Intronic CA-repeat and CA-rich elements: a new class of regulators of mammalian alternative splicing

Abstract

We have recently identified an intronic polymorphic CA-repeat region in the human endothelial nitric oxide synthase (eNOS) gene as an important determinant of the splicing efficiency, requiring specific binding of hnRNP L. Here, we analyzed the position requirements of this CA-repeat element, which revealed its potential role in alternative splicing. In addition, we defined the RNA binding specificity of hnRNP L by SELEX: not only regular CA repeats are recognized with high affinity but also certain CA-rich clusters. Therefore, we have systematically searched the human genome databases for CA-repeat and CA-rich elements associated with alternative 5' splice sites (5'ss), followed by minigene transfection assays. Surprisingly, in several specific human genes that we tested, intronic CA RNA elements could function either as splicing enhancers or silencers, depending on their proximity to the alternative 5'ss. HnRNP L was detected specifically bound to these diverse CA elements. These data demonstrated that intronic CA sequences constitute novel and widespread regulatory elements of alternative splicing.

Figure 1

Position requirements of the CA-repeat element in eNOS intron 13. (A–E) Pre-mRNAs are schematically shown on the right and are derived from exons 13–14 of the human eNOS gene with a shortened version of intron 13 containing the (CA)₃₂ splicing enhancer (gray box). The (CA)₃₂ enhancer is located at its normal intronic position (+80; A), moved upstream (+23; panel B), downstream (+175; C and D), or into exon 13 (−21; E). In addition, for CA32(+175) opt 5′ss, the 5′ splice site (5′ss; see A) was converted to the consensus sequence (opt 5′ss; see D, sequence as shown with three mutated positions underlined). Each eNOS pre-mRNA derivative was spliced in vitro, and the reaction at time 0, and after 30, 60, 90, and 120 min was analyzed by RT–PCR. The positions corresponding to pre-mRNA and product mRNAs are schematically indicated on the left, predominant splicing patterns also on the right. Note that a cryptic 5′ss is activated with the (CA)₃₂ enhancer at intron position +175 (C and D). The band marked by the asterisk (D) represents a nonspecific PCR artifact, as determined by sequence analysis. (F) Direct RNA analysis of in vitro splicing of ³²P-labeled eNOS pre-mRNA CA32(+80) and derivatives CA32(+175) and CA32(+175) opt5′ss (compare with panels A, C, and D). Time points of 45 and 90 min have been analyzed, as indicated. The positions of pre-mRNAs, spliced products, and first-exon intermediates are indicated on the right. The gray box represents the additional exon sequence incorporated as a result of use of the cryptic 5′ss. The asterisk marks most likely a degradation product of the splicing substrate, since it occurred splicing-independently and varied in intensity when different preparations of nuclear extract were used. M, pBR322/HpaII markers (sizes in nucleotides).

Figure 2

Intronic CA repeats are sufficient for splicing enhancement. (A) Schematic representation of heterologous constructs: T7 DUP 4-1 is derived from exons 1 and 2 (white boxes) of the human β-globin gene, with an artificial short exon (gray box) between two identical introns A and B. T7 DUP-CA32 contains in intron B the (CA)₃₂ sequence, T7 DUP-control a nonspecific sequence (shaded boxes). The positions of the primers used for RT–PCR assays are shown by arrows with primer names labeled. (B–D) Each of these three pre-mRNAs was spliced in vitro, and aliquots were assayed by RT–PCR at time points between 0 and 240 min (as indicated above the lanes), using primer combinations indicative of splicing of the entire intron C unit (B), of intron A (C), and of intron B (D). Positions of products corresponding to pre-mRNAs and spliced mRNAs are schematically shown on the right. Splicing efficiencies for removing different introns were quantitated and diagrammed on the right. In panel B, the efficiency of removing intron C was quantitated as the ratio of the mRNA signal without the central exon to the sum of pre-mRNA and mRNA signals. In panels C and D, the efficiencies of removing intron A and B, respectively, were quantitated as the ratio of the mRNA signal to the sum of pre-mRNA and mRNA signals (expressed as percentages). M, 100-bp ladder (Fermentas).

Figure 3

Defining the RNA binding specificity of hnRNP L by SELEX. (A) The 10-nucleotide consensus sequence. The frequency of each of the four nucleotides at any position in the consensus is represented by the letter height. Boxes mark the two highly conserved core tetranucleotides. (B) Tetranucleotide frequency in selected sequences. The first 20, most common tetranucleotide sequences, are given, in the order of their frequencies in the 108 selected sequences (heavy line, SELEX) and in the 20 sequences taken from the initial pool (thin line, control; both diagrammed as percentage of the total). (C) Characteristics of 11 SELEX-derived (clone numbers on the left) and two control sequences (with asterisks; #20 and 15). Given are the individual sequences (with high-score motifs in red, low-score motifs underlined) and the K_D values (in nM; with standard deviations, P<0.05).

Figure 4

CA-repeat and CA-rich sequences function as regulatory elements of alternative splicing. Alternative splicing of four representative candidate genes that contain intronic CA-repeat elements (MAPK10, SLC2A2, RFXANK) or a CA-rich sequence (GSTZ1) has been characterized, using minigenes diagrammed on the right. The alternatively used exon is indicated by the dark box, and intron and exon parts are shown in scale. Wild type (WT) and substituted (sub) sequence elements are given, with boundaries relatively to the 5′ss of the second intron. Vertical lines within the introns mark the positions where introns have been shortened. Minigenes were transfected into HeLa cells, and alternative splicing tested by RT–PCR (lanes WT, sub); the control transfections (lanes mock) were carried out in the absence of DNA. The positions corresponding to pre-mRNA and alternative mRNA products are indicated to the right of the gel pictures. For the GSTZ1, SLC2A2, and RFXANK minigenes, the percentages of exon inclusion with standard deviations (n=3) are given below the corresponding lanes.

Figure 5

HnRNP L binds specifically to intronic CA-repeat and CA-rich elements of candidate genes. (A) GST-hnRNP L pull-down assays. ³²P-labeled short RNAs containing the CA-repeat and CA-rich elements of candidate genes GSTZ1 and RFXANK (wild type, lanes WT, and substitution version, lanes S) as well as the positive control (CA)₂₀ (lanes CA20) were incubated with GST-hnRNP L, which had been immobilized on glutathione-Sepharose. Bound RNAs were recovered and analyzed by denaturing gel electrophoresis. For each pull-down assay, 10% of the input was loaded (left half of figure) and the total amount of recovered material (right half). M, markers (36, 67, and 76 nucleotides). (B) Anti-hnRNP L immunoprecipitation assays from HeLa nuclear extract. ³²P-labeled short RNAs containing the CA-repeat and CA-rich elements of candidate genes GSTZ1 and SLC2A2 (wild type, lanes WT, and substitution version, lanes S) as well as the positive control (CA)₂₀ (lanes CA20) were incubated in HeLa nuclear extract. Following immunoprecipitation with anti-hnRNP L antibodies, coprecipitated RNAs were analyzed by denaturing gel electrophoresis. For each assay, 10% of the input was loaded (left half of figure) and the total of immunoprecipitated material (right half). M, markers (36, 67, and 76 nucleotides). (C) Coupled UV-crosslinking/anti-hnRNP L immunoprecipitation assays. Incubations as described in panel B were subjected to UV-crosslinking, followed by anti-hnRNP L immunoprecipitation. For each assay, 10% of the reaction after crosslinking was loaded (left half of figure) and the total of crosslinked/immunoprecipitated material (right half). The arrow indicates the mobility of crosslinked hnRNP L (protein markers in kDa).