Alternative splicing attenuates transgenic expression directed by the apolipoprotein E promoter-enhancer based expression vector pLIV11.

The plasmid vector pLIV11 is used commonly to achieve liver-specific expression of genes of interest in transgenic mice and rabbits. Expression is driven by the human apolipoprotein (apo)E 5′ proximal promoter, which includes 5 kb of upstream sequence, exon 1, intron 1, and 5 bp of exon 2. A 3.8 kb 3′ hepatic control region, derived from a region ∼18 kb downstream of the apoE gene, enhances liver-specific expression. Here, we report that cDNA sequences inserted into the multiple cloning site (MCS) of pLIV11, which is positioned just downstream of truncated exon 2, can cause exon 2 skipping. Hence, splicing is displaced to downstream cryptic 3′ splice acceptor sites causing deletion of cloned 5′ untranslated mRNA sequences and, in some cases, deletion of the 5′ end of an open reading frame. To prevent use of cryptic splice sites, the pLIV11 vector was modified with an engineered 3′ splice acceptor site inserted immediately downstream of truncated apoE exon 2. Presence of this sequence fully shifted splicing of exon 1 from the native intron 1–exon 2 splice acceptor site to the engineered site. This finding confirmed that sequences inserted into the MCS of the vector pLIV11 can affect exon 2 recognition and provides a strategy to protect cloned sequences from alternative splicing and possible attenuation of transgenic expression.

phoresis, transgenic cassettes were microinjected into pronuclei of fertilized C57BL/6 (Harlan Teklad) eggs by the Transgenic Mouse Core Facility of Wake Forest University. All mice were housed in a pathogen-free animal facility in plastic cages in a temperaturecontrolled room (22°C) with a 12-h light and 12-h dark cycle. The mice were fed ad libitum a cereal-based rodent chow diet. All animal procedures were conducted in conformity with Public Health Service policy and were approved by the Institutional Animal Care and Use Committee of Wake Forest University School of Medicine.

Sequencing of PCR products
PCR products were separated by agarose gel electrophoresis, excised from the gel, and purifi ed using QiaQuick columns according to the supplier's protocol (Qiagen). PCR products were sequenced by the Biomolecular Resource Laboratory of the Comprehensive Cancer Center of Wake Forest University using an Applied Biosystems Model 3100 Genetic Analyzer.

Generation of anti-human MTP antiserum
cDNA corresponding to amino acid residues 772-874 of hMTP (precursor protein) was generated by PCR and cloned transgenic cassette to facilitate separation of the transgene from the plasmid backbone.
Inserting human MTP (hMTP) cDNA into pLIV11-NotI. The 5 ′ end of the hMTP cDNA insert was released from plasmid pRC/ neo ( 13 ) by digestion with SpeI, followed by treatment with T4 DNA polymerase. The 3 ′ end was released by digestion with MluI. pLIV11-NotI was digested with ClaI followed by treatment with T4 DNA polymerase. Linearized plasmid was digested with MluI and dephosphorylated with calf intestine alkaline phosphatase. Following gel purifi cation, the hMTP insert and pLIV11-NotI plasmid were ligated and transformed into Escherichia coli strain XL1-Blue. The resulting plasmid is named pLN-hMTP.
Inserting Drosophila MTP cDNA (dMTP) into pLIV11-NotI. The dMTP cDNA insert was separated from plasmid pCMV5 ( 10 ) by digestion with Eco RI (5 ′ end) and ClaI (3 ′ end), followed by treatment with T4 DNA polymerase. pLIV11-NotI was digested with HpaI and dephosphorylated with calf intestine alkaline phosphatase. Following gel purifi cation, the dMTP insert and pLIV11-NotI plasmid were ligated and transformed into E. coli XL1-Blue. The resulting plasmid is named pLN-dMTP.
The entire insert and fl anking regions of each plasmid construct were verifi ed by sequence analysis. The pLN-hMTP and -dMTP transgenes were separated from the plasmid backbone by digestion with NotI (5 ′ ) and SpeI (3 ′ ); pL-mLCAT was digested with SalI (5 ′ ) and SpeI (3 ′ ). Following isolation by agarose gel electro-  Table I). B: The indicated primer sets were used to amplify tail DNA from wild-type mice and transgenic founders 1, 2, 9, and 17. PCR of water serves as a negative control and PCR of plasmid pLN-hMTP serves as a positive control. Arrowheads indicate position of expected PCR products; asterisk indicates a nonspecifi c band.
RH7777 (rat hepatoma) extracts, although not in African green monkey liver or human hepatoma cell extracts ( Fig. 2A , lower). These data indicate that despite the presence of the transgene, hMTP expression was undetectable in offspring from two genotypically positive hMTP transgenic founders.

Transgenic hMTP mRNA contains 5 ′ deletions
To explore whether mRNA transcribed from the transgene was present and of the predicted structure, pLN-hMTP founder 1 liver mRNA was subjected to RT-PCR using primers that spanned the entire transcription unit. Four of the primer pairs gave rise to RT-PCR products of the expected size ( Fig. 2B , Lanes 2-5); however primer set 659+651, which spans apoE intron 1 ( Fig. 1C ), gave rise to a band of ‫ف‬ 700 bp rather than the expected 918 bp ( Fig.  2B , Lane 1). Hence, although the transgene was transcribed, the 5 ′ end of the predominant mRNA species appeared aberrant. To explore the structure of the mRNA further, the 659+651 PCR product was sequenced using antisense primer 668 (supplementary Table I  To determine if generation of the heterogeneous hMTP mRNA sequences was unique to expression in transgenic mouse liver or inherent to the construct, pLN-hMTP was transfected into HepG2 cells and RNA was into the vector pMAL-c2× (New England Biolabs) for generation in E. coli of a maltose binding protein-hMTP fusion protein. Following purifi cation via amylose affi nity chromatography, the fusion protein was injected subcutaneously into rabbits with incomplete Freund's adjuvant (Lampire Biologicals, Pipersville, PA). Whole serum was used at 1:1000 dilution for immunoblot analysis.

Generation of hMTP transgenic mice
Initial PCR analyses of tail DNA from pLN-hMTP transgenic founders using primers specifi c for the 3 ′ HCR (601 and 602; supplementary Table I Table I sequence was identical to the published apoE gene sequence used to assemble this plasmid ( 15 ) as well corresponding database entries (e.g., locus NW_001838496). To ensure that mutations affecting the 3 ′ splice acceptor site did not occur during generation of transgenic mice, the appropriate region of transgenic mouse tail DNA was amplifi ed by PCR with primers 659 and 630 and then sequenced with oligo 630 (supplementary Table I, Figs. 1A, 2C ). The 573 nucleotide sequence starting at exon 2 and extending upstream into intron 1 was also identical to the published sequence (data not shown), indicating that no mutation within or proximal to the 3 ′ splice acceptor site was responsible for the observed alternative splicing.

A consensus 3 ′ splice acceptor site positioned downstream of the apoE intron 1-exon 2 border blocks activation of cryptic splice sites within hMTP cDNA
In an attempt to protect the cDNA sequences cloned into pLIV11-NotI from being used as cryptic 3 ′ splice acceptor sites during intron 1 removal, the pLN-hMTP plasmid was modifi ed to encode an engineered 3 ′ splice acceptor site positioned downstream of the native intron 1-exon 2 junction but upstream of the hMTP cDNA to form plasmid pLNSR-hMTP ( Fig. 5A ). When this construct was transfected into HepG2 cells and RNA subjected to RT-PCR with primers 659 and 668, a PCR product size of ‫ف‬ 400 bp was observed ( Fig. 5B , lane 5), consistent with splicing at the engineered splice acceptor site. In contrast, both pLN-hMTP-transfected HepG2 cell and transgenic mouse liver RNA gave products of ‫ف‬ 260 bp, consistent with splicing at the alternative sites ( Fig. 5B , lanes 3, 4). Sequence analysis of the pLNSR-hMTP RT-PCR product revealed a homogenous sequence, consistent with splicing of exon 1 to the engineered splice acceptor site 1 ( Fig. 5A ) and retention of the initiating ATG codon (supplementary Fig. IV and data not shown). Hence, the presence of the engineered 3 ′ splice acceptor site fully shifted exon 1 splicing from the native intron 1-exon 2 junction to the new site and prevented use of the downstream cryptic splice sites diagrammed in Fig. 4A . These results confi rm that sequences cloned downstream of truncated exon 2 can dramatically affect native intron 1, 3 ′ splice acceptor site utilization and also provides a strategy for protecting cloned sequences from undergoing deletion via alternative splicing. To this end, pLIV11-NotI was modifi ed to contain the engineered 3 ′ splice acceptor site by ligating the same double stranded oligonucleotide shown in Fig.  5A into the MunI and MluI sites of pLIV11-Not1 to form the variant plasmid, pLIV11-NSR ( Fig. 5C ). cDNA sequences cloned into the MluI and/or ClaI sites of this vector should not be subjected to alternative splicing even if they contain cryptic 3 ′ splice acceptor sites.

Other cloned sequences promote alternative splicing in pLIV11
In contrast to our experience with hMTP, the vector pLIV11 has been used widely and successfully to achieve transgenic expression of a variety of genes. We attribute this to one of two possibilities: 1 ) the sequences cloned subjected to RT-PCR and sequence analysis, as described for supplementary Fig. IIIA. As observed in supplementary Fig. IIIB, the sequences obtained were identical to those observed in transgenic mice. Based on these results, it appears that when expressed in either mouse liver or transfected hepatoma cells, the pLN-hMTP plasmid gives rise to three mRNA sequences. Sequence 1 is the expected, although the least abundant sequence. Sequences 1A and 1B are the predominant species and each contain extensive deletions ( Fig. 3 ).

Transgenic hMTP mRNA 5 ′ deletions arise due to alternative splicing
The nature of the deletions observed in transgenic mouse liver and transfected HepG2 cells was consistent with alternative splicing, as each of the deletions occurred 3 ′ of AG dinucleotides, which were preceded by pyrimidine-rich domains ( 14 ) ( Fig. 3 ). The sites of putative alternative splicing, involving removal of introns 1A and 1B, are diagrammed in Fig. 4A . These two splicing events, which give rise to the majority of mature mRNA, involved removal of the initiating ATG, thereby producing a dysfunctional transcript. Although the properly spliced species is rare, such that it is barely visible in the sequencing electropherograms in supplementary Fig. IIIA and B, it can be detected by RT-PCR using the appropriate primer pair (primers 667 and 668; supplementary Table I), as observed in Fig. 4B , lanes 10-12. However, this species is present in trace abundance, as evidenced by the inability to detect the mRNA without performing high cycle number PCR using primers 659 and 668, which can detect each of the three sequences ( Fig. 4B , lane 4) and the fact that the protein product was undetectable in transgenic mouse liver ( Fig. 2A ).
A possible explanation for the preferential use of the downstream cryptic 3 ′ splice acceptor sites in pLN-hMTP is that the native intron 1-exon 2 junction underwent mutagenesis. As specifi ed under Material and Methods, the starting pLN-hMTP plasmid was sequenced, including exon 1, intron 1, and exon 2. The apoE-derived plasmid  Table I and Fig. 2C ). The multiple sequences observed in the electropherograms (supplementary Fig. IIIA, B) are depicted. E1 and E2, exons 1 and 2; MCS, multiple cloning sequence; Plasmid Sequence is sequence derived from the original hMTP expression plasmid that was carried over into pLN-hMTP; dashes indicate deleted nucleotides; arrow indicates beginning of hMTP cDNA. sequence from the 3 ′ to 5 ′ end of the transgenic mRNAs, we noted no signs of alternative splicing in NPC1L1 transgenic mouse liver (data not shown). In contrast, dMTP and mLCAT mice both showed signs of alternative splicing. In the case of dMTP, the alternative splicing occurred upstream of the initiating ATG codon; in the case of mLCAT, alternative splicing occurred within the open reading frame, although this was a minor product (supplementary Fig. V).
into the vector contain either no or very weak cryptic splice acceptor sites or in some other way do not negatively affect apoE intron 1-exon 2 splice site recognition, or 2 ) alternative splicing is taking place but removes only 5 ′ untranslated sequences, leaving the open reading frame intact. To explore the generality of our observations regarding alternative splicing, we analyzed mRNA from NPC1L1 ( 7 ), mLCAT, and dMTP transgenic mice. Using primers that Fig. 4. Alternative splicing of transgenic hMTP. A: Intron-exon structure of the hMTP transgene is depicted. Brackets and diagonal lines defi ne introns utilized during hMTP mRNA processing. Intron 1 is the native apoE intron; introns 1A and 1B arise from skipping of exon 2 and utilization of downstream cryptic 3 ′ splice acceptor sites. The sequence details found at the native and alternative 3 ′ splice acceptor sites are shown. The vertical arrows indicate the intron boundaries (lower case is used for exonic sequence). The underlined sequence shows the pyrimidine rich track, which precedes the AG dinucleotide at the 3 ′ end of the introns; the numbers indicate the number of pyrimidines/number of total nucleotides within the underlined region. The intron 1A and 1B splice acceptor sites are closely spaced; hence, the parenthetic arrow in the intron 1B sequence defi nes the splice acceptor site utilized for formation of intron 1A. B: RT-PCR of hMTP mouse liver RNA using the indicated primers. Primer pair 667 and 668 is specifi c for the properly spliced mRNA product generated by removal of native intron 1; primer pair 659 and 668 amplify all of the splicing products (arrowhead indicates PCR product corresponding to removal of native apoE intron 1); the predominant products corresponding to removal of introns 1A and 1B are similarly sized and run as a single band. HC, high cycle PCR generated by isolating initial PCR products and subjecting to a second round of PCR. Lanes 7 and 13 show PCR results using pLIV-hMTP plasmid as template with the indicated primers.  5. Insertion of an engineered 3 ′ splice acceptor site protects hMTP from alternative splicing. A: A double stranded oligonucleotide containing an engineered 3 ′ splice acceptor site was inserted into pLN-hMTP at the MunI and MluI sites, as indicated, to form plasmid pLNSR-hMTP. The branch site, pyrimidine (Pyr)-rich track, and predicted splice acceptor sites are indicated. B: pLN-hMTP (pLN) and pLNSR-hMTP (pLNSR) were transfected into HepG2 cells using FuGENE HD. RNA from transfected cells and from founder 1 (Mouse 1) liver was subjected to RT-PCR using primers 659 and 668, as described for Fig. 4B , and sized by agarose gel electrophoresis. C: Organization of vector pLIV11-NSR. Restriction enzyme sites at the 5 ′ and 3 ′ ends appropriate for separating transgene from plasmid backbone are shown (note that SalI appears twice). The 3 ′ splice acceptor site (3 ′ SAS) shown in A, was cloned into the MunI and MluI sites of pLIV11-NotI, leaving MluI and ClaI for inserting cDNAs of interest. splice acceptor sites contained within cDNA. This modification to the pLIV11 plasmid and its derivative seems advisable to ensure unattenuated expression of cloned sequences. Accordingly, for future use, a variant of pLIV11-NotI termed pLIV11-NSR was engineered as diagrammed in Fig. 5C .
Alternatively, should unmodifi ed pLIV11 vector be employed for transgenic expression, cDNA sequences should be screened and, if possible, modifi ed to remove potential cryptic 3 ′ splice acceptor sites. Although the large size of pLIV11-related plasmids gives rise to low transfection effi ciencies, making detection of protein overexpression in cultured cells diffi cult, the use of RT-PCR revealed that the pattern of splicing in mouse liver and HepG2 cells was identical (supplementary Fig. III). Hence, the sizing and sequence analyses of RT-PCR products derived from pLIV11-transfected HepG2 or other hepatoma cells can provide a reliable means of monitoring the pattern of splicing and mRNA structure prior to oocyte injection.

DISCUSSION
The 5 ′ proximal promoter elements as well as the 3 ′ HCR are critical to drive high levels of liver-specifi c expression using the vector pLIV11 and its derivatives. The exon 1-intron 1-exon 2 sequences present in this construct may also be important as in some studies, intronic sequences were shown to increase transgenic expression ( 16 ). In the current report, we demonstrated that sequences cloned downstream of the intron 1-exon 2 splice acceptor site can affect splice site utilization, resulting in exon 2 skipping and use of cryptic splice acceptor sites located within a cloned cDNA. In the case of pLN-hMTP, splicing resulted in deletion of an initiating codon and failed transgenic expression.
The reason for the poor utilization of the native apoE intron 1, 3 ′ splice acceptor site is uncertain. For the most part, the cis -acting sequences that drive splice site selection are contained within introns themselves, which are recognized by components of the spliceosome ( 14,17 ). In particular, 3 ′ splice acceptor sites adhere to a consensus that includes a conserved A positioned ‫ف‬ 18-50 nucleotides upstream of the 3 ′ end of the intron, which serves a branch point for lariat formation, a у 11 nucleotide pyrimidine-rich tract, and a conserved AG dinucleotide (18)(19)(20). Based on its close adherence to the consensus sequence, the native apoE intron 1, 3 ′ splice acceptor site would be expected to serve as an effi cient substrate. Indeed, based on database searches and available literature, no evidence of alternative splicing involving human apoE exon 2 was found. Nonetheless, alternative splicing is a common means of achieving protein diversity and is believed to affect ‫ف‬ 40-60% of human genes ( 21 ). Alternative splicing is driven in part by the presence of splicing enhancer and silencing elements, some of which are encoded within exonic sequences ( 17,(22)(23)(24)(25)(26). In this regard, several studies report that truncations of internal exons can dramatically affect 3 ′ splice site utilization. For example, progressive 3 ′ -5 ′ truncation of an internal exon in the human ␤ -globin gene from 171 to either 23, 29, or 33 nucleotides, resulted in quantitative exon skipping ( 27 ). Although small exons exist in nature, these are usually fl anked by strong intronic splicing enhancers ( 28 ). We propose, therefore, that truncation of apoE exon 2 from its native 66 nucleotide length to the 5 nucleotides present in pLIV11 could affect splice site selection by either loss of an exonic splice enhancer or, more likely, because sequences that replace the truncated exon (in this case cDNA) either negatively affect splice site selection at the native site or enhance utilization of a downstream cryptic site.
As a means to protect cloned sequences from alternative splicing, an engineered 3 ′ splice acceptor site was inserted downstream of the native apoE intron1-exon 2 junction in pLIV11 ( Fig. 5A ). Presence of this sequence resulted in a complete shift of splicing to the engineered site, confi rming the weak context of the native splice acceptor site contained within pLIV11 and providing a means to prevent opportunistic use of downstream cryptic