Assessing the Functional Characteristics of Synonymous and Nonsynonymous Mutation Candidates by Use of Large DNA Constructs

Abstract

As we identify more and more genetic changes, either through mutation studies or population screens, we need powerful tools to study their potential molecular effects. With these tools, we can begin to understand the contributions of genetic variations to the wide range of human phenotypes. We used our catalogue of molecular changes in patients with carbamyl phosphate synthetase I (CPSI) deficiency to develop such a system for use in eukaryotic cells. We developed the tools and methods for rapidly modifying bacterial artificial chromosomes (BACs) for eukaryotic episomal replication, marker expression, and selection and then applied this protocol to a BAC containing the entire CPSI gene. Although this CPSI BAC construct was suitable for studying nonsynonymous mutations, potential splicing defects, and promoter variations, our focus was on studying potential splicing and RNA-processing defects to validate this system. In this article, we describe the construction of this system and subsequently examine the mechanism of four putative splicing mutations in patients deficient in CPSI. Using this model, we also demonstrate the reversible role of nonsense-mediated decay in all four mutations, using small interfering RNA knockdown of hUPF2. Furthermore, we were able to locate cryptic splicing sites for the two intronic mutations. This BAC-based system permits expression studies in the absence of patient RNA or tissues with relevant gene expression and provides experimental flexibility not available in genomic DNA or plasmid constructs. Our splicing and RNA degradation data demonstrate the advantages of using whole-gene constructs to study the effects of sequence variation on gene expression and function.

Identifying genetic contributions to a disease phenotype requires an understanding of the molecular pathology of mutations. Genetic variants can alter protein function through amino acid substitutions and promoter function and sensitivity through alterations in binding sites. Additionally, changing intronic and exonic sequences may result in splicing or RNA-processing defects.^1,2 Such RNA-processing defects often trigger the nonsense-mediated decay (NMD) pathway, a surveillance mechanism that degrades mRNA transcripts encoding a premature termination codon (PTC).³ It is estimated that NMD accounts for as many as 30% of disease alleles.^4,5 When molecular variation in a gene is under investigation, it is imperative but often difficult to determine possible functionality and biological relevance. Specifically, RNA-processing defects are often difficult to study without the development of elaborate artificial constructs.

We recently examined the impact of RNA-processing defects that trigger NMD in patients with carbamyl phosphate synthetase I deficiency (CPSID [MIM #237300]).⁶ CPSID is an autosomal recessive disease characterized by hyperammonemia and is the result of defects in carbamyl phosphate synthetase I (CPSI), the enzyme required to catalyze the first and rate-determining step of the hepatic urea cycle. We found RNA-based physical evidence that ∼40% of the novel mutations resulting in CPSID created RNA-processing defects that trigger NMD.⁶ Unfortunately, the unavailability of cells that expressed the CPSI gene in both adequate quantity and quality hampered our study. Consequently, we were forced to eliminate almost 2/3 of the patients for whom only genomic DNA (gDNA) was available. Examining the functional mechanisms of pathogenic alleles in many diseases such as CPSID presents a formidable challenge due to the frequent lack of patient tissue for RNA analysis, as well as to the large size of the gene, which has 38 exons that span >120 kb of gDNA.⁷ Developing the means to explore these functional mechanisms is absolutely crucial for determining both the potential pathologic effects of rare variations and the milder functional variation seen in common polymorphisms. This problem is not unique to the study of CPSI mutations; indeed, it will certainly warrant considerable attention as more genetic changes become associated with a variety of common disease states.

To facilitate functional genetics studies, we developed a versatile BAC-based model system to test the effects of a diverse set of intronic and exonic mutations in a CPSI whole-gene construct. Essential elements of this system included normal BAC replication in bacteria, episomal eukaryotic replication, antibiotic selection in eukaryotic cells, and the expression of green fluorescent protein (GFP). Each element was incorporated into a vector that is easily exported to other BAC constructs. This BAC-based expression system, coupled with other recent BAC engineering and mutagenesis improvements,^8,9 provides an efficient way to test genetic variants, irrespective of type, location, or gene size. This expression system is widely applicable to many experiments that examine both coding and noncoding sequences, and it is particularly useful for assaying mutations that affect RNA processing because of the large size of genomic inserts allowed by BACs. Additionally, studying mutations within their wider sequence context should enable us to determine their mechanistic effect with greater accuracy. Here, we describe the construction and testing of this system and demonstrate its utility as an effective eukaryotic cell model in the study of the role of genetic variations with both known and unknown molecular effects identified in patients with CPSID.

Material and Methods

Plasmid and BAC Engineering

We inserted the pEHG plasmid into the backbone of BAC RP11-349G4, using Cre/loxP recombination, as described elsewhere.¹⁰ To provide ubiquitous expression of the normally hepatically expressed CPSI in our test cells (MRC-5V2 cells),¹¹ we inserted the human cytomegalovirus (CMV) immediate-early promoter (Towne strain) upstream of CPSI. We modified pCMV/Bsd (Invitrogen) by inserting a KpnI site at position 1937 by site-directed mutagenesis (Stratagene QuickChange). Then, KpnI and XhoI were used to create a complementary insertion site for an oligo containing the tetracycline resistance cassette flanked by FRT sites from plasmid pTet/FRT.¹² A 5′ homology arm (Invitrogen) containing 50 bp of sequence homologous to the BAC backbone was then synthesized with NotI and NheI overhangs to direct insertion of the homology arm immediately 5′ of the Tet/FRT sequence. A 3′ homology arm (Invitrogen) containing 50 bp of sequence homologous to the CPSI gene was synthesized, beginning with the transcription start site of CPSI (AF154830), with RsrII and XmaI overhangs to direct insertion of the homology arm downstream of the CMV promoter. The BAC+pEHG construct was transformed into EL250 Escherichia coli cells, and a monoclonal colony was induced for recombination by a temperature shift from 32°C to 42°C before electroporation of the constructed CMV vector.¹³ Recombinants were selected with tetracycline, after which the tetracycline cassette was removed by arabinose induction of Flp recombinase. Site-directed mutagenesis to introduce point mutations into the wild-type (wt) BECC (BAC+pEHG+CMV+CPSI) construct was performed using double homologous recombination with galK selection after transfection of the wtBECC construct into SW102 cells.⁹ The first homologous recombination step is to add the galK gene at the position of the desired mutation, with the use of positive selection for colonies able to metabolize galactose. The second homologous recombination step is to replace the galK cassette with the desired mutation sequence, with the use of negative selection against colonies that retain galK and therefore toxically metabolize 2-deoxygalactose, which produces a toxin in the presence of galK expression. All restriction enzymes were purchased from New England Biolabs. BAC purifications were performed using the NucleoBond BAC Maxi AX500 kit (BDBiosciences).

Verification of BAC Engineering

Pulse-field gel electrophoresis (PFGE) and fingerprinting revealed no unwanted or global rearrangements of the vector during the retrofitting or homologous recombination steps. PFGE was performed after a 3-h digestion of BAC DNA with NotI (New England Biolabs) in a 1% agarose 0.5% tris-acetate EDTA gel run at 6 V for 16 h, with an initial switching time of 0.2 s ramped to a final switching time of 22 s. Fingerprinting was performed by digesting BAC DNA with BamHI (New England Biolabs) for 3 h and then running the samples on a 1% agarose 1% tris-borate EDTA gel for 16 h at 35 V. Direct sequencing was also performed to verify proper insertion of the CMV promoter upstream of CPSI. Single-strand conformation polymorphism (SSCP) analysis for mutation detection in the coding region of CSPI revealed that no unwanted rearrangements or mutations occurred in the coding region of the gene during construct manipulations with the use of a previously published protocol and primer pairs.^7,14 All exons, as well as the 5′ and 3′ UTRs, were screened in a total of 42 PCRs. To increase mutation detection rates, direct sequencing was performed instead of SSCP for 12 of the 42 PCRs. In addition, changes between the BECC construct and a control DNA sample detected by SSCP were analyzed by direct sequencing, and all changes were shown to be different alleles of polymorphisms reported elsewhere.⁷

Sequencing

Fluorescent sequencing was performed by GenHunter Corporation and the Vanderbilt University Medical Center Core Facility, with use of BigDye chemistry from Applied Biosystems. We performed radioactive dideoxynucleotide sequencing, using the Thermo Sequenase Radiolabeled Terminator Cycle Sequencing Kit (USB) in accordance with the manufacturer's protocol, with 200 ng DNA.

Transfections and Stable Cell-Line Generation

Transfections were performed using the LID (Lipofectin:Integrin:DNA) technique⁸ with peptide 6. We used 8 μl lipofectin (Invitrogen), 8 μg peptide 6, and 4 μg of each BAC construct in a 2L:2I:1D ratio. Lipofectin was diluted in 400 μl Optimem (Gibco) and was incubated 45 min before the addition of peptide 6 and DNA, which was also diluted in 400 μl Optimem. Reactions were incubated for 10 min, during which time all cells were washed with 3 ml Optimem (no serum). All transfection reactions were increased to a 4-ml total volume with Optimem before dropwise placement on 5-cm dishes containing cells at 60% confluence. After a 16-h incubation, the transfection media were replaced with Dulbecco's modified Eagle medium (DMEM) and 10% fetal bovine serum for 72 h. After 72 h, hygromycin selection was added at a concentration of 125 μg/ml. After 2 wk under hygromycin selection, cells were flow-sorted for GFP expression. All GFP-positive cells were cultured under hygromycin selection as polyclonal cell lines.

Western Blotting

Cells were lysed by sonication and were quantified using the BCA Protein Estimation Kit (Pierce). After a 5-min denaturation step, equal microgram quantities of cell lysates were electrophoresed in a 6% polyacrylamide gel for 45 min at 200 V. Proteins were transferred to a nitrocellulose membrane. The anti-CPSI (product number sc-10516 [Santa Cruz]) primary antibody was incubated overnight at 4°C at a final dilution of 1:50. The bovine antigoat immunoglobulin G (IgG)–alkaline phosphatase (AP) (product number sc-2351 [Santa Cruz]) secondary antibody was incubated (1:2,000 dilution) at room temperature before alkaline phosphate staining (Bio-Rad).

gDNA Isolation and PCR

We used the Wizard Genomic DNA Purification Kit (Promega) to isolate DNA from all cell lines. To determine the presence of BAC DNA in transfected cell lines, we performed PCR with an upper primer complementary to the CMV promoter—UCMV (5′-CCATCCACGCTGTTTTGACCTC-3′)—and with a lower primer complementary to the sequence in the first exon of CPSI—L173 (5′-CCAGTCTTCAGTGTCCTCA-3′). The amplification thermoprofile was an initial denaturation step of 4 min at 95°C, 40 cycles of 30 s at 95°C, 30 s at 67°C, and 30 s at 72°C, and a final hold at 72°C for 10 min.

RNA Isolation, RT-PCR, and Northern Blotting

To isolate RNA, we used the RNeasy Midi Kit, including the optional RNase-free DNase set, following the manufacturer’s protocol (Qiagen). Reverse transcriptions were performed using 2 μg or 5 μg of total RNA and the TaqMan Reverse Transcription Reagents Kit (Applied Biosystems). The amplification thermoprofile was an initial denaturation at 65°C for 5 min, then a primer-annealing phase at 27°C for 10 min, an extension phase at 42°C for 45 min, and an enzyme denaturation step at 95°C for 5 min. We performed PCR, using the cDNA templates produced from 2 μg RNA to amplify desired regions of CPSI, for the determination of splicing changes from the intronic mutations. After visualization on a 2% agarose gel, bands were purified using the Wizard PCR Cleanup Kit (Invitrogen) and then were sequenced. PCR primers were designed at least one exon away from the mutation in question, to eliminate gDNA contamination. We amplified cDNA, using UBacT344A (5′-CTGCTCAGAATCATGACC-3′) and L1361 (5′-GGCTTCGGTAAGACTGATGT-3′) for c.1210-1G→T and U393 (5′-AGGACAGATTCTCACAATGG-3′) and L829 (5′-CTAGCAGGCGGATTACATTG-3′) for c.652-3T→G. For northern blotting, 10 μg of total RNA was used in a 1.2% agarose formaldehyde gel that was run for 2.5 h at 3 V/cm. RNA was transferred to a Hybond N+ membrane by use of the Turboblotter apparatus (Schleicher and Schuell) and was UV crosslinked. The membrane was prehybridized in Church buffer (0.25 M Na₂HPO₄, 7% SDS, and 50 μg/ml sheared salmon sperm DNA) at 65°C for 1 h, and hybridization occurred overnight at 65°C with either a CPSI or cyclophilin probe that was previously radiolabeled with the Prime-It RmT Random Primer Labeling Kit (Stratagene) and diluted to 1 million counts per min.

Quantitative PCR

We calculated relative RNA expression levels, using quantitative PCR with cDNA templates synthesized from 5 μg RNA. We performed these reactions using TaqMan (Applied Biosystems) technology on an automated platform with the ABI PRISM Detection System. To measure CPSI expression, we used the Hs00919484_m1 assay that spans the exon 3–exon 4 junction and the Hs00919480_m1 assay that spans the exon 34–exon 35 junction. We used an E-GFP probe as the endogenous control (part number 4331348 [Custom Taqman Gene Expression Assay Service]). All experiments were performed in triplicate. “No RT” samples were used as a PCR control measure for gDNA contamination. We performed ΔΔCt analysis, a description of relative RNA expression levels, by first calculating the difference between the average cycle time (Ct) value of each target CPSI sample and the average Ct value of the corresponding endogenous GFP sample at the 0.200 fluorescence threshold. Standardizing the CPSI Ct values to the GFP transcripts serves as a control, since GFP expression remains constant, and can also account for variation in BAC copy number, since both genes are present in a 1:1 ratio and are driven by a CMV promoter. These calculations were then expressed in relation to the calibrator (wild-type cell line in fig. 5A and untreated cell lines in fig. 6), which was arbitrarily set at an expression level of 100% (fig. 5A), or onefold (fig. 6).

Figure 5. .

ΔΔCt analysis showing relative CPSI RNA expression levels after qRT-PCR. A, Expression levels of CPSI RNA in each mutant cell line relative to wild type, set at 100% expression. Black bars indicate relative CPSI RNA levels measured from the exon 3–exon 4 TaqMan probe, and white bars indicate relative CPSI RNA levels measured from the exon 34–exon 35 TaqMan probe. SE is represented. B, Northern blot of RNA from each transfected cell line, showing the same decrease in RNA levels in each mutant cell line as in panel A. Only the wtCPSI cell lines have a visible band of exogenous, BAC-derived CPSI (between the number [#] signs). Only the HepG2 hepatic cell line shows endogenous CPSI expression (marked by an asterisk [*]). An arrow indicates a nonspecific band detected by the CPSI probe. A cyclophilin probe was used as a loading control. The BAC-derived CPSI transcript has an expected difference in size from endogenous CPSI (shown in the HepG2 lane) because of CMV promoter swapping that altered the size of the 5′ UTR by ∼100 bp.

Figure 6. .

ΔΔCt analysis of qRT-PCR data after siRNA-mediated knockdown of UPF2. Each panel graphs the relative expression levels of CPSI transcripts after UPF2 siRNA or vehicle-only treatments in the respective cell line as compared with untreated cells (set at a value of 1). Gray bars indicate relative levels measured from the exon 3–exon 4 TaqMan probe, and black bars indicate relative levels measured from the exon 34–exon 35 TaqMan probe. Error bars represent SE.

NMD Inhibition and Western Blotting

We used small interfering RNA (siRNA) to knock down UPF2 and to therefore block NMD, as performed elsewhere.¹⁵ Cells were 75% confluent in a 10-cm dish and were treated for 3 consecutive d. SiQuest transfection reagent was used according to the manufacturer’s protocol, at a final concentration of 5 μl/ml of DMEM (Mirus). siRNAs were used at a final concentration of 70 nM (Dharmacon). The siRNA sequence for UPF2 was published elsewhere,¹⁵ and the commercially available siCONTROL RISC-Free kit (Dharmacon) used to show knockdown was specific for UPF2, since this siRNA sequence does not target any known human genes. Western blotting was performed using anti-hUPF2, anti-α-tubulin (Abcam product number ab15246), and anti-CPSI (product number sc-10516 [Santa Cruz]) primary antibodies, as well as the electrochemiluminescent rabbit IgG (Amersham NA934) and bovine anti-goat IgG-AP (product number sc-2351 [Santa Cruz]) secondary antibodies. Western blot densitometry quantification was performed on blots repeated in triplicate, with use of the QuantityOne software (BioRad).

Results

Patient Mutations

Here, we have identified the mutation mechanisms of two exonic and two intronic mutations originally identified by genomic mutation screens of patients with CPSID.^6,16 The c.1893T→G substitution is a nonsense mutation located in exon 16, and the c.2388C→A mutation is a synonymous change in exon 19. Initially, we did not believe that this silent mutation was pathogenic. However, we were unable to identify another mutation on this patient allele. The two intronic substitutions, c.652-3T→G in intron 5 and c.1210-1G→T in intron 10, are both located at the 3′ splice acceptor site. Each mutation was unique to the patient with disease, since no others were detected in a screen of >200 additional chromosomes.

Construction and Verification of the BAC Model System

To study these and other CPSI mutations in a full-gene context, we modified a BAC construct to create a platform for mutation testing. Figure 1A shows the construction of this BECC vector. First, we used Cre/loxP recombination to retrofit BAC RP11-349G4 (containing the full CPSI gene) with vector pEHG. This vector contains episomal retention elements from Epstein-Barr virus (EBV)—specifically, the latent origin of replication (OriP) and the EBV nuclear antigen 1 (EBNA-1)—as well as the hygromycin resistance antibiotic marker and a constitutive GFP reporter (E-GFP) driven by the CMV promoter.^17,18 Figure 1B shows a PFGE illustrating the insertion of one copy of this 11-kb vector. Second, homologous recombination allowed simultaneous deletion of the sequence 5′ of CPSI and insertion of the CMV promoter directly upstream of the first exon, thereby producing constitutive expression of this hepatic-specific gene. The CMV promoter was inserted using a Tet-CMV targeting construct flanked by BAC homology arms (fig. 1A). Multiple assays verified the integrity of this construct (see the “Materials and Methods” section). The completed wild-type construct (wtBECC) is 170 kb in size, of which 150 kb is gDNA from chromosome 2 containing the complete CPSI gene and 11 kb is from pEHG (fig. 1C).

Figure 1. .

BECC model system. A, Schematic representation of the modifications to the RPCI11-349G4 BAC construct. The homologous recombination vector was used to simultaneously insert the CMV promoter upstream of CPSI and to delete the genomic sequence upstream of CPSI. The tetracycline resistance gene flanked by FRT sites was used for positive selection of the recombination event but was then removed from the final BAC construct. Use of the loxP sites on the pEHG vector and the BAC allowed sequence-specific integration of these two vectors. B, PFGE of NotI-digested constructs before (−) and after (+) the addition of a single copy of pEHG after Cre/loxP recombineering in the BAC. Asterisk (*) indicates the 11-kb size change in the BAC library vector corresponding to pEHG. C, Schematic of the 170-kb BECC vector after both modifications.

We next performed site-directed mutagenesis to introduce the c.652-3T→G, c.1210-1G→T, c.1893T→G, and c.2388C→A mutations into the BECC vector. These mutations were introduced using two-step homologous recombination with positive and negative galK selection.⁹ Various control assays, including PFGE and sequencing, verified that only the desired point mutations were incorporated (data not shown).

We then created polyclonal, stable cell lines for each of the BECC constructs and examined expression of the exogenous genes in cultured cells. After transfection into the MRC-5V2 immortalized human lung fibroblast cell line,^8,11 we obtained initial 20% transfection efficiencies as determined by visual inspection of GFP-expressing cells. GFP-positive cells demonstrating stable transfection were selected by collection through flow-sorting followed by continuous culture under hygromycin selection. The MRC+wtCPSI polyclonal cell line was passaged >85 times during a 1.5-year period under continuous hygromycin selection with no visual loss of GFP expression. To illustrate the presence of BAC-derived CPSI DNA in each cell line, we specifically amplified exogenous CPSI DNA, using the BAC-specific CMV promoter sequence and CPSI exon 1 sequence (fig. 2A). PCR revealed the presence of BAC-derived CPSI DNA only in the transfected cell lines (fig. 2B). Furthermore, we performed western blotting to verify the presence of exogenous CPSI protein specifically in the MRC+wtBECC cell line. Our results showed that, after transfection, not only did the BECC constructs create stable transformants that expressed GFP and were resistant to hygromycin, but the wild-type construct also produced CPSI (fig. 2C).

Figure 2. .

PCR demonstrating exogenous CPSI DNA, specifically in transfected cells. A, Representation of primer locations for identifying the presence of exogenous CPSI DNA. The forward primer is homologous to the CMV promoter and is specific for BAC-derived CPSI. B, PCR on a 2% agarose gel, demonstrating the presence of exogenous CPSI gDNA only in the MRC cell lines transfected with the indicated BAC construct. no tfn = No transfection. C, Western blot that uses a CPSI-specific antibody, showing the presence of CPSI protein specifically in the MRC cells transfected with the wtCPSI construct.

RT-PCR and Sequencing of Splicing Mutations

To determine the effect of both intronic mutations on splicing, we performed RT-PCR on RNA isolated from the c.652-3T→G, c.1210-1G→T, and wild-type transfected cell lines. Because the reverse transcription protocol that we used was previously proved sensitive enough to pick up low-level CPSI transcripts from a nonhepatic cell line,¹⁶ it was important to distinguish between BAC-derived and low-level endogenous MRC-5V2 CPSI transcripts.

c.652-3T→G

We first illustrate that the BECC model system mimics the original patient sequence data for c.652-3T→G. Whereas patient gDNA showed a heterozygous mutation at the −3 position of intron 5 (fig. 3A), patient cDNA (from a fibroblast cell line) showed a heterozygous 2-bp frameshift beginning in exon 6, indicating that the mutation activated an aberrant splice site (fig. 3B and 3E). To verify that the BECC platform reproduced what we observed directly from the patient, we performed RT-PCR on the MRC+wtCPSI and MRC+c.652-3T→G cell lines. Because a size difference indicating a splicing change from the c.652-3T→G mutation was not detectable on an agarose gel (fig. 3C), the RT-PCR products were radioactively sequenced to visualize the presence of aberrant transcript and to separate endogenous and exogenous transcripts. The sequence from the MRC+c.652-3T→G cell line mimicked the patient data by revealing the same 2-bp AG dinucleotide splicing alteration (fig. 3D and 3E). This comparison of data gathered from a patient cell line and from the BECC-transfected cell lines validated the use of BECC for studies on additional mutations, including those that affect RNA processing.

Figure 3. .

RT-PCR products from c.652-3T→G. A, Patient gDNA sequence revealing a G→T point mutation in intron 5 (arrow). B, Patient cDNA revealing a 2-bp insertion in exon 6 from one allele. The large arrow indicates the start of the insertion, and the small arrows indicate other residues in the frameshift. C, RT-PCRs from the BECC platform. Polyclonal wtCPSI, monoclonal wtCPSI, and c.652-3T→G polyclonal cell lines were tested on a 2% agarose gel. D, Dideoxyradionucleotide sequence of the RT-PCR products shown in panel A reveals the presence of a 2-bp frameshift specifically in the mutant cell line, which mimics the original patient data. The large arrow indicates the start of the frameshift, and the small arrows indicate other residues in the frameshift. E, Schematic of the splice site change caused by the T→G mutation (underlined nucleotides). The wild-type sequence is identified by an asterisk (*), the intronic sequence is in lowercase letters, and the exonic sequence is boxed and in uppercase letters.

c.1210-1G→T

To aid RT-PCR analysis of the c.1210-1G→T mutation and to distinguish between exogenous and endogenous CPSI transcripts, we were able to exploit the presence of a nearby 3-bp polymorphism with alternate alleles present in MRC-5V2–derived and BAC-derived CPSI. This polymorphism, T344A, corresponds to either nucleic ACC or GCT at position c.1136–1138.⁷ The forward primer used in these PCRs contained homology specific to the BAC allele at the 3′ end, preventing amplification of any endogenous CPSI transcripts. Comparison of RT-PCR products from the MRC+wtCPSI and MRC+c.1210-1G→T cell lines showed a difference in product size, with a larger band from the MRC+c.1210-1G→T cell line, indicating a splicing change (fig. 4A). Subsequent sequencing revealed the inclusion of 32 bp from the 3′ end of intron 10 in the spliced transcript (fig. 4B).

Figure 4. .

RT-PCR products from c.1210-1G→T. A, RT-PCR products from a polyclonal wtCPSI and c.1210-1G→T polyclonal cell line on a 2% agarose gel. The size difference between the wild-type and mutant products indicates activation of a cryptic splice site in the mutant cell line. B, Fluorescent sequence of the RT-PCR products shown in panel A reveals the insertion of 32 bp of intron 10 into the spliced message specifically in the mutant cell line. The box highlights the location of the c.1210-1G→T mutation.

Quantitative RT-PCR

To examine the effect of each mutation on RNA stability, we next performed ΔΔCt analysis from quantitative RT-PCR (qRT-PCR) assays, to determine the relative expression levels of CPSI transcript in each transfected cell line. These ΔΔCt calculations describe CPSI transcript levels that have been standardized to GFP transcript levels (both are present in a 1:1 ratio, and both are expressed from a CMV promoter) and then expressed in relation to the wild-type “calibrator” sample. We used three probes to measure either CPSI or GFP transcripts in each transfected cell line. One CPSI probe annealed to the transcript at the exon 3–exon 4 junction, upstream of all tested mutations. The other spanned the exon 34–exon 35 junction, downstream of all tested mutations. ΔΔCt analysis revealed that, for each mutation, CPSI expression levels were much lower than wild type, except for c.652-3T→G measured by the exon 3–exon 4 probe (fig. 5A, black bars). The relative expression differences between each mutant and each wild-type sample, with use of the CPSI exon 3–exon 4 probe, varied significantly. However, all mutants demonstrated a similar drop in relative expression levels with use of the CPSI exon 34–exon 35 probe (fig. 5A, white bars). A northern blot corroborated these data by showing that exogenous CPSI was present in both the wild-type polyclonal and monoclonal cell lines but not in any of the mutants (fig. 5B). Taken together, these data suggest that each mutation causes a decrease in CPSI RNA levels. It also suggests that degradation of RNA by this mechanism may slant toward the 3′ end of the transcript.

Inhibition of NMD

To determine if the decreased CPSI expression was due to degradation via NMD, we used siRNA to inhibit UPF2, which has been shown elsewhere to be sufficient to prevent NMD.^15,19,20 We demonstrated the siRNA-mediated specific knockdown of UPF2 in each transfected cell line by a western blot analyzed with densitometry; in all cases, the UPF2/α-tubulin ratio is considerably lower in the UPF2 siRNA-treated cells than in either untreated cells or those treated with vehicle alone or a nonspecific siRNA control.

qRT-PCR after siRNA-mediated inhibition of NMD showed increased levels of CPSI transcript when compared with vehicle-only and untreated controls (fig. 6). All data were again expressed as a ΔΔCt, representing a fold increase in expression when compared with the calibrator (untreated) sample set at an expression value of 1. After UPF2 siRNA treatment, the CPSI fold increase observed in every mutant cell line was greater (6-fold to 230-fold) than the CPSI fold increase in the wild-type cell line (3-fold). These results indicate that the CPSI mutations cause RNA instability—at least partially—by eliciting the NMD pathway, since knockdown of UPF2 increases the relative expression levels of each mutant CPSI transcript when compared with untreated cells. Although the relative increase varied depending on the cell line and probe used (possibly because of varying degrees of siRNA-mediated NMD inhibition or unique properties of each mutation, or both), these results strongly indicate a critical role for NMD in the degradation of these CPSI mutant transcripts.

Discussion

Understanding the effects of genetic variants associated with clinical phenotypes is a difficult yet vital undertaking. To this end, we created a novel system as a platform for determining the molecular function of any genomic change in CPSI and subsequently used this model to determine the functional significance of identified exonic and intronic mutations. To a BAC clone, we added a modular insert to create a eukaryotic expression vector, modified CPSI for ubiquitous expression under the control of the CMV promoter, and introduced specific mutations. In this article, we show—to our knowledge, for the first time—the effect of each of these mutations on the CPSI transcript, by demonstrating that each mutation causes transcript instability and degradation through NMD. In addition, our novel studies show the activation of aberrant splice sites for the two intronic mutations tested.

The BECC model system, coupled with other recent BAC engineering improvements,^8,9 provides an efficient way to test a heterogeneous set of mutations from any gene in eukaryotic cells irrespective of mutation type, location, or gene size. With a single modification, any BAC clone can be retrofitted with pEHG, which contains the necessary elements for long-term retention, selection, and construct tracking in mammalian culture cells.^17,18 The EBV elements located on pEHG, OriP, and EBNA-1 have been shown elsewhere to be the only components necessary for the episomal replication of constructs containing them, and they have been successfully exploited in the construction of other episomally replicating eukaryotic vectors.^17,21–24

This model successfully overcomes limitations of classical methods for assessing mutation function. It was important to test mutations in the physiological context of the complete CPSI gene on a single construct, but the large size of CPSI prevented the use of conventional methods, which have significant size restrictions. In addition, because many CPSI mutations are intronic or are suspected to affect splicing, the inclusion of introns was imperative.⁷ “Minigene” construction, although it allows intron inclusion, is not feasible for large genes such as CPSI. Other means for transferring and expressing large segments of exogenous DNA in eukaryotic cells have been developed, such as the use of BACs with viral delivery^25,26 or bacterial transfer,²⁷ yeast artificial chromosomes,²⁸ and human artificial chromosomes.²⁹ However, we desired to design a nonviral, versatile expression system that requires only the addition of an 11-kb vector, to ensure stable transfections, and that is capable of undergoing precise manipulations through homologous recombination. The BAC-based model system is limited only by the identification of a BAC containing the complete gene of interest and an appropriate cell line for vector expression. However, given the recent improvements in BAC engineering, it is entirely feasible to modify a BAC clone to meet size specifications or to join gene segments from multiple BACs together.³⁰ Furthermore, various transfection methods can be used to introduce large constructs in a spectrum of cell types.^8,31,32 As shown here, promoter swapping permits gene expression in an otherwise nonexpressing cell type.

We used the BECC vector to examine the mechanistic effect of four CPSI mutations—c.652-3T→G, c.1201-1G→T, c.1893T→G, and c.2388C→A—that were each thought to affect transcript stability. Introducing each mutation in the BECC platform had several advantages beyond validating the model system. First, since patient samples are usually obtained from individuals who are heterozygous for different mutations, we were able to isolate the mutation in question for study in an experimental system. Second, these mutations could be maintained in a system without the limited shelf-life typically seen for patient tissue samples, which may be difficult to obtain. Third, we were able to perform follow-up manipulations that are not always possible or reliable in hepatic samples or patient cell lines.

As shown in figure 5, the qRT-PCR probe specific for the 3′ end of the transcript shows relatively equal large decreases in detectable CPSI message in all mutant cell lines when compared with wild type. In all cases, the 5′ CPSI probe detects higher levels of CPSI transcript, but these levels are all less than those in wild type except c.652-3T→G. In each case, the observation that more mutant transcript was detected with the CPSI exon 3–exon 4 probe, which lies upstream of all tested mutations, may indicate degradation primarily from the 3′ end. Although it is possible that there are alternate splicing or polyadenylation isoforms that recognize the exon 3–exon 4 but not the exon 34–exon 35 probe, no such isoforms have been described for CPSI. Ongoing and extensive testing performed in our laboratory with a variety of techniques has been unable to identify alternate isoforms of CPSI that exclude the exon 34–exon 35 junction. Our qRT-PCR data, coupled with a survey of all mutations known to create PTCs in patients with CPSID,⁶ support the hypothesis that a significant number of pathogenic CPSID mutations result in RNA-processing defects.

Specific Mutation Observations

Both intronic mutations, c.652-3T→G and c.1210-1G→T, affect the splice acceptor site at the 3′ end of the intron by disrupting the 3′ splice site consensus (Y)_nNYAG|G sequence.³³ For both mutations, sequencing of RT-PCR products revealed the recognition of an alternate AG dinucleotide as the 3′ splice site. The c.652-3T→G mutation creates a premature AG dinucleotide sequence that was recognized as the 3′ end of the intron in both the patient and the transfected MRC5-V2 cell lines (fig. 3). The c.1210-1G→T mutation abolished the endogenous AG dinucleotide sequence at the 3′ end of the intron, causing use of an upstream AG (fig. 4). Both mutations result in a splicing change that causes a frameshift, thereby illustrating the importance of the AG dinucleotide in proper 3′ intron definition.

The c.1893T→G mutation creates a TAG stop codon in place of tyrosine at amino acid position 590. This mutation provides the most direct measure of the presence of NMD, since the stop codon is formed independently of any cryptic splicing events. When NMD was inhibited through siRNA-mediated knockdown of UPF2, the transfected cell line containing this mutant construct had the largest fold increase in CPSI transcript (fig. 6). A possible explanation for this observation is that there are separable mechanisms for eliciting NMD based on the type of mutation, since no frameshift occurred to create this PTC. It is also possible that various cryptic splice sites are recognized in transcripts with intronic mutations which, although undetectable by RT-PCR, alter not only the number of transcripts containing PTCs but also the position of the PTC, which may have an effect on recognition by the NMD pathway.

Finally, the BECC system was particularly useful for determining the mechanistic effect of the c.2388C→A mutation. This synonymous change (S755S) in exon 19 was not originally thought to cause loss of function. However, after a thorough genotyping of CPSI in this patient, no other genomic changes could be found on this allele. This result, coupled with the knowledge that many missense and silent mutations are pathogenic at the RNA level if they reside in an exonic splicing enhancer,^34,35 leads us to propose that c.2388C→A either disrupts a splicing enhancer or creates a splicing silencer binding motif. As shown in figures 5 and 6, this mutation resulted in a large drop in transcript that was restored by NMD inhibition. For this silent mutation to be sensitive to NMD, it must create a PTC, which would be achieved if exon 19 were skipped or if a cryptic splice site were activated that caused a frameshift due to partial exon inclusion. These data add to the increasing evidence that, although they do not alter actual protein structure, many mutations, such as silent and missense mutations, negatively affect the processing of the transcript.³⁵ This particular mutation underscores the utility of the BECC system for testing the functional effects of seemingly innocuous genetic variants. This system can be used to test elements thought to be important in processing exon splice enhancers.

Summary

Our data indicate that RNA-processing defects must always be considered when determining the functional significance of genetic changes because of the numerous ways this pathway can be engaged at the sequence level. Accounting for them in patient mutation screens is imperative yet difficult, because patient RNA samples are not always available or able to be manipulated for the detection of splicing changes and allelic dropout from NMD. To facilitate such studies, we report in this article the creation of a novel model system to determine the mechanism of mutations in CPSI that can be applied to other genes as well. Whereas this model system can be employed to test any sequence change, including those in regulatory sequences, its size capabilities make it particularly useful for assaying mutations that affect RNA processing. This system not only permits expression studies in the absence of patient samples but also allows experimental manipulations that could not otherwise be performed, such as inhibiting NMD to demonstrate its role in CPSI mutation pathology. We will continue to use this model to examine effects of other mutations in CPSI, as well as to determine possible functional effects of identified polymorphisms, since this BAC-based platform will surely prove helpful as we strive to unravel the many mysteries of functional genetic changes.