Frameshift Mutants of β Amyloid Precursor Protein and Ubiquitin-B in Alzheimer's and Down Patients

See allHide authors and affiliations

Science  09 Jan 1998:
Vol. 279, Issue 5348, pp. 242-247
DOI: 10.1126/science.279.5348.242


The cerebral cortex of Alzheimer's and Down syndrome patients is characterized by the presence of protein deposits in neurofibrillary tangles, neuritic plaques, and neuropil threads. These structures were shown to contain forms of β amyloid precursor protein and ubiquitin-B that are aberrant (+1 proteins) in the carboxyl terminus. The +1 proteins were not found in young control patients, whereas the presence of ubiquitin-B+1 in elderly control patients may indicate early stages of neurodegeneration. The two species of +1 proteins displayed cellular colocalization, suggesting a common origin, operating at the transcriptional level or by posttranscriptional editing of RNA. This type of transcript mutation is likely an important factor in the widely occurring nonfamilial early- and late-onset forms of Alzheimer's disease.

In Alzheimer's disease (AD) and Down syndrome (DS) patients, intracellular and extracellular deposits of proteins in tangles, neuropil threads, and neuritic plaques are correlated with neuronal dysfunction leading to dementia (1). In particular, the familial types of AD have been investigated thoroughly and are due to mutations in genes located on chromosomes 1, 14, and 21, and the apolipoprotein E genotype (chromosome 19) is a risk factor (2). However, at least 60% of AD patients do not have a family history of the disease (3). For these frequently occurring, sporadic cases, a more general mechanism must exist, ultimately leading to neuronal degeneration.

Messenger RNA editing is a means of producing phenotypic variability (4). Moreover, we have identified another type of mutation in vasopressin transcripts (5). Homozygous Brattleboro rats have a single base deletion in the vasopressin gene, and newborn rats do not have a functional vasopressin mRNA and protein. Surprisingly, functional RNA and protein are found in a small but increasing proportion of hypothalamic cells as the animals age (6). This apparent reversion is due to a dinucleotide deletion (ΔGA) within GAGAG motifs of the mutant RNA (5). Thus, genetic information in neurons is not stable but subject to modification through an as yet unknown mechanism. We surmised that the opposite process may take place in other neuronal genes, resulting in mutant transcripts from wild-type genes, and so we looked for dinucleotide deletions in two genes associated with the pathogenesis of AD. The genes encoding β amyloid precursor protein (βAPP) and ubiquitin-B (Ubi-B) protein (1, 7) each contain several GAGAG motifs.

In βAPP mRNA, seven GAGAG motifs are present in regions corresponding to exons 4, 6, 9, 10, and 14. Because three motifs are clustered in exons 9 and 10, this part of the transcript, encoding a putative growth-promoting domain (8), was selected for the detection of a +1 frameshift mutation resulting in truncated βAPP (βAPP+1) with a novel COOH-terminus (Fig.1A). In two of the three repeats of Ubi-B mRNA, a single GAGAG motif is present (Fig. 1A). The predicted +1 frameshift results in an aberrant COOH-terminus of Ubi-B of the first or second repeat (Ubi-B+1). As a result, the glycine moiety essential for multiubiquitylation (9) would be lacking. To examine the occurrence of the predicted +1 proteins, we generated antibodies to the novel COOH-termini of βAPP+1 and Ubi-B+1 and used them to evaluate the presence of the abnormal proteins in tissue sections of cerebral cortex from AD, DS, and control patients by immunocytochemistry (10) and immunoblot analysis (11) and to assess reading frame mutations by selecting cDNA clones expressing +1 immunoreactivity.

Figure 1

(A) Partial nucleotide and amino acid sequences of βAPP and Ubi-B expressed in the wild-type (WT) and +1 reading frame (+1 protein). Shaded nucleotides represent GAGAG and CTCT motifs. Peptide sequences printed in bold were used for immunization (31). Two antibodies were raised to Ubi-B+1 [Y-Q and R-Q (indicated by lines; Ubi-B1 and Ubi-B2, respectively); both have 11 amino acids]. For βAPP, there are seven GAGAG motifs, and the predicted molecular mass of the truncated proteins is 38 kD (32). For Ubi-B, there are two GAGAG motifs (33), and the predicted molecular mass of the truncated protein is 11 kD (monomer). Ubi-B is expressed in the brain (34). The inverted solid triangle indicates the exons 9 and 10 junction. (B) Sequence gels showing a GA deletion in βAPP transcripts (exons 9 and 10) (left) and a GT and CT deletion in Ubi-B transcripts (right). (C) Colocalization of Ubi-B+1 mRNA (a) in Ubi-B+1 immunoreactive cells (b) in the temporal cortex of an AD patient. In (a), the section was counterstained with hematoxylin. Bar, 10 μm. (c to e) Immunoblots showing (c) an intense immunoreactive band at 38 kD stained with the βAPP+1 antibody in the temporal cortex of a DS patient, (d) the same band stained with an NH2-terminal βAPP antibody (3H5), and (e) the frontal cortex of a young control, in which no reaction was visible. In (d), the 38-kD band can be seen, but, in the absence of previous immunoprecipitation, degradation products of βAPP were detected as well.

Immunoreactivity for βAPP+1 and Ubi-B+1 was prominent in early- and late-onset AD cases and even more prominent in DS patients compared with controls matched for age, sex, postmortem delay, and duration of fixation (12) (Fig.2 and Table1). When the three brain areas studied were taken together, βAPP+1 immunoreactive structures were present in 71% and Ubi-B+1 immunoreactive structures in 100% of the AD patients (12). In young controls and one nondemented DS patient devoid of neuropathology in the frontal and temporal cortices and hippocampus, no Ubi-B+1immunoreactivity was found (12). When Ubi-B+1immunoreactivity was found in elderly, nondemented controls (>72 years), their neuropathological diagnosis revealed the presence of some plaques and tangles (12). Furthermore, no βAPP+1 and Ubi-B+1 immunoreactivities were found in the substantia nigra and striatum of 11 patients suffering from Parkinson's disease, except for the striatum and temporal cortex of one patient with AD neuropathology.

Figure 2

Neuritic plaques. (A and B) βAPP+1 and (C) Ubi-B+1immunoreactivities in the frontal cortex of an AD patient are present in neuritic plaques. In (A), neuropil threads can be seen (arrowheads); s, sulcus. In (B) and (C), a higher magnification shows immunoreactivity in dystrophic neurites and neuropil threads (arrowheads). The core of the plaques is unstained as seen in (B) and (C). Ubi-B+1 immunoreactivity was obtained with two different antibodies (that is, Ubi-B1 and Ubi-B2). Ubi-B2 was most immunoreactive and used for all immunohistochemistry. (D toH) Neurofibrillary tangles. Ubi-B+1 (D to F) and βAPP+1 (G and H) immunoreactivities in tangle-shaped structures in the temporal (D) and frontal (E to G) cortex and the hippocampus (CA1 area) (H) of two different AD patients (D and E to H). Neuropil threads are indicated with arrowheads. Bar in (A) and (D), 50 μm; bar in (B) and (C) and (E) to (H), 20 μm.

Table 1

Immunoreactivities in the human frontal and temporal cortices and hippocampus for βAPP and Ubi-B, for which the mRNA is expressed in the +1 reading frame (resulting in βAPP+1and Ubi-B+1 protein). Tissues were obtained from controls and neuropathologically confirmed AD and DS cases (12). Immunoreactivity present in tangles, dystrophic neurites, and neuritic plaques of patients is expressed as a percentage of the total number of patients studied.

View this table:

In the frontal and temporal cortices and the hippocampus of AD and DS patients, both βAPP+1 and Ubi-B+1immunoreactivities were present in neurofibrillary tangles, neuropil threads, and dystrophic neurites (Fig. 2). In many cases, βAPP+1 and Ubi-B+1 immunoreactivities coexisted (Fig. 3, A and B), especially in neurons located in layers 2, 3, and 5 (Fig. 2, A and D). The βAPP+1 or Ubi-B+1 immunoreactive tangles and neuropil threads formed a considerable subpopulation of classical Bodian silver and Alz-50– or MC-1–stained neuropathological structures (Fig. 3, C to F). The presence of βAPP+1 and Ubi-B+1 in a subpopulation of Alz-50 or MC-1 immunoreactive neurons excludes the possibility that they cross-react with hyperphosphorylated tau protein. In a subpopulation of the wild-type βAPP and Ubi-B immunoreactive neurons, accumulation of βAPP+1 and Ubi-B+1 immunoreactivities was found in neurofibrillary tangles. βAPP+1 immunoreactivity was often found in restricted areas of the sections, whereas Ubi-B+1 immunoreactivity was distributed much more widely throughout the section and was present in a higher percentage of AD patients. In the hippocampus, intense βAPP+1 and Ubi-B+1 immunoreactivities were prominent in neurofibrillary tangles present in pyramidal cells of CA1 (Fig. 2H) and the subiculum, whereas, in the more intensely stained cases (12), immunoreactivity was frequently seen as cytoplasmic staining in CA4, CA3, CA2, and the hilus. In the entorhinal cortex tangle, staining was often observed in the pre-α layers (13) and concentrated in cellular islands.

Figure 3

Colocalization of (A, C, andE) βAPP+1 and (B) Ubi-B+1 immunoreactivities, with (D) Bodian- and (F) Alz-50–stained cell bodies (large arrowheads) in consecutive sections of the frontal cortex (A to D) and subiculum (E and F) of an AD patient. Alz-50 (F) stains neurons and neuropil threads more abundantly than βAPP+1 (E). Small arrowheads, Alz-50–positive neurons and βAPP+1–negative neurons; *, capillary. Bar, 20 μm.

To characterize the immunoreactive products detected in immunocytochemistry by molecular size, we used immunoblots of homogenates of the temporal cortex of AD and DS patients to reveal the presence of immunoreactive proteins with sizes [38 kD for βAPP+1 (Fig. 1C) and 11 kD for Ubi-B+1(14, 15)] predicted by the open reading frames of transcripts with a dinucleotide deletion. In young, nondemented controls, these +1 protein bands were absent. The eukaryotically (βAPP+1) or bacterially expressed (Ubi-B+1) recombinant proteins corresponding to the +1 mutant proteins served as positive controls (14, 15). Antibodies to wild-type sequences of βAPP (Fig. 1C) and Ubi-B, located in the unaffected region near the COOH-terminus with the +1 reading frameshift, revealed bands of the same size.

Preimmune and solid-phase adsorbed antisera showed no reaction in paraffin sections or immunoblots. In contrast to the βAPP+1 antibody to a region in exons 9 and 10, two βAPP+1 antisera to regions of exons 14 and 18 and a βAPP+2 antiserum (part of exon 17) did not display immunocytochemical staining in any of the sections that were positive for βAPP+1.

To establish the nature of the mutations resulting in the truncated βAPP+1 and modified Ubi-B+1 proteins, we cloned cDNAs generated by polymerase chain reaction (PCR) into an expression vector (16). Immunoscreening of reverse transcriptase (RT)–PCR products generated from βAPP cDNAs of young and elderly AD and DS patients and subsequent sequencing of immunopositive clones revealed that, in all AD and DS cases examined, a GA deletion was present in either exon 9 or exon 10 (Fig. 1B and Table2). In exon 9, a higher frequency of mutations was found than in exon 10. In one AD patient, GA deletions were found in both exons 9 and 10 in separate mRNA molecules. No immunopositive clones were found in nondemented young and elderly control patients.

Table 2

Immunoscreening and sequencing of cDNA of βAPP and Ubi-B for dinucleotide deletions in the cortex and hippocampus of AD and DS patients and nondemented control patients.

View this table:

In addition, another dinucleotide deletion (ΔGT) was found in the first repeat of Ubi-B. This mutation is located directly adjacent to the GAGAG motif in young and elderly AD and DS patients (Fig. 1B and Table 2). The frequency of Ubi-B+1 immunopositive clones was much higher than with βAPP. In a young control, no immunopositive clones were found, whereas in the elderly control displaying neuropathology and Ubi-B+1 immunoreactivity, not only was a GT deletion found in the first repeat, but a CT deletion was also found in the third repeat. A CT deletion was also found in a DS patient. In four patients, dinucleotide deletions were found in both βAPP and Ubi-B transcripts as expected from the colocalization of βAPP+1 and Ubi-B+1 (Fig. 3, A and B).

To exclude PCR and cloning artifacts as a possible explanation for the mutation, we confirmed the presence of mutated RNA by in situ hybridization (17). PCR artifacts were also excluded by genomic PCR (see below). The GA deletion in βAPP+1 and the GT deletion in Ubi-B+1 transcripts were both detected by stringent in situ hybridization with an oligoprobe discriminating between wild-type and mutant (ΔGA) βAPP (17) and (ΔGT) Ubi-B (Fig. 1C).

The mutations in βAPP and Ubi-B proteins could be caused at two different levels, either by a deletion in the DNA or by a transcriptional defect or editing mechanism in the RNA. The problem in detecting mutations in the DNA is the fact that a tissue sample inevitably contains not only the immunoreactive cortical layers 2, 3, and 5 but also the other layers and a part of the white matter. Thus, only about one in 10,000 cells would carry a mutation. Hence, we developed two PCR protocols that would be sufficiently sensitive to detect this frequency of mutation (18).

With the assumption that the human haploid genome contains 3 × 109 base pairs (bp) for the RT-PCR (19) and the genomic PCR for βAPP, an equivalent of 50,000 cells (that is, 0.5 μg of total RNA and 300 ng of genomic DNA, respectively) was assayed. Given the estimate of one immunopositive cell out of 10,000, we should have found about five out of 50,000 positive clones. In the immunoscreening assay after the RT-PCR of the RNAs, we found a minimum of 2 and a maximum of 12 positive clones, resulting in a mean of 6 (Table 2). However, we were not successful in finding immunopositive clones after the genomic DNA PCR in two AD and two DS patients. A total of 400,000 clones were negative with immunoscreening, which also excluded PCR artifacts in the cDNA screening. A similar result was obtained with the RT-PCR and genomic PCR for Ubi-B of one of the DS patients: 100,000 clones were negative with immunoscreening. A comparable amount of RNA and genomic DNA resulted in 2 to 138 positive clones (Table 2) after RT-PCR compared with none after genomic PCR. In a best case estimate of the variables of the genomic PCR-immunoscreening approach, if screening 100,000 colonies we should have found five positives [for both βAPP and Ubi-B, on the basis of the assumption of a heterozygous genotype (6)]. Because we found no positive clones and in a worst case estimate this method would not be sufficiently sensitive, we developed a more sensitive method.

A direct discrimination between mutant and wild-type DNA was based on the use of primers specific for the mutation. Using 5′ oligonucleotides ending in the expected deletions and 3′ oligonucleotides, which hybridize 100 and 674 bp downstream of the βAPP and Ubi-B mutations, respectively, we reached a detection level of 10 copies of mutant cDNA mixed into 500 ng of genomic DNA (80,000 cells expressing βAPP or Ubi-B). Under the same sensitive conditions, that is, one copy of the mutant DNA in 8000 cells, genomic PCR was performed on human brain DNA derived from an AD, a DS, and a control patient.

These experiments repeatedly failed to detect any amplification products for either βAPP+1 or Ubi-B+1. We repeated the PCR 10 times, checking at least 5 μg of genomic DNA from each patient (that is, 1.6 × 106 copies of βAPP and Ubi-B), but none of the PCRs showed a specific amplification product for the mutation. A PCR with oligonucleotides hybridizing to the wild-type sequences of βAPP and Ubi-B gave the expected products. Although it is difficult to base conclusions on a negative result, the mixing control experiments show that there is only a very small chance that a positive DNA amplification product was not detected. Thus, it is likely that frameshift mutations introduced in the transcripts and not in the DNA are responsible for the +1 proteins observed by immunocytochemistry.

Here, in the cerebral cortex of AD, DS, and control patients, two novel βAPP- and Ubi-B–derived proteins generated by mutations of βAPP and Ubi-B transcripts were detected. The mutations were in all instances a dinucleotide deletion (ΔGA or ΔGT) occurring preferentially in or adjacent to GAGAG motifs. Because we also found a CT deletion in a CTCT motif of Ubi-B transcripts, it seems that other dinucleotide deletions in simple dinucleotide repeats occur as well. The much higher frequency of βAPP+1 and Ubi-B+1 proteins in AD patients compared with their age- and sex-matched controls indicates that transcript mutation is a critical factor for initiating neuropathology in nonfamilial forms with early- and late-onset AD.

The absence of these +1 proteins in patients with Parkinson's disease, except for one patient who also showed AD neuropathology, suggests that these +1 proteins correlate strongly with AD. Furthermore, the +1 proteins occurred in areas known to be severely affected in AD [for example, CA1 and the subiculum in the hippocampus (20)]. The DS patients revealed intense βAPP+1 and Ubi-B+1 immunoreactivities, except for one DS patient, who did not display any neurodegeneration in the three areas studied and did not suffer from dementia (12). Consistent with the idea that the transentorhinal cortex is an early target for neuropathological changes in AD (13), in the nondemented DS patient, βAPP+1, Ubi-B +1, and Alz-50 immunoreactivities coexisted in Bodian-stained tangles in cellular islands of the pre-α layers. The age-related presence of Ubi-B+1 immunoreactivity in the hippocampus of nondemented controls indicates that the Ubi-B+1 peptide may contribute to initial stages of neurodegeneration in AD. The Ubi-B+1protein may therefore be a valuable diagnostic tool for the early detection of AD.

Because enhanced transcriptional activity may be correlated with the presence of +1 proteins (5), it is possible that a transcript mutation resulting in βAPP+1 and Ubi-B+1 proteins occurred. This idea is best illustrated in DS, in which βAPP gene expression is much higher than expected on trisomy 21 alone (21-23). On the other hand, the lower frequency of βAPP+1 protein in AD patients (Table1) is consistent with the fact that their βAPP transcript levels are not essentially increased (24, 25). The high frequency of Ubi-B+1 protein in AD and DS is in accordance with overexpression of the Ubi-B gene (7, 26).

The coexisting βAPP+1 and Ubi-B+1proteins, as well as other +1 proteins, may impair neuronal functioning and amplify or induce neuropathology in an as yet unknown manner. For instance, the Ubi-B+1 molecules may be responsible for the lack of multiubiquitylation of the hyperphosphorylated tau-rich neurofibrillary tangles that accumulate during the long period of neurodegeneration (9). One explanation for these findings is that the Ubi-B+1 molecules are unable to bind to lysine residues in target molecules, because they lack the COOH-terminal glycine residue in the first repeat, which is essential for subsequent multiubiquitylation and activation of the proteasomal machinery (27). This process does not seem to occur efficiently in cells with tangles in AD. No Ubi-B–associated COOH-terminal hydrolase and 26S proteasomal immunoreactivity have been found in compact tangles (28, 29).

This study shows that βAPP and Ubi-B transcripts can be modified by dinucleotide deletions (ΔGA, ΔGT, or ΔCT). The GA deletion is similar to the one reported in vasopressin transcripts of the homozygous Brattleboro rat (5). The frequently mutated motif in exon 9 of the βAPP gene transcript is in fact an extended version of GAGAG (that is, GAGAGAGA) (Fig. 1). We also addressed the issue of whether the dinucleotide deletions occur at the transcript or the genomic level. Although we used two different sensitive approaches to reveal a genomic mutation, we failed to find any indication of a mutation at the genomic level. Support for the possibility of a general process of transcript mutation was provided by the present results showing that βAPP+1 and Ubi-B+1 proteins are coexpressed within the same neurons. In addition, in individual AD and DS patients, two or three different dinucleotide deletions were found in two different transcripts, which makes a genomic event unlikely. We thus tentatively conclude that these modifications may take place during or after transcription. In view of the finding that frameshift mutations occur in multiple proteins within the same neuron, we postulate that a common denominator in the transcription-propagating events is involved. The mechanism of transcript mutation (ΔGA, ΔGT, or ΔCT) is, however, unclear. It is most probably not restricted to postmitotic cells, because we were able to show that an ectopically expressed rat vasopressin transgene undergoes a similar process in dividing cells (30).

Transcript mutation may thus be a widely occurring phenomenon. In principle, each transcript containing a susceptible motif, such as GAGAG, could undergo such a process. However, postmitotic neurons are less capable of compensating for transcript-modifying activity and are thus particularly sensitive to the accumulation of frameshifted proteins. Accumulation of +1 proteins together with the consequent lack of functional proteins is probably critical for cellular functioning. Thus, during aging, single neurons may generate and accumulate abnormal proteins, consequently leading to cellular disturbances and causing degeneration. The mechanism of dinucleotide deletion at the transcript level may well underlie a number of neurodegenerative pathologies.

  • * To whom correspondence should be addressed. E-mail: f.van.leeuwen{at}


View Abstract

Navigate This Article