Alternative extracellular and cytoplasmic domains of the integrin alpha 7 subunit are differentially expressed during development.

Examination of cDNAs for the laminin-binding alpha 7 integrin subunit identified two different sequences (designated X1 and X2) coding for the variable region between the III and IV homology repeat domains near the putative ligand-binding site. Sequencing of a mouse alpha 7 genomic clone established that the X1 and X2 regions are derived by mutually exclusive alternative mRNA splicing. Reverse transcriptase-polymerase chain reaction analysis of alpha 7 mRNA indicated that the X1 and X2 isoforms were present in equal amounts in mouse skeletal myoblasts and adult heart. However, in adult skeletal muscle, the X2 variant was exclusively expressed. Amino acid sequence homologies in the III/IV segment suggest that alpha 3 and alpha 6 are also alternatively spliced at this site. We identified alternatively spliced exons in a human alpha 6 genomic clone that encode X1- and X2-like segments. Analysis of the alpha 7 cytoplasmic domain indicated that this region was also alternatively spliced and like alpha 3 and alpha 6 could exist as the A or B form. In mouse skeletal and cardiac muscle the B form of alpha 7 was strongly expressed. However, we identified alpha 7A in neonate and adult skeletal muscle but not in cardiac tissue. High levels of alpha 7A were detected in differentiating myotubes, but in proliferating myoblasts only the alpha 7B isoform was present. These results indicate that alternative splicing of alpha 7 mRNA is differentially regulated during development and generates variant integrin chains with structurally and presumably functionally unique ligand-binding and cytoplasmic domains.

Examination of cDNAs for the laminin-binding a7 integrin subunit identified two different sequences (designated X 1 and X2) coding for the variable region between the I11 and IV homology repeat domains near the putative ligand-binding site. Sequencing of a mouse a7 genomic clone established that the X 1 and X2 regions are derived by mutually exclusive alternative mRNA splicing. Reverse transcriptase-polymerase chain reaction analysis of a7 mRNA indicated that the X 1 and X2 isoforms were present in equal amounts in mouse skeletal myoblasts and adult heart. However, in adult skeletal muscle, the X2 variant was exclusively expressed. Amino acid sequence homologies in the 1111 IV segment suggest that a3 and a6 are also alternatively spliced at this site. We identified alternatively spliced exons in a human a6 genomic clone that encode X1and X2-like segments. Analysis of the a7 cytoplasmic domain indicated that this region was also alternatively spliced and like a3 and a6 could exist as the A or B form. In mouse skeletal and cardiac muscle the B form of a7 was strongly expressed. However, we identified a7A in neonate and adult skeletal muscle but not in cardiac tissue. High levels of a7A were detected in differentiating myotubes, but in proliferating myoblasts only the a7B isoform was present. These results indicate that alternative splicing of a7 mRNA is differentially regulated during development and generates variant integrin chains with structurally and presumably functionally unique ligand-binding and cytoplasmic domains.
Integrins are a family of transmembrane heterodimeric receptors that mediate cell-extracellular matrix and cell-cell interactions (reviewed in Ref. 1). Each integrin is composed of a noncovalently associated a and P subunit. This pairing of OL and 8 chains is apparently required for both the transport of the receptor to the cell surface and the formation of the extracellular ligand-binding site located near the amino-terminal region of the integrin subunits. To date the CA33834, CA51884, DE10356, and DE10306. The costs of publication * This study was supported by National Institutes of Health Grants 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(sj reported in thispaper has been submitted to the GenBankm/EMBL Data Bank with accession number(s) L23421, L23422, and L23423.
(1 TO whom correspondence should be addressed University of The diversity of the integrin supergene family of receptors has been highlighted by the observations that individual a subunits can pair with multiple fl subunits (1, 2). In addition, several integrin subunits have been shown to be alternatively spliced at the cytoplasmic domain. They include 013, a6, 81, 83, and 84 (3)(4)(5)(6)(7)(8). The A cytoplasmic forms of a3 and a6 have been reported to be phosphorylated on serine and tyrosine residues suggesting a possible differential regulatory and functional mechanism for the two forms (9,10). In addition, alternative splicing in the extracellular domain of PS2a subunit has also been described (11). It was suggested that this alternative splicing in the extracellular domain may be important in determining the specificity and affinity of integrin receptors for their ligands (11).
Previously, we reported that human and mouse melanoma cells express a unique a61 integrin complex that bound to laminin yet contained an a subunit biochemically distinct from all known a subunits (2,12,13). The human receptor as well as the mouse homolog were purified, and the a chain because of its unique NH2-terminal amino acid sequence was designated as a7 (13). This novel integrin complex, a7p1, has been shown to bind to the E8 fragment of laminin and mediate cell adhesion to this ligand (13). Recently, the cDNA for the rat homolog of a7 has been isolated, and its expression was shown to be developmentally regulated during skeletal myogenesis (14,15).
In the present study we show that a7 is alternatively spliced both in the extracellular and cytoplasmic domains. The two extracellular isoforms, designated X1 and X2, of a7 are derived from the mutually exclusive splicing of the a7 mRNA. The X1 and X2 a7 mRNAs are differentially expressed in heart and skeletal muscle. We also show that the alternative splicing in the cytoplasmic domain of a7 gives rise to A and B variants that show similarity to those previously identified for the a 3 and a6 isoforms (3)(4)(5). Furthermore, the expression of the a7A form appears to be developmentally regulated in skeletal muscle. The existence of alternative splicing in a7 transcripts provides for a diverse set of integrin isoforms with structural and potentially functional differences. Murine C2C12 and rat L8 skeletal myoblasts (from Dr. Douglas Cooper, UCSF) were grown in media composed of 20% M199, 80% Dulbecco's minimal essential medium (H-21), and 10% horse serum. To induce myotube formation the C2C12 cells were plated at a density of 5 X 10' cells/cm2 on plastic dishes. The cultures were grown for 2 days before replacing the growth medium with Dulbecco's mimimal essential medium containing 2% fetal calf serum. After 5 days >70% of the cells had fused to form myotubes as determined by phase microscopy.
RNA Isolation and Northern Blot Analysis-Total RNA from mouse tissues was isolated by the guanidinium-isothiocynate/phenol method (16, 17). RNA samples (15 jtg) were electrophoresed in a 1.2% agarose gel containing formaldehyde. After electrophoresis, the RNA was transferred to nylon membranes (Amersham Hybond N) by capillary blotting and fixed to the filter by exposure to UV light. The RNA was hybridized with a 700-bp' PCR-amplified a7 cDNA fragment (see next paragraph) labeled with 32P using a Multiprime DNA labeling kit (Amersham Corp.). Hybridizations were carried out at 42 "C in 50% formamide, 5 X SSC, 5 X Denhardt's, 0.1% SDS, and 300 rg/ml salmon sperm DNA. Filters were washed twice in 1 X SSC, 0.1% SDS at room temperature and once at 65 "C in 0.1 X SSC, 0.1% SDS. Filters were exposed to x-ray film at -80 "C with intensifying screens.
Reverse Transcription and Polymerase Chain Reaction-Total RNA from cell lines, mouse neonates, and tissues was used for reverse transcription-polymerase chain reaction (RT-PCR). Briefly, the RNA sample (5 pg) was heated to 75 "C and then mixed with Moloney murine leukemia virus RNase Hreverse transcriptase (Life Technologies, Inc.), 1 mM dNTPs, 40 pmol of 3' primer, 20 units of RNasin (Promega Corp.), 10 mM dithiothreitol, and 10 X PCR buffer (Perkin-Elmer Cetus Instruments) to a final volume of 20 ~1 .
The reaction was incubated at 42 "C for 90 min, heated at 95 "C for 5 min, and then cooled on ice. The reverse transcriptase cDNA mixture was usually subjected to 30 amplification cycles (unless otherwise indicated) in 10 X PCR amplification buffer (Perkin-Elmer Cetus) containing 0.2 mM dNTPs, 40 pmol of the 5' primer, 40 pmol of the 3' primer, and 2 units of Taq polymerase (Boehringer Mannheim). For each set of PCR reactions a negative control, which contained the mixture above but lacked template, was always included. After denaturation for 5 min at 94 "C, the PCR reaction was amplified in a Perkin-Elmer Cetus thermocycler. Typically, cycles consisted of a 1min denaturation at 94 "C, 30-s annealing at 50-55 "C, and a 2-min extension at 72 "C. After the final cycle the reaction was extended an additional 10 min at 72 "C and stored at 4 "C. Amplified fragments were analyzed in ethidium bromide-stained agarose gels.
The design and synthesis of the 5' primer, used in the amplification of a 700-bp a7 PCR fragment probe that was subsequently used for Northern analysis and screening of the cDNA libraries, were based on the NHz-terminal amino acid sequence of the murine a7 subunit (13). A degenerate primer (A2AR), which is complementary to one of the highly conserved repeat regions (18), the sixth domain, of the a subunits, was used as the 3' primer. Primers used in the amplification of X1 were anti-sense primer 5'-CTATCCTTGCGCAGAATGAC-3' and sense primer 5'-GCCAGGGTGGAGCTCTG-3'; primers for amplification of X2 were anti-sense primer 5"CTATCCTTGCGC-AGAATGAC-3' and sense primer 5"GTGACCAACATTGATAG-CTC-3'; primers used in the amplification of the a7A and a7B cytoplasmic variants were anti-sense primer 5'-CACCGGATGTCC ATCAGGAC-3' and sense primer 5"GCTGCTCAGAGATGCATC c-3'.
PCR fragments from the above reactions were routinely subcloned into pCR-I1 as per the manufacturer's instructions (TA cloning kit, Invitrogen, San Diego, CA) and sequenced to confirm the identity of each PCR product.
Amplification and Cloning of the 5' End of a7 cDNA-Rapid amplification of cDNA ends as previously described (19) was used to isolate and clone the 5"untranslated sequences and signal sequence of the a7 cDNA. Briefly, a reverse transcriptase reaction, as described above, was performed using 5 jtg of total RNA from K1735 cells and 50 pmol of a 20-mer antisense primer (5'-CCTTCTGCACCGT-TAGCTCC-3'). This primer is located -200 bp 3' of the mature protein's amino-terminal sequence, FNLDVM. Excess primer was removed by using a Centron 100 spin filter (Amicon Corp.), and the reverse transcriptase cDNA was polyadenylated with 1 mM ATP and 10 units of Terminal d Transferase (Boehringer Mannheim) for 5 min at 37 "C. A nested antisense primer (5'-CTGTCTCCTCTCGGC-3') along with a 5' anchor primer (5"AAGGATCCGTCGACAA-GCTTGAATTCATACG-(T)11-3') was used to PCR amplify the 5' a7 cDNA. This PCR product was cloned into pCR-I1 (TA cloning ' The abbreviations used are: bp, base pair(s); kb, kilobase pair@); RT-PCR, reverse transcriptase-polymerase chain reaction; PCR, polymerase chain reaction. kit, Invitrogen) and sequenced in both directions.
Screening of cDNA and Genomic Libraries-Standard techniques were used to screen and analyze libraries (20). In brief, recombinant plaques (5 X 10') from a hZap-I1 phage vector cDNA library derived from 12-week-old mouse (BALB/c) hearts (Stratagene Cloning Systems, La Jolla, CA) were plated in duplicate and transferred to Hybond-N nylon membranes (Amersham Corp.). The membranes were probed with a 3ZP-labeled 700-bp PCR-amplified a7 cDNA fragment as described under "Reverse Transcription and Polymerase Chain Reaction." Recombinant plaques (5 X lo6) from a X g t l l cDNA library (gift of Dr. Larry Kedes, University of Southern California) constructed from differentiated C2C12 mouse myoblasts were screened with the same probe. A human heart cDNA library in pcDNA-1 (Invitrogen) was also screened with the PCR-generated a7 probe. Positive clones from the mouse heart cDNA library were plaque-purified after three successive rounds of screening, and pBluescript-SK vectors containing the cDNA inserts were excised and sequenced. The cDNA inserts from plaque-purified positive clones of the myoblast cDNA library were subcloned into pGEM-3Zf (Promega Corp.) and sequenced. Three positive clones from the murine heart library, one positive clone from the myoblast library, and one positive clone from the human heart library were identified and sized by restriction digestion and electrophoresis in 1% agarose gels. The sizes of the three cDNA inserts from the murine heart library, cl-2b, cl-7a, cl-lla, were 0.6,3.5, and 3.0 kb, respectively. The single clone, cl-3m, from the myoblast library was determined to be 0.7 kb in size. The human heart cDNA clone was 1.2 kb in size and, by sequence analysis, found to contain sequences for X2 and repeat regions I11 and IV.
Approximately 1 X lo6 plaque-forming units of a mouse spleen genomic library constructed in a XFix I1 phage vector (Stratagene) were screened with a 32P-labeled 700-bp cDNA probe. This probe contained contiguous sequences in repeat region 111, X2 exon, and repeat region IV of the a7 subunit. Positive clones were plaquepurified, restriction enzyme-digested, and analyzed by Southern blob using the probe above and standard techniques (20). Restriction fragments of interest were subcloned into pBluescript and sequenced. A cDNA probe that contained sequences in repeat region 111, X1 exon, and repeat region IV of the a6 subunit was used to screen a human leukocyte genomic library in EMBL 3 ( g i f t from Dr. Robert Wade, University of Maryland). Positive clones were analyzed as described above.
The cDNA probes used in the screening of the libraries were labeled with [(u-~'P]~CTP using a Multiprime DNA labeling kit (Amersham Corp.). Hybridizations of nylon filters from each library were performed at 42 "C overnight in 50% deionized formamide, 6 X SSPE, 2 X Denhardt's, 0.5% SDS, and 100 pg/ml denatured salmon sperm DNA. The membranes were washed twice in 2 X SSC/O.5% SDS at room temperature and then twice for 20 min at 55-60 "C in 0.1 X SSC and 0.1% SDS. The membranes were then exposed to x-ray film at -80 "C with intensifying screens.
Sequence Analysis-All nucleotide sequences were determined by the dideoxy chain termination methods (21) for double-stranded DNA, using dATP-5'-[a-35S]thiophosphate (Amersham Corp.) and a modified T7 DNA polymerase (Sequenase 2.0; U. S. Biochemical Corp.). Primers for sequencing and PCR were synthesized and purified at the Biomolecular Resource Center, University of California at San Francisco. Greater than 80% of both strands were sequenced from mouse heart cDNA clones l l a and 7a. Sequences were obtained from both strands of all other cDNA, PCR, rapid amplification of cDNA ends, and genomic clones. Any ambiguous areas were resequenced using the nucleotide deoxyinosine triphosphate. Sequence data were analyzed using DNA Strider 1.0 and MBIR programs (Baylor College of Medicine).

RESULTS
Nucleotide and Deduced Amino Acid Sequence of Mouse Heart a7 cDNA-We have previously identified and characterized a novel laminin-binding integrin designated u7Pl (2, 12, 13). Northern analysis indicated that heart tissue, and to a lesser extent skeletal muscle, expressed a7 mRNA (data not shown). To study the a subunit in more detail we isolated the cDNA for a7 using a mouse heart library and C2C12 myoblast library (see "Experimental Procedures"). The nucleotide and deduced amino acid sequence for the a7 subunit, derived from overlapping mouse heart cDNA clones, is shown in Fig. 1. The approximately 3700-nucleotide sequence of the a7 cDNA included one open reading frame, 455 bp comprising the 3'untranslated region, and a presumed polyadenylation and termination recognition signal, ATTAAA, which precedes the poly(A) tail.
The rat a7 cDNA, recently reported by Song et al. (15), lacked the 5"untranslated sequences and the nucleotides encoding the signal sequence. Because a full-length cDNA clone could not be identified from the two libraries screened here, rapid amplification of cDNA ends (see "Experimental Procedures") was used to synthesize and clone the 5' portion of the a7 cDNA. DNA sequence analysis of this fragment demonstrated that it contained 170 bp of the 5"untranslated region of a7 as well as the nucleotides encoding the signal sequence ( Fig. 1

q~t~~~~t~~~t t t~~~~c~a q a~~~a~c~t~q a~a q a c c t p p p p c t c c c~c q c q c g a c q a t t t c c t c q c a gPtaqctpptcc~caqcagcp0qaa~actqaaqaggPgtgctgaqqtgaaaqattgPaaqacc~aaa ~t P p o c~p 0 . t t c~~~~~~~t t t t~t~~~~~~~~~p p~a t c t a~t~t~t t~t t~c t t c c t t q c t c q c~a c t q t t c t t~c c a c c w c t a t M~R 1 p R C D F L R P P G I Y Y L I T S L L A G L F L P P A I A ~t c~~t c t~~t g t g a t p g p t q c c a t a c q c a a p p a~a q a a c c c~c a q c c t a t t c~c t t t t~C C C t g C~C C~8 C 8~t t~C~~C C C~~C C C F N L D V M G A I R K E G E P G S L F G F S V A L E R Q L Q P R~ c a q a q c t g~c t g c t g q~q c p c~c c c c~g P c c c~c~C~~C C~p p~C 8 q c a~C~~8 C C q C~C C P g~~C C~C~~~~C~ Q S W L L V G A P Q A L A L P Q Q Q A N R T G G L F A C P L S L E E T D C Y R V D I D R G A N V Q K E S K E N Q W L G V S V R S Q G cca~caaqattqttacptqtqcacaccqatatgaqtctcqacaqagaqtppacca~ctttp0aqactc~atgtgat~ccqctgctttg~ A G G K I V T C A E R Y E S R Q R V D Q A L E T R D V I G R C F V
Ctl.qCC.op.CCtppCCatCCq~atqaqC~atWt~aqtpp8aqttCtqtqa~CppCCCCa~CC8tqa8Caqtt~ttC~tC8q

actqagatccactttctqaaqcaappctqcpptcaaqata8qatctgtc8qagcaacctccagctcqaqc~gtaccagttctqttccaopatcagcqac T E I E F L K Q G C G Q D K I C Q S N L Q L E R Y Q F C S R I S D I C l q~q t t C C 8 q q C t C t g C C C 8~a t C t P g~t~~~P g~C C g C C C t g t t C g C~t~~q t~C a q C C q t t C 8 t 8 q q C C t g g~~C t g a C~q t C~C C~~C T E F Q A L P M D L D G R T A L F A L S G Q P F I G L E L T V T N q i p B C l g~t~t 8 C a q~g +~8 C a t~a C~q~8 q C t~~C~q C a g~8 p p~q~g C~~W~q 8 a C C~~~C~~q t C~q~
C a~C . C r q C t q C C~C C t t C t C C C C t q~C a q t C a C t~C C t c p t C t t t g P g P C t C C a p 0 8 8 C C t~t~~C~8~~t t q C t t t t~q~C C~C~t t a~a~a q~a q a c a g g c a c c a t c c a g a p 0 a q t a a c~g c a a c t c c c a q~a~c t c c q a c q c a c a c c c  a7 subunit (13,15). The sequences surrounding the presumed initiator codon for a7, GATCCCATGG, are very similar to that identified in the mouse and human a6 subunit (3,22). The context of such initiator sequences is considered rare and is thought to play a role in modulating the yield of proteins that exhibit such unfavorable initiation sites (23). We have subsequently confirmed the rapid amplification of cDNA end results by sequencing the corresponding region contained within a 17-kb mouse genomic clone containing the a-promoter region.2 Translation of the a7 cDNA indicated that the amino terminus of the mature protein is proceeded by a signal sequence of 34 amino acids. The mature protein is comprised of 1102 amino acids, which differs from the 1106 amino acids reported by Song et al. (Ref. 15 and see below). The extracellular domain consists of 1002 amino acids followed by a single hydrophobic stretch of 24 amino acids that represents the transmembrane domain. Two putative protease cleavage sites, RRRRE and RRQ, are located in the extracellular domain ( Fig. 1, underlined residues). In addition, there are only five potential N-linked glycosylation sites (Fig. 2, arrow marked residues). The cytoplasmic domain (77 amino acids residues) of this a7 cDNA clone shows strong homology to the a3 and a6 B cytoplasmic variants (3,5 ) indicating that this clone is the B cytoplasmic isoform of the a7 subunit (see below).

K E E K T G T I Q R S N W G N S Q W E G S D A E P I L A A D W E P qaqctqggtcctqatqgacatccgP~ccagccactgcctaacatctgctg+cccaaqcccagcc~~ttccttt~cccttttccaaga~ctcc
The overall amino acid homology of a7 is the highest with the a6 subunit (47% identity) followed by a3 (37% identity; Fig. 2). When the deduced amino acid sequences of a3, a5, and a6 are aligned with that of a7 (Fig. 2), regions of strong and weak homology, especially in the extracellular domain, are revealed. All a subunits contain seven conserved repeat domains (I-VII) in the extracellular segment, which share 25-52% identity (24,25). The homologies of the repeat domains in a6 and a7 are particularly striking, averaging >66% identity when the two subunits are compared. Most of the variable region segments of a6 and a7, which are located between the repeat domains, also show strong homology. These segments of a6 and a7 between repeats 1/11, II/III, IV/V, V/VI, and VI/ VI1 have an average identity of 50% (ranging from 34.5 to 65%).
Evidence for Alternative Splicing of a7 in the Extracellular Domain-Sequence analysis of the heart a7 cDNA revealed an unusual 43-amino acid stretch between repeats III/IV in a7 that was essentially nonhomologous to the corresponding region in a6 and differed in length by four amino acids (Fig.  2, amino acid residues 194-227). All four a7 cDNA clones from two different libraries contained identical sequence overlaps in this region. Furthermore, a human a7 cDNA clone from a heart library also encoded the same exact sequence in this segment between the 111 and IV repeats (see below). The recently published sequence for rat a7, derived from an L8 myoblast cDNA library, had a completely different sequence for the III/IV repeat segment, which in contrast to our sequence was highly homologous at the amino acid level to the corresponding region in a6 (15). Additionally, this region in the L8 myoblast cDNA was 4 amino acids longer than that in the heart a7 cDNA. A second region of approximately 19 amino acids (residues 45-64) located in repeat region I of the heart a7 cDNA was also strikingly different in the L8 myoblast a7 cDNA. Our sequence has been confirmed in both a library-derived a7 cDNA clone and a genomic a7 clone (15). The remaining amino acid sequence deduced from the heart a7 cDNA shared greater than 91% identity with that reported for the L8 myoblast a7 cDNA (15). The lack of homology in the variable regions of a6 and a7 in the III/IV segment and B. L. Ziober, et al., manuscript in preparation. the identification of the two nonhomologous sequences for the III/IV segment in the L8 myoblast and heart a7 cDNAs suggested that the sequence of the III/IV segment is encoded by two alternatively spliced exons. This view is supported by the finding that the aPS2 fly integrin subunit is also alternatively spliced in this same variable region ( 1 1 ) .
To prove the existence of alternative splicing in the III/IV segment of a7 it is important to determine if separate exons for the regions in question are contained within the ~y7 gene. Therefore, a mouse genomic library in XFix I1 was screened with a mouse cDNA fragment, which contained sequences that overlapped the 111-IV repeat domain. One positive clone was isolated, plaque-purified, and determined to be -17 kb in size. A subclone (designated clone G9), which by Southern blot analysis was shown to contain sequences between repeats 111-IV, was examined further.
Clone G9 was directly sequenced to identify the intron/ exon organization of the region spanning repeats I11 and IV in the a7 gene (Fig. 3, A and B). The sequence for repeat I11 domain is contained in a single exon of 194 bp, which is followed by a -500-bp intron. Adjacent to this intron is an open reading frame of 132 bp that encodes an amino acid sequence (43 residues) with a high degree of homology to the published sequences of the III/IV segments of human, mouse, and chicken a6 and is identical to the corresponding region reported for the rat L8 myoblast a7 (3,15,22,26). We have designated this exon X1, for extracellular (Fig. 3, A and B, and Fig. 4 ) . Exon X1 is followed by nearly 400 bp of intronic sequences and a second exon. This exon (designated X2) encodes for a 39-amino acid segment that is completely distinct from the X1 sequence and the III/IV segment of a6 but is identical to the connecting segment between the repeat domains obtained from the mouse heart cDNA clones (Fig.  4). Thus, X1 codes for four more additional amino acids than the X 2 exon; depending on which exon is spliced in, these result in mature proteins of 1106 and 1102 amino acids, respectively. Exon X2 is separated from a 205-bp exon that encodes repeat domain IV by a -900-bp intron. Each of the intron/exon boundaries analyzed conformed to the consensus 5' and 3' splice sequences, GT/AG (27) (Fig. 3B). These results establish that the alternative forms of a7 obtained from the L8 myoblast and heart cDNAs arise through the mutually exclusive splicing of either one of two adjacent but distinct exons, X1 or X2.
Evidence of Alternative Splicing of a6 in the Extracellular Domain-The high degree of overall homology of a6 and a7 (47% identity) suggests that the a6 subunit is spliced in a similar fashion as a7. Using a similar approach as that for a7 we isolated a human genomic clone for the a6 subunit that contained the III/IV segment. Exon X1 was identified by sequencing this clone with a primer derived from the sequences between repeats III/IV of the a6 cDNA ( 3 , 5 , 2 2 ) . We identified the X2 exon in a6 by sequencing the same clone with a primer generated from the nucleotides that encode the common amino acid residues, NYSL (Fig. 4 ) . ' Alignment of the amino acid sequences for the X1 and X2 isoforms of a6 and a7 from different species revealed some potentially important conserved motifs for the alternatively spliced isoforms (Fig. 4 ) . Between a6 and a7 the X1 exon is over 70% identical. Two major highly conserved, but distinct, regions with consensus sequences, DDGPYEXGGE and LXPVPANSYL, are located in the X1 segments of a6 and a7; a shorter sequence, XARE, at the beginning of the exons is also conserved (Fig. 4). In the X 2 segments of a6 and a7 two conserved regions with consensus sequences, (S)DPDQXVYK and NSYL, can be identified (Fig. 4) a3B (281, human a6B (3,60), human a5 (61), and mouse a7 are compared by computer-generated alignments.
To maximize homologies, gaps were introduced, and alignments were optimized slightly by hand. The seven conserved repeat domains are boxed and numbered (Roman numerals), the transmembrane domains and conserved sequences, GFFKR, are boxed. All cysteine-conserved residues are marked and potential glycosylation sites are indicated with arrows.

EXON XZ
...  consensus regions in X1 are distinct from those in X2 suggests that these may be sites of functional divergence for the two isoforms.

A P R A N H K G A V V I L R K D S A T R L I P E T T G T G C T G T C T C G G G A G C G C C T G A C C T C T G G C T T T G O C T A G V V L S G E R L T S G F G Y S L A V T D L N
Expression of the X1 and X2 Extracellular Alternative Splice Form in a7 mRNA-To estimate the relative abundance of X1 and X2 alternatively spliced forms we used RT-PCR analysis. After isolation, total RNA from various tissues and cultured cell lines was reverse transcribed, and the reaction was divided in two and then amplified by PCR with a primer specific to exon IV and a primer specific to either of the alternative exons (X1 or X2). The primer site in X1 was near the 5' end of the exon, while the binding site of the primer to X2 was near the middle of this exon (see "Experimental Procedures"). Thus, when amplified, the X1-IV primer pair resulted in a product of -220 bp, while primer pair X2-IV generated a 200-bp fragment. In addition, because the binding sites of the primer pair are located in separate exons, potential amplification of contaminating DNA is identified as bands >1 kb. These fragments were readily identified when electrophoresed through a 2% agarose gel (Fig. 5 ) . This assay as-sumes that the efficiency of PCR amplification for these two sequences is similar and that saturation of the products has not occurred using limited rounds of PCR amplification. The latter was confirmed by analyzing the products after 22, 25, and 30 cycles of amplification (data not shown).
Adult mouse heart and lung strongly express both mRNA forms of a7 with somewhat higher levels of X2 (Fig. 5). In contrast, for adult mouse skeletal muscle the X2 transcript was exclusively expressed (Fig. 5 ) . Neither a7 variant could be detected in RNA isolated from mouse small intestine (Fig.  5). Overall, adult mouse skeletal muscle consistently had lower levels of PCR-generated X2 fragments than cardiac muscle; this agrees with the results for total a7 RNA transcripts in these tissues assessed by Northern analysis (data not shown). Proliferating rat L8 and mouse C2C12 skeletal myoblasts expressed approximately equal levels of X1 and X2 (Fig. 5). PCR analysis indicated only a slight increase in the ratio of X2/X1 in C2C12 myotubes (not shown). Identical results were obtained for all tissue samples and cultured cells when the PCR reactions were repeated using poly(A+) RNA.   I11 and IV of a3, a6, and a7 integrins. The deduced amino acid sequences of the variable regions containing the alternatively spliced exons X1 or X2 between conserved repeats I11 and IV from the hamster and human a 3 cDNAs (~~3 x 2~" and C Y~X~~" , respectively (28, 29, 30)), chicken and mouse a6 cDNAs ( (~6 x 1 '~ and a6Xlm0, respectively (27,22)), rat a 7 cDNA (a7X1" (15)), human a6 exons X1 and X2 (a6Xlh" and a6X2h"),2 human a 7 cDNA (a7Xlh"; see "Experimental Procedures"), and mouse a7 exons X1 and X2 (a7Xlmo and a7X2"" from Fig. 3 B ) are compared by alignment. Amino acid residues are shoded if two or more integrin segments within X1 or X2 groups are either identical or have conservative substitutions. Procedures" (amplification 25 cycles) on mouse lung tissue, mouse small intestine, mouse heart, mouse adult skeletal muscle, mouse C2C12 myoblasts, and rat L8 myoblasts. The -220-bp fragments correspond to the X1 extracellular domain isoform, while the -200bp fragments correspond to the X2 isoform. The PCR products were separated on a 2.0% agarose gel and stained with ethidium bromide. Marker is 100-bp ladder (200-400-bp markers are shown).
T o verify the PCR results, products were routinely subcloned into pCR-11, and the ends of each fragment were sequenced. In addition, to confirm that the mRNA for a 7 X2 variant encoded a protein product, a polyclonal antiserum, Ab915, directed against the unique peptide sequence in the X2 domain was generated. This antiserum reacted strongly with the a 7 chain, obtained from lysates of various cell types, after analysis by SDS-polyacrylamide gel electrophoresis and Western blotting (data not shown). Evidence for Alternative Splicing in the Cytoplasmic Domain of a7"Alternative mRNA splicing has recently been reported in the cytoplasmic domains of the a7-related integrins, a3 and 06, to produce the A and B forms (3,5). Comparison of the amino acid sequence, derived from the heart cDNA clones, for the cytoplasmic domain of a 7 with that of a6 and a3 indicates that this is the B isoform homolog (Fig. 6C). Using an approach similar to that used for a6 and a3 (3, 5), we investigated whether an alternative A cytoplasmic domain existed for a7.'A set of PCR primers specific for a 7 was designed with the 5' primer corresponding to nucleotides 3198-3218 and the 3' primer directed to nucleotides 3546-3566 (Fig. 6B). This set of primers flank a region 5' to the transmembrane domain and to the end of the coding sequences of a7. A series of rodent tissues and cell lines was examined by RT-PCR using these primers. A major band of -370 bp was usually present after RT-PCR, to varying degrees, in a number of tissues and cell types examined (Fig.  6A). This PCR product represented the B form of a7 as determined by sequencing analysis. A second band of -480 bp, which was determined to be homologous to the A cytoplasmic isoform, was present in differentiated C2C12 mouse myoblasts, adult mouse muscle, and whole mouse neonates (Fig. 6A). RNA from proliferating undifferentiated L8 and C2 myoblasts also exhibited the A isoform albeit a t much lower levels, while adult heart and mouse melanoma cells (K1735) had undetectable levels of the A form after amplification. An additional band, which ran slightly above the 480-bp band in cells expressing the A isoform, was shown to be a heteroduplex resulting from re-annealing single strands of the 370-and 480-bp fragments? A similar hybrid band was seen in a 3 and a 6 (3, 5). The 480-bp fragment was subsequently cloned, sequenced, and found to encode a unique amino acid sequence (57 residues) that was distinct from a7B yet showed strong homology to the cytoplasmic A isoforms of a3 and a6 (Fig. 6, B and C). As with a6A and a3A, the A isoform of a 7 represents an insertion immediately after the transmembrane domain (Fig. 6B). However for a3 and a6, the entire cytoplasmic A isoform appears to be generated by a putative exon that is either spliced in or out (3,5). In contrast, the A form of a 7 is encoded by a contiguous sequence, which includes not only the 113-bp alternatively spliced segment but also a 59-bp frame shifted sequence that is shared by the a7B form.
Alignment of the cytoplasmic domain sequences for the A and B forms of a3, a6, and a7 reveals major areas of sequence similarity within each variant group (Fig. 6C). While all a chains share a homologous GFFKR sequence near the transmembrane domain, there is little if any cross-isoform identity. The A forms have extensive homology particularly between a3 and a6 and, to a lesser extent, a7.   number L23421. C, alignment of the amino acid sequences of the a3, a6, and a 7 A and B cytoplasmic isoforms. The deduced amino acid sequences of the related A and B isoforms of a3, a6 (3,5), and a 7 are compared. Similar residues, either identical or conservative substitutions, in two or more a chains are shaded. Conservative changes are defined as in Fig. 4. Amino acid residues common to a3, a6, and a 7 isoforms are marked ( I ).

Alternative Splicing in a7 Integrin mRNAs 26781
interest because it contains a high density of charged and mostly acidic amino acids. Further, toward the COOH-terminal end are several additional areas of sequence overlap, which are also charged residues. The long additional segments found in the COOH terminus of a7A and B share no homology to the isoforms of a3 and a6 or to each other. Finally, a7A has several potential tyrosine and serine phosphorylation sites that may be functionally important.

DISCUSSION
We have provided evidence that the a7 integrin subunit is alternatively spliced in both the extracellular and cytoplasmic domains. This allows for the generation of variant a7 subunits that are structurally and presumably functionally distinct. Furthermore, the pattern of alternative splicing at both sites appears to be developmentally regulated and tissue specific. This is the first example of alternative splicing in the ligandbinding domain in a vertebrate a integrin subunit. On the other hand, the alternative spliced segments at the cytoplasmic domain of a7 are analogous to that previously described for the A and B forms of a3 and a6, confirming that the three subunits belong to a closely related integrin subfamily. The differential pattern of alternative splicing of a7 during skeletal muscle development indicates that different regulatory mechanisms are operational for X1/X2 and A/B isoforms.
Alternative Splicing of a7 in the Extracellular Domain-The evidence supporting the existence of the alternatively spliced region between conserved repeats 111-IV in the extracellular domain is substantial. Sequences corresponding to the X1 and X2 segments have been cloned from mouse (Fig.  l), human (Fig. 4), and rat cDNA libraries (15). More importantly, we have confirmed the presence and ordering of the X1 and X2 segments in a genomic clone (Fig. 3). It is clear that the two different extracellular isoforms of a7 arise through the mutually exclusive splicing of either one of two adjacent exons, X1 or X2. Finally, we have detected the expression of the X1 and X2 forms by RT-PCR in mRNA isolated from tissues and cultured cells.
The process of alternative splicing in the III/IV variable region appears to be developmentally regulated. The evidence for this was provided by estimating the relative levels of X1 and X2 in transcripts from tissues and cell lines using RT-PCR (Fig. 5). In adult mice, the relative ratio of X1 to X2 was tissue specific with some tissues lacking significant levels of a7 mRNA (e.g. intestine) while others had comparable levels of X1 and X2 (e.g. heart) or a predominance of X2 (lung). Undifferentiated myoblast cells had similar levels of X1 and X2, but only the X2 form was detected in mature skeletal muscle.
Two other a subunits have recently been shown to be alternatively spliced in the extracellular domain. A 102-bp exon was determined to be either spliced in or out of the transcripts for aIIb mRNA (31). This splicing event occurred in the extracellular domain just 5' of the transmembrane region. The Drosophila aPS2 subunit that complexes with the BPS subunit and binds vitronectin (32) has been shown to be alternatively spliced between domains I11 and IV (11). However, unlike a7, alternative splicing in aPS2 involves splicing into an internal splice acceptor site, which results in the deletion of a portion of the exon. These examples of alternative splicing in the extracellular domain of several different subunits suggest that such processing of a transcripts may be more widespread than previously believed. In support of this conclusion, we found that a6 (Fig. 4) is also alternatively spliced in a similar manner between the 111-IV domains to yield XI-and X2-like isoforms. The published a3 subunit contains an X2-like sequence (Fig. 4) at the III/IV segment, which suggests that this integrin may also be alternatively spliced at this site.
We can only speculate as to how alternative splicing at X1/ X2 could influence integrin function. We hypothesize that this region is involved in defining ligand specificity and/or affinity. Several observations support this idea. The 111-IV domain segment neighbors the metal-binding sites (see Fig.  3), a region known to be essential for maintaining integrin subunit association, active receptor conformation, as well as ligand specificity and affinity (reviewed in Ref. 1). Also, the I domain insert, which is believed to be important in ligand binding (33), borders the 111-IV splice site. Furthermore, a large portion of the NHp-terminal region including the 111-IV segment and metal-binding domains of the a subunit has been implicated in ligand binding. In these studies, with avo3 and aIIbP1 as model integrins, photoactivatable cross-linker RGD peptides identified a number of sites that interacted with the probes (34, 35). Finally, recent work with aV/aIIb chimeras indicates that both the NH2-terminal segments and the metalbinding sites are important for ligand-binding specificity.4 Because all integrin subunits are highly homologous, it is likely that the ligand-binding site for a7 is also defined in this region.
a6 and a7 are the only D l integrins that bind laminin exclusively (13, 36-38). Both a6 and a7 bind specifically to the E8 fragment of A-chain laminin. Yet the two integrins differ remarkably in their affinity for laminin. Whereas a7pl efficiently binds to immobilized laminin, a6pl binds poorly and readily elutes with physiological salt concentrations (2, 3, 12, 13, 39). We suspect that the alternatively spliced Xl/ X2 segment is one region that is important in defining these activities. Preliminary evidence obtained with RT-PCR indicates that the X1 form of a6 is more commonly expressed in cultured cells, and consequently it is this form that has been identified in cDNA libraries (3, 5) also derived from cultured cells. For the a7 subunit isolated from melanoma cells, it is the X2 form that predominates, and it is this form that binds with high affinity to laminin as detected with an anti-X2 antibody: Moreover, X1 and X2 may differ not only in their affinity for laminin but in their specificity for laminin isoforms. This possibility is particularly important because of the number of tissue-specific laminin isoforms that have recently been identified, including s-laminin (40), merosin (41), s-merosin (42), k-laminin (43), and kalinin or epiligrin (44,45). It is now evident that integrin receptors bind differently to these macromolecules. Both a 3 and a6 have been shown to bind different laminin isoforms including A-chain laminin (30, 38, 46-48) and epiligrin/kalinin (45, 49, 50). Moreover, a3 can apparently interact with fibronectin (30, 51, 52), collagens (30, 51, 52), and entactin (53) and is also involved in cell-cell adhesion (54). Such a diverse range of ligand binding properties could be related to the existence of X1/X2 isoforms for a6 and probably for a3.
Alternative Splicing of a7 in the Cytoplasmic Domain-The cytoplasmic domains of integrins interact with the cytoskeleton and potentially other cytoplasmic proteins. It is through this interface with the cytoplasmic compartment that signals are transmitted to the cell. In addition, it is now apparent that functionality of the extracellular domain of the integrin is influenced by "inside-out signaling" (55 the existence of different cytoplasmic splice variants could be important in regulating the quality and strength of signal input from the extracellular space and vice versa. The a7 cytoplasmic domain sequence encoded by clones isolated from the heart cDNA library described here is the B isoform of a7. The a7A cytoplasmic isoform has strong homology to the A isoforms of a3 and n6 (3, 5), but alignment of the A and B forms of a3, a6, and a 7 shows only mild homology (Fig. 6C). The cytoplasmic domain of a7A and a7B is generated somewhat differently than for the A and B isoforms of a3 and a6. For a3A and a6A, the cytoplasmic domains are derived from a single putative alternatively spliced exon, which encodes the unique A isoform domain, a translation stop codon, and 3"untranslated sequences. In contrast, the A isoform of a7 is encoded by a contiguous sequence that includes not only the 113-bp alternatively spliced fragment but also a 59-bp segment that is also shared with the a7B isoform. Insertion of the alternatively spliced fragment results in a frame shift that gives rise to the translation stop codon for a7A.
How the A and B cytoplasmic variants of a3, a6, and a7 may regulate function of the integrins has not yet been established. Hogervorst et al. (9) have recently demonstrated that the A cytoplasmic domains of a6 and a3 are phosphorylated on serine and weakly on tyrosine residues, while the B isoforms are not phosphorylated (9). It has been shown that phosphorylation of the a6A inhibits ligand binding activity (50) and that a6A expression is associated with embryonic development (22). We have shown here that the progression from myoblast to myotubes, to neonate, and finally to adult muscle is paralleled by increases in a7A. This suggests that for a7, isoform switching may be an important mechanism in skeletal muscle differentiation. Such regulatory mechanisms could be associated with ligand binding, cytoskeletal interactions, signal transduction, or a combination of these events (9). The similarity of a6 with a7 and the number of consensus sites for serine and tyrosine phosphorylation in the a7A cytoplasmic domain (57) indicate that, like a6A, this isoform may also be regulated by phosphorylation. Finally, whether X1 or X2 is preferentially spliced with A or B has yet to be determined. However, it has been shown that either X1 or X2 can be spliced together with the B cytoplasmic isoform (15) ( Fig. 1).
The finding that alternative splicing at the extracellular and cytoplasmic domains is found in the same regions of a6 and a7 (and probably a3) makes it likely that there is a functional role for the proteins that are encoded by these mRNA isoforms. Furthermore, that the amino acid sequences in both X1 and X2 are highly conserved among several different species also argues for an important operational role for these domains. Alternative splicing in numerous other proteins has been shown to be a way of conferring multiple protein functions from a single gene (reviewed in Refs. 58 and 59). As shown here a7 is spliced both in the extracellular and cytoplasmic domain. Such splicing has the potential for multiple variant forms that could modulate function as needed whether for ligand affinity or specificity (Xl/X2) or cytoplasmic interactions and signaling (A/B).