Occurrence of Multiple Aberrantly Spliced mRNAs upon a Donor Splice Site Mutation That Causes Familial Lipoprotein Lipase Deficiency*

A donor splice site mutation was found in the lipo- protein lipase (LPL) gene of a patient with familial LPL deficiency. The mutation, a G 4 A substitution, occurred at the first nucleotide of intron 2. Northern blot analysis of total RNA from the patient showed strikingly low levels of LPL-specific mRNAs. Using the polymerase chain reaction, the LPL mRNA splicing was analyzed in detail. The results demonstrated that no normal splicing occurred at the authentic splice site; rather a cryptic splice site 18 bases upstream from the mutation site was preferentially utilized. Although the resulting alteration in mRNA was a minute in-frame 18-base deletion, the amount of the abnormal tran- script was only %2 that of the normal. In addition to this major cryptic splice site, we also identified multi- ple minor sites which were utilized at extremely lower efficiencies. Unexpectedly, one of these minor sites was also used as an alternative splice site in the normal subject at a comparably low efficiency. The sequences of these minor cryptic The concentration of plasma apolipopro- tein C-11, a co-factor of LPL, 9.8 mg/dl (normal 3.6 -C 1.0). LPL enzyme mass was measured by a sandwich enzyme-linked immuno- sorbent assay developed recently in our laboratory.’ The enzyme-linked immunosorbent assay utilizes two separate polyclonal antibod- ies raised against a synthetic peptide of 16 amino acids corresponding to the N terminus of the mature human LPL and against a recom- binant whole human LPL. No immunoreactive mass was detectable

A donor splice site mutation was found in the lipoprotein lipase (LPL) gene of a patient with familial LPL deficiency. The mutation, a G 4 A substitution, occurred at the first nucleotide of intron 2. Northern blot analysis of total RNA from the patient showed strikingly low levels of LPL-specific mRNAs. Using the polymerase chain reaction, the LPL mRNA splicing was analyzed in detail. The results demonstrated that no normal splicing occurred at the authentic splice site; rather a cryptic splice site 18 bases upstream from the mutation site was preferentially utilized. Although the resulting alteration in mRNA was a minute in-frame 18-base deletion, the amount of the abnormal transcript was only %2 that of the normal. In addition to this major cryptic splice site, we also identified multiple minor sites which were utilized at extremely lower efficiencies. Unexpectedly, one of these minor sites was also used as an alternative splice site in the normal subject at a comparably low efficiency. The sequences of these minor cryptic sites possessed many of the characteristics common to those of other normal splice sites, indicating that even such minor sites should have also been selected according to the general rules for splice site selection. These results demonstrate that upon mutation, a broad spectrum of cryptic splice sites is activated in vivo at the sites' respective efficiencies.
Lipoprotein lipase (LPL)' is a principal determinant for the hydrolysis of triglycerides in plasma lipoproteins and is necessary for the supply of free fatty acids to tissues for nutrition and energy source. LPL is a glycoprotein synthesized in most extrahepatic tissues and is anchored to the luminal surface of the capillary endothelium by membranebound heparan sulfate (1). The cDNA coding for LPL was first cloned from human (2) and bovine (3). Human LPL is composed of 448 amino acids in its mature form (2), and the corresponding gene has a span of 30 kilobases comprising 10 exons (4). A congenital defect in the enzyme, known as familial LPL deficiency, is a rare autosomal recessive disorder usually diagnosed by a reduced level of postheparin plasma * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. LPL activity and characterized by marked hypertriglyceridemia (5). Recently, point mutations were identified in the LPL genes of some patients (6)(7)(8)(9)(10)(11). In this paper we report a donor splice site mutation identified in the LPL gene of a patient with familial LPL deficiency.
Splice site mutations are common causes of a number of human genetic disorders because of a disturbance in the normal processing of pre-mRNA (12). We found a G -+ A substitution at the beginning o f intron 2 in the LPL gene of an LPL-deficient patient. The change converts the invariant GT motif of the 5"donor splice site into AT. Similar types of substitutions have been reported previously in many other human genetic diseases (13)(14)(15)(16)(17), but the resulting changes in the mRNA transcripts were quite varied. The amounts of the corresponding mRNAs range from markedly decreased (13) to normal levels (17). The pattern of aberrant splicing varies; it includes the activation of cryptic splice sites, exon skipping, and retention of an unspliced intron in the mature transcripts (17). Precise characterization of the abnormally spliced transcripts is often difficult in uiuo, especially when the amount of the aberrant transcript is extremely minute (18). Experiments with transfected cells have been used for the analysis of the abnormal transcripts (13). However, there is evidence to suggest that the gene expression of transfected cells is not always the same as that of normal cells in vivo (19).
In the present study we carried out the precise characterization and quantitation of aberrantly spliced mRNA obtained from an LPL-deficient patient. We also demonstrated various cryptic splice sites that were activated upon mutation and discuss factors that affect splice site selection in vivo.

MATERIALS AND METHODS
Subject-The patient, whose parents are first cousins, was first diagnosed as LPL deficient at the age of 34 years, when he had a bout of acute pancreatitis secondary to massive chylomicronemia. His fasting serum triglyceride concentration was 4,928 mg/dl (normal: 40-150), and total cholesterol concentration was 568 mg/dl (normal: 130-220). Postheparin plasma LPL activity was 0.8 pmol of free fatty acids/ml/h (normal: 6.4 k 2.1), which was measured 10 min after a bolus injection of heparin (10 units/kg) after an overnight fast as described previously (20). The concentration of plasma apolipoprotein C-11, a co-factor of LPL, was 9.8 mg/dl (normal 3.6 -C 1.0). LPL enzyme mass was measured by a sandwich enzyme-linked immunosorbent assay developed recently in our laboratory.' The enzymelinked immunosorbent assay utilizes two separate polyclonal antibodies raised against a synthetic peptide of 16 amino acids corresponding to the N terminus of the mature human LPL and against a recombinant whole human LPL. No immunoreactive mass was detectable M. Kawamura, unpublished data. in the post heparin phsma of the patient.
(;rnomic ('loning nnd /INA Srqucncinfi-Gene fragments that rover exons :X-<) were ohtained from the genomic lihrary constructed previously for the patient (21). To ohtain gene fragments containing the remaining rxons, gene amplification was performed hv the polvmerase chain reaction (1Y;R) ( 2 2 ) with oligonucleoticle primers that were svnthesized according to the published sequence of the human IZPL gene (44). Ilirectlv cloned gene fragments were srrhclonetl into the MI3 vector mp18 or mp19 and sequenced hv the dideoxv method ( 2 3 ) with Sequenase (ti. S. Riochemicnl Corp.). Direct sequencing of amplified IlNA was performed utilizing an unequal ratio o f primers l'or the second I'CII (244). Alter the first I'CR, l /~( * ! of the reartion was amplilied again, together with an unequal ratio (1OO:l) of the same primers as used in the first reaction. The single-stranded DNA from the second I'CR was used for sequencing after chloroform extraction and filtration using a (lentriron 100 microconrentrator (Amicon ('orp., I)anvers, M A ) .
T h e cells were then placed in fresh I i l ' M I l(i40 medium supplemented with lor; autologous serum and cultrrred for 1 week. T h e medirrm was changed twice. The differentiated macrophages were lvsed with a 5.5 M guanidinr thiocyanate homogenization h f f e r c o n t a i n i n g 25 mM sodium citrate (pH 7.01, O.5";, N-lauroylsarcosine, and 0.2 M 2rnercaptoethanol. Total cellular RNA was isolated from thr cell lysates by the p~anidine thiorvanate method (26).
Hvhridization with ij-actin cI)NA prohe was carried out similarly.

RESULTS
In the LPL gene of the patient, the G 4 A substitution was found at the first nucleotide of intron 2 (Fig. 1A). T h i s G + A transition was the only pathological change that was found during sequence analysis of all coding exon and adjacent intron sequences. Homozygosity for the mutation was confirmed by direct, sequencing of six products independently amplified by PCR. Northern blot analysis of total RNA from monocyte-derived macrophages revealed a strikingly decreased level of L P L m R N A in the patient although the levels of @-actin mRNA were the same for normal suhject and the patient (Fig. 1R). Despite  resulting from an alternative utilization of polvadenvlat ion signals (2), were detected in both cases.
T o investigate further the LPL mRNAs in the patient, PCR technology was adopted for amplification of the regions of int,erest within the transcripts using a procedure descrihed previously (28). Shortly after reverse transcription of mRNA with random primers (hexamers), the cDNA product was subjected to the first PCR with only a 5"outer primer ( p r i m~r A in Fig. 2). Suhsequently, %,w of the first reaction product was amplified again hv the second PCR; the target regions were amplified by a pair of appropriate internal primers. Fig.  3 shows the results of a second PCR with a pair of oligonucleotide primers (primers C and I in Fig. 2 ) . Staining with ethidium bromide showed a single major hand of the expected size for the normal subject, indicating that the PCR accurately amplified the target region of 242-hase pair (bp) length. In contrast, the same PCR gave rise to 224 hp, an 18-hp shorter fragment for the patient. Since our previous findings had indicated that there is no deletion in the coding region of the patient's gene, this ahnormal fragment most likely reflects the aberrant splicing event in the mRNA of the patient. Southern hlot analysis, performed with a human 1,PL cDNA prohe as described previously (21 ), confirmed that both the 242-and 224-hp fragments were indeed from LPI, mRNAs (Fig. 3). These results suggested that the ahnormal 224-hp fragment was prohablv derived from the misspliced 1,PL mRNA of patient as depicted in the rizht p n n d of Fig. 3.
The amount of the abnormal transcript was measured precisely by the competitive PCR method, which was recently developed for the quantitation of cvtokine mRNA hv Gilliland et al. (29). Since hoth the 242 and 224-bp fragments can be co-amplified with the same set of primers and their sizes are only slight,ly varied, co-amplification of the two fragments should occur in a concentration-dependent manner. As shown in Fig. 4, the same amount of 242-and 224-hp fragments was ohtained when cDNAs of the normal suhject and the patient were mixed at a ratio of 1:12. The results indicate that the amount of the ahnormal transcript is only ahout of that of the normal. Direct sequencing of the 224-hp fragment identified a newly created boundary between exons 2 and 3 , which was located 18 bases upstream from the authentic houndary (Fig. 5 )    t h e 224-bp fragment in the lane corresponding to the patient (Fig. 3). Repeated PCR amplifications of this DNA fraction with primers C and I failed to amplify any specific fragment.
The results, therefore, excluded the possibility of exon skipping, which joins exons 1 and 3, and activation of cryptic splice sites upstream from the major cryptic site. Next,, we performed a second PCR with a new pair of primers D and I ( Fig. 2), which indicated the occurrence of downstream splicing sites by evading the major cryptic site involved in the formation of the 224-bp fragment. After the second PCH  Fig. 2). However, as shown in Fig. 7 (loncs I . .'I. 4 . a n d 5 ). thc  results clearly denied the possibility of further downstream splicing as well as the possibility of retention of intron 2 in the mRNA.
For reference, we also examined the LPL mRNA prepared from normal subjects. The results unexpectedly demonstrated low levels of aberrant splicing in the normal genes. As shown in Fig. 7 (lane 2 ) , the PCR-amplified cDNA from the normal gene contained an aberrant transcript derived from an alternative splice site. T h e site (+250) was identical to one of the minor sites identified in the mutant gene, and the levels of utilization oft he +250 sit.e appeared comparable in both cases. In addition, t.he occurrence of downstream splicing (farther than +387), which was not detected in the patient, was observed in the normal subjects (data not shown). This finding indicates that low levels of aberrant splicing could occur in uiuo even in the presence of the normal splice site. Fig. 8 illustrates the splicing events t.hat occur in the normal and mutant genes as well as the estimated relative amount of each transcript.
Splicing is mediated by the initial interaction of primary transcripts with U1 small nuclear RNA (snRNA) (121. T o gain insight into the mechanism of splice site selection in uiuo, we compared the seven cryptic sites described above with other GT-containing sequences in intron 2, with respect to the degree of nucleotide homology, location, and their freeenergy changes (80) upon binding to U1 snRNA. The degree of homology to the other known 5' splice sites was rated according to the report of Shapiro and Senapathy (31). The results are shown in Fig. 9 along with other possible cryptic site sequences that contain GT dinucleotide. The cryptic sites generally showed more favorable values for splice site selection than the other sequences in the two parameters (homology score and free-energy value). It should be noted that the major cryptic site, although showing unexpectedlv unfavorable values for these two parameters, contained the GT dinucleotide nearest to the authentic splice site.

DISCUSSION
In the present study we showed that the LPL mRNAs in this patient were misspliced by activation of nearby cryptic splice sites, and their amounts were markedly decreased to %? of the normal level. These abnormalities explain the molecular basis of the LPL-deficient state in the patient.
In some splicing mutations, two or more cryptic sites are activated simultaneously (17, 19). Furthermore, competition among cryptic sites has been demonstrated recently for the transfected cells by a superb mutagenic technique ( 3 2 ) . These findings led us to h.ypothesize that a series of candidate sequences for cr-yptic sites might be competitively utilized in uiuo although this competition would be undetectable hv conventional methods. To test this h-ypothesis we attempted to detect minor transcripts present in vioo in the affected cells of this patient with the PCR method. The experiments successfully amplified four minor transcripts to a visible state in gels (Fig. 6). The selection of the minor cryptic sites would not be accidental but would rather follow the same rule that applies to major sites. Indeed, the sequences of all the minor sites showed a degree of favorable values similar to those of other established splice sites in two parameters (Fig.  9). In addition, the sequence analysis of the subcloned PCR product revealed the presence of two additional rarer transcripts, indicating the possibility that a number of transcripts could exist i n vivo at barely detectable levels. These results enabled us to establish a model for 5"cryptic site selection i n vivo whereby a spectrum of sequences can compete with each other for activation and can be utilized at their respective efficiencies. The resultant transcripts are quantitatively so diverse that only a few can normally be found and only some can be detected by PCR, but the majority are undetectable. Unexpectedly, similar low levels of aberrant splicing event were observed in normal gene expression. In normal subjects, it was found that the cryptic site (+250) was used for alternative splicing as well. Although we have not yet analyzed for other genes, such low levels of alternative splicing may be associated with normal gene expressions in uiuo. Another unexpected finding was that the utilization of the alternative site (+250) was not augmented by the disruption of the authentic splice site in the patient, suggesting that such low levels of alternative splicing should not be directly linked to the activation of cryptic splice sites caused by splicing mutations.
Recently, a number of nonsense mutations have been reported to cause a considerable reduction in the corresponding mRNA (33, 34). A reduction in mRNA associated with splicing mutations (13, 35) might be explained partly by a similar mechanism ( i e . mRNA instabilization) because roughly twothirds of the splicing mutations should alter the reading frame of mRNA, which would introduce a stop codon downstream. The concept was supported by the fact that the splicing mutations with normal levels of mRNA usually cause no frameshift in the coding sequence (17, 35). In contrast, the major change in the LPL mRNA of the patient was shown to be a simple in-frame deletion of 18 nucleotides (Fig. 5); however, the amount of the abnormal transcript was greatly reduced to about %Z of the normal level (Fig. 4). This reduction can hardly be explained by the extent of the structural change of the transcripts since the deletional alteration in the present patient is small and thus is unlikely to affect mRNA stability seriously. One possible explanation is that the binding between the sequences of the major cryptic site in this patient and U1 snRNA may be weak. In fact, the predicted free-energy value of the major cryptic site (-2.8 kcal/mol) is not a particularly favorable value for stable binding to U1 snRNA (Fig. 9). Since such binding is one of the rate-limiting steps in RNA processing, instability of the binding would be disadvantageous for sufficient production of mature transcripts.
These results also provide some insight into the factors that affect splice site selection. As has been shown in other studies (17), we found a significant difference in the average values of the two parameters between the seven cryptic site sequences and other candidate sequences (average homology score: 73.6 uersus 58.2%; average free-energy value: -5.1 uersus -1.7 kcal/mol) (Fig. 9). In the individual case, however, there are some interesting contradictions. For example, the major cryptic site (68.6% and -2.8 kcal/mol) is less favorable than the other minor sites (74.4% and -5.5 kcal/mol on average). In addition, utilization of the sequence GTGGTGAGG a t position +38 could not be detected despite the high values in the two parameters. Therefore, in addition to these two parameters, there must be other undetermined factors crucial for splice site selection i n uiuo. In the selection of the cryptic splice sites, the distance from the authentic site must be another important determinant. In fact, in the present study we could never detect the activation of the cryptic site farther than +387 (Fig. 7). The overall RNA structure of the candidate sequences may also be another important factor which should be considered.
In this study, we demonstrated that one major and multiple minor cryptic sites can be activated a t their respective efficiencies in uiuo. Since even such minor cryptic splice sites apparently follow the common rules for splice site selection, extensive analysis of these sites will provide new insight into the mechanism governing splice site selection.