Similar Alternative Splicing of a Non-homologous Domain in PA4-Amyloid Protein Precursor-like Proteins*

The pA4-amyloid protein precursor (APP) is a transmembrane glycoprotein that is the source of the charac- teristic PA4amyloid deposits found in Alzheimer brains. It is expressed in several isoforms generated by alterna- tive splicing of exons 7,8, and 16, of which the leukocyte-derived APP mRNh lacking exon 16 are significantly expressed in non-neuronal tissues, but not in neurons. The recent finding of APP-like proteins prompted us to analyze alternative splicing of the nearest relative of APP, the amyloid protein precursor homologue (APPH) or amyloid precursor-like protein 2 (APLP2). We were able to show that there are two alternatively spliced inserts, Le. the Kunitz protease inhibitor domain and a 12-amino-acid-encoding region on the NH,-terminal side of the transmembrane domain, which is part of the region of highest divergence between APP and APLP2/ APPH. Analysis of the tissue-specific differential expres- sion of the resulting four APLP2/APPH mRNA isoforms revealed that isoforms lacking the latter, non-homolo- gous insert are highly expressed in non-neuronal tissues, but only weakly in neurons. While this resembles the tissue-specific alternative splicing of exon 16 of APP, expression of the Kunitz protease inhibitor-encoding exon of APLP2IAPPH

The pA4-amyloid protein precursor (APP) is a transmembrane glycoprotein that is the source of the characteristic PA4amyloid deposits found in Alzheimer brains. It is expressed in several isoforms generated by alternative splicing of exons 7,8, and 16, of which the leukocytederived APP m R N h lacking exon 16 are significantly expressed in non-neuronal tissues, but not in neurons. The recent finding of APP-like proteins prompted us to analyze alternative splicing of the nearest relative of APP, the amyloid protein precursor homologue (APPH) or amyloid precursor-like protein 2 (APLP2). We were able to show that there are two alternatively spliced inserts, Le. the Kunitz protease inhibitor domain and a 12-amino-acid-encoding region on the NH,-terminal side of the transmembrane domain, which is part of the region of highest divergence between APP and APLP2/ APPH. Analysis of the tissue-specific differential expression of the resulting four APLP2/APPH mRNA isoforms revealed that isoforms lacking the latter, non-homologous insert are highly expressed in non-neuronal tissues, but only weakly in neurons. While this resembles the tissue-specific alternative splicing of exon 16 of APP, expression of the Kunitz protease inhibitor-encoding exon of APLP2IAPPH is abundant in both neuronal and non-neuronal tissues and thus differs from APP. Because of the similar regulation of alternative splicing of exon 16 of APP and the described APLP2/APPH insert, and because of structural similarities of the sequences and the predicted secondary structures, a functional homology of alternatively spliced isoforms of APP and APLP2/ APPH is suggested.

~ ~~
A characteristic feature ofAlzheimer's disease is the cerebral deposition of pA4-amyloid (1,2). The pA4 peptide consists of maximally 43 residues and is derived from a larger transmembrane glycoprotein termed the pA4-amyloid protein precursor (APP)' (3)(4)(5) by a partially characterized metabolic pathway * This work was supported by Grants SFB 317 and 258 from the Deutsche Forschungsgemeinschaft, the Bundesminister fur Forschung und Technologie of Germany, the Metropolitan Life Foundation, the Wurttemberg (to K. B.), the National Health and Medical Research Fonds der Chemischen Industrie, the Forschungsschwerpunkt Baden-Council of Australia, the Vietorian Health Promotion Foundation, and the Aluminium Development Corporation ofAustralia (to C. L. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisenent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The aforementioned lengths of the APP isoforms are correct under the assumption that all exons other than exon 7 and 8 are being expressed. Recently, however, another alternatively used splice site was identified, involving the 54 bp of exon 15. This exon codes for 18 amino acids preceding the NH, terminus of the PA4 region of APP by 16 amino acids. The APP transcripts excluding exon 15 were first discovered in peripheral leukocytes and immunocompetent cells of the brain and are, therefore, denoted leukocyte-derived APP (L-APP) mRNAs (19,20). By means of a quantitative polymerase chain reaction assay from reverse-transcribed RNA(RT-PCR), it was then shown that L-APP mRNA isoforms are ubiquitiously expressed with the exception of neurons (21). In peripheral rat tissues, L-APP mRNA isoforms represent between 25% (skeletal muscle) and -70% (aorta, pancreas) of total APP transcripts. All four possible APP mRNA isoforms without exon 15 were shown to exist, i.e. L-APP752, L-APP733, L-APP696, and L-APP677. In the rat central nervous system, where L-APP expression accounts for about 4% of total APP mRNA, non-neuronal tissues like meninges, plexus choroideus, and brain vessels also showed significant expression of L-APP transcripts. While primary cultured glial cells equally contain high portions of L-APP mRNA, primary cultured neurons were shown not to express detectable levels of L-APP transcripts (21).
A distinct physiological role for the ubiquitously distributed APP has not yet been determined. Several putative functions have been ascribed for the transmembrane andlor secreted APPs (discussed in Ref. 20). Apart from the putative function of the KPI insert (exon 7) as protease inhibitor, however, the functional significance of alternative splicing of exon 8 and 15 has not been elucidated yet.
Recently, cDNAs were isolated coding for proteins related to APP. From a mouse brain library, a cDNA was cloned that encodes a protein whose predicted amino acid sequence is 42% ase inhibitor; PCR, polymerase chain reaction; RT-PCR, PCR of reverse transcribed RNA, bp, base pair(s). identical to that of APP (22). This 653-amino-acid protein, which was initially termed the amyloid precursor-like protein (APLP) and is now denoted as APLP1, appears to be similar to APP in overall structure (Fig. 1). The amino acid homologies are concentrated within two regions of the extracellular part and the cytoplasmatic domain, while the region on the NH,terminal side of the transmembrane domain shows the highest divergence.
More recently another full-length cDNA of this emerging multigene family was identified. This molecule, which also shares the overall domain organization with APP, was termed amyloid precursor protein homologue (APPH) (23) or amyloid precursor-like protein 2 (APLP2) (24) (Fig. 1). As identified by Sprecher et al., it is 763 amino acids in length and encodes a KPI domain like APP770, L-APP752, APP751, and L-APP733. The homology between APP and APLPSIAPPH is even higher than between APP and APLP1. As with APLP1, there is no significant similarity in a region NH,-terminal to the transmembrane domain.
A partial-length cDNA had been isolated from a rat sperm library even before this (25) that appears to be identical with the COOH-terminal part of rat APLPSIAPPH (Fig. 1). The sequence of the encoded protein YWK-I1 is highly homologous to human APLP2/APPH including a very high similarity regarding the region on the NH,-terminal side of the putative transmembrane domain.
Besides APLP1, another murine cDNA encoding for an APPlike protein was recently identified, which was called CDE1binding protein (26). By comparing its sequence to human APLPB/APPH, we realized that both sequences are highly homologous to each other. Of the 511 amino acids encoded by the published murine sequence, -90% are identical to human APLP2/APPH. However, the first nucleotide of the murine sequence corresponds to position 551 of the coding sequence of human APLPS/APPH. As minor differences, there seem to be two missing nucleotides around position 16 and a surplus nucleotide at position 1479. More importantly, there is a deletion of 168 nucleotides at position 70/71 which exactly represents the KPI sequence. And, there is also another region of 12 amino acids which seems to be deleted. This is encoded by nucleotides 1919-1945 (residues 613-624) of APLP2/APPH, and this sequence is also part of the YWK-I1 protein. This region was also deleted in several human APLP2IAPPH cDNA clones analyzed (23, 24). Taken together, the described sequences of human APLP2I APPH, murine CDEl-binding protein, and rat YWK-I1 seem to represent the species-specific versions of the same protein. The differences between the murine and the human APLPPIAPPH sequences indicate the existence of a t least two alternatively spliced regions, residues 309-364 of APLPBIAPPH representing the KPI domain and residues 613-624, which are part of the region on the NH,-terminal side of the transmembrane domain in which APP and APLP2/APPH most strongly differ.
This seems to constitute a striking parallel to the above described alternative splicing of APP. Therefore, we performed a detailed analysis of alternative splicing and tissue-specific expression of APLPIIAPPH mRNAs in the rat.

MATERIALS AND METHODS
ICFssue Preparation and Primary Cell Culture-Perfused tissues were obtained from 7or 8-month-old rats as described (21). Primary neuronal cultures and a mixed culture of neurons, astrocytes, and some microglial cells were prepared from septal cells of E18 rat brain and analyzed at 8 days in vitro. Cell culture conditions and preparation of astrocyte-enriched cultures have also been described (21). Isolation of RNA, Reverse Transcription, and RT-PCR-Total RNA extraction and preparation of cDNA using oligo(dT),, as primer and Super-ScriptT" reverse transcriptase (Life Technologies, Inc.) were performed as described (21,27). One-hundredth of this cDNA preparation was amplified by PCR (28) using the protocol given in Ref. 21. For PCR analysis without labeling the products, 35 cycles were performed; for radioactively labeled PCR products, 26 cycles were usually performed.
Analysis and Sequencing of PCR Products-PCR products were analyzed either directly on 1.5% agarose gels by ethidium bromide staining or by denaturing polyacrylamide gel electrophoresis (4.0%) and autoradiography. For quantitation of individual bands, gels were analyzed on a PhosphorImager (Molecular Dynamics) by overnight exposure (21). The figures shown were prepared either from corresponding autoradiograms (Kodak X-Omat A R ) or, for ethidium bromide-stained gels, by means of an electronic video imaging system (21).
For sequencing, unlabeled PCR products (sm1838 and ar1877) were purified by agarose gel electrophoresis (1.5%) and subsequent electroelution of the excised gel fragments. One-twentieth of the isolated products were reamplified using the same primer pair (25 cycles). Onefiftieth of the reamplification products was then used for thermal cycle sequencing using the "fmol DNA sequencing kit" (Promega) with 33Plabeled sm1838, sm11099, and ar1877.

Identification of Four Alternatively Spliced Isoforms of
APLP2IAPPH in Rat Tissues-On the basis of our knowledge of alternative splicing of APP, we decided to analyze the alternative splice patterns of one of the other, newly identified members in the emerging family of APP-like proteins. So far, a protein termed APPH or APLP2 has been described as the natively spliced APLPB/APPH mRNA

FIG. 2. RT-PCR assay to detect alter-
isoforms. a, the two potential alternatively spliced inserts of APLP2/APPH are shown. Bottom, the series of PCR reactions employed in this study is depicted. The following primer pairs were used: PCR A, sh1051/ahm11099; PCR B, sm11099/ar11805; PCR C, sm18381 or11805; PCR D, sm11099/ar11877; PCR E, sm1838/ar1877; and PCR F, sr1809/ ar12342. Tm indicates the sequence coding for the transmembrane part of APLPP/APPH (gray). b, PCR products obtained with cortical and thymus cDNA preparations. Total RNA was extracted from perfused rat brain and thymus, and 2 pg were reverse transcribed and aliquots amplified for 35 cycles using the primer pairs described in nearest relative, of which only the human full-length cDNA is available yet. However, we preferred to examine perfused rat tissues. For these, the tissue-specific alternative splicing pattern ofAPP mRNAs has recently been described in great detail (21).
To analyze alternative splicing of rat APLPBIAPPH, we performed a series of PCR analyses from reverse-transcribed RNA (Fig. 2a). On the basis of the published sequences of rat YWK-I1 (251, murine CDEl-binding protein (261, and human APLP2I APPH (231, we selected primer pairs A-F to especially identify any alternative splicing of either the KPI domain or the 12amino-acid region (on the NH,-terminal side of the transmembrane domain) which were both lacking in the described mouse sequence. For analysis, 2 pg of RNA from perfused cortex and thymus were reverse transcribed, and aliquots were amplified in 35 cycles. After electrophoretic separation of the PCR products, we got the results shown in Fig. 2b. Alternative splicing of the KPI region was analyzed by PCR analysis A and C, both of which include the KPI domain-encoding sequence. With both cortex and thymus cDNA, each of these two primer pairs produced two fragments differing in length by about 170 bp. As indicated by PCR analysis B, which gave only one PCR product, this indeed seems to be caused by alternative splicing of the KPI region (for this, the expected product lengths are PCR A, 1071 and 903 bp; PCR C, 989 and 821 bp; PCR B, 728 bp, calculated on the basis of the human APLPBIAPPH cDNA sequence (23)).
Using PCR analysis D and F, cDNA regions encoding more COOH-terminal parts of APLPS/APPH were analyzed. Two different PCR products were amplified with both primer pairs differing by -40 bp. Again, it was proven by PCR analysis B that the occurrence of these two isoforms was due to the small region which the PCR products of both primer pairs D and F have in common. This region was defined to include the 36-bp sequence in which the described mouse and the human APLP2I APPH cDNA differ (expected length of PCR products: PCR D, 800 and 764 bp; PCR F, 555 and 519 bp).
Finally, with PCR analysis E, alternative splicing of both the KPI and this 36-bp region ofAPLP2IAPPH was detected simultaneously. Three different bands representing four PCR products with the expected lengths were observed (1061 and 1015 bp (unresolved on this gel), 893 bp, and 857 bp).
Sequencing Proves Alternative Splicing of Inserts Encoding the KPI Domain and the More COOH-terminal Located Region of 12 Amino Acids-The four different isoforms detected by PCR analysis E (primer pair sm1838 and ~1 1 8 7 7 ) were further analyzed by sequencing. We therefore amplified these four APLP2IAPPH cDNA fragments from a cDNA preparation of rat cardiac muscle and purified them by electrophoresis and subsequent electroelution. They were then reamplified using the same primer pair and their purity checked on the gel shown in Fig. 3a. Other aliquots were then used for thermal cycle sequencing, proving the presence of the KPI domain-encoding sequence (168 bp) in fragments 1 and 2 and its absence in fragments 3 and 4 (Fig. 3b). On the other hand, only fragments 1 and 3, but not fragments 2 and 4 contain a 36-bp insert at their 3'-end (Fig. 3c). This alternatively spliced sequence exactly represents the 36-bp region omitted in the murine CDE1binding protein cDNA. Since the full-length APLPBIAPPH isoform (corresponding to fragment 1) should consist of 763 residues as predicted from the human cDNA sequence (32), the other APLPBIAPPH isoforms (fragments 2 4 ) were calculated to consist of 751, 707, and 695 residues. These four different isoforms were therefore denoted as APLP2IAPPH763, L-APLP2IAPPH751, APLP2IAPPH707, and L-APLP2IAPPH695. In analogy to APP denotation, we introduced the prefix "L-" to selectively label those isoforms lacking the alternatively spliced 12-amino-acid region on the NH,-terminal side of the transmembrane domain. As will be shown later, alternative splicing of this insert exhibits extensive parallels to alternative splice patterns of APP transcripts regarding exon 15.
Rat Sperm Membrane Protein YWK-11 and Murine CDEIbinding Protein Are the Species-specific Versions of Human APLPZIAPPH-We then continued sequencing of the four described PCR products to obtain their complete sequence. Apart from the alternatively spliced regions, this sequence was identical for all four PCR fragments. The last 152 nucleotides are completely identical to the YWK-I1 cDNA sequence published. The protein sequence intermediate to the two alternatively spliced inserts is highly homologous to the human APLP2I APPH sequence (97%) and to the murine CDEl-binding protein (98%). We therefore conclude that rat APLPBIAPPH as ana- I lyzed by us is identical to rat YWK-11, representing the rat version of human APLPBIAPPH and murine CDEl-binding protein. The major differences between the published sequences of CDEl-binding protein and human APLP21APPH are explained by alternative splicing of two inserts, as demonstrated above for rat APLPB/APPH. Consequently, the published murine CDEl-binding protein sequence is identical to a partial sequence of the murine L-APLP2/APPH695 isoform.
No APLP2IAPPH Equivalent to the Alternatively Spliced Exon 8 of APP Was Observed-Our results indicate that APLP2/APPH and APP not only resemble each other in their domain structure, but also in alternative splicing. The question arises whether the alternatively spliced exon 8 of APP (expressed in APP770, L-APP752, APP714, and L-APP696) also has an equivalent expressed in one or more of the alternatively spliced APLPBIAPPH mRNA isoforms. However, no homologue to exon 8 of APP was detected in any of the four rat heart APLP2/APPH transcripts analyzed. No further APLPBIAPPH mRNAs were detected within this cDNA preparation nor in a large number of other rat organs and tissues including brain, as described later. Therefore, in contrast to APP, either the rat APLPZ/APPH gene does not contain the MRC OX-2 domainencoding exon, or this region is constitutively spliced during APLP2/APPH hnRNA processing.
RT-PCR Assay for Quantitation of APLP2JAPPH mRNA Isoforms-For quantitation of alternatively spliced APLP2/ APPH mRNAs by RT-PCR, the primer pair sm1838 and cur1877 was employed after radioactive labeling of cur1877 on its 5'-end with polynucleotide kinase and [y-""P]ATP. Initially we tested the PCR amplification rate of the different products for different cycling times, using a cDNA pool from perfused rat hippocampus (Fig. 4). Four different bands were detected corresponding to APLP2/APPH763, L-APLP2/APPH751, APLP2/ APPH707, and L-APLP2/APPH695 (Fig. 4a). Thus, all four APLPB/APPH isoforms detected in rat cardiac muscle are also observed in rat hippocampus. The amount of radioactivity incorporated in the individual bands was then measured and the logarithms plotted as a function of the PCR cycle number (Fig.  4b). As illustrated, the slopes of the curves remain constant through 32 cycles with a similar amplification efficiency for all four APLPSIAPPH mRNA isoforms (17,211. Therefore, the initial relative amounts of the different APLP2/APPH transcripts in the sample should be identical to the numerical values as determined for the corresponding PCR products. As was shown for a similar RT-PCR assay for quantitation of APP mRNA isoforms, the relative amounts of individual mRNA isoforms obtained by use of such a method are highly accurate and well reproducible (21).
All Four Alternatively Spliced APLP2JAPPH mRNA Isoforms Are Ubiquitiously Expressed in Peripheral Rat llssues-For a detailed analysis of APLP2/APPH mRNA isoforms in peripheral tissues, the aforementioned quantitative PCR assay was used. The corresponding autoradiograms are shown in Fig.  5, a and b. In general, four different bands are visible. As described, the upper two bands correspond to the KPI domainencoding APLPS/APPH transcripts, while the second from the cardiac muscle cDNA preparation. Initial amplification was performed as described (PCR analysis E , Fig. 2). PCR products were separated by agarose gel electrophoresis, individually electroeluted, and reamplified. Aliquots were compared by gel electrophoresis with the initial mixture of amplified cDNA fragments. b and c, sequence analysis of the individual APLPB/APPH cDNA fragments. Aliquots of the purified and reamulified framents 1 4 from a were sequenced by thermal cycle sequencing using sm1838 ( b ) or ar11877' (c) as sequencing primer. top and the lowest band are due to the two L-APLPBIAPPH transcripts. All four APLPB/APPH mRNA isoforms are expressed in all peripheral tissues analyzed, including the two L-APLPPIAPPH isoforms. This is very similar to L-APP expression which was recently also found to occur in all of these peripheral rat tissues (21). Quantitation of the four cDNA fragments revealed that in peripheral tissues L-APLPB/APPH expression varies between 35% (skeletal muscle) and 92% (epididymis, followed by uterus and ovary) of total APLPB/APPH mRNAs (Table I). Comparison with results obtained for selected unperfused tissues showed that the results are virtually independent of perfusion (data not shown). This is in agreement with the only very weak levels of PCR products observed for blood cDNA, which was also the case for pancreatic cDNA. Differential Expression of APLP2IAPPH mRNA Isoforms in the Central Nervous System of the Rat-We then examined the distribution of APLPX/APPH transcripts in the rat central nervous system and related tissues (Fig. 5c, Table I). In three different brain regions analyzed (cerebellum, cortex, and hippocampus), the majority of APLPB/APPH mRNAs is made up by the KPI motif encoding APLPB/APPH763 transcript (about 80%), but all three other APLPB/APPH mRNA isoforms are also detectable. L-APLPBIAPPH transcripts contribute between 5 and 13%. A higher portion of L-APLPBIAPPH mRNA isoforms (26%) was observed in spinal cord. This tissue is unique in expressing APLPB/APPH707 mRNA to about 44%. This is the highest value measured for this isoform among all rat tissues examined. A relatively low content of KPI domain-encoding  Fig. 4. A water control was performed which included all steps of RNA extraction, cDNA synthesis, and PCR, and no detectable signals were observed (data not shown). GI, gland. c, RT-PCR analysis of APLP2/ APPH mRNAs in tissues of the central nervous system, in related tissues, and in primary cultured brain cells. Septal neurons were cultured for 8 days in vitro in a serum-free medium containing cytosine arabinoside, and mixed brain cells indicate septal cells cultured for the same period in a serum-containing medium without cytosine arabinoside. Sp., spinal; N. trig., trigeminal nerve; N. opt., optic nerve, P1. chor. I V , choroid plexus from the IVth ventricle, P1. brach., plexus brachialis. isoforms was also observed for the trigeminal nerve; in contrast to spinal cord, however, this tissue mainly expresses L-APLPB/ APPH695 mRNA.
In all neural tissues examined other than the described brain regions, which includes cranial, spinal, and peripheral nerves, significantly higher relative amounts of L-APLPBIAPPH mRNAs were observed. For non-neuronal tissues such as meninges and choroid plexus, the APLP2/APPH splicing pattern resembles the typical expression pattern of peripheral organs. This includes expression of L-APLPBIAPPH transcripts of about 70%.
Primary Cultured Astrocytes and Neurons Are Similar in Their Expression of KPI-APLP2IAPPH mRNAs, but Differ in Regard to L-APLP2IAPPH mRNA Expression-For comparison of neuronal with non-neuronal cells of the central nervous system, primary cultures of E18 rat septal cells were prepared. As shown in Fig. 5c, primary cultured septal neurons predominantly expressed APLPB/APPH763 mRNA (about 86%) ( Table  TABLE I Differential expression of APLPZIAPPH mRNAs in peripheral tissues, brain tissues, and primary cultured brain cells Radiolabeled and electrophoretically separated RT-PCR products (Fig. 5) were analyzed on a PhosphorImager and relative amounts calculated in percent. For each band, a corresponding background value was subtracted. 763, 751, 707, and 695 represent the corresponding APLPWAPPH cDNA bands. L-APPH denotes the sum of L-APLP2I APPH751 and L-APLP2/APPH695, while KF' I represents the KPI do-APPH751). I). L-APLPBIAPPH mRNA isoforms were present in these cultures only up to -7%, mainly as KPI-encoding transcript. Similar results were obtained for primary cultured neurons from other brain regions (data not shown). Therefore, the APLPBI APPH mRNA expression pattern of neurons cultured in vitro highly resembles the splice pattern obtained for intact brain regions as described above.

main-encoding mRNAisoforms (sum ofAPLP2IAPPH763 and L-APLPW
Mixed septal cultures with a significant number of proliferating non-neuronal cells showed a different APLPBIAPPH splice pattern (Fig. 5c), in which the contribution of L-APLPBI APPH mRNA isoforms was assessed to be -26%. In cultures highly enriched for astrocytes, considerable amounts of L-APLPBIAPPH transcripts were observed contributing to -79% (Fig. 5c). Mainly the KPI motif encoding L-APLPBI APPH751 mRNA was detected. Hence, primary cultured astrocytes and neuronal cells are similar in regard to predominant KPI expression, but differ in alternative splicing of the more downstream, 12-amino-acid encoding region. This leads to predominant L-APLPBIAPPH expression in astrocytes, while only marginal amounts of L-APLPBIAPPH mRNAs are expressed in neuronal cells. DISCUSSION Among the two recently identified APP-like proteins, human APLPBIAPPH is the nearest APP relative identified so far. The striking degree of amino acid sequence similarity and the conservation of overall domain structure between human APP and human APLPBIAPPH prompted us to identify the rat version of APLPBIAPPH. We analyzed the partial cDNA sequence extending from nucleotide 865 to 1875 of the coding sequence, as presumed by comparison to the human APLPBIAPPH cDNA sequence. As we were able to show for rat brain and thymus by RT-PCR analysis covering the almost complete coding sequence of APLPBIAPPH, this partial cDNA encompasses the two principal sites of alternative splicing. Four different isoforms were detected, resulting from the combination of these two alternatively spliced inserts. By analyzing the composition and the tissue distribution of these rat APLPBIAPPH mRNA isoforms, we present for the first time a detailed analysis of alternative splicing of one of the recently identified APP-like proteins. We were able to show that the previously described murine CDE1binding protein and rat sperm membrane protein YWK-I1 are the species-specific versions ofAPLP2IAPPH. The principal differences between the murine and the human APLPS/APPH sequence can clearly be explained by alternative splicing of these two inserts.
These two alternatively spliced regions were characterized as follows. First, APLPBIAPPH hnRNA contains a KPI-encoding region highly homologous to exon 7 of the APP gene. Expression of an MRC 0x2-region homologous to exon 8 of the APP gene was not detected. The second alternatively spliced insert is a 12-amino-acid encoding sequence that is part of the region with the highest divergence in comparison to APE Strikingly, APP mRNA also exhibits an alternative splicing of a similar sequence, exon 15, in exactly this non-homologous region. APP transcripts lacking this exon have been previously termed L-APP mRNAs. By analogy, we therefore denoted the four observed APLPBIAPPH mRNA isoforms as APLPBI APPH763, L-APLPBIAPPH751, APLPBIAPPH707, and L-APLPBIAPPH695 transcripts; the latter two are lacking the KPI-encoding sequence. As with APP denotation, the numbers given correspond to the presumed number of residues. However, these APLPBIAPPH isoforms have to be distinguished from APP isoforms denoted with similar numbers (i.e. 695 and 751).
We then developed a quantitative RT-PCR assay to assess tissue-specific APLPBIAPPH splice patterns. Although there are interesting variations among the organs analyzed, two principal patterns of differential expression of APLPBIAPPH mRNAs were observed. In most of the non-neuronal tissues, both from peripheral organs and from the central nervous system, including meninges, brain vessels, choroid plexus, and primary cultured astrocytes, the KPI-encoding L-APLPB/ APPH751 mRNA was the predominant isoform, followed by APLP2IAPPH763 and L-APLP2IAPPH695 mRNAs. In brain regions and primary cultured neurons, however, a significantly different expression pattern was observed. APLPBIAPPH 763 mRNA was by far the most abundant isoform here. APLPBI APPH707 was slightly more strongly expressed in these tissues than L-APLPB/APPH751 and L-APLPB/APPH695 mRNAs. Hence, L-APLPBIAPPH mRNA expression is high in peripheral organs, but low in neuronal tissues. KPI-encoding APLPB/ APPH mRNAs, however, are the predominantly expressed APLPBIAPPH transcripts in both neuronal and peripheral tissues.
When these results are compared with the corresponding APP expression patterns (21), alternative splicing of the 12amino-acid-encoding APLPBIAPPH insert is found to be even  more similar to APP alternative splicing than that of the KPIencoding region. With APP, both L-APP and KPI-encoding transcripts are highly expressed in non-neuronal tissues, while their expression in rat brain and in primary cultured neurons is quite low. Therefore, differential expression of L-APLPB/ APPH mRNAs very much resembles L-APP mRNA expression.
In primary cultured septal neurons, L-APLPB/APPH transcripts contributed to -7% of total APLPBIAPPH mRNAs. Using the same cDNA pool, L-APP mRNAs had previously been assessed to represent less than 1% of total APP transcripts. Hence, in neuronal cells, the relative amount of L-APP mRNAs is significantly lower than the corresponding contribution of L-APLPB/APPH transcripts. This was also the case in periph-eral organs: the median value of the L-APLPBIAPPH mRNA portion in the 24 peripheral tissues was determined as 79%, while the median value of L-APP mRNA in the same tissues was only 47%. Nevertheless, the "L-APP position" and the "L-APLPBIAPPH position" of a given peripheral organ in this list of tissues seems to be quite conserved. For instance, skeletal muscle exhibits the lowest relative amount of both L-APLPBI APPH (35%) and L-APP (25%) transcripts.
Surprisingly, the alternatively spliced inserts resulting in L-APP and L-APLPB/APPH formation are part of exactly that region of about 130 residues on the NH,-terminal side of their transmembrane domain in which APP and APPWAPLPB most strongly differ. Thus, the question of common features in the amino acid sequence and secondary protein structure arises which could indicate a function regulating conformation. We therefore compared the APP and APLPS/APPH sequence of this divergent region plus the transmembrane and the intracellular, COOH-terminal domain (Fig. 6a). The alternatively spliced inserts do not only differ in length (18 versus 12 residues), but also exhibit a significantly different distance to the beginning of the transmembrane region. However, in both APP and APLPS/APPH these inserts represent the COOH-terminal end of a proline-rich region (9 prolines in APP, 8 in APLPBIAPPH). Predictions for the corresponding secondary structures are illustrated in Fig. 6b (proline-residues marked by small arrowheads). For the highly homologous transmembrane and COOHterminal domains of APP and APLPZ/APPH, a very similar secondary structure and hydrophilicity is anticipated. For both proteins, the insert region is predicted to be of substantial hydrophilicity. Between the transmembrane region and the alternatively spliced insert, for both APP and APLPB/APPH, a significant a-helical region is expected (Chou-Fasman prediction). With the APLPS/APPH insert shifted to the NH, terminus, this a-helical region is also NH, terminally shifted. Both proline-rich regions (inserts plus adjacent regions on the NH,terminal side), however, contain several stretches with a high turn potential. This is emphasized by the prediction method of Gamier-Robson: the insert domains are expected to contain the most COOH-terminal turns of this part rich in turn regions, followed by a n extended domain with only a low probability for turn regions.
The described features of the protein sequences and the predicted secondary structures indicate a similar spatial organization of the non-homologous regions ofAPP and APLPB/APPH and the alternatively spliced inserts within these domains. Because of this and the similarities in the tissue-specific regulation of APP and APLPBIAPPH alternative splicing, a related function of this domain is suggested. It is well conceivable that the common structural features of the non-homologous domain are an indication of the regulation or modulation of a yet unknown functional activity of these proteins. We also postulate that deletion of this insert leading to L-APP and L-APLP2/ APPH isoforms will result in similar changes of the structural characteristics of these proteins, suggesting a functional homology of the individual alternatively spliced isoforms of APP and APLPBIAPPH. In view of the special position of neuronal alternative splicing (virtually all transcripts encode the alternatively spliced insert), the low frequency of the insert lacking L-APP and L-APLPBIAPPH isoforms might well be closely related to typical neuronal function. The relevance of this characteristic neuronal feature for the degeneration of neurons in Alzheimer's disease should be further evaluated.
Revealing the distribution and differential expression not only ofAPP, but also of other members of the family ofAPP-like proteins contributes to our efforts to understand the physiological function of these proteins. This should help to elucidate the role ofAPP and its related proteins in the neuropathogenesis of Alzheimer's disease. A careful and profound analysis of similarities and differences between APP and the other members of this family might turn out to be a tool for understanding the special role of APP in the etiology of this disease.
Note Added in Proof-While the manuscript was in proof, we sequenced all the missing parts of the rat version of APLP2/APPH. The sequence data will appear in the EMBLJGenBank and DDBJ Nucleotide Sequence Databases under the accession number X77934 and are in press (Sandbrink, R., Masters, C. L., and Beyreuther, K. (1994) Biochim. Biophys. Acta, in press). Because of a deletion of two codons and an insertion of four codons within the acidic domain of rAPLP2/APPH as compared with its human and murine homologues, the four different rAPLP2/APPH isoforms raised by alternative splicing are now expected to consist of 765, 753, 709, and 697 amino acids.