Cell shape-dependent regulation of protein 4.1 alternative pre-mRNA splicing in mammary epithelial cells.

Expression of the complex gene encoding multiple isoforms of structural protein 4.1 is regulated by alternative pre-mRNA splicing. During erythropoiesis, developmental stage-specific inclusion of exon 16 generates protein 4.1 isoforms having a fully functional spectrin-actin binding domain. Here we show that human mammary epithelial cells (HMEC), coincident with the dramatic morphological changes induced by altered culture conditions, exhibit a novel pre-mRNA splicing switch involving a new exon (exon 17B, 450 nucleotides) in the COOH-terminal coding region. 4.1 RNA expressed in proliferating HMEC adherent to culture dishes mostly excluded exon 17B, whereas 4.1 transcripts processed in nondividing suspension cultures of HMEC strongly included this exon. This pre-mRNA splicing switch was reversible: cells transferred from poly(2-hydroxyethyl methacrylate) back to plastic resumed cell division and down-regulated exon 17B expression. More detailed studies revealed complex tissue-specific alternative splicing of exon 17B and another new exon 17A (51 nucleotides). These results predict the existence of multiple 4.1 protein isoforms with diverse COOH termini. Moreover, they strongly suggest that regulation of gene expression during differentiation of epithelial cells is mediated not only by transcriptional mechanisms, but also by post-transcriptional processes such as alternative pre-mRNA splicing.

Under controlled conditions in culture, mammary epithelial cells exhibit dramatic changes in morphology and gene expression which mimic certain aspects of differentiation. Proliferative cells adherent to culture dishes assume a flattened polygonal phenotype and fail to express certain differentiated gene products characteristic of mammary epithelia in vivo. In contrast, both mouse and human mammary epithelial cells can be induced to re-express differentiated markers by culture on a complex extracellular matrix substrate or simply by culture in suspension (1)(2)(3). Such treatments cause cells to cease dividing, assume a more rounded morphology, cluster together, and reorganize the cytoskeletal and nuclear architecture. The processes of morphological and functional differentiation are thus intimately related (4,5).
Previous studies have demonstrated that one major mecha-nism for regulation of mammary epithelium-specific gene products operates at the level of transcription initiation. The murine genes encoding ␤-casein and whey acidic protein are not expressed at the RNA or protein level in cells growing on plastic; however, treatment with prolactin and extracellular matrix activates transcription of these genes (6 -8).
We are exploring whether differentiation of epithelial cells can be mediated not only by transcriptional control mechanisms, but also by post-transcriptional processes. Specifically, pre-mRNA splicing might also play a critical role in regulating gene expression in these cells. There are several ways in which this could be envisioned to operate. First, alternative splicing might directly alter the structure of a transcription factor, which could secondarily regulate a specific subset of epithelial genes. It is well known in other cell systems that alternatively spliced transcription factor isoforms may exhibit opposite effects on gene expression (for review, see Ref. 9). Alternatively, changes in splicing factor activity could mediate changes in pre-mRNA processing, thus effecting structural and functional changes in proteins even in the absence of any obvious differences in transcription of the gene.
To explore the hypothesis that differentiation-associated switch(es) in pre-mRNA splicing may play an important role in epithelial differentiation, we are using as a model system the complexly spliced gene encoding the structural protein, protein 4.1 (10). The prototypical ϳ80-kDa protein 4.1 polypeptide is a key component of the erythrocyte membrane skeleton which plays a critical role in maintaining membrane structural integrity via its stabilization of spectrin-actin linkages within the plane of the membrane skeleton (11,12). It also interacts vertically with the integral membrane proteins, band 3 (13,14) and glycophorin C (15)(16)(17), leading to the intriguing proposal that 4.1 in non-erythroid cells may link the internal actin cytoskeleton to the overlying plasma membrane. In addition, novel intracellular functions for 4.1 are suggested by recent studies showing (a) that multiple structural isoforms of protein 4.1 are encoded via a complex pattern of alternative pre-mRNA splicing (18 -21), and (b) that individual members of this family may exhibit dramatic changes in binding interactions (12,(22)(23)(24) and/or intracellular localization (25)(26)(27)(28).
Extensive studies have revealed that changes in 4.1 gene expression during erythropoiesis, mediated at the level of pre-mRNA processing, play an important role in regulating the function of 4.1 protein within the erythroid membrane skeleton. The splicing machinery in erythroid precursors excludes exon 16 from mature 4.1 mRNA, leading to synthesis of 4.1 isoforms lacking a competent spectrin-actin binding domain. However, a switch in pre-mRNA splicing patterns at a late stage of erythroid development leads to efficient exon 16 inclusion (18,20,29,30), expression of a functional spectrin-actin binding activity, and increased mechanical stability of the erythroid plasma membrane (24). The nature of the extracellular signal that induces the switch in 4.1 splicing during erythropoiesis is unclear at this time.
In this paper we have examined 4.1 expression in proliferating adherent versus quiescent suspension cultures of human mammary epithelial cells (HMEC) 1 to investigate differentiation-associated switches in pre-mRNA splicing which might alter 4.1 structure and function in epithelial cells. Unexpected complexity in expression of multiple COOH-terminal isoforms of protein 4.1 was suggested by the discovery of two novel exons encoding peptides of 17 amino acids (exon 17A) and 150 amino acids (exon 17B) in 4.1 cDNA. Remarkably, exon 17B (E17B) expression was increased dramatically upon culture of the mammary epithelial cells in suspension. This pre-mRNA splicing switch predicts the existence of novel 4.1 isoforms having extended COOH-terminal domains and suggests that these 4.1 isoforms may play novel structural role(s) in differentiated epithelial cells. Moreover, these results suggest that molecular regulation of gene expression during HMEC differentiation includes not only transcription initiation but also pre-mRNA splicing.

EXPERIMENTAL PROCEDURES
Cell Culture-Normal HMEC derived from reduction mammoplasty were grown in medium MCDB (Clonetics Corp., La Jolla, CA) supplemented with epidermal growth factor (Upstate Biotechnology, Inc., Lake Placid, NY), bovine pituitary extract (Hammond Cell Technologies, Palo Alto, CA), insulin, hydrocortisone, transferrin, and isoproterenol (Sigma) as described previously (31). Poly(2-hydroxyethyl methacrylate) (poly-HEMA) was obtained from Sigma and dissolved at 12 mg/ml in 95% ethanol. Tissue culture plates were coated at 0.8 mg/cm 2 and air dried as described earlier (4). 184 HMEC were grown until 9th passage on ordinary tissue culture plates, then trypsinized and transferred to either untreated or poly-HEMA-treated plates for 2 days. The cells plated on poly-HEMA-treated plates formed multicellular aggregates that did not attach and remained in suspension. In experiments to the test reversibility of HMEC phenotype, cells were cultured for 2 days on plastic, 2 days on poly-HEMA, and then 2 additional days on plastic again. By 48 h after replating on plastic, the majority of cells had reattached and spread.
RNA Isolation-Human reticulocyte total RNA was purified from blood of sickle cell anemia patients as described previously (32). 184 HMEC were harvested in buffered guanidine thiocyanate solution and RNA purified by cesium chloride ultracentrifugation (33).
Western Blotting-Analysis of protein 4.1 isoform expression in erythroid and HMEC was performed by techniques described previously (20,34). Proteins were quantitated using the Bio-Rad protein assay, fractionated by 8% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (35), transferred to nitrocellulose filters, and probed with affinity-purified rabbit IgG directed against synthetic peptides representing various 4.1 structural domains. Antibody 10-1 is directed against a 21-amino acid peptide in the spectrin-actin binding domain, encoded by exon 16; antibody 24-2 is directed against a 34-amino acid peptide in the COOH-terminal domain encoded by exon 19. Immunoreactive species were visualized using goat anti-rabbit IgG and enhanced chemiluminescence using reagents from Amersham Corp.

RESULTS
To characterize the 4.1 mRNAs expressed in HMEC and to assay for switches in 4.1 pre-mRNA splicing which may accompany differentiation, RNA was prepared from cells maintained under adherent (Fig. 1A) and suspension conditions (Fig. 1B) and subjected to RT-PCR analysis. Several primer pairs were utilized to amplify various coding domains of 4.1 mRNA (Fig.  2). Expression of sequences encoding the NH 2 -terminal extension and the 30-kDa membrane binding domain was examined using primers in exons 2 and 6 ( Fig. 2, lower left). No changes in 4.1 splicing patterns were observed in this region, among 4.1 mRNAs extracted from adherent proliferating HMEC (PL) or suspended quiescent HMEC (PH). RNA from both sources yielded a major product including exons 2 and 4 -6 (skipping only the noncoding exon 3) and minor products skipping exons 4 and/or 5. These products are identical to those characterized previously in studies of reticulocyte 4.1 mRNA (20).
Next, splicing patterns were examined among exons encoding the spectrin-actin binding domain, using primers in the flanking exons 13 and 17. Proliferating HMEC expressed a major product including exon 16 and a minor isoform skipping exon 16 (middle panel, lane PL). In contrast, the exon 16 skipping product was reduced to almost undetectable levels in suspension cultures of HMEC, reflecting a partial switch in the splicing pattern of 4.1 pre-mRNA (PH). This increased efficiency of exon 16 inclusion predicts that a higher proportion of the 4.1 protein in the suspended HMEC should bear a functional spectrin-actin binding domain.
Alternative splicing of 4.1 exons encoding the COOH-terminal domain was explored by amplifying 4.1 mRNA sequences with primers spanning exons 17-22. Unexpectedly, this experiment revealed a major differentiation-associated splicing switch in HMEC. Fig. 2 shows that proliferating HMEC expresses primarily 4.1 mRNA isoform(s) including exons 17-22 (lower right panel, PL). In contrast, RNA from HMEC on poly-HEMA yielded a novel ϳ750-nucleotide product not previously appreciated in studies of reticulocyte or proliferating HMEC RNA and not explained by any known exons of the 4.1 gene (PH). A less abundant ϳ800-nucleotide product was also visible in some experiments. Sequence analysis revealed that these amplified products contained novel 51-and 450-nucleotide motifs inserted between exons 17 and 18 (presented in Fig. 7 below). Characterization of genomic clones confirmed that these inserts represent new exons (data not shown), which are now assigned as exons 17A (51 nucleotides) and 17B (450 nucleotides), respectively, because of their position between the previously designated exons 17 and 18. This experiment demonstrated that a splicing switch involving E17B was correlated with the morphological changes induced by culture of HMEC in suspension. Moreover, it revealed the existence of at least two novel isoforms of 4.1 mRNA generated via expression of two previously unrecognized alternative exons.
To determine whether induction of the E17B splicing switch was reversible, HMEC were transferred from plastic to poly-HEMA and back to plastic. Under these changing culture conditions, HMEC exhibited reversible changes in morphology and proliferation (data not shown). Concomitantly, efficiency of E17B inclusion underwent a reversible splicing switch, being almost entirely excluded in proliferating HMEC but efficiently included in cells on poly-HEMA (Fig. 3A).
The specificity of the E17B splicing switch was explored by examining alternative splicing of several additional pre-mRNAs in adherent versus suspension HMEC. Fig. 3B shows that no differentiation-associated switches in pre-mRNA splicing were detected among three alternative exons of fibronectin pre-mRNA. Neither were any changes observed in pre-mRNA encoding the human homolog of Drosophila discs large (hdlg; data not shown). This finding indicates that the pre-mRNA splicing switch is quite specific and therefore likely reflects change(s) in selected splicing factor(s) rather than a more general modulation of the splicing machinery.
The results described above indicate that numerous COOH- terminal isoforms of 4.1 might be expressed by combinatorial alternative splicing of the newly discovered exons 17A and 17B, together with alternative splicing of exons 18, 19, and 20 described earlier. To explore this issue, we used PCR techniques to investigate which of the many potential isoforms are actually expressed and whether there is any tissue specificity to the splicing patterns. Fig. 4 shows the results obtained using as templates either single-stranded cDNA prepared from various mouse tissues (left) or cDNA libraries prepared from different human tissues (right). Distinct expression patterns of the COOH-terminal coding exons were evident in different tissues. Whereas every tissue assayed expressed a substantial amount of the prototypical 17/18/19/20 isoform, inclusion of exons 17A and 17B was tissue-specific. For example, isoforms with the 17/17A/18/19/20 splicing pattern were highly expressed in both skeletal muscle and heart, with moderate expression also in liver. Very low to undetectable levels of these E17A-containing isoforms were observed in other tissues. Sequence analysis of the 17A-containing doublet generated in this reaction revealed the existence of alternative splice donor sites located 15 nucleotides apart near the 3Ј end of the exon. Expression of E17B, in contrast, was found most abundantly in tissues with high content of epithelial cells, such as HMEC, small intestine, large intestine, placenta, and kidney. Similar patterns of expression of E17A and E17B were observed in both mouse and human. Again, detailed sequence analysis of 17B-containing PCR products demonstrated additional microheterogeneity because of the use of alternative splice donor sites located 39 nucleotides apart. Exon 17B therefore can be 411 or 450 nucleotides in length, depending on which splicing signals are utilized. All of these novel inserts maintain an open reading frame with the downstream exons. Fig. 5 shows the nucleotide sequence of E17A and E17B. Consistent with the cDNA data indicating the presence of alternative splice donor sites near the 3Ј end of both E17A and E17B, sequences at the internal donor sites closely match the AG/gtragt consensus sequence normally found at exon/intron boundaries. These exons possess open reading frames encoding peptides of 17/12 and 150/137 amino acids, respectively (length depending on splice donor site usage). The reading frame of succeeding exons is unaltered, so they function to insert new motifs without otherwise modifying the COOH-terminal domain. Searches of available protein data bases have not revealed significant homologies between these new peptides and any previously characterized proteins.
Immunoblot analysis of proliferating versus differentiating HMEC proteins was performed to test the hypothesis that the E17B splicing switch should mediate expression of novel 4.1 protein species. Antibodies directed against synthetic peptides from several 4.1 domains were employed in these studies. The predominant 4.1 polypeptides expressed in both red cell ghosts and in proliferating HMEC migrated at ϳ80 kDa, as detected by antibodies against the spectrin-actin binding domain (Fig. 6,  left) or the COOH-terminal 34-amino acid sequence encoded by alternative exon 19 (right). The profile in suspended HMEC contained, in addition to the 80-kDa isoforms, a prominent 4.1 immunoreactive polypeptide at ϳ110 kDa. The appearance of higher molecular weight 4.1 protein in HMEC cultured on poly-HEMA correlates with the detection by RT-PCR of 4.1 RNA transcripts containing E17B, thus supporting the hypothesis that novel high molecular weight 4.1 isoforms are generated in HMEC via a differentiation-associated splicing switch. Consistent with this hypothesis, we have recently demonstrated that this ϳ110-kDa 4.1 species (but not 80-kDa protein 4.1) unequivocally cross-reacts with antibody directed against an E17B-encoded peptide (data not shown). Moreover, this band was significantly more abundant in HMEC in suspension than in adherent cultures of HMEC. ulation at the level of transcription initiation, using expression of milk protein genes as models. The genes encoding ␤-casein and whey acidic protein are not transcribed in proliferating cells adherent to culture dishes; however, treatment of cells with a combination of extracellular matrix and prolactin results in transcriptional activation via extracellular matrix-and prolactin-dependent enhancer elements (7,8). Induction of transcription appears to require transmission of both physical and biochemical signals from the extracellular environment (1).
The major novel finding of this paper is that pre-mRNA splicing regulation represents a second important mechanism for mediating changes in gene expression resulting from changes in the physical environment of mammary epithelial cells. We have documented a switch in the splicing pattern of 4.1 RNA involving expression of the novel E17B within the COOH-terminal coding sequences of the gene. Significantly, the dramatic induction of E17B expression which accompanies HMEC morphological changes in suspension appeared both specific and reversible. Among other alternative exons assayed, only exon 16 of the 4.1 gene exhibited a modest increase in relative expression under the same conditions; other alterna-tive exons within 4.1, as well as several in the fibronectin and HDLG pre-mRNAs, were unaffected. This specificity argues that the inductive signals do not promote global changes in the constitutive splicing machinery. We hypothesize, instead, that changes in the expression or activity of specific splicing factor(s) operate to regulate selectively the splicing of a limited repertoire of pre-mRNAs, i.e. a subset of those necessary to orchestrate the morphological and functional differentiation of HMEC. Moreover, the observed reversibility of E17B expression, marked by induction in suspended cells and down-regulation in cells transferred back to adherent conditions, indicates that these signals are required not only for initiation of the splicing switch, but also for its maintenance.
An RNA splicing map showing the complex array of alternative pre-mRNA splicing pathways whereby the 4.1 gene can encode multiple 4.1 polypeptides is shown in Fig. 7. The most salient feature of this map relates to the enormous range of potential structural protein isoforms generated via combinatorial alternative splicing events. In the particular case of the 4.1 gene, dozens of isoforms with distinct COOH-terminal structures could be generated by alternative splicing patterns involving five consecutive exons (17A, 17B, 18,19,and also 20 (Ref. 36)), together with additional microheterogeneity imparted by use of alternative splice donor sites in both 17A and 17B. Although a specific function for this domain has not yet been elucidated, an important role for this region is suggested by its strong phylogenetic conservation from Drosophila (37) and Xenopus (38) to mammals. We hypothesize that the alternative splicing-mediated activation of E17B and E17A may impart some new functionality to 4.1 in differentiated epithelial and muscle cells, much as the activation of exon 16 splicing imparts a new spectrin-actin binding function to 4.1 protein in differentiating erythroid cells.
As one approach to explore the physiological significance of 4.1 isoform diversity, we have begun to characterize the tissueand development-specific expression of individual mRNA isoforms. Three pre-mRNA splicing switches that can affect 4.1 function have been identified in erythroid and epithelial cells. In the erythroid system where 4.1 expression has been most extensively studied, alterations in the structure and function of the NH 2 -terminal and spectrin-actin binding domains are mediated by alternative splicing events involving exons 2 and 16, respectively (18,20,29,30). Early erythroid precursors engineer a switch in splice acceptor site usage at the 5Ј end of exon 2, thereby skipping upstream AUG-1 and extinguishing synthesis of a high molecular weight class of 4.1 polypeptides. Mature red cells thus express almost exclusively the 80-kDa isoforms initiated downstream at AUG-2. Later during erythroblast maturation, activation of exon 16 splicing facilitates expression of a functionally essential 21-amino acid peptide within the spectrin-actin binding domain. This simple insertion of a small peptide is quite typical of the majority of alternative splicing events described in numerous genes. However, in contrast to many other examples, the functional consequences of exon 16 activation are well understood, endowing newly synthesized 4.1 isoforms with the ability to bind spectrin and actin with high affinity (12,22). This splicing switch thereby contributes to the mechanical stabilization of plasma membranes in maturing erythrocytes.
In contrast to these erythroid-specific changes in 4.1 gene expression, the splicing switch described here operates in a different cell type, involves regulation of a much larger exon, and effects structural changes in a distinct domain of 4.1 protein. Experiments with cultured HMEC initially suggested that exon 17B expression may be specific for differentiated epithelia. This hypothesis was supported by the finding that exon 17B is present in 4.1 mRNA extracted from intestine, lung, and kidney but not muscle or brain cells. In addition to its tissue-specific splicing, control of exon 17B splicing may be interesting because of the unusually large size of the regulated cassette (450 nucleotides). The average size of vertebrate internal exons is 137 nucleotides, with less than 1% exceeding 400 nucleotides (39); even fewer of these represent differentially spliced exons. Therefore, regulation of E17B splicing may involve interesting and novel mechanistic features distinct from those controlling exons 2 and 16. This is the first regulated splicing event reported to affect the COOH-terminal domain structure of 4.1.
Finally, we also observed strong expression of E17A in muscle cells derived from either skeletal muscle or heart. Muscle thus may represent a third tissue, in addition to erythroid and epithelial tissues, in which 4.1 structure and function are regulated by alternative splicing.
Regarding the possible intracellular function of novel 4.1 protein isoforms with extended COOH-terminal domains, little information is yet available. A search of available protein data bases did not reveal significant homology between the 150amino acid motif encoded by E17B and any other reported gene products. It is possible that this peptide encodes a new protein interaction site that alters 4.1 localization or affinity for other skeletal structures, thus playing a role in the fundamental reorganization of the cytoskeleton which accompanies the cell shape changes during epithelial differentiation. Alternatively, this new domain may instead serve a regulatory function to modify or mask an existing interaction site, such as the adjacent spectrin-actin binding domain.
More broadly, it is worthwhile to consider the biological impact of extreme isoform diversity of the type illustrated in Fig. 7. Is such versatility in gene expression an unusual strategy adopted by only a few specialized genes, or a general mechanism employed by many genes to facilitate expression of multiple closely related proteins? Comparably complex alternative splicing has been reported in only a few other genes, including the pre-mRNAs for the transmembrane protein CD44 for (review, see Ref. 40) and for skeletal proteins of the ankyrin family (41,42). It is tempting to speculate that structural proteins may particularly benefit from expression of multiple isoforms with subtle or not so subtle variations in binding affinity for their interacting partners.
In conclusion, we have defined a pre-mRNA splicing switch that alters the structure of an important skeletal protein during epithelial differentiation. The fundamental process of gene regulation represented by this switch has considerable biolog-ical importance in several areas. One major challenge for the future is to understand the mechanism underlying this switch, both in terms of the proximal extracellular signaling event(s) that initiate this process, as well as the distal changes in activity of the nuclear RNA splicing machinery which catalyze the switch. An equally important problem relates to the function of the novel 4.1 polypeptides in epithelial cell biology. Future studies will be aimed at exploring the function of this novel protein 4.1 isoform, to understand its potential role in the morphological differentiation characteristic of epithelial cells.