The Neurotoxic Carboxy-Terminal Fragment of the Alzheimer Amyloid

The Journal of Neuroscience, May 1992, Q(5): 1679-1667

The Neurotoxic Carboxy-Terminal Fragment of the Alzheimer Amyloid Precursor Binds Specifically to a Neuronal Cell Surface Molecule: pH Dependence of the Neurotoxicity and the Binding

Michael R. Kozlowski,l Athena Spanoyannis,’ Susan P. Manly,’ Seth A. Fidel,2 and Rachael L. Neve2 ‘Bristol-Myers Squibb Pharmaceutical Research Institute, Waliingford, Connecticut 06492 and 2Department of Psychobiology, University of California, Irvine, California 92717

One of the hallmarks of Alzheimer’s disease neurodegen- precursor protein (PAPP) is an integral membrane protein with eration is the accumulation of deposits of amyloid in neuritic the P/A4 peptide spanningthe border between the extracellular plaques and in the cerebral vasculature. Recent studies have domain and the transmembraneregion. ,f3APPin cultured cells implicated carboxy-terminal fragments of the Alzheimer am- is rapidly processedfrom a membrane-bound form to a secreted yloid precursor protein (BAPP) in the processes of amyloi- form (Weidemann et al., 1989); pulse-chasestudies show that dogenesis and neurodegeneration. In particular, the car- the secretion of /3APP leaves behind a stable 11.5 kDa fragment boxy-terminal 104 amino acids of BAPP (@APP-C104) have consisting of the transmembrane region and the cytoplasmic been shown to cause amyloid-like fibrils when expressed in domain (Oltersdorf et al., 1990). This normal processingof non-neuronal cells and to cause the degeneration of neu- @APP is a membrane-associatedevent that occurs within the ronal cells. These data suggest that it may play a role in the extracellular domain of PIA4 (Esch et al., 1990; Sisodia et al., development of the progressive neuropathology of Alzhei- 1990). Hence, the generation of ,&/A4 in Alzheimer’s disease mer’s disease. We hypothesized that @APP-Cl04 may cause appears to be the result of abnormal cleavage. However, the the degeneration of neurons by interacting with a cell surface molecular mechanismsof amyloidogenesisand the role of amy- receptor. In the present report, we show that /lAPP-Cl04 loidogenesis in the development of Alzheimer’s diseasepa- synthesized in vitro binds specifically and with high affinity thology remain unclear. to the surface of NGF-treated PC1 2 cells. Both the cell sur- Recent evidence has implicated the 100-104 amino acid car- face binding and the neurotoxicity of BAPP-Cl04 are pH boxy-terminal fragment of @APP in the processesof amyloi- dependent and are not inhibited by tachykinins. Mutational dogenesisand neurodegeneration.This fragment, which spans analysis suggests that both the binding and the neurotoxicity the @/A4 and cytoplasmic domains, has a tendency to self-ag- are dependent at least in part on the presence of a tyrosine gregate (Dyrks et al., 1988). Moreover, the expression of this residue that is a potential site of phosphorylation at the car- carboxy-terminal pAPP fragment in primate cellshas been shown boxy terminus of the fragment. to lead to the production of a 16 kDa protein that aggregates and accumulatesinto deposit-like structures (Wolf et al., 1990) The deposition of amyloid in neuritic plaques and along the and that results in the formation of amyloid-like fibrils (Ma- walls of the cerebral vasculature is a key feature of Alzheimer’s ruyama et al., 1990). diseasepathology (Terry et al., 1981; Glenner, 1983). The prin- We have shown that this samecarboxy-terminal @APP frag- cipal component of amyloid is a 4 kDa protein termed @IA4 ment is neurotoxic (Yankner et al., 1989). PC12 cells transfected (Glenner and Wong, 1984a,b; Masters et al., 1985) that is prob- with a retroviral recombinant expressingthe carboxy-terminal ably derived from one or more of four identified precursor pro- 104 amino acids of pAPP (formerly termed AB 1, Yankner et teins with predicted lengths of 695 (Kang et al., 1987) 751 al., 1989; now termed PAPP-C104) degeneratewhen induced (Ponte et al., 1988; Tanzi et al., 1988) 770 (Kitaguchi et al., to differentiate into neuronal cells with NGF. Moreover, con- 1988) and 714 amino acids (Golde et al., 1990) respectively. ditioned medium from thesecells is toxic to differentiated neu- DNA sequence(Kang et al., 1987) and biochemical (Dyrks roblastoma cells (Slave et al., 1989) and to neuronsbut not non- et al., 1988; Selkoeet al., 1988) studiessuggest that the amyloid neuronal cells in primary rat hippocampal cultures (Yankner et al., 1989). The neurotoxicity can be removed from the medium by immunoabsorption with an antibody to PAPP-Cl04 (Yank- Received Aug. 27, 1991; revised Nov. 20, 1991; accepted Nov. 26, 1991. ner et al., 1989), suggestingthat PAPP-Cl04 is secretedby the We are grateful to Drs. Paul Rosenberg, Rim Neve, and Kathryn Ivins for transfected cells and is neurotoxic. We have recently demon- extensive helpful discussions, Dr. William Mobley for the generous gift of &NGF, and Dr. Dale Schenk for the BX6 antibody. We are pleased to recognize the strated that brain transplants of cells expressingPAPP-Cl04 assistance of M. P. Rosser and A. Longden in culturing the cells used in the binding causespecific neuropathology in vivo (Neve et al., in press).We assay, and of E. Hall in assisting with these assays. In addition, the assistance of have also shown that transgenic mice expressingPAPP-Cl04 Dr. A. H. Balch in the statistical analysis of the cytotoxicity data was invaluable. This work was supported by California State Department of Health Services in the brain manifest intracellular and neuropil deposition of Contract 90-00198, a grant from Bristol-Myers Squibb, a Metropolitan Life Foun- @IA4 protein, as well as other abnormalities that resembleas- dation award, and NIH Grants HD18658, NS28406, and BRSG SO7 RR07008 (R.L.N.). pects of Alzheimer’s diseasepathology (Kammesheidt et al., Correspondence should be addressed to Rachael L. Neve at the above address. 1991; and A. Kammesheidt and R. L. Neve, unpublished ob- Copyright 0 1992 Society for Neuroscience 0270-6474/92/l 2 1679-09%05.00/O servations). 1680 Kozlowski et al. l pH Dependence of BAPP-Cl04 Neurotoxicity and Binding

- 21.5

1.0 10.0 21.5 CONCENTRATION OF @APP-Cl04 14.3 Figure 2. Representative curve showing the inhibition of ?S+‘APP- Cl04 binding to NGF-treated (5 d) PC12 cells by @APP-C104.

Figure 1. SDS-PAGE (A) and immunoblot (B) analysis of PAPP-C 104 transfected cells were regularly examined for the presence of the trans- made by in vitro transcription-translation. In B, the first lane shows fected RNA (neo in the case of the DO transfectants: BAPP-Cl04-neo reactivity ofantibody BX6 with @APP-ClO4, the second lane is a control fusion in the‘case of the BAPP-Cl04 transfectants; see Yankner et al., with no primary antibody. 1989). To assay the toxicity of conditioned media from NIH 3T3 transfectants, the growth medium of the transfectants (DMEM plus 5% bovine The neurotoxicity of this carboxy-terminal /3APP fragment calf serum; HyClone) was replaced with the PC 12 medium for 2 d when suggeststhat it may play a role not only in amyloidogenesisbut the 3T3 cell transfectants were at approximately 50% confluence. At also in the development of the progressiveneuropathology of the time that the conditioned media were harvested, RNA was isolated Alzheimer’s disease.We have therefore sought to determine the from the 3T3 cell transfectants and analyzed by Northern blots to confirm that all the transfectants were expressing equivalent amounts of mechanismby which PAPP-Cl04 causesthe degeneration of mRNA from the appropriate transfected recombinant. PC1 2 cells were neurons. We report here that PAPP-Cl04 synthesized in vitro seeded at 1O4 cells/well in polylysine-coated 24-well plates (Costar) and binds specifically and with high affinity to the surfaceof NGF- the conditioned media (diluted 1: 1 with PC1 2 medium) containina 50 treated PC12 cells. Both the cell surface binding and the neu- @ml &NGF were applied and replaced every other day. To test-the ability of tachykinins to inhibit BAPP-Cl04 toxicity, each tachykinin rotoxicity of/3APP-C 104are pH dependent.Mutational analysis analyzed was included in the conditioned medium at 20 PM, and re- suggeststhat both the binding and the neurotoxicity are depen- placed each time the conditioned medium was replaced. dent at least in part on the presenceof a tyrosine residue that Preparation andcharacterization of@APP-ClOlhnd 35S-fiAPP-C104. is a potential site of phosphorylation at the carboxy terminus The B&I-SmaI franment of the BAPP-695 cDNA (bo 1769-2959 ac- of the fragment. cording to the sequence of Kang et al., 1987), or of the appropriate mutant, was cloned into BamHI-SmaI-digested pGEM3 (Promega Bio- tee). The resulting recombinant was linearizied with ClaI (to make sense Materials and Methods BAPP-Cl04 RNA using SP6 RNA polymerase) or Sal1 (to make anti- Cell culture. PC1 2 cells for the binding assays were grown in RPM1 1640 sense pAPP-Cl04 RNA using T7 RNA polymerase) and used as tem- (GIBCO) with the addition of 10% heat-inactivated horse serum (Hana plate to make 5’-capped RNA in vitro using Promega Biotec reagents Biological) and 5% fetal bovine serum (J.R. Scientific). T-175 ‘flasks and protocol. RNA (20 Irg) was typically synthesized in a 50 ~1 reaction (Falcon) were seeded with 5 x 10’ cells/ml in 60 ml of growth medium using 5 fig of template. Two micrograms of this RNA were used in a containing 10 @ml NGF, An additional 60 ml of this medium were 50 ~1 wheat germ lysate in vitro translation reaction together with 3.8 added every 2 d until the cells were harvested (usually at 6 d). Where WCi of ?S-methionine (Amersham; 1200 Ci/mmol). The concentration indicated, PC12 cells were cultured as above except in the absence of of @APP-Cl04 in the lysates was determined by SDS-polyacrylamide NGF. Cells were collected bv centrifuaation at 138 x .e for 8 min. The gel electrophoresis of an aliquot of the in vitro translation product, supematant was discarded and the pellet resuspended in 5 ml of the followed by excision of the radiolabeled band from the dried and au- growth medium. The cells were triturated three times through a 23 gauge toradiographed gel, solubilization of the band in H,O,, and quantitation needle attached to a 10 cc syringe to provide a single cell suspension of the radioactive protein in the band. The concentration of BAPP-C 104 or, at most, small cell aggregates. Trypan blue (Polyscience Inc.) dye in the band was extrapolated from these data, using the assumption that exclusion indicated over 99% viable cells. during the in vitro synthesis of 35S-@APP-C104, 35S-methionine was PC1 2 cells for the toxicity assays were grown in the presence of 10% incorporated at all of the five methionine residue positions (specific CO, in Dulbecco’s modified Eagle’s medium (DMEM) (Sigma), 5% heat- activity, 11,000 Ci/mmol) in the polypeptide. The ?j-methionine was inactivated horse serum (HyClone), and 101 fetal bovine serum replaced with 80 PM methionine to make unlabeled BAPP-C104. A (HyClone) containing 50 @ml j3-NGF (gift of Dr. William Mobley). l/ 10 vol of 100 mM N-acetyl-D-glucosamine (Sigma) was added to each Cells were seeded at lo4 cells/well in polylysine-treated 24-well plates completed in vitro translation reaction to block the binding activity of (Costar). The medium was replaced every 2 d. When the pH dependence wheat germ agglutinin before the reaction was aliquotted and stored at of the toxicity was examined, a given pH was maintained by adding 25 - 80°C. The in vitro translation products were thawed only once for use mM HEPES to the medium and adjusting the medium to the desired in the binding assays. Multiple freeze-thaw cycles, which resulted in pH at 36°C (the temperature at which the cells were maintained). The aggregation of in vitro translated BAPP-Cl04 (S. A. Fidel and R. L. pH of the medium in the wells was rechecked during the course of the Neve, unpublished observations), also resulted in loss of binding activity experiment. Four or more replicates of each experiment were plated, of the ?S-BAPP-C104. ?S-PAPP-Cl04 was 90% bindable to NGF- one field in each well containing 100-300 cells on day 0 was defined treated PC12 cells in conditions of binding site excess. for cell counts. We returned to the same field for cell counts on sub- Receptor binding assay. For BAPP-C 104 binding experiments, PC1 2 sequent days. Viability of the NGF-treated cells was determined on the cells, grown and prepared as described above, were counted, collected basis of phase brightness of cell bodies, intactness of neurites (some by low-speed centrifugation (138 x g), and resuspended in PBS (pH cells became phase-dark due to spreading on the polylysine substrate 7.4) with 1% BSA and 1% glucose (PBG). In the experiments in which but extended neurites in response to NGF, these cells were counted as the pH dependence of the binding was tested, pH values for the PBS viable as long as they had neurites), and adherence to substrate. All the ranged from 7.4 to 8.0. A 0.08 ml volume of the cell suspension, con- The Journal of Neuroscience, May 1992, 12(5) 1691

& ?, loo- 4 /” 0.10 0 1 0 50 100 150 200 TIME (min.) Figure 3. 35S-~APP-C104 association with (open circles) and dissoci- ation from (solid circles) NGF-treated (5 dl PC 12 cells. Data shown are from one of two experiments in which‘similar results were obtained. A KD value was not determined with the kinetic data since a 0 time off rate could not be measured using the current binding protocol (see Materials and Methods).

taining 2 x lo6 cells, was mixed with 0.01 ml of 35S-@APP-C104 and 0.01 ml of PBG, or PBG containing unlabeled ,f3APP-C104. For kinetic 0-l : : : : : : : I and inhibition experiments, the concentration of YS-BAPP-Cl04 was 0.0 0.2 0.4 0.0 0.8 1.0 1.2 1.4 1.6 either 25 PM or 50 PM. For saturation experiments, the ligand concen- BO"ND (frnol/lO~ CSIS) tration ranged from 15 PM to 6 nM. The mixture was incubated at 4°C Figure 4. Representative saturation isotherm (top) and Scatchard plot for 3 hr, unless otherwise indicated. At the end of the incubation period, (bottom) of YS-pAPP-Cl04 binding to NGF-treated (6 d) PC12 cells. the cells were pelleted by centrifugation at 138 x g for 90 set, the The lines were computer generated by the LUNDON 1 program (see Ma- supematant (containing the unbound ligand) was removed, and 0.2 ml terials and Methods) and represent the best interpretation of the data. of fresh, ice-cold PBG was added. The cells and fresh buffer were briefly agitated, and again centrifuged and suspended in fresh, ice-cold buffer as above. Final&, the cells were pelleted by centrifugation, filtered through translated in vitro in a wheat germ lysate. SDS-PAGE (5-20% BSA (I%)-treated filters using a Tomtec (Orange, CT) 96-well harvester, gradient) of 3sS-labeled PAPP-Cl04 made in this in vitro tran- and washed with four 1 ml rinses of ice-cold PBS with 1% BSA. The duration of the period between resuspension of the cells in fresh, ice- scription-translation system revealed that the radiolabeled cold buffer and filtration of the cells was less than 10 min. In the dis- product migrated as an approximately 15 kDa band (Fig. IA). sociation experiments and to collect a sample of bound ligand, the cells To show that this band represented PAPP-C104, the in vitro were incubated in the buffer added as the second wash for up to 3 hr translation extract was analyzed by immunoblot with an anti- to allow the bound ligand to dissociate. The cells were then pelleted by centrifugation and the supematant (containing the ligand that had dis- body (termed anti-BX6; Benowitz et al., 1989; Oltersdorfet al., sociated) was removed. The cells were then rewashed and harvested as 1990) directed against a bacterial fusion protein expressing PAPP- described above. Radioactivity retained by the filters was measured by Cl04 (Fig. 1B). The anti-BX6 immunoreacted with an approx- scintillation counting using a Betaplate counter (LKB). Nonspecific imately 15 kDa band in an in vitro translation of the sense PAPP- binding was taken as that occurring in the presence of excess (3.2-18 Cl04 RNA (first lane); no reactivity was seen with secondary nM) PAPP-ClO4. Saturation data were analyzed using LUNDON 1 binding analysis software (Lundeen and Gordon, 1986). The binding model that antibody alone (second lane). best fitted the experimental data was determined by F test (p < 0.05). In vitro mutagenesis of @APP-004. In order to make mutants of Specific binding of /3APP-Cl04 to the surface of dlflerentiated PAPP-C104, we used oligonucleotide-directed in vitro mutagenesis as PC12 cells described by Kunkel(1985). To change serine 603 to an alanine, a 20mer representing the reverse complement of bases 1795-l 8 14 (according to Because PAPP-C 104 is toxic to NGF-differentiated PC1 2 cells, the sequence of Kang et al., 1987) that directs the alteration of the initial we tested the binding of the in vitro synthesized PAPP-Cl04 T of the serine codon to a G was used. To change tyrosine 687 to a fragment to the surface of PC 12 cells that had been treated with phenylalanine, we constructed a 19mer representing the reverse com- NGF. Y+pAPP-Cl04 binding to the differentiated PC12 cells plement of bases 2053-207 1 that directs the appropriate A to T alter- was inhibitable by unlabeled PAPP-Cl04 (Fig. 2). The inhib- ation. The EcoRI-ClaI subfragment of @APP-C 104 was s&cloned into the phagescript vector SK M 13 (Stratagene) and introduced into the itable fraction of the binding accounted for 40-60% of the total Escherichia coli dut, ung strain CJ236. The uracil-containing phage DNA binding. The IC,, value for the inhibitable binding was 1.7 + produced in this strain was used as template in the in vitro mutagenesis 0.7 nM (n = 5). Inhibitable binding reached a maximum after reaction, which was transformed into E. coli MVll90 for selection 3 hr of incubation and was completely dissociable (Fig. 3). Sat- against the parental strand. The mutated EcoRI-ClaI subfragment was used to replace the wild-type fragment in the DO recombinant, after uration experiments indicated the presence of a single class of which presence of the desired mutation was confirmed by sequence binding sites with a KD value of 0.8 1 f 0.37 nt.4, in approximate analysis. agreement with the IC,, value determined above, and a B,,,,, value of 0.37 f 0.05 fmol/106 cells (n = 3; Fig. 4). Results Binding was not significantly inhibited by other peptides, in- Synthesis and analysis of PAPP-Cl04 cluding a number of tachykinins (Table 1). Non-sense 35S-la- The portion of the /3APP cDNA encoding BAPP-Cl04 was cloned beled material synthesized from antisense PAPP-C 104 RNA did into the plasmid vector pGEM3, in which the cloning site is not bind to NGF-treated PC 12 cells. flanked by binding sites for SP6 and T7 RNA polymerases. The To assess the stability of 35S-PAPP-Cl04 in the assay con- sense (or antisense control) transcript was synthesized and then ditions, samples of both free ligand at the end of the incubation 1682 Kozlowski et al. l pH Dependence of BAPP-Cl04 Neurotoxicity and Binding

Table 1. Inhibition of %+3APP-C104 binding by various agents

Inhibitor G @APP-Cl04 fragments ,8APP-Cl04 1.7 nM j3APP-C 104 MutY Inactive at 14 nhi Tachykinins SubstanceP Inactive at 1 PM Eledoisin Y Physalaemin ” Kassinin I, Neurokinin A I, Neurokinin B I Other peptides ACTH Inactive at 10 PM Bradykinin 0 Endothelin- 1 ,, Neurotensin Oxytocin I, Pentagastrin ” Vasoactiveintestinal peptide Neuropeptide Y c Somatostatin ” II

Fibroglast growth factor ,, Secretin v Leupeptin ,, Helodermin I, Peptide YY Y Figure 5. Characterization of %-@APP-Cl04 by SDS-PAGE analysis. All substances were tested in triplicate. The left lane shows the ligand at the end of the binding assay. The middlelane gives the initial profile of the ligand. The right lane shows the profile of the ligand that had dissociated from binding to the receptor (see Materials and Methods).The last two lanes contain 14C-l&&d commercialmolecular weight markers(Amersham). The molecular lecular massof 15 kDa (Fig. 5). This representsmonomer BAPP- massesof the markersare also shown. C104. A small amount of material was also present in two additional diffuse bands. One migrated at approximately twice the molecular weight of the major band and possibly represents period (unbound) and ligand that had bound to the cells and a dimer of BAPP-C104. The other band had a much higher been released(see Materials and Methods) were examined by molecuiar weight and may be an aggregate.The gel profile of SDS-PAGE. Before addition to the assay,most of the radio- the unbound ligand at the end of the incubation period was activity was contained in a single band with an apparent mo- identical to that of the ligand before incubation, indicating that the ligand had not been degraded.Similarly, the ligand released from binding had the sameprofile asthe ligand originally added, with the possibleexception of a slight reduction in the amount of high molecular weight material. These data show that the binding of PAPP-Cl04 ligand does not causeits modification. Inhibition of binding of 35S-pAPP-C104to NGF-treated PC12 cells by conditioned mediumfrom @APP-Cl04-transfected PC12 cells Having identified a binding site for %-PAPP-Cl04 on the surface of NGF-treated PC12 cells, we then carried out a seriesof experiments to ascertain whether the binding of PAPP-C 104 to thesesites could account for its neurotoxicity. Becausethe de- generationcaused by PAPP-C 104occurs gradually over a period DAYS OF EXPOSURE TO NGF of 11 d, it wasnot feasibleto synthesizethe amount of in vitro- Figure 6. Increasein 35S-flAPP-C104binding to NGF-treatedPC12 translated ,f3APP-C104 necessaryto maintain a reasonablecon- cells as a function of the duration of the exposure to NGF. Values are centration of the peptide in the cultures for the full courseof mean f SEM oftwo to four experiments. Values from some experiments run at lower ligand concentrations were normalized to the standard each experiment. Hence, we carried out, in parallel with the concentration(100 PM). Time 0 representscells that werenot exposed binding studies, cell culture studies utilizing PC12 cells stably to NGF. transfectedwith a retroviral recombinant expressingpAPP-C 104 The Journal of Neuroscience, May 1992. 72(5) 1683

-1 0 1 2 3 4 5 6 7 8 9 10 11 12 DAYS OF EXPOSURE TO NGF

DO - PC12 p APP - Cl04 - PC12#1

Figure 7. A, NGF-induced cell loss in two independently isolated @APP-C 104 PC1 2 cells transfectants, # 1 (opensymbols) and #2 (solidsymbols). The cells were cultured in serum-containing medium (circles)or serum-free medium (triangles).Values are mean f SEM (n = 4) cell number as a Percentage of control (DO-PC12 cells). Control cell number increased from 256 + 7 on day 0 to either 681 f 3 (serum-containing medium) or 429 + 18 (serum-free medium) on day 11. In all cases, the decline in cell number in the @APP-Cl04 transfected cell lines was statistically significant (t test, p < 0.05). B, Depiction of control (DO-PCl2) and BAPP-Cl04 PC12 #l cells after 6 d of treatment with NGF, showing cell loss and retraction of neurites in the BAPP-Cl04 transfected PC12 cells.

(Yankner et al., 1989) or using conditioned medium from cells expressing PAPP-C 104 (seeMaterials and Methods). Dependenceof both binding and neurotoxicity on exposureof We first showed that binding of 35S-/3APP-C104to NGF- cells to NGF treated PC12 cells was inhibited by conditioned medium from We had previously reported (Yankner et al., 1989) that PC12 two different isolatesof PAPP-C104-transfectedPC1 2 cells,with cells stably transfected with a retroviral recombinant expressing half-maximal displacementof inhibitable binding occurring at PAPP-Cl04 degeneratewhen causedto differentiate into neu- dilutions of 1: 10 to 1:32. Medium from cells transfected with ronal cells with NGF. Thus, the neurotoxicity appearsonly after the vector alone did not inhibit binding. treatment of the cells with NGF. We therefore tested the effect 1664 Kozlowski et al. l pH Dependence of BAPP-Cl04 Neurotoxicity and Binding

appearance of vacuolar inclusions in the cell body, retraction of neurites, and loss of attachment of the cells to the substrate. -- Q In many cases(see Fig. 9B), degeneratingcells clumped together

gf 70 -. /\ % before floating into the medium; extensive fasciculation of pro- cesseswas also evident among the remaining population of cells 2 so-- I that did not show retraction of net&es. 4 o‘o\ OYA 0 Dependenceof both binding and neurotoxicity on pH w 0 A secondvariable affecting binding was pH. A marked effect of lo- “0 30 -_ 0 A-O- d pH on binding was observed between pH 7.7 and 7.9 (Fig. 8A). 0 0 -104 I Between pH 7.4 and 7.6, binding was relatively constant, or 7.3 7.4 7.5 7.6 7.7 7.0 7.9 0.0 0.1 slightly declining. At approximately pH 7.7, there was a sharp A PH increase in binding followed by a total loss at pH 7.8 to 7.9. Inhibitable binding was again seenat pH 8.0. Becausethe 35S-PAPP-C104binding to NGF-treated PC12 cells was pH dependent, we examined the pH dependenceof NGF-induced degenerationof BAPP-C 104-transfectedcells. Our initial studies of neurotoxicity in these cells had been carried with the medium at pH 7.3 (seeabove). We then compared the PAPP-Cl04 neurotoxicity in medium at pH 7.2 with that in medium at pH 8.2 (Fig. 9). Becausewe had noted that cell loss was more rapid in serum-containing than in serum-free medium, we choseto carry out theseand subsequentexperiments in the presenceof serum. PAPP-Cl04 transfectants displayed no detectable degeneration at pH 8.2 even after 9 d of treatment with NGF, whereasneurodegeneration of the transfectants was B pH OF MEDIUM apparent by day 6 of NGF treatment at pH 7.2. BecauseHEPES was usedto maintain the 8.2 pH, we included a HEPES control Figure 8. A, Dependenceof J.?+@APP-Cl04binding to NGF-treated (6 d) PC12 cells on pH. Resultsfrom three experimentsare shown at pH 7.2 to show that HEPES alone is not sufficient to inhibit separately(dlfirent symbols).pH valueswere adjusted by 0.1 pH unit PAPP-C 104 neurotoxicity. or lessto bringthe curvesinto register.Maximal bindingwas 498 dpm To examine in greater detail the pH dependence of PAPP- in two experimentsand 110dpm in the third (triangles).B, Effect of C 104 toxicity, we exposed control and experimental transfected pH on NGF-inducedcytotoxicity in BAPP-Cl04 PC12 #l cells(circles) PC12 cell lines to NGF in media at pH ranging from pH 7.2 and BAPP-Cl04 PC12 #2 cells(squares) cells. Values are as in Figure 7. Cell numberis expressedas a percentageof control (DO-PC12)cell (medium at pH lower than 7.2 was slightly toxic to control cells) numberafter 11d culturingin the presenceof NGF. Significanttoxicity to pH 8.2 in 0.1 pH unit increments (Fig. SB). It is clear that wasfound at pH valuesof 7.7 or less(isotonic regression, p < 0.05), the neurotoxicity of PAPP-Cl04 is dependent upon pH and is but not at pH valuesof 7.8 or greater.Control cell numberon day 11 almost completely inhibited at pH 7.8 or above. Notably, bind- (603 + 14)did not vary with pH. ing of BAPP-C 104 to NGF-treated PC 12 cells is virtually lost at pH 7.8 and 7.9 (Fig. 8A).

of NGF on ,f3APP-Cl04 binding to PC12 cells, and found that Lack of inhibition of @APP-Cl04 binding and neurotoxicity by the amount of inhibitable 35S-j3APP-C104 binding was depen- tachykinins dent upon the duration of exposure of the PC12 cells to NGF. As stated above, binding of BAPP-C 104 to NGF-treated PC 12 PC12 cells that were not exposed to NGF showed little inhib- cells was not significantly inhibited by a number of tachykinins itable binding, whereasthose cultured in the presenceof NGF (Table 1). We tested the ability of the tachykinins listed in Table for 4 or 6 d showed progressively greater amounts of binding 1 (with the exception of neurokinin A) to inhibit @APP-Cl04 (Fig. 6). neurotoxicity. We had previously transfected NIH 3T3 cells To confirm and extend our earlier results (Yankner et al., with the samePAPP-C 104 retroviral recombinants and control 1989) demonstrating the neurotoxicity of /?APP-Cl04 in cul- vector that we usedto transfect the PC 12 cells (Yankner et al., ture, we evaluated the effects of NGF on PAPP-ClO‘Ltrans- 1989). We had shown that conditioned medium from the PAPP- fected cells in serum-containing and serum-free medium. Two C 104 transfected 3T3 cells, but not from the vector-transfected independently isolated PAPP-C 104 PC 12 cell transfectants 3T3 cells, was toxic to NGF-differentiated PC12 cells. To ex- (termed ABl-PC12 #l and #2 in Yankner et al., 1989) and a amine the effect on ,&APP-C104 neurotoxicity ofthe tachykinins, control vector-transfected PC 12 cell line (termed DO-PC 12 and we added pAPP-C 104 NIH 3T3-conditioned medium contain- describedin Yankner et al., 1989) were treated with NGF in ing NGF, or control conditioned medium containing NGF to- serum-containing medium (seeMaterials and Methods) or in gether with 20 PM of each tachykinin (as describedin Yankner serum-freemedium (Ex-Cell30 1, JRH Biosciences),both at pH et al., 1990), to PC12 cells. The cells exposed to PAPP-C104- 7.3. The results, shown in Figure 7, show that PAPP-C104- conditioned medium, but not the cells exposed to control transfected PC 12 cells, but not DO-transfected PC 12 cells, grad- conditioned medium, degenerated over a period of 8 d; the ually degeneratein the presenceof NGF. The degeneration of degeneration was not inhibited by the presenceof any of the the /3APP-C 104-transfectedPC 12 cells wascharacterized by the tachykinins tested (n = 12 for each condition; data not shown). The Journal of Neuroscience, May 1992, C(5) 1595

DAYS OF EXPOSURE TO NGF

l3APP - Cl04 - PC12#1pH 7.3 fi APP - Cl04 - PC12#2pH 7.3 fi APP - Cl04 - PC12#1pH 8.2 B Figure 9. A, Comparisonof NGF-inducedtoxicity in @APP-Cl04transfected cells grown at pH 7.2 versuspH 8.2. BAPP-Cl04PC12 #l cells werecultured in the presenceof NGF at eitherpH 7.2 in phosphate-bufferedmedium (circles) or HEPES-bufferedmedium (squares), or at pH 8.2 in HEPES-bufferedmedium (triangles). Valuesare as in Figure 8. Therewas a significantdecline in cell numberwith time in both buffersat pH 7.2 (t test,p < 0.05),but no changeat pH 8.2. Similarresults were obtained with PAPP-Cl04 PC12#2 cells.Control cell (DO-PC12)numbers increasedfrom 263 + 22 (phosphate-bufferedmedium), 156 ? 20 (HEPES,pH 7.2) and269 + 12(HEPES, pH 8.2) on day 0 to 353+ 8, 221 2 55, and 821 +- 115,respectively, on day 9. B, Depictionof @APP-Cl04PC12 #l and #2 at pH 7.3 (HEPES),and of @APP-Cl04PC12 #l at pH 8.2 (HEPES)after 9 d of exposureto NGF.

mRNA levels in nonmutated transfectants were isolated and Mutational analysis of PAPP-Cl04 neurotoxicity and binding their conditioned media assayedfor neurotoxicity by adding the By comparing the sequenceof /3APP-Cl04 with that of other media to NGF-treated PC 12cells (seeMaterials and Methods). moleculeswith known functions, we identified at leasttwo sites The conditioned media from the BAPP-Cl04 MutS 3T3 cells within @APP-C104 that may play a role in its binding and neu- wasas toxic to NGF-treated PC 12 cells aswas the BAPP-C 104- rotoxicity. The amino terminus of BAPP-Cl04 contains the conditioned medium (n = 8 for eachcondition; data not shown). ttipeptide aspartic acid-setine+lycine (residues 603-605 of Thus, the replacementof serine 604 with an alanine in PAPP- @APP-695),which is conserved in the active site of serinepro- Cl04 has no effect on its toxicity. teases(Hartley, 1970). Roberts (1986) has suggestedthat @/A4 The carboxy terminus of /3APP-C104 contains a tyrosine (687 or a similar @APP fragment may have serineprotease activity in PAPP-695) that lies within a sequencehomologous to that and causeneurodegeneration by virtue of suchactivity. If @APP- surrounding a phosphorylated tyrosine in integrins and some Cl04 has serineprotease activity, and if this activity is impli- plasmamembrane receptors (reviewed by Tamkun et al., 1986). cated in its neurotoxicity, it should be possibleto neutralize its This homology suggeststhat tyrosine 687 in BAPP-C 104 might neurotoxicity by removing the serine residue in the putative be important to its function. We usedsite-directed mutagenesis serineprotease active site. We usedoligonucleotide-directed in to replace the tyrosine with a phenylalanine, and selectedtwo vitro mutagenesisto changethe serineat residue604 (usingthe stable NIH 3T3 cell transfectants (termed BAPP-Cl04 MutY) numbering system of Rang et al., 1987, for PAPP-695) to an expressingthe mutant mRNA at appropiate levels. The con- alanine. The mutant was termed PAPP-Cl04 MutS. Two in- ditioned medium of thesetransfectants was not toxic to NGF- dependentstable NIH 3T3 cell transfectantsexpressing this mu- treated cells, although conditioned medium of a nonmutated tant mRNA at levelsequivalent to or exceedingthe PAPP-Cl04 PAPP-Cl04 3T3 cell transfectant tested at the same time was 1666 Kozlowski et al. * pH Dependence of /3APP-Cl04 Neurotoxicity and Binding

of small aggregatesof ligand at each site. If the extent of aggregation increaseswith increasedpH, the binding domainsof the ligand may becomeinaccessible, resulting in decreasedreceptor occupancy. Finally, when the aggregatesbecome large enough, the effect of low receptor occupancy would be offset by the binding of large aggregatesof the ligand at each site, resulting in the return of measurableamounts of binding. It is surprising that the binding site for PAPP-Cl04 doesnot recognize tachykinins. P/A4 (the amyloid encoding sequenceof PAPP) has also been reported to be neurotoxic and to have its 04 : : : : : : : : : : I toxicity reversed by various tachykinin peptides (Yankner et -1 0 1 2 3 4 5 6 7 8 9 10 al., 1990). BecausePIA4 is contained within the ,L?APP-Cl04 DAYS OF EXPOSURE domain, it seemsreasonable to expect that both peptideswould have a similar site of action and therefore would be inhibited Figure 10. Comparisonof the cytotoxicity to NGF-treatedPC12 cells by similar agents. The present data do not support this idea. of conditionedmedium from NIH 3T3 cells transfectedwith either Tachykinins inhibit neither PAPP-Cl04 binding nor its neu- @APP-Cl04(circles) or @APP-Cl04MutY (squares).Values are as in Figure8. Control(DO-PC1 2) cellnumber increased from 158+ 14on rotoxicity to NGF-treated PC 12 cells. day0 to 236+- 16on day 9. Conditionedmedium from a secondBAPP- Mutational analysisof PAPP-Cl04 neurotoxicity and binding Cl04 MutY NIH 3T3 transfectantgave similar results (data not shown). indicates that alteration of a serinein a putative serineprotease active site within the P/A4 coding sequence,at the amino terminus of pAPP-C104, does not affect its neurotoxicity. How- toxic to the NGF-treated cells(Fig. 10). In accordancewith these ever, alteration of a tyrosine located nine residues from the carboxy terminus of PAPP-C 104 (tyrosine 687) abolishesboth data, @APP-C104 MutY madein an in vitro transcription-trans- the neurotoxicity of the molecule and also its ability to inhibit lation systemdid not significantly inhibit ?S-PAPP-Cl04 bind- binding of ?S-PAPP-Cl04 to NGF-treated PC12 cells. This ing to NGF-treated PC 12 cells at concentrations up to 14 nM (n = 3; Table l), indicating that this mutant is at least 2.5 log tyrosine is contained within a sequenceof amino acids homol- ogousto those surrounding phosphotyrosinesin integrins and units lesspotent in its binding affinity than is PAPP-C104. some growth factor receptors, suggestingthat phosphorylation Discussion of tyrosine 687 may be a requirement for the binding and neurotoxic activities of PAPP-C 104. The present results show that the 35S-fiAPP-C104binding site Degeneration of PC12 cells exposedto PAPP-Cl04 is char- expressedby NGF-treated PC 12 cells has several receptorlike properties. First, it binds the ligand reversibly, as shown in the acterized by the appearanceof vacuolar inclusions in the cell body, accompaniedby retraction of neurites and lossof adher- dissociationexperiments. Second, it specifically recognizesthe enceof cells to the substrate.The cells that remain tend to clump ligand in that binding is inhibited by unlabeledPAPP-Cl04 but together, and to display fasciculation of processes.Primary rat not inhibited by other peptides, including tachykinins. Fur- cortical or hippocampal neuronstreated with @APP-ClOkon- thermore, a small modification of @APP-C104 (the mutation of ditioned medium exhibit the same pattern of degeneration(A. tyrosine 687) abolishesbinding activity. Third, PAPP-Cl04 is Spanoyannisand R. L. Neve, unpublished observations).These not altered after binding as shown by SDS-PAGE analysis of characteristics suggestthat PAPP-Cl04 may causecell lossby the ligand, suggestingthat the binding site is not an enzyme or, at least,not an enzyme that metabolizesBAPP-C 104under these virtue of its interference with cell-substrateinteractions during processoutgrowth. Isolation and identification of the cell surface assayconditions. Finally, it has high (nanomolar) affinity and receptor to which PAPP-Cl04 binds will aid us in defining the low capacity, as would be expected for a receptor. mechanismby which this molecule causesneurodegeneration. Three of the present results provide circumstantial evidence that this binding site mediates the toxic effects of PAPP-Cl04 References on differentiated PC12 cells. First, the binding site is much more BenowitzLI, RodriguezW, PaskevichP, Mufson EJ, Schenk D, Neve prevalent on PC12 cells rendered susceptibleto the toxic effects RL (1989) The amyloid precursor protein is concentrated in neu- of PAPP-C 104by treatment with NGF than on nontreated cells, ronal lysosomes in normal and Alzheimer disease subjects. Exp Neu- to which the peptide is not toxic (Yankner et al., 1989). The rol 106:237-250. increasein PAPP-C 104binding following NGF treatment is not Dyrks T, Weidemann A, Multhaup G, Salbaum JM, Lemaire H-G, Kang J, Mtiller-Hill B, Masters CL, Beyreuther K (1988) Identifi- due to the increasedcell surface area that accompaniesdiffer- cation, transmembrane orientation and biogenesis of the amyloid A4 entiation since, in the present study, the cells are not attached precursor of Alzheimer’s disease. EMBO J 7:949-957. and therefore do not extend neurites. Second, a loss of both Esch FS, Keim PS, Beattie EC, Blather RW, Culwell AR, OltersdorfT, binding and toxicity occursnear pH 7.8. Third, a mutation that McClure D, Ward PJ (1990) Cleavage of amyloid @ peptide during constitutive processing of its precursor. Science 248: 1122-l 124. eliminatesthe neurotoxic efficacy of PAPP-Cl04 (tyrosine 687) Glenner GG ( 1983) Alzheimer’s disease: the commonest form of amy- also abolishes its ability to bind to the site identified in this loidosis. Arch Path01 Lab Med 107:281-282. study. Glenner GG, Wong CW (1984a) Alzheimer’s disease: initial report of The relationship between binding and pH may be due to a the purification and characterization of a novel cerebrovascular am- pH-related aggregation of the ligand. The increase in ,f3APP- yloid protein. Biochem Biophys Res Commun 120:885-890. Glenner GG, Wong CW (1984b) Alzheimer’s disease and Down’s Cl 04 binding to NGF-treated PC 12 cells at pH 7.7 could result syndrome: sharing of a unique cerebrovascular amyloid protein. Bio- from low-order aggregation of the ligand, which would have them Biophys Res Commun 122:1131-l 135. little effect on receptor occupancybut would result in the binding Golde TE, Estus S, Usiak M, Younkin LH, Younkin SG (1990) Ex- The Journal of Neuroscience, May 1992, f2(5) 1687

pression of j3 amyloid protein precursor mRNAs: recognition of a Roberts E (1986) Alzheimer’s disease may begin in the nose and may novel alternatively spliced form and quantitation in Alzheimer’s dis- be caused by aluminosilicates. Neurobiol Aging 7:561-567. ease using PCR. Neuron 4~253-261. Selkoe DJ, Podlisny MB, Joachim CL, Vickers EA, Lee G, Fritz LC, Hartley BS (1970) Homologies in serine proteases. Philos Trans R Sot OltersdorfT (1988) @-Amyloid precursor protein of Alzheimer dis- Lond ]Biol] 257:77-86. ease occurs as 110-l 35-kilodalton membrane-associated proteins in Kammesheidt A, Hohamnn CF, Brandt P, Ivins KJ, Spanoyannis A, neural and non-neural tissues. Proc Nat1 Acad Sci USA 85:7341- Neve RL (199 1) Mechanisms of neurotoxicity of carboxyterminal 7345. fragments of the Alzheimer amyloid precursor protein. Sot Neurosci Sisodia SS, Koo EH, Beyreuther K, Unterbeck A, Price DL (1990) Abstr 17: 1447. Evidence that @-amyloid protein in Alzheimer’s disease is not derived Kang J, Lemaire H-G, Unterbeck A, Salbaum MJ, Masters CL, Grze- by normal processing. Science 248:492-495. schik KH, Multhaup G, Beyreuther K, Muller-Hill B (1987) The Slave I, Dawes LR, Neve RL, Korf B (1989) A fragment of the Alz- precursor of Alzheimer’s disease amyloid A4 protein resembles a cell heimer amyloid precursor is toxic to differentiated neuroblastoma surface receptor. Nature 325~133-736. cells. Sot Neurosci Abstr 15: 1375. Kitaguchi N, Takahashi Y, Tokushima Y, Shiojiri S, Ito H (1988) Tamkun JW, DeSimone DW, Fonda D, Pate1 RS, Buck C, Horwitz AF, Novel precursor of Alzheimer’s disease amyloid protein shows pro- Hynes RO (1986) Structure of integrin, a glycoprotein involved in tease inhibitory activity. Nature 331530-532. the transmembrane linkage between fibronectin and actin. Cell 46: Kunkel TA (1985) Rapid and efficient site-specific mutagenesis with- 271-282. out phenotypic selection. Proc Natl Acad Sci USA 82:488492. Tanzi RE, McClatchey AI, Lamperti ED, Villa-Komaroff L, Gusella JF, Lundeen JE, Gordon JH (1986) Computer analysis of binding data. Neve RL (1988) Protease inhibitory domain encoded by an amyloid In: Receptor binding in drug research (O’Brien R, ed), pp 3 149. New protein precursor mRNA associated with Alzheimer’s disease. Nature York: Dekker. 331:528-530. Maruyama K, Terakado K, Usami M, Yoshikawa K (1990) Formation Terry RD, Peck A, DeTeresa R, Schechter R, Horoupian DS (198 1) of amyloid-like fibrils in COS cells overexpressing part of the Alz- Some morphometric aspects of the brain in senile dementia of the heimer amyloid protein precursor. Nature 347:566-569. Alzheimer type. Ann Neural 10: 184-192. Masters CL, Simms G, Weinman NA, Multhaup G, McDonald BL, Weidemann A, Konig G, Bunke D, Fischer P, Salbaum JM, Masters Beyreuther K (1985) Amyloid plaque core protein in Alzheimer CL, Beyreuther K (1989) Identification, biogenesis, and localization disease and Down syndrome. Proc Nat1 Acad Sci USA 82:42454249. of precursors of Alzheimer’s disease A4 amyloid protein. Cell 57: Neve RL, Kammesheidt A, Hohmann CF (in press) Brain transplants 115-126. of cells expressing the carboxyterminal fragment of the Alzheimer Wolf D, Quon D, Wang Y, Cordell B (1990) Identification and char- amyloid precursor cause specific neuropathology in viva Proc Nat1 acterization of C-terminal fragments of the @-amyloid precursor pro- Acad Sci USA, in press. duced in cell culture. EMBO J 9:2079-2084. OltersdorfT, Ward PJ, Henriksson T, Beattie EC, Neve R, Lieberburg Yankner BA, Dawes LR, Fisher S, Villa-Komaroff L, Oster-Granite I, Fritz LC (1990) The Alzheimer amyloid precursor protein: iden- ML, Neve RL (1989) Neurotoxicity of a fragment of the amyloid tification of a stable intermediate in the biosynthetic/degradative precursor associated with Alzheimer’s disease. Science 245:4 17420. pathway. J Biol Chem 265~4492-4497. Yankner BA, Dully LK, Kirschner DA (1990) Neurotrophic and neu- Ponte P, Gonzalez-DeWhitt P, Schilling J, Miller J, Hsu D, Greenberg rotoxic effects of amyloid p protein: reversal by tachykinin neuro- B, Davis K, Wallace W, Lieberburg I, Fuller F, Cordell B (1988) A peptides. Science 2501279-282. new A4 amyloid mRNA contains a domain homologous to serine protease inhibitors. Nature 331:525-527.