FIRMA: a method for detection of alternative splicing from exon array data.

MOTIVATION: Analyses of EST data show that alternative splicing is much more widespread than once thought. The advent of exon and tiling microarrays means that researchers now have the capacity to experimentally measure alternative splicing on a genome wide level. New methods are needed to analyze the data from these arrays. RESULTS: We present a method, finding isoforms using robust multichip analysis (FIRMA), for detecting differential alternative splicing in exon array data. FIRMA has been developed for Affymetrix exon arrays, but could in principle be extended to other exon arrays, tiling arrays or splice junction arrays. We have evaluated the method using simulated data, and have also applied it to two datasets: a panel of 11 human tissues and a set of 10 pairs of matched normal and tumor colon tissue. FIRMA is able to detect exons in several genes confirmed by reverse transcriptase PCR. AVAILABILITY: R code implementing our methods is contributed to the package aroma.affymetrix.

The 13-kD FK506 binding protein, FKBP13, interacts with a novel homologue of the erythrocyte membrane cytoskeletal protein 4.1.

We have identified a novel generally expressed homologue of the erythrocyte membrane cytoskeletal protein 4.1, named 4.1G, based on the interaction of its COOH-terminal domain (CTD) with the immunophilin FKBP13. The 129-amino acid peptide, designated 4.1G-CTD, is the first known physiologic binding target of FKBP13. FKBP13 is a 13-kD protein originally identified by its high affinity binding to the immunosuppressant drugs FK506 and rapamycin (Jin, Y., M.W. Albers, W.S. Lane, B.E. Bierer, and S.J. Burakoff. 1991. Proc. Natl. Acad. Sci. USA. 88:6677- 6681); it is a membrane-associated protein thought to function as an ER chaperone (Bush, K.T., B.A. Henrickson, and S.K. Nigam. 1994. Biochem. J. [Tokyo]. 303:705-708). We report the specific association of FKBP13 with 4.1G-CTD based on yeast two-hybrid, in vitro binding and coimmunoprecipitation experiments. The histidyl-proline moiety of 4.1G-CTD is required for FKBP13 binding, as indicated by yeast experiments with truncated and mutated 4.1G-CTD constructs. In situ hybridization studies reveal cellular colocalizations for FKBP13 and 4.1G-CTD throughout the body during development, supporting a physiologic role for the interaction. Interestingly, FKBP13 cofractionates with the red blood cell homologue of 4.1 (4.1R) in ghosts, inside-out vesicles, and Triton shell preparations. The identification of FKBP13 in erythrocytes, which lack ER, suggests that FKBP13 may additionally function as a component of membrane cytoskeletal scaffolds.

Mechanochemistry of protein 4.1's spectrin-actin-binding domain: ternary complex interactions, membrane binding, network integration, structural strengthening.

Mechanical strength of the red cell membrane is dependent on ternary interactions among the skeletal proteins, spectrin, actin, and protein 4.1. Protein 4.1's spectrin-actin-binding (SAB) domain is specified by an alternatively spliced exon encoding 21 amino acid (aa) and a constitutive exon encoding 59 aa. A series of truncated SAB peptides were engineered to define the sequences involved in spectrin-actin interactions, and also membrane strength. Analysis of in vitro supramolecular assemblies showed that gelation activity of SAB peptides correlates with their ability to recruit a critical amount of spectrin into the complex to cross-link actin filaments. Also, several SAB peptides appeared to exhibit a weak, cooperative actin-binding activity which mapped to the first 26 residues of the constitutive 59 aa. Fluorescence-imaged microdeformation was used to show SAB peptide integration into the elastic skeletal network of spectrin, actin, and protein 4.1. In situ membrane-binding and membrane-strengthening abilities of the SAB peptides correlated with their in vitro gelation activity. The findings imply that sites for strong spectrin binding include both the alternative 21-aa cassette and a conserved region near the middle of the 59 aa. However, it is shown that only weak SAB affinity is necessary for physiologically relevant action. Alternatively spliced exons can thus translate into strong modulation of specific protein interactions, economizing protein function in the cell without, in and of themselves, imparting unique function.

Neurobehavioral deficits in mice lacking the erythrocyte membrane cytoskeletal protein 4.1.

The erythrocyte membrane cytoskeletal protein 4.1 (4.1R) is a structural protein that confers stability and flexibility to erythrocytes via interactions with the cytoskeletal proteins spectrin and F-actin and with the band 3 and glycophorin C membrane proteins. Mutations in 4.1R can cause hereditary elliptocytosis, a disease characterized by a loss of the normal discoid morphology of erythrocytes, resulting in hemolytic anemia [1]. Different isoforms of the 4.1 protein have been identified in a wide variety of nonerythroid tissues by immunological methods [2-5]. The variation in molecular weight of these different 4.1 isoforms, which range from 30 to 210 kDa [6], has been attributed to complex alternative splicing of the 4.1R gene [7]. We recently identified two new 4.1 genes: one is generally expressed throughout the body (4. 1G) [8] and the other is expressed in central and peripheral neurons (4.1N) [9]. Here, we examined 4.1R expression by in situ hybridization analysis and found that 4.1R was selectively expressed in hematopoietic tissues and in specific neuronal populations. In the brain, high levels of 4.1R were discretely localized to granule cells in the cerebellum and dentate gyrus. We generated mice that lacked 4.1R expression; these mice had deficits in movement, coordination, balance and learning, in addition to the predicted hematological abnormalities. The neurobehavioral findings are consistent with the distribution of 4.1R in the brain, suggesting that 4.1R performs specific functions in the central nervous system.

An isoform-specific mutation in the protein 4.1 gene results in hereditary elliptocytosis and complete deficiency of protein 4.1 in erythrocytes but not in nonerythroid cells.

Multiple protein 4.1 isoforms are expressed in a variety of tissues through complex alternative pre-mRNA splicing events, one function of which is to regulate use of two alternative translation initiation signals. Late erythroid cells express mainly the downstream initiation site for synthesis of prototypical 80-kD isoforms; nonerythroid cells in addition use an upstream site to encode higher molecular mass isoform(s). In this study, we examined the effects of a 5' gene rearrangement in a family with hereditary elliptocytosis and complete deficiency of erythrocyte 4.1 protein on 4.1 isoform expression in erythroid vs. nonerythroid cells. Patient 4.1 mRNAs from reticulocytes, fibroblasts, and B lymphocytes were amplified by reverse transcriptase/polymerase chain reaction techniques and shown to exhibit a 318-nucleotide deletion that encompasses the downstream AUG, but leaves intact the upstream AUG. Immunoblot analysis revealed a total deficiency of 4.1 in patient red cells and a selective deficiency of 80-kD isoform(s) but not high molecular weight 4.1 in patient nonerythroid cells. Thus, the 4.1 gene mutation in this family produces an isoform-specific deficiency that is manifested clinically in tissue-specific fashion, such that red cells are affected but other cell types are unaffected because of tissue-specific differences in RNA splicing and translation initiation.

Protein 4.1R-deficient mice are viable but have erythroid membrane skeleton abnormalities.

A diverse family of protein 4.1R isoforms is encoded by a complex gene on human chromosome 1. Although the prototypical 80-kDa 4.1R in mature erythrocytes is a key component of the erythroid membrane skeleton that regulates erythrocyte morphology and mechanical stability, little is known about 4.1R function in nucleated cells. Using gene knockout technology, we have generated mice with complete deficiency of all 4.1R protein isoforms. These 4.1R-null mice were viable, with moderate hemolytic anemia but no gross abnormalities. Erythrocytes from these mice exhibited abnormal morphology, lowered membrane stability, and reduced expression of other skeletal proteins including spectrin and ankyrin, suggesting that loss of 4. 1R compromises membrane skeleton assembly in erythroid progenitors. Platelet morphology and function were essentially normal, indicating that 4.1R deficiency may have less impact on other hematopoietic lineages. Nonerythroid 4.1R expression patterns, viewed using histochemical staining for lacZ reporter activity incorporated into the targeted gene, revealed focal expression in specific neurons in the brain and in select cells of other major organs, challenging the view that 4.1R expression is widespread among nonerythroid cells. The 4.1R knockout mice represent a valuable animal model for exploring 4.1R function in nonerythroid cells and for determining pathophysiological sequelae to 4.1R deficiency.

Deciphering the nuclear import pathway for the cytoskeletal red cell protein 4.1R.

The erythroid membrane cytoskeletal protein 4.1 is the prototypical member of a genetically and topologically complex family that is generated by combinatorial alternative splicing pathways and is localized at diverse intracellular sites including the nucleus. To explore the molecular determinants for nuclear localization, we transfected COS-7 cells with epitope-tagged versions of natural red cell protein 4.1 (4.1R) isoforms as well as mutagenized and truncated derivatives. Two distant topological sorting signals were required for efficient nuclear import of the 4.1R80 isoform: a basic peptide, KKKRER, encoded by alternative exon 16 and acting as a weak core nuclear localization signal (4.1R NLS), and an acidic peptide, EED, encoded by alternative exon 5. 4.1R80 isoforms lacking either of these two exons showed decreased nuclear import. Fusion of various 4.1R80 constructs to the cytoplasmic reporter protein pyruvate kinase confirmed a requirement for both motifs for full NLS function. 4.1R80 was efficiently imported in the nuclei of digitonin-permeabilized COS-7 cells in the presence of recombinant Rch1 (human importin alpha2), importin beta, and GTPase Ran. Quantitative analysis of protein-protein interactions using a resonant mirror detection technique showed that 4.1R80 bound to Rch1 in vitro with high affinity (KD = 30 nM). The affinity decreased at least 7- and 20-fold, respectively, if the EED motif in exon 5 or if 4.1R NLS in exon 16 was lacking or mutated, confirming that both motifs were required for efficient importin-mediated nuclear import of 4.1R80.

Computational analysis of candidate intron regulatory elements for tissue-specific alternative pre-mRNA splicing.

Alternative pre-mRNA splicing is a major cellular process by which functionally diverse proteins can be generated from the primary transcript of a single gene, often in tissue-specific patterns. The current study investigates the hypothesis that splicing of tissue-specific alternative exons is regulated in part by control sequences in adjacent introns and that such elements may be recognized via computational analysis of exons sharing a highly specific expression pattern. We have identified 25 brain-specific alternative cassette exons, compiled a dataset of genomic sequences encompassing these exons and their adjacent introns and used word contrast algorithms to analyze key features of these nucleotide sequences. By comparison to a control group of constitutive exons, brain-specific exons were often found to possess the following: divergent 5' splice sites; highly pyrimidine-rich upstream introns; a paucity of GGG motifs in the downstream intron; a highly statistically significant over-representation of the hexanucleotide UGCAUG in the proximal downstream intron. UGCAUG was also found at a high frequency downstream of a smaller group of muscle-specific exons. Intriguingly, UGCAUG has been identified previously in a few intron splicing enhancers. Our results indicate that this element plays a much wider role than previously appreciated in the regulated tissue-specific splicing of many alternative exons.

Mouse erythroid cells express multiple putative RNA helicase genes exhibiting high sequence conservation from yeast to mammals.

RNA secondary structure is a critical determinant of RNA function in ribosome assembly, pre-mRNA splicing, mRNA translation and RNA stability. The 'DEAD/H' family of putative RNA helicases may help regulate these processes by utilizing intrinsic RNA-dependent ATPase activity to catalyze conformational changes in RNA secondary structure. To investigate the repertoire of DEAD/H box proteins expressed in mammals, we used PCR techniques to clone from mouse erythroleukemia (MEL) cells three new DEAD box cDNAs with high similarity to known yeast (Saccharomyces cerevisiae) genes. mDEAD2 and mDEAD3 (mouse DEAD box proteins) are > 95% identical to mouse PL10 but exhibit differential tissue-specific expression patterns; mDEAD2 and mDEAD3 are also approx. 70% identical (at the aa level) to yeast DED1 and DBP1 proteins. Members of this DEAD box subclass contain C-terminal domains with high content of Arg, Ser, Gly and Phe, reminiscent of the RS domain in several Drosophila and mammalian splicing factors. mDEAD5 belongs to a second class related to translation initiation factors from yeast (TIF1/TIF2) and mammals (eIF-4A); this class contains a novel conserved peptide motif not found in other DEAD box proteins. Northern blotting shows that mDEAD5 is differentially expressed in testis vs. somatic tissues. Thus, mouse erythroid cells produce two highly conserved families of putative RNA helicases likely to play important roles in RNA metabolism and gene expression.

Regulation of alternative pre-mRNA splicing during erythroid differentiation.

Although the mature enucleated erythrocyte is no longer active in nuclear processes such as pre-mRNA splicing, the function of many of its major structural proteins is dependent on alternative splicing choices made during the earlier stages of erythropoiesis. These splicing decisions fundamentally regulate many aspects of protein structure and function by governing the inclusion or exclusion of exons that encode protein interaction domains, regulatory signals, or translation initiation or termination sites. Alternative splicing events may be partially or entirely erythroid-specific, ie, distinct from the splicing patterns imposed on the same transcripts in nonerythroid cells. Moreover, differentiation stage-specific splicing "switches" may alter the structure and function of erythroid proteins in physiologically important ways as the cell is morphologically and functionally remodeled during normal differentiation. Derangements in the splicing of individual mutated pre-mRNAs can produce synthesis of truncated or unstable proteins that are responsible for numerous erythrocyte disorders. This review will summarize the salient features of regulated alternative splicing in general, review existing information concerning the widespread extent of alternative splicing among erythroid genes, and describe recent studies that are beginning to uncover the mechanisms that regulate an erythroid splicing switch in the protein 4.1R gene.

Posttranslational uptake and processing of in vitro synthesized ornithine transcarbamoylase precursor by isolated rat liver mitochondria.

The mitochondrial matrix enzyme ornithine transcarbamoylase (OTCase; ornithine carbamoyltransferase; carbamoylphosphate:L-ornithine carbamoyltransferase, EC 2.1.3.3) is encoded by a nuclear gene on the X chromosome, synthesized on cytoplasmic ribosomes, and translocated across both mitochondrial membranes. Using specific immunoprecipitation, we presented evidence previously that the primary in vitro translation product of OTCase in rat liver is a polypeptide about 4000 daltons larger than the "mature" OTCase augment subunit purified from homologous mitochondria. In this report we augment the immunological identification of this cell-free translation product (pOTCase) with structural information and show, by electrophoresis of proteolysis products, that pOTCase is structurally similar to mitochondrial OTCase. Moreover, we now demonstrate that, when pOTCase is incubated posttranslationally with isolated rat liver mitochondria, it is converted to the size of mature OTCase and is sequestered within the mitochondria in such a way that it becomes resistant to externally added proteases. Such posttranslational processing is catalyzed specifically by the mitochondrial fraction of rat liver cells and is dependent both on the duration of incubation with mitochondria and on the amount of mitochondrial protein added. We conclude that pOTCase is indeed the bona fide precursor of mitochondrial OTCase and that use of this simplified cell-free system will facilitate analysis of OTCase biogenesis at both the cellular and the molecular level.

Pre-ornithine transcarbamylase. Properties of the cytoplasmic precursor of a mitochondrial matrix enzyme.

Biogenesis of the mitochondrial matrix enzyme, ornithine transcarbamylase, has been shown to begin with synthesis on cytoplasmic ribosomes of a precursor, designated pre-ornithine transcarbamylase, which is approximately 4000 daltons larger than its corresponding mitochondrial subunit, followed by post-translational uptake and proteolytic processing of the precursor to its mature counterpart by mitochondria. We now report initial studies on the structure and properties of preornithine transcarbamylase. When this precursor is labeled at the NH2 terminus with N-formyl[35S]methionine and processed by mitochondria, no label is recovered with the mature subunit. This demonstrates that the amino acid extension which is characteristic of the precursor and which is removed during mitochondrial processing is NH2-terminal. This NH2-terminal extension is found intact in two peptides produced by limited proteolysis of the labeled precursor. Moreover, this amino acid extension modifies the behavior of the precursor during immunoprecipitation in the presence of ionic detergents and plays a critical role in facilitating uptake of the precursor by mitochondria.

In vitro synthesis of a putative precursor of mitochondrial ornithine transcarbamoylase.

Ornithine transcarbamoylase (OTCase; ornithine carbamoyltransferase; carbamoyl phosphate:L-ornithine carbamoyltransferase, EC 2.1.3.3), a major mitochondrial matrix enzyme in ureotelic animals, is synthesized on cytoplasmic ribosomes and translocated across both mitochondrial membranes to the matrix. In an attempt to identify the primary translation product (or an early intermediate) that is the substrate for this transport process, we translated rat liver polysomal RNA in vitro by using the rabbit reticulocyte lysate system. Immunoprecipitation of the [35S]methionine-labeled translation mixture was performed by using monospecific OTCase antiserum and the immunoadsorbent Staphylococcus aureus. Approximately 0.3% of total trichloroacetic acid-insoluble 35S-labeled material was specifically precipitated. Analysis of the precipitate by fluorography of a dried sodium dodecyl sulfate/polyacrylamide gel showed a single major translation product whose mobility corresponded to a polypeptide of 43,000 daltons, a value approximately 4000 daltons greater than that noted for the "mature" OTCase subunit isolated from rat liver. This translation product was not precipitated by preimmune rabbit serum, and excess unlabeled mature OTCase competed with it for interaction with OTCase antiserum. These results suggested that rat liver OTCase, like a number of other cytoplasmically synthesized organellar proteins, is initially made as a larger precursor that contains an amino acid sequence necessary to confer on OTCase its transport properties. The potential application of these findings to the study of inherited complete OTCase deficiency in humans is discussed.

Mechanochemistry of the alternatively spliced spectrin-actin binding domain in membrane skeletal protein 4.1.

Protein 4.1's interaction with the erythroid skeletal proteins spectrin and actin and its essential role in regulating membrane strength are both attributable to expression of an alternatively spliced 63-nucleotide exon. The corresponding 21-amino acid (21-aa) cassette is within the previously identified spectrin-actin binding domain (10 kDa molecular mass) of erythroid protein 4.1. This cassette is absent, however, in several isoforms that are generated by tissue- and development-specific RNA splicing. Four isoforms of the 10-kDa domain were constructed for comparative assessment of functions particularly relevant to red cells. In vitro translated isoforms containing the 21-aa cassette, denoted 10k21 and 10k19,21, were able to bind spectrin, stabilize spectrin-actin complexes, and associate with red cell membrane. Isoforms replacing or lacking the 21-aa cassette, 10k19 and 10k0, did not function in these assays. A bacterially expressed fusion protein with glutathione-S-transferase, designated GST-10k21, congealed spectrin-actin into a network in vitro as found with purified protein 4.1. Additionally, incorporation of GST-10k21 into mechanically weak, 4.1-deficient membranes increased mechanical strength of these membranes to normal. GST-10k19 did not function in these assays. These results show that the 21-aa sequence in protein 4.1 is critical to mechanical integrity of the red cell membrane. These results also allow the role of protein 4.1 in membrane mechanics to be interpreted primarily in terms of its spectrin-actin binding function. Alternatively expressed sequences within the 10-kDa domain of nonerythroid protein 4.1 are suggested to have different, yet to be defined functions.

Multiple protein 4.1 isoforms produced by alternative splicing in human erythroid cells.

Protein 4.1 is a multifunctional structural protein located in the erythrocyte membrane skeleton and in many nonerythroid cells. Molecular characterization of cloned protein 4.1 sequences from human reticulocytes has revealed the existence of multiple transcripts of the protein 4.1 gene that may encode a family of closely related protein isoforms. Several independently isolated cDNAs were sequenced and demonstrated to encode four different protein 4.1 species having identical primary sequences, except for the presence or absence of discrete peptides in the 8-kDa spectrin/actin binding domain (21 amino acids) and near the carboxyl terminus (43 and 34 amino acids). The same four protein 4.1 isoforms were detected when reticulocyte protein 4.1 mRNA sequences were reverse transcribed into cDNA and enzymatically amplified in vitro by using protein 4.1-specific oligonucleotide primers and the polymerase chain reaction. The finding of multiple protein 4.1 isoforms raises the possibility that the many binding functions ascribed to protein 4.1 may reside in distinct structural isoforms. Since only a single protein 4.1 gene appears to be expressed in erythrocytes, it is likely that these isoforms are produced by alternative mRNA splicing from a common protein 4.1 pre-mRNA. Multiple RNA splicing pathways are thus operative in the protein 4.1 gene even within a single cell lineage, human erythroid cells.

Tissue- and development-specific alternative RNA splicing regulates expression of multiple isoforms of erythroid membrane protein 4.1.

Protein 4.1, a multifunctional structural protein originally described as an 80-kDa component of the erythroid membrane skeleton, exhibits tissue- and development-specific heterogeneity in molecular weight, subcellular localization, and primary amino acid sequence. Earlier reports suggested that some of this impressive heterogeneity is generated by alternative RNA splicing (Conboy, J. G., Chan, J., Mohandas, N., and Kan, Y. W. (1988) Proc. Natl. Acad. Sci. U. S. A. 85, 9062-9065; Tang, T. K., Leto, T., Marchesi, V. T., and Benz, E. J. (1990) J. Cell Biol. 110, 617-624). We have now completed a systematic analysis of 4.1 mRNA isoforms expressed in erythroid cells, and have generated an "alternative splicing map" which summarizes diagrammatically a multitude of polypeptide isoforms potentially generated by combinatorial splicing of nine alternative exons. Complex 5' splicing events yield mRNA isoforms that may initiate translation at different sites and thus generate elongated or truncated NH2 termini; elongated approximately 135-kDa and prototypical approximately 80-kDa species were detected in both erythrocytes and T-lymphocytes, but in very different ratios. Among the functional domains of 4.1 responsible for interaction with other membrane skeletal elements, four variants of the 10-kDa spectrin-actin-binding region and four variants of the putative 30-kDa glycophorin-binding region are predicted. Developmentally controlled alternative RNA splicing in the spectrin-actin-binding region may help regulate remodeling of membrane architecture and mechanical properties that occur during erythropoiesis.

Cloning of mDEAH9, a putative RNA helicase and mammalian homologue of Saccharomyces cerevisiae splicing factor Prp43.

Yeast splicing factor Prp43, a DEAH box protein of the putative RNA helicase/RNA-dependent NTPase family, is a splicing factor that functions late in the pre-mRNA splicing pathway to facilitate spliceosome disassembly. In this paper we report cDNA cloning and characterization of mDEAH9, an apparent mammalian homologue of Prp43. Amino acid sequence comparison revealed that the two proteins are approximately 65% identical over a 500-aa region spanning the central helicase domain and the C-terminal region. Expression of mDEAH9 in S. cerevisiae bearing a temperature-sensitive mutation in prp43 was sufficient to restore growth at the nonpermissive temperature. This functional complementation was specific, as mouse mDEAH9 failed to complement mutations in related splicing factor genes prp16 or prp22. Finally, double label immunofluorescence experiments performed with mammalian cells revealed colocalization of mDEAH9 and splicing factor SC35 in punctate nuclear speckles. Thus, the hypothesis that mDEAH9 represents the mammalian homologue of yeast Prp43 is supported by its high sequence homology, functional complementation, and colocalization with a known splicing factor in the nucleus. Our results provide additional support for the hypothesis that the spliceosomal machinery that mediates regulated, dynamic changes in conformation of pre-mRNA and snRNP RNAs has been highly conserved through evolution.

Ca(2+)-dependent and Ca(2+)-independent calmodulin binding sites in erythrocyte protein 4.1. Implications for regulation of protein 4.1 interactions with transmembrane proteins.

In vitro protein binding assays identified two distinct calmodulin (CaM) binding sites within the NH(2)-terminal 30-kDa domain of erythrocyte protein 4.1 (4.1R): a Ca(2+)-independent binding site (A(264)KKLWKVCVEHHTFFRL) and a Ca(2+)-dependent binding site (A(181)KKLSMYGVDLHKAKDL). Synthetic peptides corresponding to these sequences bound CaM in vitro; conversely, deletion of these peptides from a 30-kDa construct reduced binding to CaM. Thus, 4.1R is a unique CaM-binding protein in that it has distinct Ca(2+)-dependent and Ca(2+)-independent high affinity CaM binding sites. CaM bound to 4.1R at a stoichiometry of 1:1 both in the presence and absence of Ca(2+), implying that one CaM molecule binds to two distinct sites in the same molecule of 4.1R. Interactions of 4.1R with membrane proteins such as band 3 is regulated by Ca(2+) and CaM. While the intrinsic affinity of the 30-kDa domain for the cytoplasmic tail of erythrocyte membrane band 3 was not altered by elimination of one or both CaM binding sites, the ability of Ca(2+)/CaM to down-regulate 4. 1R-band 3 interaction was abrogated by such deletions. Thus, regulation of protein 4.1 binding to membrane proteins by Ca(2+) and CaM requires binding of CaM to both Ca(2+)-independent and Ca(2+)-dependent sites in protein 4.1.