Identification and differential expression of a novel alternative splice isoform of the beta A4 amyloid precursor protein (APP) mRNA in leukocytes and brain microglial cells.

The gene for the beta A4-amyloid precursor protein (APP) consists of 19 exons which code for a typical N- and O-glycosylated transmembrane protein with four extracellular domains followed by the transmembrane domain and a short cytoplasmic domain. The beta A4-amyloid sequence is part of exons 16 and 17. Several APP isoforms can be generated by alternative splicing of exons 7 and 8, encoding domains with homologies to Kunitz-type protease inhibitors and the MRC OX-2 antigen, respectively. The mechanism by which the pathological beta A4 is generated is unknown, it is however a critical event in Alzheimer's disease and is distinct from the normally occurring cleavage and secretion of APPs within the beta A4 sequence. We report here for the first time considerable APP mRNA expression by rat brain microglial cells. In addition we showed by S1 nuclease protection and polymerase chain reaction analysis of reverse transcribed RNA (RT-PCR) that T-lymphocytes, macrophages, and microglial cells expressed a new APP isoform by selection of a novel alternative splice site and exclusion of exon 15 of the APP gene. This leads to a transmembrane, beta A4 sequence containing APP variant, lacking 18 amino acid residues close to the amyloidogenic region. The use of this novel alternative splice site alters the structure of APP in close proximity to the beta A4 region and thus may determine a variant, potentially pathogenic processing of leukocyte-derived APP in brain.

The gene for the BA4-amyloid precursor protein (APP) consists of 19 exons which code for a typical Nand O-glycosylated transmembrane protein with four extracellular domains followed by the transmembrane domain and a short cytoplasmic domain. The BA4amyloid sequence is part of exons 16 and 17. Several APP isoforms can be generated by alternative splicing of exons 7 and 8, encoding domains with homologies to Kunitz-type protease inhibitors and the MRC OX-2 antigen, respectively. The mechanism by which the pathological BA4 is generated is unknown, it is however a critical event in Alzheimer's disease and is distinct from the normally occurring cleavage and secretion of APPs within the BA4 sequence. We report here for the first time considerable APP mRNA expression by rat brain microglial cells. In addition we showed by S1 nuclease protection and polymerase chain reaction analysis of reverse transcribed RNA (RT-PCR") that T-lymphocytes, macrophages, and microglial cells expressed a new APP isoform by selection of a novel alternative splice site and exclusion of exon 16 of the APP gene. This leads to a transmembrane, @A4 sequence containing APP variant, lacking 18 amino acid residues close to the amyloidogenic region. The use of this novel alternative splice site alters the structure of APP in close proximity to the BA4 region and thus may determine a variant, potentially pathogenic processing of leukocyte-derived APP in brain.
Alzheimer's disease (AD)' is characterized by the intracerebral deposition of large amounts of PA4-amyloid, a 4-kDa insoluble breakdown product of a larger transmembrane span-ning PA4-amyloid precursor protein (APP) (1-7). The ubiquitously expressed APP is posttranslationally processed by an as yet unidentified protease, resulting in the release of the extracellular part of APP (8,9). This cleavage occurs within the @A4 sequence, leaving behind a non-amyloidogenic carboxyl-terminal transmembrane and cytoplasmic part of APP.
Generation of PA4 protein from its precursor therefore requires a processing event which is distinct from the normally occuring secretion of APP. The APP gene consists of 19 exons which are alternatively spliced into five different products, named according to their length in amino acids (APP695, APP714, APP751, APP770, and APRP563, for BA4-amyloid precursor related protein) (3-5, 10-15). APRP563 is a rare non-amyloidogenic secreted APP isoform. APP751 mRNA is different from AP695 mRNA by inclusion of exon 7, coding for a 57-amino acid domain with homology to Kunitz-type protease inhibitors (KPI), APP770 mRNA includes exons 7 and 8, the latter coding for a 19-amino acid large domain with homology to the MRC OX2 antigen found on thymocytes (10-12, 16). The exon 7-containing APP mRNA isoforms are predominantly expressed in the periphery, whereas APP695 mRNA is the predominant splice form in neurons (10,17). A number of studies have indicated that alternative splicing of exons 7 and 8 in APP mRNAs is changed in brain during aging and possibly during AD (18)(19)(20)(21)(22)(23). This led to the suggestion that an altered splicing pattern of APP mRNA might be a risk factor for AD. Additional alternative splicing more proximal to the PA4 region has not been studied. Since such an additional splicing event could alter the amyloidogenicity of APP, we investigated further possible alternative splicing immediately adjacent to the PA4 protein encoding exons 16 and 17 in various tissues and cell lines.
Here we describe the identification of a novel splice site selection of the APP gene. The novel splicing event fuses exon 14 to exon 16, thereby excluding the 54-bp large exon 15. It was demonstrated by S1 nuclease protection analysis and RT-PCR", that only cells of hematopoietic origin make use of this novel splice event resulting in exclusion of exon 15. Furthermore it was shown for the first time that brain microglial cells express marked amounts of APP mRNA, as well as this new form of APP mRNA, designated leukocytederived APP (L-APP) mRNA.

Isolation of Hematopoietic
Cells-Fresh human peripheral mononuclear blood leukocytes (PMBL) were obtained from buffy coats of healthy or AD donors by Ficoll-Hypaque gradient density centrifugation (24, 25). Cells were cultured at a density of 2 x lo6 cells/ml in Dulbecco's modified Eagle's medium (GIBCO/BRL, United Kingdom), containing 1 g/liter glucose, penicillin (50 units/ml), streptomycin (40 pglml), and 10% (v/v) fetal calf serum (GIBCOPRL). Stimulation of the nonadherent resting T-lymphocytes was performed with 10 pg/ml phytohemagglutinin (PHA, Sigma) for different time intervals. B-cells were isolated from the PMBL fraction using a monoclonal antibody against the CD19 antigen coated with magnetizable polystyrene beads (Dynabeads Dynal, Norway). CD4+ and CD8+ T-cells were isolated using a monoclonal antibody coated with magnetizable polystyrene beads (Dynabeads, Dynal, Norway). Monocytes were separated from lymphocytes by adherence to the plastic dishes. Adherent monocytes were either cultivated overnight in medium (RPMI 1640,10% human AB serum), recovered from the dishes by vigorous pipetting, and resuspended in fresh medium and placed in rectangular Teflon bags (Biofolie 25, Heraeus, Germany). At days 5 and 8 fresh medium was added to the bags. At the latter time point a 100% pure macrophage culture could be recovered by sedimentation of nonadherent cells (26). Macrophages were stimulated with either lipopolysaccharide (Salmoneh abortus equi, generous gift of Chris Galamos, MPI fiir Immunologie, Freiburg, Germany; 100 ng/ml), or y-interferon (500 units/ml, Bachem, Germany) for 8 h. Microglial cells were prepared from newborn rat brain as described previously and cultured for 2-4 weeks without further stimulation (27,28).
Isolation of RNA-Total RNA was prepared as described (29). Poly(A+) RNA was isolated by oligo(dT)-cellulose affinity chromatography (Pharmacia), following the manufacturer's protocol. Yield and quality of RNA preparations were determined by spectrophotometry and agarose gel electrophoresis (30). S1 Probes and Nuclease Protection Analysis-An 826-bp fragment, spanning the KPI-and OX-2 coding domain from position +320 to position +1135 of the APP770 cDNA, was subcloned into the phage Ml3mpl8 which resulted in the recombinant phage containing probe 1 (3, 31) ( Fig. 1). Probe 2 was constructed by subcloning a RsaI restriction fragment from position +1109 to +2181 of the APP770 cDNA into the phage M13mp19. A uniformly labeled single-stranded DNA probe was synthesized by annealing 1 pmol of M13 phagespecific universal oligonucleotide primer to 250 ng of single-stranded template M13 DNA. This primer was extended with 1 unit of Klenow polymerase (Boehringer Mannheim) in the presence of 25 mM dATP, dGTP, d W P , 2 mM dCTP, and 40 pCi of [ c~-~' P ]~C T P (3000 Ci/ mmol, Amersham) for 15 min at 37 "C. Extended products were digested with Sac1 (for probe I), HindIII (for probe 2), PvuII (for probe 3), and with BamHI (for probes 4 and 5) for 45 min. The labeled single-stranded DNA probes were purified on a 5% denaturing polyacrylamide gel, cut out, and eluted in 300 mM NaCl, 30 mM Tris/ HC1, pH 7.5, 3 mM EDTA, pH 8.0, at 65 "C for 20 min. S1 probes were ethanol-precipitated and resuspended in DEPC-H20 to 50,000 cpmlpl. RNA (amounts see figure legends) was hybridized overnight (75% formamide, 0.4 M NaC1, 20 mM Tris/HCl, pH 7.4, and 1 mM EDTA) at T,,, +5 "C to probe 1 or at T,,, +3 "C (calculated for longest protected fragment) to probes 2-5. S1 nuclease digestion (Boehringer Mannheim, 1200 units/sample) was performed for 2 h at 37 "C (0.3 M NaC1,3 mM ZnSO,, 60 mM NaAc, pH 4.5,0.5 pg of denatured calf thymus DNA). The resulting products were ethanol-precipitated and resolved on 5% denaturing sequencing gels (32). Autoradiography was performed using Kodak X-OmatN AR films, and appropriately exposed x-ray films with signals in the linear responsive range were used for densitometry on a Hirschmann Elscript 400 densitometer.
Southern and Northern Hybridization-RNA was run in denaturing 1% agarose, formaldehyde gels, and blot-transferred to Gene-Screen Plus (Du Pont) nylon membranes following the manufacturer's protocol (35). RT-PCR products were analyzed by 2.5% agarose gel electrophoresis and visualized by ethidium bromide staining. Agarose gels were transferred to GeneScreen Plus (Du Pont) nylon membrane following the manufacturer's protocol. Filters with RNA and DNA were hybridized with a SacI/RsaI L-APP cDNA restriction fragment using [cY-~'P]~CTP (Amersham) and nick translation (Boehringer Mannheim) (36). The buffer used for prehybridization (1 h) and hybridization contained 0.5 M Na2HP04 (pH 7.4), 7% sodium dodecyl sulfate, 1% bovine serum albumin. Hybridization was performed overnight at 65 "C, washed three times in 40 mM Na2HP04, pH 7.4,1% sodium dodecyl sulfate at 65 "C, and exposed to a Kodak X-AR film. Exposure times for Southern hybridization were approximately 1 h, for Northern hybridization 4 days at -80 "C ( Fig. 5A) or 16 h at -80 'C (Fig. 5B).

RESULTS AND DISCUSSION
Human PMBL were isolated, and the nonadherent Tlymphocytes were stimulated for 5 days with PHA. Total RNA was prepared and subjected to S1 nuclease protection analysis with probe 1, which scans from exon 3 to exon 9 of APP770 cDNA (Fig. 1). Three protected fragments were obtained using probe 1; they corresponded to splice products which contained neither the KPI nor the MRC-OX2 exons (APP695 mRNA), the KPI exon alone (APP751 mRNA), or both the KPI plus the MRC-OX2 exons (APP770 mRNA) ( Fig. 2A). The ratio of alternatively spliced APP mRNAs was approximately 41712 of APP695/APPl51/APP770. To study additional alternative splicing more proximal to the PA4 region in APP, we used S1 nuclease protection analysis with a probe spanning the region from exon 9 to exon 17, which codes for the COOH-terminal part of the 0A4 sequence and part of the cytoplasmic domain. Two protected fragments with probe 2 were observed (Fig. ZB, lane 3). The longer protected fragment corresponded to transmembrane coding APP695, APP751, and APP770 mRNA splice forms (5, 10-12). The second protected fragment, obtained with probe 2, was about 280 bases shorter and suggested the presence of a novel form of APP mRNA. Subsequently, probes 3-5 were used in S1 nuclease protection analysis to determine that the  (Fig. 2B, lanes 5, 7, and 9). This shorter signal was approximately 20-35% of the total, indicating this mRNA comprises a significant fraction of the APP mRNA made in T-lymphocytes. Fragments corresponding to APP-related transcripts (APRP563) were not seen (14). One can therefore exclude the presence of this non-amyloidogenic APP splice form in T-cells. To elucidate the region of divergence, we wished to sequence the novel APP cDNA. APP mRNA of T-lymphocytes was reverse-transcribed with APPspecific 3"primer into single-stranded cDNA and amplified by 34). The PCR products were subcloned into a Bluescript vector (pSK+, Stratagene) and sequenced with T7 polymerase.
We found an APP cDNA form in which exon 14 was spliced to exon 16, and exon 15 was excluded (Fig. 3). This alternative splicing event, fusing exon 14 to exon 16, generates a glycine codon, which is also present when exon 15 is fused to exon 16. Thus neither the reading frame, nor the amino acid at the junctional border, is altered. S1 nuclease protection analysis, sequencing, and PCR analysis revealed no further alternative splicing involving exons 3-18, including the pA4-amyloid sequence (exons 16 and 17), the transmembrane domain (exon 17), and the cytoplasmic domain (exons 17 and 18).
Tissue Specificity of Novel Splicing Form-To determine whether the novel APP mRNA is expressed in other tissues, we applied PCR to cDNAs from various human, mouse, and rat tissues and cell lines. We used oligonucleotide primers flanking the sequence of exons 14-18; the corresponding PCR products were 462 and 408 bp, for cDNAs with and without exon 15, respectively. The 462-bp PCR product was detected in all tissues and cell lines examined, which were HeLa and neuroblastoma cells, astrocytoma, mouse cerebellar primary cell culture, human temporal cortex, and hippocampus of control and Alzheimer cases, kidney, skeletal muscle and fetal thymus tissue, and all of the T-lymphocytes. In contrast the 408-bp PCR product was only found in T-lymphocytes and rat microglial cells (Fig. 4A, lanes [18][19][20][21][22][23] and 4C). Macrophages were also shown to contain considerable amounts of the novel APP splice form, as demonstrated by S1 nuclease protection analysis (Fig. 2C). B-lymphocytes demonstrated weak APP as well as L-APP mRNA expression in the S1 nuclease protection analysis only after 8 days exposure time, compared with the 36-h exposure time shown in Fig. 2C, lane 3. The specificity of both signals in the RT-PCR was confirmed by Southern blot analysis with a human L-APP cDNA probe comprising exons 14-17 (Fig. 4B). Both PCR products which were seen in leukocytes were APP-specific, since they gave signals after stringent wash conditions (Fig. 4B, lanes [18][19][20][21][22][23]. We refer to this new splice form as L-APP, since all leukocytes examined showed L-APP mRNA expression. Lymphatic organs, such as spleen and lymph nodes were also positive for L-APP mRNA (data not shown). It can be assumed that this is probably due to the leukocytes residing in these organs, rather than an expression of the lymphatic organ by itself. L-APP mRNA is expressed in both subtypes of CD4+ and CD8+ T-lymphocytes (Fig. 4A). In regard to Alzheimer's disease, there seems to be a reduction in the relative expression level of L-APP mRNA in the two cases examined (Fig. 4A, lanes 19 and 20). However this is far from conclusive and remains to be determined in a further more detailed statistically relevant analysis.
Size of Novel APP Splicing Form-Analysis of leukocyte RNA already showed that L-APP is an amyloidogenic (containing the PA4 sequence) transmembrane PA4 precursor protein isoform ranging at least from exon 3 to exon 18. In order to determine the full length of the corresponding novel APP mRNA, Northern blot hybridizations were performed (30). Total RNA from activated human T-lymphocytes, the CD4+, CD8+ subtypes, and rat microglial cells were separated by agarose gel electrophoresis and blotted to a nylon filter. The filter was hybridized with a 1.05-kb EcoRI restriction fragment, comprising the 3'-part of the APP cDNA starting within exon 16. The hematopoietic cells showed a strong signal of APP mRNA in the range of 3.4 kb (Fig. 5, A and B). HeLa cells showed a roughly three times stronger expression of a similar 3.4-kb APP mRNA (Fig. 5A).
Thus, L-APP mRNAs in hematopoietic cells are approximately the same size as transmembrane APP mRNAs and are distinguished only by the omission of exon 15. L-APP mRNA generates transmembrane APP isoforms, which are expected to be 18 amino acids shorter (54 bp) than the previously described Alternative Splicing of L-APP in Leukocytes-In order to determine the relative amounts of L-APP mRNA versus total coding APP mRNA in resting lymphocytes and lymphocytes stimulated with PHA, we treated cells with PHA for various times, isolated RNA, and conducted S1 nuclease protection analysis with probe 2 (Fig. 6). A change in the ratio of the two splice isoforms can be observed. Resting human T-lymphocytes showed approximately 40% (+5%; n = 2) L-APP mRNA compared with about 20% (*lo%; n = 3) L-APP mRNA in short term stimulated human T-lymphocytes. After several days of PHA treatment, the relative amounts of L-APP increased again (Fig. 6, lanes 6 and 7). Thus we observed a regulation of alternative splicing of L-APP during activation of T-lymphocytes in vitro. We hypothesize that the regulation of L-APP might therefore be important in the functional transition of leukocytes during their stimulation through inflammatory agents in vivo. Functional Implications-We were able to demonstrate for the first time APP mRNA expression in the microglial cells. Microglial cells are of mononuclear phagocyte lineage and are thought to migrate into the brain during development of the  (37). They share a number of surface antigens with peripheral macrophages and are thus clearly distinguished from all other glial cells as a class of potential immunoeffector cells in the brain. They have therefore been referred to as the "resident macrophages of the brain" (37). Activation of microglial cells during neuro-degeneration in the brain results in active production of APP as determined by APP immunoreactivity and immunoprecipitations of APP? The precise role and fate of this new contributor of APP to the APP pool in brain has to be shown in further studies. Microglial cells are characterized by a high respiratory burst activity potential (28). In the light of the proposed role of redox potential as one possible factor for amyloidogenesis, microglial cells are an attractive brain cell to study APP synthesis and degradation pathways in more detail.
L-APPs are the first splice products of the APP gene for which a highly restricted expression pattern has been established. L-APP is generated by exclusion of a short exon in a region not previously thought to be involved in alternative splicing. L-APP mRNA represents a substantial fraction of total APP mRNA in leukocytes and microglial cells and is readily detected with S1 nuclease protection analysis. It has still to be determined if the new splice product contains the KPI and OX2 exons as would be expected in leukocytes (see Fig. 2A). Alternative splicing of APP could then give rise to at least three L-APPs (Fig. 7). Expression of L-APP and the ubiquitous APP forms APP695, APP751, and APP770 appears to be correlated with the functional state of leukocytes. L-APP is high in nonadherent and low in adherent leukocytes.
The presence of a novel APP isoform (L-APP) in lymphocytes or cells of the mononuclear phagocyte system such as macrophages or microglial cells might be necessary to allow the rapid transistion of the dual functions of nonadherence and adherence in the immune system (38). This is in accordance with the proposed role of APPs in cell adhesion and attachment in the nervous system and in the periphery (39).
Exclusion of exon 15 changes the molecular conformation of APP at a site that might be critical for recognition and binding of APP processing and possibly BA4-amyloid gener-R. Banati, J. Gehrmann, U. Monning, C. Czech, G. Konig, K. Bayreuther, and G. W. Kreutzberg, manuscript in preparation. ating proteases. Future protein expression of L-APP and sequencing of a proteolytic breakdown products will determine if the physiological cleavage of APP, which occurs inside the amyloidogenic region and prevents amyloid formation, is conserved in L-APP. BA4 deposits in AD are infiltrated by microglial cells, which are, however, absent in early lesions (40,41). Whether the microglial cells in brain are a substrate for a potentially amyloidogenic processing of APP isoforms has yet to be determined. The investigation of L-APP expressed in situ will help to follow the fate of microglial APP/ L-APP and to clarify the suggested pathogenetic role of microglial cells in AD.