks the end of the 90-bp deletion, and this coincides with the 3' junction of the 2.2 splice region defined for human erythroid ankyrin (16). The amino acid number (act#) for the first residue of the segment is given. coincides with the 3' junction of the human 2.2 splice (16,43). The amino acid sequences of E r l and Cb14/11 in this region are aligned in Fig. 4B.
In the COOH-terminal region, three alternatives have previously been described (i) the presence of a highly acidic 33amino acid segment, (ii) the replacement of this segment with 32 amino acids overall basic in charge, or (iii) the absence of both the acidic and basic segments (15,16). We have found a fourth alternative in the Cb12 clone, i.e. the use of both the acidic and basic segments in the same transcript. A graphic representation of the four alternatives is given in Fig. 5A.  18), and this work, the alternative segments can be defined as follows: a 33-amino acid segment (A); a 24-amino acid segment (B); and a segment (C) containing coding sequence for 8 amino acids, a second stop codon (C2), and an indeterminate amount of 3"untranslated sequence. The upstream stop codon (Cl) is contained within the basic segment and would be utilized when both or just the basic segments are present (Fig. 5A, lines 1 and 2). The downstream stop codon (C2) would be utilized when neither or just the acidic segment is present (Fig. 5A, lines 3  and 4). The COOH-terminal alternatives found in the clones used in this study are illustrated in Fig. 2B. The amino acid sequences for the Er18 and Cb12 alternatives are given in Fig.  5B.
The last area o f difference detected was in the 3"untranslated region. The reticulocyte clone (Er18) overlaps and extends beyond the existing E r l sequence to a poly(A) tail 1027 bp from the stop codon (Cl) (Fig. 3). A polyadenylation signal occurs just 12 bp upstream from the Er18 poly(A) addition site (this corresponds to bp 6910 in Cb14/11). The cerebellar clones Cbll and Cb12 terminate at a poly(A) tail 2244 bp beyond C1. A third polyadenylation signal is located at bp 7525 of Cb14/11, but no clones that utilize this site were found. A homology comparison of the 3"untranslated sequence reported by Lux et al. (16) for human reticulocyte ankyrin with the corresponding region of Cb14/11 (bp 5955-7433) showed 68% overall homology, with six regions showing %O%. One particularly long region (172 bp) occurred around the Er18 polyadenylation signal and averaged 87%. The CA dinucleotide repeat found at base pair 6304 in the sequence of Lux et al. (16) was not found in Cb14/11. Small Transcripts of Spleen and Skeletal Muscle-The anemic spleen clone (Er4) has a completely unique 3' end. This clone begins in the 89-kDa domain at bp 1307 of Erl (bp 1579 of Cb14/11) and is identical to Erl until bp 4487 (bp 4783 of Cb14/11), where Er4 substitutes 340 bp of unique sequence. This sequence encodes 38 amino acids and includes 226 bp of 3"untranslated region, a polyadenylation signal, and a poly(A) tail (Figs. 2B and 6). The predicted peptide has a molecular weight of 4451 and a PI of 6.29 and is very hydrophobic in nature, especially at the COOH terminus. The point of divergence with E r l coincides with a splice junction identified in the human erythroid ankyrin gene (43). Assuming the same 5' end as that of Erl, the predicted size of the Er4 transcript would be at least 4.5 kb. A search of the GenBank Data Bank did not detect any sequences homologous to the Er4 unique region.
Several experiments were done to verify Er4 as an Ank-1 transcript. A 272-bp PCR fragment was amplified from the Er4 unique region using the oligonucleotide primer pair indicated in Fig. 6. The fragment (Er4/272) was 32P-labeled and used as a probe on DNA and RNA filter blots. Fig. 7A shows a comparison of Hind111 and EcoRI restriction fragments detected in mouse genomic DNA by the mAnk-1 and Er4/272 probes. The Er4/272 probe hybridizes to a subset of the fragments detected by mAnk-1, indicating that the Er4 unique sequence is within the Ank-l gene. On RNA blots of total cellular RNA, Er4/272 detects an -5.0-kb transcript in anemic spleen, the tissue from which the clone was isolated, but not in the other tissues examined (reticulocytes, cerebellum, and skeletal muscle) (Fig. 7B). Using the same two oligonucleotide primers described above, we were able to reversetranscribe total cellular RNA and to amplify by PCR (29) the Er4 unique region from anemic spleen RNA, but not from reticulocytes, cerebellum, or skeletal muscle RNA (Fig. 7C). These data strongly support the identity of Er4 as a transcript of the Ank-I gene.

DISCUSSION
Our analysis of transcripts from the mouse Ank-I gene indicates the potential for a greater complexity of ankyrin protein structure and function than formerly appreciated. Several new features of Ank-1 transcripts are described. Foremost among these is the discovery of multiple NH2 termini and the possibility of gene regulation by multiple promoters. Alternative polyadenylation also occurs, and it is likely that this accounts for the difference in size between the 7.5-and 9.0-kb transcripts. Somewhat surprisingly, the 3"untranslated regions of mouse and human showed relatively high homology. This could indicate a functional role for some 8124. The lines in which these features of the amino acid sequence appear are indicated by asterisks. The relevant sequences are shown in boldface type. They are the NHp-terminal amino acids that differ from Erl, the 8-amino acid insert at residue 825, the position of the 30amino acid deletion at residue 1649, and the amino acids present in the variable COOH-terminal region starting at residue 1816.

A.
Erythroid Ankyrin Gene Transcripts 1 Er18 FIG. 5. Graphic representation of known alternative COOH-terminal sequence patterns found among Ank-Z transcripts. A, the alternative sequence segments are represented as boxes A-C, as described under "Results." Open boxes indicate coding regions; hatched boxes indicate 3'-untranslated regions. The lines connecting the boxes show the patterns of sequence segments used. CI and C2 indicate the alternative stop codons, and the asterisks indicate the stop codon used in each case. B, shown are the amino acid sequences for the two new mouse alternatives described in this report. The segments begin with residue 1813 of Cb14/11. Asterisks indicate the first residue of each segment.

FIG.
6. Nucleotide and deduced amino acid sequences of clone Er4 unique region. The segment of Er4 shown here starts at bp 4187 of Erl (bp 4483 of Cb14/11). The asterisk indicates the predicted start of the 55-kDa regulatory domain. The stop codon, the CA dinucleotide repeat, and the polyadenylation signal of the unique region are underlined. The oligonucleotide primer pair used to amplify by PCR the unique region are ouerlined. The arrows indicate the direction of primer extension. The unique amino acid sequence appears in boldface type. segments of this region, possibly affecting transcript stability (44) and/or subcellular localization (45).
Our data also define differences among Ank-1 transcripts that involve changes in small segments of coding sequence. We detected, in cerebellar transcripts, a 24-bp insert at the junction of the 89-and 62-kDa domains. This would insert a neutral hydrophobic peptide into a basically charged hydrophilic region and couId be functionally significant to the structure of the domain junction. Additional alternatives occur within the regulatory domain. Although no examples of the 2.2 splice are known in mouse, a 90-bp deletion is found within the 2.2 region. In addition to the three described previously, a fourth alternative at the COOH terminus was found. It should be noted that each of these differences would produce transcripts (and proteins) of nearly equal size. The data predict that ankyrin transcripts and proteins that migrate as a single band when separated electrophoretically are, in fact, a complex mixture of alternative forms.
We believe that the most likely explanation for the sequence variations we have documented is alternative splicing. The observation that the sequences adjacent to the variable regions are identical to known Ank-1 sequences and the coincidence of the junctions of several of our sequence alternatives with splice junctions identified in the human ANKl gene support this explanation. It should be emphasized, however, that our data offer only indirect evidence for alternative splicing and cannot rule out the possible involvement of other genes. Because of the lack of sequence data, this is especially true for the small transcripts from skeletal muscle and heart.
Several ankyrin proteins smaller than the usual 210 kDa have been detected in mammalian tissues (see Ref. 46; for review, see Refs. 20 and 21). Once thought to be degradation products of the 210-kDa ankyrin, some are now known to be functional isoforms. Proteins of 195 and 186 kDa found in human red blood cells have been shown to be missing parts of the regulatory domain and to have different binding affinities for spectrin and band 3 (22). The 195-kDa protein is produced by post-translational calpain cleavage of a 20-kDa segment from the carboxyl terminus. The 186-kDa protein is produced from an alternative transcript missing the 2.2 region (16). The 5.0-kb transcript we described from erythrogenic spleen (Er4) is interesting in this context. This mRNA codes for 89-and 62-kDa domains identical in sequence to that known for ANK-1; however, most of the normal regulatory domain has been replaced by a unique segment coding for 38 amino acids. This transcript suggests that, as in human, the mouse normally produces small ankyrin proteins.
The three small transcripts found in skeletal muscle and heart suggest a situation unique among known ankyrin transcripts because they appear to diverge significantly in the 5' and central functional domains. The evidence supporting their identity as ankyrin transcripts is currently limited to high stringency RNA hybridization blot analysis. Although these data are not proof that these transcripts are derived from the Ank-1 gene, they do establish a relatedness, and as such they are very interesting. Relevant to our finding of smaller transcripts in mouse skeletal muscle and heart is the fact that a 3.4-kb ankyrin transcript was previously detected in chicken myotube and cardiac muscle poly(A)+ RNAs using an independently isolated chicken erythroid ankyrin cDNA  2 and 4 ) . B, RNA filter blot of total cellular RNA probed with '*Plabeled mAnk-1 (lanes 1-4) and Er4/272 (lunes 5-8). Lanes 1 and 5, 4 pg of reticulocyte RNA; lanes 2 and 6, 2 pg of anemic spleen RNA; lanes 3 and 7, 2 pg of cerebellar RNA; hnes 4 and 8 , 4 pg of skeletal muscle RNA. Band sizes were estimated relative to 28 S and 18 S RNAs. C, ethidium bromide-stained agarose gel (1.5%) of the products generated by reverse-transcribed PCR amplification (29) from total cellular RNA using the Er4 unique region oligonucleotide primer pair indicated in Fig. 6. Lone I, 123 as a probe (47). It is interesting to speculate that one of these small transcripts could be responsible for the 43-kDa protein whose putative function is to bind P-spectrin to the cytoplasmic domain of the muscle acetylcholine receptor in the neuromuscular junction (25). A clear description of the primary structure of these transcripts should aid our understanding of the structure and function of the spectrin-based membrane skeleton in muscle cells and resolve the question of the gene(s) responsible for their production.
Our results indicate that the Ank-I gene may produce other transcripts in addition to those described here. It is not clear, for example, whether the bands (-3.5 and 2.0 kb in size) seen in anemic spleen (Fig. 1, lane I ) are the same as those seen in skeletal muscle. Also, there is a small (cl.0-kb) band detected in anemic spleen by the Er4/272 unique region probe (Fig. 7B, lune 6). It is unlikely that all of the Ank-I transcript alternatives have been discovered. In addition, it is unlikely that all tissues expressing Ank-I transcripts have been found. The potential for diversity of transcripts from just the Ank-I gene alone is truly remarkable. This sort of diversity has implications for ankyrin function and lends support to the notion of ankyrins as adaptors that mediate and regulate interactions hetween integral mernhrane proteins and other cytoskeletal and cytoplasmic elements (for review, see Refs. 20 and 21). It will he important to determine whether the Ank-I transcripts we have identified are translated into functional proteins. Our data should aid protein work by alerting investigators to the potential complexity of forms and by defining segments useful for preparing the necessary imrnunological probes.