tau Exon 10 expression involves a bipartite intron 10 regulatory sequence and weak 5' and 3' splice sites.

tau mutations that deregulate alternative exon 10 (E10) splicing cause frontotemporal dementia with parkinsonism chromosome 17-type by several mechanisms. Previously we showed that E10 splicing involved exon splicing enhancer sequences at the 5' and 3' ends of E10, an exon splicing silencer, a weak 5' splice site, and an intron splicing silencer (ISS) within intron 10 (I10). Here, we identify additional regulatory sequences in I10 using both non-neuronal and neuronal cells. The ISS sequence extends from I10 nucleotides 11-18, which is sufficient to inhibit use of a weakened 5' splice site of a heterologous exon. Furthermore, ISS function is location-independent but requires proximity to a weak 5' splice site. Thus, the ISS functions as a linear sequence. A new cis-acting element, the intron splicing modulator (ISM), was identified immediately downstream of the ISS at I10 positions 19-26. The ISM and ISS form a bipartite regulatory element, within which the ISM functions when the ISS is present, mitigating E10 repression by the ISS. Additionally, the 3' splice site of E10 is weak and requires exon splicing enhancer elements for efficient E10 inclusion. Thus far, tau FTDP-17 splicing mutations affect six predicted cis-regulatory sequences.

binding domain, generates tau isoforms with four MT-binding repeats (4R tau) compared with isoforms without E10 that have only three MT-binding repeats (3R tau).
The FTDP-17 splicing mutations in E10 and I10 indicate that multiple cis-acting regulatory elements control E10 inclusion. Previous work demonstrated that E10 splicing is complex, involving exonic and intronic regulatory elements that interact to modulate use of the weak E10 5Ј splice site (21). The exonic elements include three nonredundant exon splicing enhancer (ESE) sequences located within the first 45 bases of E10. These are a 5Ј SC35-like sequence, a polypurine enhancer (PPE), and a 3Ј A/C-rich enhancer (ACE). Additional E10 elements include an exon splicing silencer (ESS) and two adjacent 9-nucleotide positive sequences immediately upstream of the 5Ј splice site. The I10 intronic regulatory element is the intron splicing silencer (ISS) located downstream of the 5Ј splice. FTDP-17 mutations affect six of these cis-acting regulatory sequences (see "Discussion").
In previous work, multiple cis-acting E10 regulatory elements were identified that interact with the weak 5Ј splice site and the ISS (11,21). Here, the role of flanking intronic se-quences on E10 splicing regulation is examined. The results show that both I9 and I10 sequences regulate E10 inclusion. In I10, a novel bipartite regulatory element composed of the inhibitory ISS and an adjacent intron splicing modulator (ISM) sequence affect E10 inclusion. The data are consistent with the ISS functioning as a linear element rather than through a secondary structure as previously proposed (7,9). The E10 3Ј splice site, like the 5Ј splice site, is also weak and its activity is dependent on E10 ESEs. Thus, there is a complex interplay between I9, E10, and I10 regulatory sequences that maintains the normal ratio of 4R versus 3R tau. FTDP-17 mutations disrupt this complex regulatory system by multiple mechanisms leading to distinct disease phenotypes.

EXPERIMENTAL PROCEDURES
Plasmid Construction and DNA Mutagenesis-Splicing was assayed using a derivative of vector pSPL3 called pSPES (21). pSPL3 contains an HIV genomic fragment with a multiple cloning site (MCS) between highly truncated tat exons 2 and 3 fused to rabbit ␤-globin coding sequences (22). pSPES was generated from pSPL3 by deleting a 500-bp fragment between the EcoRV (nucleotide 1050) and StuI (nucleotide 1550) sites, removing a previously reported cryptic exon within the tat intron. Construct hN contains human tau E10 with 33 and 51 bp of flanking introns inserted into the MCS of pSPES as previously described (21). In hN and other similar tau constructs, the weak E10 5Ј splice site activates a 60-bp cryptic exon between the MCS and the downstream tat/globin hybrid exon 3. Modifications within tau sequences were performed by PCR mutagenesis (11,21). Construct E10860 is identical to hN but with longer flanking I9 (567 bp) and I10 (190 bp) sequences. The E10860 insert was generated by PCR from PAC clone 4139 using forward primer I9AF (5Ј-GGACTAGTGAGACT-GAAGCCAGACTCCTAGATT-3Ј) and reverse primer I10AR (5Ј-ATG-CATCCTCACACTGGGAACAGTGGACCATG-3Ј). The PCR product was blunt-end ligated into the MCS of pSPES.
Cell Culture, Transfection, and RNA Isolation-COS-7 cells were cultured and transfected as previously described (11). PC12 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% horse serum and 5% fetal bovine serum (Invitrogen) and seeded onto poly-D-lysine-coated six-well plates to achieve 70% confluence 1 day prior to transfection. Primary cortical neurons were isolated from post-natal day 1 Sprague-Dawley rat pups cultured for 5-7 days prior to transfection as previously described (23). PC12 cells (5 ϫ 10 5 ) were transfected using 2 g of plasmid DNA and 10 l of LipofectAMINE in 1 ml of Opti-MEM (Invitrogen) for 5 h at 37°C (5% CO 2 ), after which 1 ml of Dulbecco's modified Eagle's medium supplemented with 20% horse serum and 10% fetal bovine serum was added. Primary cortical neurons were transfected with 1 g of plasmid DNA mixed with 10 l of the polycationic liposomal reagent DOSPER (Roche) and incubated at room temperature for 30 -45 min in a total of 60 l of HEPES-buffered saline (137 mM NaCl, 4.8 mM KCl, 0.56 mM Na 2 PO 4 , and 21 mM HEPES, pH 7.3). The DNA/DOSPER mixture was added to 1 ml of Neurobasal medium (Invitrogen) containing 0.32 mM glutamine. Culture medium was removed and stored before adding the DNA/DOSPER/Neurobasal mixture (23). After 5 h, conditioned medium was added back to allow neurons to recover. RNA was isolated as previously described (11).
Quantitation of tau E10 Splicing by Reverse Transcription-PCR (RT-PCR)-E10 splicing was assayed by RT-PCR as previously reported (11,21). The E10Ϫ and E10ϩ products as amplified by RT-PCR from transfected cells are 261 and 354 bp, respectively. For each mutant construct, values presented are the average of at least three different transfection experiments with the normal tau E10 or ␤E2 construct also transfected in parallel. Statistical comparisons were made using a two-tailed Student's t test. Criteria for significance were calculated using a Bonferroni correction for multiple comparisons by dividing the initial p value of 0.05 by the number of comparisons made.

RESULTS
Intronic FTDP-17 mutations are clustered in an I10 inhibitory element designated the ISS (Fig. 1A). These mutations dramatically increase E10 inclusion in tau transcripts, demonstrating that I10 sequences regulate E10 inclusion. Whether I10 sequences downstream of E10ϩ16 also contribute to E10 splicing regulation is unknown. Furthermore, the role of I9 sequences in E10 inclusion has not been examined. Two mechanisms for ISS regulation of E10 splicing have been proposed. One is that, normally, the ISS is part of a stem loop (Fig. 1B) that blocks access of U1 snRNP to the 5Ј splice site, thereby inhibiting use of this splice site and E10 inclusion (7,9). I10 FTDP-17 mutations disrupt base pairing in the stem, resulting in increased availability of the 5Ј splice site and increased E10 inclusion. Three variations of the stem loop that differ in stem length have been proposed (7,9,11), but previous studies show that only the shortest predicted 18-nucleotide stem loop (Fig.  1B) is a candidate for regulating E10 splicing (17,21,24). The second hypothesis is that the ISS functions as a linear sequence that binds trans-acting factors and that FTDP-17 mutations in I10 alter this protein-RNA interaction (11,21). Here, experiments were designed to: 1) define the sequences required for ISS function, 2) identify additional I10 regulatory elements, 3) distinguish between the stem loop and linear models, and 4) determine whether I9 contributes to E10 splicing regulation. E10 splicing regulation was evaluated by transfecting COS-7, PC12, and rat primary cortical neuron cultures with vector pSPES containing E10 and portions of I9 and I10 inserted into an HIV tat intron between tat exons 2 and 3. E10 splicing was quantitated by RT-PCR assays of RNA from transfected cells.
I10 Sequences Required for Regulated Splicing-To determine the I10 sequences needed for regulated E10 splicing, a series of constructs were generated that differ only in the amount of I10 included (6B-37B, Fig. 1A), and cloned into the XhoI/BamHI sites of the MCS of vector pSPES. The FTDP-17 mutation E10ϩ3 was incorporated into a second set of constructs (6B3A to 37B3A) to determine whether the regulation of E10 splicing paralleled that observed in vivo in FTDP-17 patients with I10 mutations. The shortest constructs (6B and 6B3A) yielded only partial E10 inclusion, probably because positive downstream elements are missing (Fig. 1, C and D). Splicing was not influenced by the E10ϩ3 mutation. The 12B and 12B3A constructs showed near-constitutive splicing, presumably because I10 nucleotides 7-12 contain a positive sequence and the ISS is missing. Construct 18B has reduced E10 inclusion, indicating that I10 nucleotides 13-18 contain part or all of the ISS. These sequences restore mutation-sensitive regulated E10 splicing (18B3A). Longer constructs (24B, 37B, and hN) yielded varied amounts of E10 inclusion percentages, indicating additional regulatory element(s) exist downstream of the ISS.
I10 Has Antagonistic Regulatory Sequences-To further refine the locations of the ISS and the positive regulatory sequences adjacent to the ISS, a series of deletions in I10 were constructed, beginning at position 7 and extending to 42 (Fig.  1B). Deletion ⌬7-10 removes 4 of 6 nucleotides from the loop sequence. The stem loop hypothesis predicts this deletion would destabilize the proposed secondary structure, causing E10 to oversplice. However, E10 inclusion was decreased when these 4 nucleotides were removed (Fig. 2). Thus, the deleted nucleotides contain a positive element, a conclusion that is consistent with the results with constructs 6B and 12B (Fig. 1).
Additional constructs were tested to refine the locations of the ISS and the downstream positive element. Deletion ⌬7-11, designed to define the 5Ј end of the ISS, slightly increased E10 inclusion compared with ⌬7-10 ( Fig. 2). It appears that ISS function is partially compromised in ⌬7-11, which compensates for loss of the positive sequence located between positions 7 and 10. Thus, the functional ISS sequence begins at I10 position 11, a conclusion supported by FTDP-17 mutation FIG. 1. Deletion analysis of tau I10. A, the 3Ј end of tau E10 used in construct hN is shown by an open box, the intronic sequence as lowercase letters, and the 5Ј splice site as 5Јss. I10 deletion constructs are shown below hN. Each construct was inserted into the BamHI site (underlined) in the MCS of pSPES. BamHI sites were added to the 3Ј end of each I10 deletion construct to facilitate cloning into pSPES. FTDP-17 mutations are shown above the sequence. The ISS and ISM regulatory sequences from positions 11-18 and 19 -26, respectively, are shown below the I10 sequence by solid bars. B, the E10 -I10 boundary is shown as a previously proposed stem loop structure (7,9). Exon nucleotides are in capital letters and intronic nucleotides in lowercase letters. The locations of deletions and substitutions used in subsequent experiments are shown. C, representative autoradiograph of E10 splicing assays. D, quantitation of E10 splicing assays. Each bar represents the mean of at least three separate transfection experiments and 100% is the sum of E10Ϫ and E10ϩ. Error bars are standard deviations. A corrected significance criterion of p Ͻ 0.003 was used, and p values for comparison of each construct to normal human E10 are indicated with the following symbols: ‡, p Ͻ 1 ϫ 10 Ϫ3 ; , p Ͻ 1 ϫ 10 Ϫ4 ; †, p Ͻ 1 ϫ 10 Ϫ5 ; , p Ͻ 1 ϫ 10 Ϫ7 ; ϩ, p Ͻ 1 ϫ 10 Ϫ8 . Other comparisons are indicated using lines connecting the bars for the constructs being compared.
E10ϩ11A, that causes an increase in E10 splicing (25). Deletion ⌬17-18, designed to define the 3Ј end of the ISS, increased E10 inclusion. Thus, these 2 nucleotides are part of the ISS. The results from these deletions, from I10 FTDP-17 mutations, and the constructs in Fig. 1 (compare 12B, 18B, and 24B) define the ISS element as the sequence from 11 to 18. Additionally, these results demonstrate that there is positive element between 19 and 26.
To determine whether the ISS and the positive 19 -26 element interact, the ISS alone or both elements were deleted in mutant ⌬12-18 and ⌬12-24, respectively. Constitutive E10 incorporation was observed in both ⌬12-18 and ⌬12-24 ( with ⌬12-18 and ⌬12-24), it does not behave as a splicing enhancer per se, but rather as a modulator of ISS inhibition of E10 inclusion. Thus, the E10 19 -26 element is designated the intron splicing modulator (ISM). Further evidence supporting the existence of the ISM is the recent identification of an FTD mutation, E10ϩ19G (Fig. 1A). 2 This mutation decreases E10 inclusion (Fig. 3), although not as dramatically as removing the entire ISM (⌬19 -26, Fig. 2). It also appears that a purine at position 19 (mutants E10ϩ19G and E10ϩ19A, Fig. 3) compromises ISM function more effectively than a pyrimidine (E10ϩ19T).
Loop Nucleotides 8 and 10 Increase E10 Inclusion-The decrease in E10 splicing observed in mutant ⌬7-10 indicates this FIG. 2. Defining cis-acting sequences in I10. A, representative autoradiographs of E10 splicing assays in COS-7, PC12, and rat P1 neurons. B, quantitation of E10 inclusion. Bar graphs are as in Fig. 1. A corrected significance criterion of p Ͻ 0.004 was used for COS-7 and PC12 cells and p Ͻ 0.02 for rat primary neurons. Comparisons are for hN to each construct for a given cell type, and p values are as in Fig. 1 and as follows: sequence is not neutral. Intronic sequences immediately downstream of the 5Ј splice site consensus sequence (Fig. 4A) interact with splicing machinery components (26,27). Both U1 and U6 snRNA directly hybridize with the consensus 5Ј splice site sequence and these interactions extend to intron position 8 for U1 snRNA and positions 6 -10 for U6 snRNA ( Fig. 4A; Refs. 28 -31). Intron position 7 of tau protein is an A that is compatible with the corresponding U nucleotide in both U1 and U6 snRNAs. However, the C at I10 position 8 is incompatible with the corresponding A residue in both snRNAs. To increase base pairing between these snRNAs and the tau I10 5Ј splice site, nucleotides at positions 8 and 10 were individually mutated to complementary residues U and G, (mutants c(ϩ8)t and t(ϩ10)g, respectively; Fig. 4). E10 inclusion was enhanced in c(ϩ8)t, but remained unaltered in t(ϩ10)g only in COS-7 cells (Fig. 4). Double mutant c(ϩ8)t/t(ϩ10)g showed a slightly higher incorporation of E10, suggesting that U6 snRNA association with I10 is more stable when both I10 positions 8 and 10 are complementary. Thus, the E10 5Ј splice site extends to E10ϩ10 and deletion ⌬7-10 weakens 5Ј splice site interactions with splicing machinery resulting in reduced E10 inclusion.
The ISS Sequence Can Regulate a Heterologous Exon-To distinguish between the stem loop and linear models for ISS action, the 8-nucleotide ISS was inserted 11 nucleotides downstream of human ␤-globin exon 2 (␤E2) (Fig. 5). ␤E2 is flanked by 42 and 14 nucleotides of globin intron sequences and inserted into the MCS of pSPES (GE2, Fig. 5A). If the linear model is correct, the ISS should function in this heterologous context independent of secondary structure proposed at the tau E10/I10 junction. Although ␤E2 is a constitutive middle exon, it is included with 76% efficiency in GE2 (Fig. 5, B and C), because of the suboptimal nature of the flanking HIV tat splice sites (32). Previously, we demonstrated that ISS function requires a weak 5Ј splice site (21). Consistent with this previous work, when the ISS was tested with the normal ␤⌭2 5Ј splice site in GE2/ISS, the ISS did not alter exon inclusion (Fig. 5, B and C), nor did the addition of FTDP-17 mutation E10ϩ14 in the ISS (GE2/ISSϩ14) affect ␤E2 inclusion. When the ␤E2 5Ј splice site was weakened at the Ϫ1 position in GE25Ј (Fig. 5A), making it identical to the tau E10 5Ј splice site, total levels of ␤E2 inclusion remained unaltered for GE25Ј as in GE2 (Fig. 5, B and C). However, 16 and 21% of total E2ϩ transcripts were larger than expected in COS-7 and PC12 cells, respectively. This larger product, designated ␤E2aЈ, was sequenced and the results showed that the ␤E2 5Ј splice site is not used. Rather, an in-frame cryptic 5Ј splice site 39 nucleotides downstream of the ␤E2 5Ј splice site is used. Because the total amount of ␤E2 inclusion does not change when GE2 and GE25Ј are compared, use of a cryptic donor site for ␤E2aЈ is at the expense of the normal 5Ј splice site. When the ISS sequence was inserted (GE25Ј/ISS) total ␤E2ϩ transcript levels did not change (Fig.  5C), but the cryptic 5Ј splice site was greatly favored (89 and 93% of ␤E2ϩ transcripts in COS-7 and PC12 cells, respectively) (Fig. 5D). Compromising the ISS with FTDP-17 mutation E10ϩ14 (GE25Ј/ISSϩ14) restored ␤E2 splicing to the pattern observed for GE25Ј in COS-7 cells and to a lesser extent in PC12 cells. These experiments demonstrate that the ISS inhibits use of a 5Ј splice site in a heterologous context, strongly supporting the ISS linear model and that the ISS consists of 8 nucleotides (positions [11][12][13][14][15][16][17][18]. ISS Function Is Position-independent-The effect of ISS position on E10 splicing was investigated by inserting it at different locations flanking the 5Ј splice site (Fig. 6A). When the ISS was put into E10, a T nucleotide was added to the 8-nucleotide ISS so that the inserted sequence did not introduce a frameshift. When the ISS was 4 nucleotides upstream of the 5Ј splice site (E10-ISS, Fig. 6, A and B), E10 inclusion was virtually abolished, presumably because of the presence of 2 functional ISS copies. Inactivation of the E10 ISS copy using the E10ϩ14 mutation (E10-ISSϩ14) restored E10 splicing to al- most normal levels (compare hN and E10-ISSϩ14). Inactivation of both ISS copies (construct E10-ISSϩ14/ϩ14) resulted in increased but not constitutive E10 inclusion relative to hN. Thus, the ISS maintains partial inhibitory function in spite of mutation E10ϩ14, as also observed in ␤E2 (see GE25Ј/ISSϩ14, Fig. 5D). When only the E10 ISS was present and the I10 ISS was deleted (construct E10-ISS/⌬12-18), E10 inclusion was comparable with the normal construct hN. Mutating the E10 ISS (constructs E10-ISSϩ14/⌬12-18 and E10-ISSϩ14/⌬12-24) again resulted in elevated E10 inclusion at near constitutive levels.
The ISS was also tested at locations more distant from the 5Ј splice site. Insertion into E10 at the Ϫ11 position resulted in inhibition of E10 inclusion in both COS-7 and PC12 cells, but the effect was not altered by the E10ϩ14 mutation (data not shown). The inhibition observed is probably because of the disruption of a previously identified positive element at this location (21). Insertion of the ISS alone at I10 locations 31 and 40, or both the ISS and ISM at I10 position 40 did not alter E10 inclusion (data not shown). Thus, the ISS functions as a copy number-dependent linear element that can function when in close proximity to the 5Ј splice site.
The E10 3Ј Splice Site-Mammalian 3Ј splice sites have an invariant AG at the 3Ј end of the intron, a polypyrimidine tract (PPT), and a BPS that includes the branch point A nucleotide 15-50 nucleotides upstream of the exon (Fig. 7A). To see if the tau I9 PPT is suboptimal, the number of consecutive pyrimidines in the PPT was increased (Ϫ7T-11T and Ϫ3T-7T; Fig. 7 A and B). Both constructs showed constitutive E10 inclusion in COS-7, PC12, and rat P1 neuronal cells, indicating that the E10 3Ј splice site is weak. This conclusion was confirmed by replacing the 33-nucleotide tau I9 sequence in hN with a 42nucleotide fragment containing the 3Ј splice site from human ␤-globin intron 1 (construct Glo3Јss). Near constitutive splicing occurred in COS-7 and PC12 cells. The ␤-globin sequence was used because it is a well studied splice site from a constitutive exon with a strong BPS at Ϫ36 to Ϫ42 and a branch point A at Ϫ37 (33).
tau E10 has 2 candidate branch point A nucleotide at Ϫ23 and Ϫ27 (Fig. 7A). To identify the tau BPS, both candidate A nucleotides were individually neutralized by changing each to a G (Ϫ23G and Ϫ27G, Fig. 7A). E10 inclusion was almost completely abolished in both constructs (Fig. 7B), and neither A can compensate for the absence of the other. The strength of the BPS is determined primarily by its ability to hybridize to the U2 snRNA sequence 5Ј-GUAGUA-3Ј. Both the candidate proximal and distal BPSs defined by Ϫ23A and Ϫ27A, respectively, appear weak with a 4/7 match against the degenerate mammalian consensus BPS 5Ј-UNCURAC-3Ј (where N is any nucleotide, R is a purine, and the branch point A is underlined). However, the proximal BPS may provide a stronger template for base pairing with U2 snRNA (Fig. 7A).
Utilization of the weak E10 5Ј splice site requires E10 enhancer sequences (21). To ascertain whether use of the weak E10 3Ј splice site is also dependent on exon enhancer sequences, mutants Ϫ3T-7T and Ϫ7T-11T were combined with deletion E⌬5 in Ϫ3T-7T/E⌬5 and Ϫ7T-11T/E⌬5, respectively. Mutant E⌬5 (FTDP-17 mutation ⌬280K) is an in-frame 3-nu- FIG. 5. ISS function in a heterologous construct. A, exon/intron junction sequences for ␤E2 constructs. The ␤E2 sequences are in uppercase letters and the ␤-globin intron 2 sequences in lowercase letters. For cloning purposes, modifications (arrows) were introduced into the normal globin sequence in construct GE2, which destroy and create a BamHI (underlined) site in ␤E2 and ␤-globin intron 2, respectively. Mutation at ␤E2 position Ϫ1 in GE25Ј is shown in italics, a substitution that weakens the ␤E2 5Ј splice site sequence changing it from the normal ␤E2 sequence of 5Ј-AGgtgagt-3Ј to 5Ј-ATgtgagt-3Ј, which is identical to the tau E10 5Ј splice site sequence. The inserted ISS sequence in GE2/ISS is represented in italics and is located 11 nucleotides downstream of the ␤E2 5Ј splice site. The c to t alteration (FTDP-17 mutation E10ϩ14) at the fourth nucleotide in the ISS is underlined in GE2/ISSϩ14 and GE25Ј/ISSϩ14. B, representative autoradiographs for splicing assays. Arrows indicate E2ϩ and E2Ϫ fragments. Normal and cryptic E2ϩ spliced products are represented by a and aЈ, respectively. C, quantitation of splicing assays showing total ␤E2 levels in COS-7 and PC12 cells. The bar graphs are as described in Fig. 1 cleotide deletion within the PPE splicing enhancer element. This deletion results in almost complete inhibition of E10 inclusion (11,21). Strengthening the PPT immediately adjacent to the 3Ј AG could not compensate for the loss of the PPE in COS-7 cells, as E10 splicing is abolished in mutant Ϫ3T-7T/ E⌬5, whereas E10 splicing is maintained at normal levels in PC12 cells (Fig. 7B). When the PPT adjacent to the BPS is strengthened, constitutive E10 splicing was observed in both COS-7 and PC12 cells despite the lack of PPE function in mutant Ϫ7T-11T/E⌬5. These results show that the PPE interacts with the weak 3Ј splice to control E10 inclusion.
Effects of Distal I9 and I10 Sequences on E10 Inclusion-To determine whether distal intron sequences regulate E10 inclusion, I9 and I10 sequences were extended from 33 and 51 nucleotides (hN) to 567 and 190 nucleotides, respectively, in E10860 (Fig. 8A). In human I9, a tandem-repeat sequence is located at nucleotides Ϫ21 to Ϫ55 and Ϫ69 to Ϫ116 upstream of E10 (Fig. 8; Ref. 34). Mouse has only 1 repeat (34). Construct hN contains 11 nucleotides of the 3Ј end of the repeat closest to E10 and includes both BPS with branch point AЈs at Ϫ23 and Ϫ27. The distal I9 repeat present in construct E10860 also contains a sequence identical to the E10 BPS with the equivalent branch point A at position Ϫ81. There was no difference in E10 inclusion when hN and E10860 were compared. Thus, additional regulatory sequences were not apparent when extended I9 and I10 sequences were tested, although the presence of offsetting positive and negative regulatory elements cannot be excluded. DISCUSSION In MAPT, a complex hierarchy of enhancer and silencer sequences in E10, I9, and I10 determines E10 alternative splic-FIG. 6. ISS function within E10 sequences. A, tau sequence for the E10/I10 junction. E10 and I10 sequences are in uppercase and lowercase letters, respectively. The normal ISS sequence is indicated by a filled oval and mutant ISSϩ14 by a hollow oval with a t in the center. Construct E10-ISS contains a second ISS copy in E10, which replaces E10 sequences between Ϫ5 and Ϫ10 nucleotides upstream of the 5Ј splice site. Insertion of an in-frame normal or mutant ISS copy at the same position without replacing E10 sequences also show similar splicing effects (data not shown). In subsequent mutants, the ISS within E10, I10, or both was neutralized by a point mutation or deletion. Filled triangles indicate deletion of the ISS (⌬12-18) or ISS and ISM (⌬12-24). Values to the right of the constructs represent the average percentage of inclusion of E10ϩ transcripts from three transfection experiments with standard deviations (S.D.). A corrected significance criterion of p Ͻ 0.008 and p Ͻ 0.016 was used for COS-7 and PC12 cells, respectively. Significance levels for each construct compared with normal E10 (hN) are given with symbols as in Fig. 1 and 2 ing. Earlier work demonstrated that multiple ESE sequences and an ESS in E10, a weak 5Ј splice site, and the ISS in I10 regulate E10 inclusion (11,21). Here, we identify an additional I10 regulatory element, the ISM, and demonstrate that a weak 3Ј splice site also contributes to E10 splicing regulation. The ISM interacts with the ISS, a weak 5Ј splice site, and E10 regulatory elements to control E10 inclusion. Likewise, the weak 3Ј splice site interacts with cis-acting elements in E10. The ISS and ISM were defined in experiments using transiently transfected COS-7, PC12, and rat primary neurons, with most results being qualitatively and quantitatively similar in all three cell types. Thus, the regulatory elements studied here are not neuron-specific and are generalizable to different cell types.
The ISS and ISM form a novel 16-nucleotide bipartite regulatory sequence immediately downstream of the weak E10 5Ј splice site. The ISS and ISM function antagonistically in regulating E10 splicing, as they exert opposite effects on E10 inclusion when individually inactivated (Fig. 2). Several lines of evidence indicate that the ISM and the ISS elements are discrete. First, the ISS sequence alone, without the ISM, can function independent of position on either side of the 5Ј splice site (Fig. 6). Second, the ISS can function in a heterologous setting (Fig. 5). Third, the ISS inhibits E10 inclusion when the ISM is inactivated by deletions ⌬19 -22 and ⌬23-26 (Fig. 2) and by mutations (Fig. 3). In contrast, the ISM does not appear to act without the ISS, and functions to mitigate the ISS-mediated repression of E10 inclusion (Fig. 2). FIG. 7. Characterization of the E10 3 splice site. A, human I9 sequence in hN. Shown are components of the 3Ј splice site that include the invariant intronic 3Ј-AG dinucleotide, which is separated from an upstream BPS by a PPT. The BPS is usually located 15-50 nucleotides upstream of the 3Ј cleavage site and is known to base pair with the U2 snRNA sequence AUGAUG. Arrows above the human I9 sequence indicate nucleotides altered in experimental constructs. Candidate branch point A nucleotides at positions Ϫ23 and Ϫ27 (underlined) were individually neutralized to a "g" in Ϫ23G and Ϫ27G, respectively. To strengthen the PPT, purine residues at positions Ϫ3, Ϫ4, and Ϫ7 as well as at positions Ϫ7, Ϫ10, and Ϫ11 were converted to "t" residues in mutants Ϫ3T-7T and Ϫ7T-11T, respectively. The box shows a comparison between the human BPS and the loosely conserved mammalian BPS consensus (N, any nucleotide; R, purine) and the U2 snRNA sequence. B, autoradiograph of RT-PCR products showing E10 inclusion.
The close proximity of the ISM, ISS, and the 5Ј splice site suggests that these three elements interact. As shown previously, ISS function requires that the adjacent 5Ј splice site be suboptimal (21). Here we show in E10 constructs that the ISS can function in a position-independent fashion as long as it is close to the 5Ј splice site (Fig. 6), but will not function if the distance to the 5Ј splice site too great. Thus trans-acting factors that bind to the ISS may directly hinder U1snRNP or other components from correctly interacting with the 5Ј splice site, and therefore inhibit E10 inclusion. Results compatible with this hypothesis were obtained when the ISS was inserted downstream of a weak 5Ј splice site in ␤E2 globin constructs (Fig. 5). The ISS inhibited the use of the attenuated ␤E2 5Ј slice site. Instead, an in-frame cryptic 5Ј splice site located 39 nucleotides downstream was utilized (Fig. 6). Interestingly, this cryptic 5Ј splice site has a much weaker consensus sequence than the attenuated ␤E2 5Ј splice site. The cryptic and weakened 5Ј splice sites show a 5/8 and a 7/8 match, respectively, with the mammalian 5Ј splice site consensus. Thus, the ISS inhibits use of a weak 5Ј splice site, thereby favoring splicing at the next available cryptic site, even though the cryptic site has a weaker consensus sequence.
The ISM does not appear to directly affect use of the 5Ј splice site. Instead, the ISM has a positive effect on E10 splicing only in presence of the ISS. Thus, the ISM directly inhibits the ISS function of repressing use of the E10 5Ј splice site (Fig. 2). A plausible mechanism for ISM function is that factors associating with the ISM sterically hinder the stability or association of ISS-specific inhibitory factors, thus allowing indirect enhancement of E10 inclusion. 3 The work presented here and previously (11,21) is consistent with ISS acting as a linear sequence rather than through a secondary RNA structure such as a stem loop (Fig. 1B). The following arguments support this conclusion. 1) The 8-nucleo-tide ISS can function in a heterologous setting to inhibit use of a weak 5Ј splice site (Fig. 5), a sequence context that does not form the proposed stem loop. 2) The ISS can be placed in E10 near the 5Ј splice site and function to inhibit splicing (Fig. 6). 3) Although FTDP-17 mutations are proposed to disrupt the hypothetical stem loop, some (17) but not all compensatory changes designed to restore stem loop base pairing act to restore normal ISS function (11). 4) Double mutants containing I10 FTDP-17 mutations (E10ϩ12, E10ϩ13, E10ϩ14, and E10ϩ16) and a 3-nucleotide deletion in the E10 ESE (E⌬5; Ref. 21) give variable E10 inclusion levels even though these I10 mutations alone disrupt the stem loop to a comparable degree (21). 5) Both human and mouse E10/I10 junctions form stem loops of comparable stability, yet normal or chimeric human and mouse constructs do not support a stem loop model for E10 regulation either in transfected COS-7 cells (21) or in rat P1 neurons. 3 Additionally, preliminary UV-cross-linking experiments indicate that nuclear factors associate with normal but not mutant ISS and ISM RNA sequences. 3 Further work is necessary to characterize these factors including their roles in regulating E10 splicing.
The ISS interaction with the 5Ј splice site components is not dominant. The antagonistic action of the ISS on 5Ј splice site recognition and/or use can be over-ridden either by strengthening an E10 ESE, or by strengthening the 5Ј splice site sequence (21). Likewise, strengthening the 3Ј splice site also overrides ISS action (Fig. 7), demonstrating the complex interaction between the different regulatory elements.
The E10 5Ј Splice Site-The tau E10 5Ј splice site is inherently weak as it deviates from the mammalian consensus based on complementation with the 5Ј end of U1 snRNA (Fig. 4A). The predicted result is that U1 snRNA binds the tau E10 5Ј splice site weakly as recently confirmed by in vitro association studies (35). U1 snRNP, aided by other non-snRNP splicing factors, associates transiently with the 5Ј splice site early during spliceosome assembly and serves to commit the pre-mRNA 3  to the splicing pathway (36). Although this interaction alone does not determine the position or even use of the 5Ј splice site, it ensures at an early step proper fidelity in 5Ј splice site selection (31). More importantly, the 5Ј splice site interaction with U1 snRNP is replaced by subsequent interactions with U5 and U6 snRNPs in a process that defines the final position of the 5Ј splice site. An invariant sequence ACAGAG within U6 snRNA interacts with intron positions 5-10 downstream of the 5Ј cleavage site (Fig. 4A; Ref. 37). For tau, previous studies showed that strengthening the E10 5Ј splice site either by incorporating FTDP-17 mutations S305N or E10ϩ3 (9,11,18) or replacing it with a constitutive 5Ј splice site from human ␤-globin exon 2 (21) caused E10 to oversplice. Here, results from ⌬7-10 and substitution analyses in I10 just downstream of the conserved 5Ј splice site sequence (Figs. 2B and 4C) indicate that nonconserved I10 sequences from positions 7 up to 10 are part of the weak 5Ј splice site sequence. Thus, an interesting feature regarding E10 splicing regulation is that both U1 and U6 snRNPs are predicted to interact weakly with the normal E10 5Ј splice site region, as substitutions within I10 positions 8 and 10 that are thought to strengthen U1 snRNP and U6 snRNP interactions increase E10 inclusion.
Within the pre-mRNA/spliceosome complex, U6 snRNA is thought to possess the catalytic activity for cleaving 5Ј and 3Ј splice sites (38). Although our work predicts a weak interaction between the E10 5Ј splice site and U6 snRNA, not all substitutions that are predicted to increase U6 snRNA association with the 5Ј splice site result in increased E10 splicing (t(ϩ10)g in COS-7 cells; Fig. 4C). One explanation is that this particular substitution may alter the association of other splicing factors within the I10 region. For example, interaction of the yeast U1 snRNP protein Nam8p with intron positions 6 -13 is required for efficient 5Ј splice site recognition by U1 snRNP, especially when the 5Ј splice site is weakened (26). Several other snRNPspecific and non-snRNP splicing factors have also been crosslinked to the 5Ј splice site (27). Splicing is most enhanced in double mutant c(ϩ8)t/t(ϩ10)g in both COS-7 and PC12 cells, suggesting possible cooperative effects when both U1 and U6 snRNA associations with a weak 5Ј splice site are strengthened. The important contributions of other cis-acting splicing sequences on E10 5Ј splice site use must also be taken into account as seen by mutations or substitutions elsewhere in E10, I9, or I10 sequences that dramatically affect the outcome of E10 splicing (21). These splicing regulatory sequences in association with trans-acting splicing factors help in stabilizing (or destabilizing) spliceosome interactions with the pre-mRNA template.
The E10 3Ј Splice Site Is Regulatory-The 3Ј splice site is weak, as shown when the PPT is strengthened by increasing the number of consecutive pyrimidines or where the entire I9 sequence is replaced with a constitutive globin 3Ј splice site (Fig. 7B). Most enhancer-dependent introns are associated with a weak or suboptimal PPT (Ref. 39; reviewed in Ref. 40), which when strengthened overcomes the need for an enhancer. ESEs not only aid in the recognition of weak splice sites (reviewed in Ref. 41), but some also function in overcoming the effects of splicing silencers (42). This is certainly true for mutant Ϫ3T-7T (compare Ϫ3T-7T to double mutant Ϫ3T-7T/E⌬5, Fig. 7B). Strengthening the PPT adjacent to the non-consensus BPSs in Ϫ7T-11T allows for efficient recognition of the 3Ј splice site independent of enhancer and silencer sequences (compare Ϫ7T-11T and Ϫ7T-11T/E⌬5, Fig. 7B). Early in the splicing pathway, splicing factors such as the 35-kDa U2 auxiliary factor (U2AF 35 ) associates with the 3Ј AG, which in turn stabilizes the interaction of U2AF 65 with the PPT. U2AF 65 in turn strengthens the recruitment of U2 snRNP to the BPS. These interactions are critical in the case of weak 3Ј splice sites, where SR proteins in association with ESEs increase the affinity of U2AF and U2 snRNP for the upstream 3Ј splice site. The presence of splicing silencer sequences further complicates these interactions. The PPT in Ϫ7T-11T may provide a stronger template for U2AF 65 interaction than in Ϫ3T-7T. Although 3 purines are converted to pyrimidines (all uridines) in both Ϫ3T-7T and Ϫ7T-11T, they do differ in the number of consecutive pyrimidines. Construct Ϫ7T-11T has 17 continuous pyrimidines adjacent to the BPS, whereas Ϫ3T-7T has a 10-and a FIG. 9. Model for tau E10 splicing regulation. Shown are putative splicing factor interactions with exonic and intronic splicing regulatory sequences that modulate use of the weak E10 5Ј and 3Ј splice sites. E10 splicing elements include three ESE elements (SC35-like, PPE, and ACE) at the 5Ј end of E10 and two ESE elements at the 3Ј end of E10, as well as an 18-nucleotide ESS sequence between these two ESEs. I10 regulatory sequences include the bipartite regulatory ISS and ISM element. SR factors usually in association with ESE elements recruit and stabilize U1 and U2 snRNP interactions with the 5Ј and 3Ј splice sites, respectively. These exon-bridging interactions ultimately help in defining the exon. For tau E10 definition, the positive roles of the ESEs (blue arrows with ϩ) and ISM are critical because of weak 3Ј and 5Ј splice sites as well as the inhibitory ESS and ISS elements (red arrows with Ϫ). The positive role of the ISM on E10 splicing is indirect and ISS-dependent. Known disease mutations are shown at the bottom. 7-nucleotide pyrimidine stretch interrupted by two guanidines. The difference in Ϫ3T-7/E⌬5 splicing between COS-7 and PC12 cells (Fig. 7) may be a result of the fact that PC12 are neuronal cells and COS-7 are not. The difference could also be caused by the recently identified splicing factor U2AF 26 that is particularly enriched in brain and shares a similar function with U2AF 35 in recruiting U2AF 65 to weak 3Ј splice sites (43). Thus, the E10 PPE is required not only for recruitment of splicing factors to the weak 3Ј splice site, but also needs to overcome the effects of splicing silencer sequences within E10 and I10. E10 inclusion also involves the use of at least two overlapping and weak BPSs with branch point A nucleotides at nucleotides Ϫ23 and Ϫ27 upstream of E10. In addition to the weak PPT, multiple weak BPSs also contribute to the weak 3Ј splice site in regulating E10 splicing.
Model for Regulation of tau E10 Splicing-A model for tau E10 splicing regulation is proposed (Fig. 9), where multiple weak interactions between factors binding to sequences in E10, I9, and I10 ultimately help in defining E10. Rather than any single sequence, multiple enhancer and silencer sequences modulate use of the weak 5Ј and 3Ј splice sites that border E10. In both constitutively and alternatively spliced exons, SR proteins interact with ESE elements and stabilize interactions between the splicing machinery (snRNPs) and splice sites across the length of the exon, thus defining the exon (44). For example, ESE-bound SR factors SC35 and ASF/SF2 interact directly with U1 snRNP-specific factor U1 70K and stabilize U1 snRNP association with the 5Ј splice site (36,39,45). At the 3Ј splice site region, the U2AF 65 and U2AF 35 subunits of the U2AF heterodimer interact with the PPT and 3Ј-AG, respectively. These 3Ј splice site associations are stabilized through ESE-associated SR factors that directly interact with U2AF 35 . U2AF 65 in turn stabilizes U2 snRNP interaction with the BPS. The involvement of ESEs is particularly important when either splice site is weak. In the case of tau E10, both splice sites are weak, and thus multiple ESE sequences are required for E10 inclusion (21). In addition to their role in exon definition and splice site selection, SR and other proteins may also antagonize splicing inhibitory factors in association with splicing silencer sequences, such as the ESS and ISS sequences in tau E10 splicing. An additional level of complexity is presented by the juxtaposition of the 5Ј splice site with the ISS and ISM elements in I10.
The proposed interactions of ESE and ESS sequences with weak 3Ј and 5Ј splice sites for MAPT E10 are consistent with the exon definition model for splicing (44). According to this model, components of the splicing machinery in association with the 5Ј and 3Ј splice sites communicate across the exon to define the exon/intron boundary. This interaction across the exon is mediated and strengthened by SR proteins bound to enhancer sequences in the exon. Weakened interactions with splice sites at either end of the exon allow exon skipping.
E10 Regulation and FTDP-17-Work in cell culture models using a number of different cell lines has demonstrated that E10 is regulated by at least eight cis-acting elements in the context of weak 3Ј and 5Ј splice sites. Evidence supporting the function of these elements in vivo comes from the identification of FTDP-17 mutations in six of the cis-acting sequences (Fig. 9). Mutations include intronic mutations in the ISS (7,9,16), a missense mutation (N279K) and a deletion mutation (⌬280K) in the PPE (11,16,19), a silent mutation (L284L) in the ACE (11), a deletion mutation (⌬296N; Ref. 46), a silent mutation (N296N; Ref. 47) and a missense mutation (N296H, Ref. 48) in the ESS, and a new mutation E10ϩ19 2 in the ISM. FTDP-17 mutations S305N, S305S, and E10ϩ3, which are in the 5Ј splice site consensus sequence, provide in vivo evidence that the 5Ј splice site is functionally suboptimal. The fact that these splicing mutations cause severe neurodegenerative disease demonstrates the functional significance of these regulatory elements.