Inhibition of glutaminyl cyclase prevents pGlu-Aβ formation after intracortical/hippocampal microinjection in vivo/in situ

Authors


Address correspondence and reprint requests to Hans-Ulrich Demuth, Probiodrug AG, Weinbergweg 22, Biocenter, D-06120 Halle (Saale), Germany. E-mail: hans-ulrich.demuth@probiodrug.de or Stephan von Hörsten, University of Erlangen-Nürnberg, Franz-Penzoldt-Center, Experimental Therapy, Palmsanlage 5, 91054 Erlangen, Germany. E-mail: stephan.v.hoersten@ze.uni-erlangen.de

Abstract

Modified amyloid β (Aβ) peptides represent major constituents of the amyloid deposits in Alzheimer’s disease and Down’s syndrome. In particular, N-terminal pyroglutamate (pGlu) following truncation renders Aβ more stable, increases hydrophobicity and the aggregation velocity. Recent evidence based on in vitro studies suggests that the cyclization of glutamic acid, leading to pGlu-Aβ, is catalyzed by the enzyme glutaminyl cyclase (QC) following limited proteolysis of Aβ at the N-terminus. Here, we studied the pGlu-formation by rat QC in vitro as well as after microinjection of Aβ(1–40) and Aβ(3–40) into the rat cortex in vivo/in situ with and without pharmacological QC inhibition. Significant pGlu-Aβ formation was observed following injection of Aβ(3–40) after 24 h, indicating a catalyzed process. The generation of pGlu-Aβ from Aβ(3–40) was significantly inhibited by intracortical microinjection of a QC inhibitor. The study provides first evidence that generation of pGlu-Aβ is a QC-catalyzed process in vivo. The approach per se offers a strategy for a rapid evaluation of compounds targeting a reduction of pGlu formation at the N-terminus of amyloid peptides.

Abbreviations used
AD

Alzheimer’s disease

APP

amyloid precursor protein

amyloid-β

BSA

bovine serum albumin

DAB

diaminobenzidine

FDD

familial Danish dementia

HEK

human embryonic kidney cell

PBS

phosphate-buffered saline

pGlu

pyroglutamate

QC

glutaminyl cyclase

RT

room temperature

SDS

sodium dodecyl sulfate

TBS-T

Tris-buffered saline Tween

Amyloid β (Aβ) peptides are the major constituent of amyloid plaques, one of the pathological hallmarks of Alzheimer’s disease (AD). Aβ is produced from the large transmembrane precursor protein (APP) by consecutive catalysis of β- and γ-secretase generating the N- and C-terminus, respectively (Selkoe 2001). Numerous studies of the past decade revealed Aβ42 as highly amyloidogenic and crucial for development of the disease (Jarrett et al. 1993; Iwatsubo et al. 1994; McGowan et al. 2005). Also, various N-terminal modifications have been described for the Aβ peptides. Besides N-terminal truncation, the aspartic acid residues at positions 1 and 7 are isomerized and racemized in amyloid deposits (Mori et al. 1992; Roher et al. 1993; Iwatsubo et al. 1996). Most prominent modifications are related to N-truncated peptides, which carry a pyroglutamyl-modified N-terminus, which is formed by cyclization of glutamic acid at positions 3 and 11 (Saido et al. 1995), while the process of limited proteolysis resulting in Aβ(3–40) is unknown. However, those pyroglutamate (pGlu)-Aβ species readily built up to 50% of the total amyloid in senile plaques and vascular deposits (Lemere et al. 1996; Saido et al. 1996; Kuo et al. 1997). In addition, recent evidence suggests that the presence of pGlu-modified peptides correlates with the severity of the disease, and, in particular, pGlu-Aβ is detected in subjects carrying mutations in the presenilin genes 1 and 2 (Russo et al. 2000; Miravalle et al. 2005; Guntert et al. 2006). Furthermore, the presence of pGlu-modified peptides has been described in diffuse plaques of Down syndrome subjects, indicating an early formation in an AD-like deposition process (Iwatsubo et al. 1996; Saido et al. 1996).

Thus, in spite of the established contribution of pGlu-amyloid in the disease process, the formation of the pGlu-residue from glutamic acid itself is still a subject of speculation. To this end, the cyclization of glutamyl residues via a spontaneous release of water from glutamic acid has been suggested (Hashimoto et al. 2002). In contrast, more recently, we were able to show that the enzyme glutaminyl cyclase (QC), which is involved in the pGlu-formation from N-terminal glutamine of neuropeptides, also is capable to catalyze the formation of pGlu-residues from N-terminal glutamic acid in Aβ species in vitro (Schilling et al. 2004). Subsequently, this was supported by findings showing that QC-inhibitors suppress the pGlu-Aβ formation in mammalian cell culture (Cynis et al. 2006).

Consequently, as a next step, we here studied the faith of human Aβ species after intracortical microinjection in rats trying to monitor the formation of pGlu-Aβin vivo, potentially resulting in a rapid screening model system for QC-inhibitor lead candidates.

Materials and methods

Aβ peptides

Aβ(3–40), Aβ(1–40), and pGlu-Aβ(3–40) were synthesized as described previously (Schilling et al., 2006). For sample preparation the crude peptides were treated with hexaflouro isopropanol (HFIP) and dissolved in acetonitrile-containing water. Insoluble material was removed by centrifugation prior to injection into the HPLC. Preparative HPLC was performed with a gradient of acetonitrile in water (10–65% acetonitrile, 0.04% trifluoro acetic acid over 40 min) on a 250 × 20 Luna RP18 column (Phenomenex, Aschaffenburg, Germany). Peptide-containing fractions were further purified to get oligomer-free amyloid peptide: After lyophylization, 1–4 mg of amyloid peptide were dissolved in a buffer consisting of 10 mM Tris, 2.5 mM dithiothreitol, 5 mM EDTA, 2% sodium dodecyl sulfate (SDS), pH 9.0, followed by centrifugation and injection of the supernatant to a 150 × 4.6 Source 5RPC ST column (5 μm, Amersham, Uppsala, Sweden). A gradient mixed from solvent A (0.1% NH3 in H2O, pH 9) and solvent B (60% acetonitril, 40% solvent A) was used for purification. The resulting peptide could be easily solved in 0.15 M Tris buffer, pH 8.8 at a concentration of 10 mg/mL. Peptide purity and identity was confirmed by analytical HPLC (150 × 4.6, 5 μm Source or Gemini) and matrix-assisted laser-desorption ionization mass spectometry (MALDI-MS).

Cloning and characterization of rat QC

The cDNA sequence of rat QC was obtained from nucleotide entry XM_233812. Mature rat QC was isolated from cDNA of PC12 cells using the primers 5′-CGCTGTTGCCTGGACGCAGG-3′ (sense) and 5′-ATATAAGCTTTTACAAGTGAAGATATTCC-3′ (antisense). The cDNA was cloned into the yeast expression vector pPICZαB, which is suited for expression of recombinant proteins in the yeast Pichia pastoris. The system was already applied for isolation of murine QC (Schilling et al. 2005). Rat QC was purified to apparent homogeneity from the supernatant after fermentation of Pichia, applying two chromatographic steps, an expanded bed absorption on a cation exchange resin followed by a hydrophobic interaction chromatography step on a Butyl-Sepharose resin (see also supplementary information). Usually, approximately 20 mg of protein was isolated from a typical 5 L fermentation. QC activity was assayed as described previously (Schilling et al. 2002, 2003a,b).

Animals, surgery and injection procedure

Female Sprague–Dawley rats (12 weeks of age, 220 ± 14 g BW; four per group and time point) were used. Injection of pGlu-Aβ(3–40) served as positive control and to establish injection coordinates, injection, fixations, immunohistochemistry and ELISA. For each compound, two time points (24 and 48 h) were tested and in each case the contralateral side served as negative control. Animals were anesthetized using i.p. injection of Dormitor® (Pfizer, Karlsruhe, Germany) and Ketamine® (Albrecht, Aulendorf, Germany) (0.09 + 0.3 mL/270 g BW), the skull was shaved, a midline incision set, bregma targeted, and coordinates [bregma (Z): −4.0 mm; lateral right (X): −3.0 mm; caudal (Y): −3.0 mm] according to the Paxinos Atlas were used for deep cortical/intrahippocampal injection. Surgery and microinjections were conducted as previously described (Kask et al. 2001), the general approach being adapted from Shin et al. (1997), and all research and animal care procedures were approved by the local authorities and performed according to international guidelines for the use of laboratory animals. The lyophilized Aβ samples were freshly solubilized to 10 mg/mL in 0.15 M Tris buffer, pH 8.8, filtered (0.22 μm), immediately before each injection. Five microliters of the solution was injected into the right side of the deep cerebral cortex and/or hippocampus. In case of the co-injection experiments, Aβ injection was preceded by injection of either vehicle (5 μL, control) or PBD150 (5 μmol, dissolved in saline).

Cultivation, transfection, and immunostaining of rat QC in HEK293 cells

Human embryonic kidney cell line (HEK) 293 was cultured in Dulbecco’s modified Eagle’s medium, 10% fetal bovine serum in a humidified atmosphere of 5% CO2 at 37°C. For transfection, cells were cultured in six-well dishes, grown until 80% confluency and transfected by incubation in a solution containing Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) and the respective plasmids according to the manufacturer’s manual. Appropriate growth media replaced the solution after 5 h and cells were grown overnight. For subsequent immunostaining, cells were grown in six-well dishes containing a coverslip. One day after transfection, cells were washed twice with Dulbecco–phosphate-buffered saline (PBS) (Invitrogen) and fixed using ice-cold methanol for 10 min at −20°C, followed by three washing steps of Dulbecco–PBS for 10 min at 23°C. The slices were treated sequentially with 0.5% H2O2, 0.3% TX-100, and 2% normal goat serum for 10 min, 10 min, and 1 h, respectively. Application of the primary QC-specific antibody [1 : 2000 in 1% bovine serum albumin (BSA)] was followed by visualization of immunoreactive products with the avidin–biotin complex kit (Vector ABC Elite, Burlingame, CA, USA) using the substrate diaminobenzidine (DAB). For immunoblotting, proteins were loaded onto a Tris–glycine 4–20% gradient SDS–polyacrylamide electrophoresis gel (Serva, Heidelberg, Germany), separated and transferred onto a nitrocellulose membrane under semi-dry conditions. The membrane was blocked using 3% (w/v) dry milk in Tris-buffered saline Tween (TBS-T) [20 mM Tris/HCl; pH 7.5; 500 mM NaCl, 0.05% (v/v) Tween-20]. QC was detected applying primary antibody A01 (1 : 750 in TBS-T; Abnova Inc., Taipei City, Taiwan). Blots were developed using horseradish peroxidase-conjugated secondary antibodies and the SuperSignal West Pico System (Pierce, Rockford, IL, USA) according to the manufacturer’s protocol.

Histochemistry and immunohistochemistry

After different post-injection survival times, the brains were removed from the rats, fixed in 70% ethanol/0.15 M NaCl and embedded in paraffin as reported by Shin et al. (1993, 1994). Paraffin blocks were sectioned at 6 and 10 μm thickness through the injection side, which is visible by inspection by eye during cutting. After de-paraffination and mounting, sections were treated for 30 min with formic acid, 10 min with H2O2 (3%), followed again by a TBS rinse. This was either followed either by cresyl violet (Nissl) and congo red stainings according to standard procedures, or by incubation (6 h at 4°C) with primary antibodies for immunohistochemistry, the latter using several different antibodies that specifically recognize different epitopes of Aβ.

For immunohistochemistry on paraffin sections, the 4G8 (Chemicon, Temecula, CA, USA) [1 : 100 (mouse) in PBS + 5% rabbit serum + 0.3% TX-100] and anti-N3 (pE) (Assay Designs, Ann Harbor, MI, USA) (1 : 500 rabbit, in PBS + 5% Goat serum + 0.3% TX-100) antibodies were used. The primary antibodies were detected by anti-mouse bio (1 : 200 in PBS + 5% rabbit serum + 0.3% TX-100) or anti-rabbit bio (1 : 200 in PBS + 5% goat serum + 0.3% TX-100) antibodies. The immunoreactive products were visualized with the avidin–biotin complex kit (Vector ABC Elite) consisting of the following steps as previously described (von Horsten et al. 2003): A Tris–HCl rinse was followed by ABC kit (according to instruction), TBS-T rinse, substrate (DAB) solution for 15 min, and a series of isopropanol and xylol steps, finally completed by coverslipping in paramount or eukitt media. For immunohistochemistry of QC in the rat brain, animals were deeply anesthetized as described above and transcardially perfused through the ascending aorta with 200 mL ice-cold 0.9% NaCl, followed by at least 300 mL ice-cold fixation solution that consisted of 4% paraformaldehyde in 0.1 N PBS. Brains were removed, post-fixed for 4 h in the fixation solution, cryoprotected in 30% sucrose solution (5 days) and shock-frozen in petrolether. Coronal cryosections (30 μm) were processed free-floating for substrate-based immunohistochemistry (3′,3-diaminobenzidine tetrahydrochloride solution, DAB). Between incubation steps, all washing of sections were done three times (each 10 min) in 0.1 N PBS. Briefly, for inhibition of endogenous peroxidase, sections were incubated in 1.5% H2O2 in 20% v/v methanol, 0.1 N PBS (5 min, 23°C). This was followed by pre-incubation in 5% BSA solution (0.3% Triton, 0.1 N PBS, 45 min, 23°C) and consecutive incubation with the polyclonal mouse anti-QPCT antibody (A01, Abnova Inc.) (diluted 1 : 1000 in 0.3% Triton, 0.1 N PBS, overnight, 4°C). Subsequently, sections were incubated in donkey anti-mouse antibody biotinylated (Jackson Immuno-Research Inc., West Grove, PA, USA) (1 : 200, diluted in 0.1 N PBS, 2% BSA, 1 h, 23°C) and followed by incubation in the avidin–biotin complex kit (30 min, 23°C) (Vector ABC Elite). Finally, sections were reacted with DAB as a chromogen (SK-4100, Vector Laboratories) (6 min, 23°C), mounted onto object slides, air-dried and coverslipped with DPX Mountant (Fluka Chemie GmbH, Buchs, Switzerland). Pre-absorption experiments: for evaluation of the specificity of the QC antibody A01, the immunogenic peptide Q01 was spotted (2 μL) in 10-fold excess onto a nitrocellulose membrane. Membranes were repeatedly incubated in the antibody working dilution. Efficacy of pre-absorbed antibodies was tested by visualization of bound antibodies onto the membrane using a suitable peroxidase-coupled secondary antibody and processing of chemiluminescence by conventional enhanced chemiluminescence plus detection system (Amersham).

For double-labeling immunofluorescence, sections were blocked in 10% normal donkey serum containing 0.3% Triton X-100 in PBS for 1 h, and incubated with the 4G8 (1 : 200 in PBS + 5% rabbit serum + 0.3% TX-100) and anti-N3 (pE) (1 : 1000 in PBS + 5% goat serum + 0.3% TX-100). Following subsequent washes, sections were labeled using Cy-2- and Cy-3-coupled secondary antibodies (1 : 200, 2 h, 23°C) (Jackson Immuno-Research Laboratories), each diluted in PBS and 3% normal donkey serum. In some case 4′, 6-diamidino-2-phenylindole (DAPI) nuclear staining was applied according to standard procedures. After washing with PBS, sections were mounted in Vectashield-DAPI mounting medium (Vector).

All sections were analyzed using a Nikon light microscope (Eclipse 80i; Nikon, Tokyo, Japan), Nikon objectives (Plan Apo, VC series), motorized specimen stage for automatic sampling (Märzhäuser, Wetzlar, Germany), electronic microcator (Heidenhain, Traunreut, Germany), a dedicated Nikon HiSN fluorescence system, a Nikon cooled DS-5Mc camera, and imaging software (Stereo Investigator; MicroBrightField, Williston, VT, USA). Controls consisted in omitting primary antibodies and yielded lack of specific stainings.

ELISA analysis

Sequential Aβ extraction was performed, essentially as described (Kawarabayashi et al. 2001). At every stage of the extraction procedure, the homogenized tissue was subjected to sonification (Bandelin sonoplus, Berlin, Germany). Briefly, frozen hemispheres without cerebellum were homogenized in 2.5 mL TBS using a dounce homogenizer, sonificated and centrifuged at 75 500 g for 1 h. The supernatant was then subjected to two further homogenizations in 2.5 mL 2% SDS and 0.5 mL formic acid. The final fraction was neutralized using 19.5 mL 1 M Tris, pH 9.0. The vast majority (> 90%) of Aβ was extracted by 2% SDS. The assay samples were diluted 1 : 10 at least prior to ELISA analysis.

Aβ(x–40) sandwich ELISA (IBL, Hamburg, Germany) was performed according to the manufacturer’s manual. pGlu-Aβ(3–40) was assayed applying a novel established ELISA. The horseradish peroxidase-conjugated antibody directed against pGlu-Aβ (IBL) was applied in combination with microplates that were pre-coated with an Aβ40 – detecting antibody known from the Aβ(x–40) assay kit. Ratios of each individual data set (treated/untreated) were evaluated using unpaired Student’s t-tests. Significant values are shown in figures.

Results

Isolation and enzymatic characterization of rat QC

The sequence of the putative rat QC was obtained by database mining. On the basis of the primary structure of human QC, nucleotide entry XM_233812 was identified encoding the rat QC (Fig. S1). Rat and human QC share 83% sequence identity. The mature rat QC was isolated from a cDNA library of PC12 cells, expressed in P. pastoris and purified to homogeneity from the supernatant after fermentation (Fig. S2).

As revealed by the conversion of various peptide substrates and dipeptide surrogates, rat QC displays a very similar catalytic specificity compared with human QC (Table 1). The functional homology of the QC proteins from mammalian origins is also reflected by the inhibition by imidazole derivatives (Table 1). Virtually identical Ki-values were obtained with several imidazoles and the specific QC-inhibitor PBD150. The inhibition mode was linear competitive for all compounds (not shown).

Table 1.   Kinetic parameters of substrate conversion and inhibition of rat QC
CompoundRat QCHuman QCa
KM (mM)kcat (s-1) Ki (mM)kcat/KM (mM−1*s−1)kcat/KM (mM/s) Ki (mM)
  1. Reactions were carried out in 0.05 M Tris/HCl, pH 8.0 at 30°C.

  2. aData from Refs Schilling et al. 2003a, b and Buchholz et al. 2006.

  3. bSubstrate inhibition.

  4. QC, glutaminyl cyclase; ND, not determined.

Substrates
 H-Gln-AMC0.049 ± 0.0014.0 ± 0.24.5 ± 0.4b81.5 ± 0.498 ± 2ND
 H-Gln-βNA0.050 ± 0.00412 ± 11.3 ± 0.1b234 ± 21294 ± 61.21 ± 0.07b
 H-Gln-Gly-OH0.30 ± 0.013.0 ± 0.2 10.1 ± 0.653 ± 1
 H-Gln-Gln-OH0.20 ± 0.018.1 ± 0.2 41 ± 3140 ± 2
 H-Gln-Glu-OH0.51 ± 0.057.7 ± 0.1 15 ± 1 58 ± 1
 H-Gln-Gly-Pro-OH0.19 ± 0.017.3 ± 0.1 39 ± 3195 ± 7
 H-Gln-Phe-Arg-His-NH20.05 ± 0.0116 ± 1 324 ± 53
Inhibitors
 Imidazole0.116 ± 0.0040.103 ± 0.004
 Benzimidazole0.130 ± 0.0040.138 ± 0.005
 N-Methylimidazole0.013 ± 0.0010.030 ± 0.001
 Benzylimidazole0.0035 ± 0.0010.0071 ± 0.0003
 PBD1501 × 10−4 ± 5 × 10−6 6 × 10−5 ± 2 × 10−7

The characterization of the newly isolated QC from rat clearly demonstrates conserved enzymatic mechanisms and modes of inhibitor binding, thus enabling QC-inhibitor studies, which exhibit highly predictive potential for the human target enzyme.

Immunohistochemical analysis of rat QC

Initial expression analysis of a bovine QC revealed transcript levels in all brain regions (Pohl et al. 1991). To prove the expression of rat QC in the cortex, which was intended as injection site of the Aβ peptides, QC was immunohistochemically labeled using a polyclonal antibody for human QC. The reactivity with recombinant rat QC was verified applying western blot analysis of HEK293 cells expressing rat QC (Fig. 1a). Furthermore, the specificity of the antibody was characterized applying immunocytological labeling of the fixed cells (Fig. 1c) and also in rat cortex (Fig. 1e). Applying a pre-absorption with the antigenic QC-related peptide, the staining was abolished (Fig. 1b, d, f). Thus, the data clearly support a high selectivity of the antibody for QC and demonstrate that the procedure is suitable for immunodetection of rat QC. In fact, a prominent immunostaining was obtained with the polyclonal QC-antiserum in the cortex of the rat (Fig. 2a). The pattern clearly suggests a high expression of QC in neurons of several cell layers of the cortex and hippocampus (Fig. 2b). Higher magnifications give rise to the assumption that the protein is localized in soma and axons of the expressing cells (Fig. 2c, d). This, in turn, suggests a secretion of the QC into the extracellular space, which was also suggested from other studies in the bovine hypothalamus (Bockers et al. 1995).

Figure 1.

 Characterization of a polyclonal antibody raised against human glutaminyl cyclase (QC) used in this study. (a) Western blot analysis of proteins present in conditioned media and cells of line human embryonic kidney cell (HEK) 293. Naturally, HEK293 expresses only minor amounts of QC activity (Cynis et al. 2006). Lanes are (from the left to the right): 1, recombinant human QC expressed in yeast; 2, concentrated medium of HEK cells expressing rat QC; 3, cell pellet following expression of rat QC in HEK cells; 4, cell pellet of mock-transfected HEK cells. (b) Illustration of the pre-absorption event performed before IHC staining depicted in images (d) and (f). The peptide Q01 was spotted on a nitrocellulose membrane and the antibody working solution applied consecutively for three times. The pre-absorption was followed by enhanced chemiluminescence staining of the spots. (c) Immunohistochemical staining (antibody A01, DAB) of HEK cells, which transiently express rat QC. QC-positive cells are identified by the dark-brown coloration. (d) Cells as described in (c), with pre-absorbed antibody using the QC-derived peptide Q01. (e and f) Immunohistochemical localization of QC in the rat brain representing deeper cortical layers. Incubation of cryosections with the QC antibody A01 shows a QC immunoreactivity in the cortex. Pre-absorption of the primary antibody with the appropriate antigenic peptide decreases staining intensity and served as negative control.

Figure 2.

 Immunohistochemical staining of glutaminyl cyclase (QC) in cortical layers and hippocampal formation of the rat brain (a and b). A representative brain section (bregma −2.04 mm according to the brain atlas of Paxinos and Watson, 2007) showed strong QC immunoreactivity within several cortical regions such as the primary motor cortices, retrosplenial granular cortex and cingulum (b, upper rectangle). In addition, moderate QC immunoreactivity was anatomically defined in subcortical regions including pyramidal cell layers and CA3 field of the hippocampal formation (b, lower rectangle). Under a higher magnification, immunostaining for QC was not restricted to cell somata within cortical neurons, but was also observed in axons and processes (c). Punctuate, cytoplasmatic staining was observed in hypocampal neurons (d). Scale bars: (a) 1000 μm, (b) 100 μm, (c and d) 10 μm.

Establishment of a pGlu-Aβ(3–40) ELISA

To quantify the formation of pGlu-Aβ(3–40) after injection of Aβ(1–40) and Aβ(3–40), an ELISA was established based on commercially available antibodies and microplates (IBL, Hamburg, Germany). In the assay, the microplate-bound primary antibody binds to the C-terminus of Aβ. The secondary antibody (monoclonal N3pE, IBL) recognizes specifically the N-terminus of pGlu-Aβ. The Aβ-standard was synthesized and the peptide concentration determined by spectrophotometry. A fairly linear dependence of absorption was obtained in the concentration range between 1 and 500 pg/mL (Fig. 3). In addition, the cross-reactivity of the ELISA against Aβ(1–40) and Aβ(3–40) was assessed, revealing a reactivity for Aβ(3–40) and Aβ(1–40) of < 0.5% (Fig. 3, inset) compared with pGlu-Aβ(3–40). Thus, the method was well suited for the intended trial.

Figure 3.

 Establishment of a pGlu-Aβ(3–40) specific ELISA. For analysis, a pGlu-Aβ specific monoclonal antibody and coated plates specific for Aβ40 were used (both IBL). The Aβ peptides were synthesized as described in Materials and methods and the concentration verified by quantification of the absorbance at 275 nm. A fairly linear dependence of the absorbance was observed in a concentration range between 0 and 130 pg/mL. Moreover, as illustrated in the inset, the ELISA did not show significant cross-reactivity against Aβ(1–40) (dark circles) and Aβ(3–40) (squares), suggesting reliable quantification of pGlu-Aβ(3–40) under the experimental settings of the deep cortical injection of human Aβ.

Immunohistochemistry of pGlu-Aβ after deep cortical injection of pGlu-Aβ(3–40), Aβ(3–40) and Aβ(1–40)

To investigate the pGlu-Aβ formation under in situ conditions, human Aβ(1–40) and human Aβ(3–40) were injected into the deep cortex of the rat. The brains were removed and the Aβ composition analyzed in histochemical stainings applying monoclonal pGlu-Aβ antibody, monoclonal 4G8 antibody detecting total Aβ and thioflavin S. Corresponding specific immunoreactivity was found in the deep cortical areas as well as in the cingulum, alveolus of the hippocampus, external capsule, fibriae of the hippocampus as well as rarely in hippocampal fissure, CA1 and CA2 of the hippocampus. As can be observed in Fig. 4, showing the evaluation of a representative injection experiment, 24 h after injection, prominent stainings with antibody 4G8 were observed, indicating a slow degradation of the Aβ40 peptides. In the first experiments, the assay was established using pGlu-Aβ(3–40) as standard peptide. All Aβ peptides presumably form aggregates after injection, as evidenced by the congo red staining. Interestingly, the most intense 4G8 and congo red stainings were observed following injection of pGlu-Aβ(3–40). A pGlu-Aβ immunoreactivity was not observed after injection of Aβ(1–40). In contrast, pGlu-Aβ was detected 24 h after injection of Aβ(3–40), indicating a partial conversion of the N-terminal glutamic acid into pyroglutamic acid. The apparent differences in the stainings especially using the 4G8 and thioflavin S, e.g. highest intensities for pGlu-Aβ, lowest for Aβ(1–40), might be caused by rapid aggregation of the peptide and generation of degradation-resistant deposits. Those differences have been proven in in vitro experiments and potentially influence the staining intensity. The detection of pGlu-Aβ following injection of Aβ(3–40) strongly supports a catalyzed pGlu-Aβ formation under in vivo conditions.

Figure 4.

 Dark field microscopy of representative sections of rat brains (magnification 40×) stained with congo red (left column), 4G8 (middle column), and Aβ(N3pE) (right column) 24 h after injection of either pGlu-Aβ(3–40) (upper row), Aβ(1–40) (middle row), and Aβ(3–40) (lower row). Sections are rotated 90°, and the direction of injection is indicated by gray arrows. Small panels represent higher magnification (200×) of the corresponding section. The left column (congo red) illustrates that congophilic ‘red-orange’ (orange arrows) deposits were found after injection of all three Aβ peptides. However, the ‘typical’ change of color in polarized light (Ph2) was most convincingly observed after injection of pGlu-Aβ(3–40) (green/orange deposits in upper left small panel). The middle column illustrates that 4G8 immunoreactivity (yellow arrows) was also detected after injection of all three Aβ peptides. The right column [Aβ(N3pE)] illustrates that pGlu-Aβ immunoreactivity (yellow arrows) was observed after injection of pGlu-Aβ(3–40). Although not as prominent, pGlu-Aβ immunoreactivity was also observed following injection of Aβ(3–40) peptides (see small upper right and lower right panels with yellow arrows). Positive stainings for pGlu-Aβ were never observed after injection of Aβ(1–40) (right column, panel in the middle).

Immunohistochemistry of pGlu-Aβ after deep cortical injection of Aβ(3–40) and co-injection of QC-inhibitor PBD150

A primary aim of the study was to prove, whether the cyclization is enzyme catalyzed, presumably by endogeneous QC in vivo. Therefore, vehicle followed by Aβ(3–40) were injected into left hemisphere, as previous experiments revealed pGlu-Aβ within 24 h under these conditions. Into the other, right hemisphere, however, the QC-inhibitor PBD150 (1 or 5 μmol) and consecutive Aβ(3–40) were co-injected. A rapid generation of pGlu-Aβ was important, because in control experiments, a clearance or metabolism of the inhibitor was suggested within 48 h (not shown). To avoid repeated injections of the inhibitor, the pGlu-Aβ formation following application of Aβ(3–40) was analyzed 24 h after injection of Aβ and PBD150. Representative sections of four brains after injection of 5 μmol PBD150 into the right hemisphere are illustrated in Fig. 5. In the left hemispheres, where Aβ(3–40) was injected, prominent DAB staining was observed, illustrating significant pGlu-Aβ formation. In the contralateral hemispheres, a markedly reduced staining is evident, suggesting an attenuated pGlu-Aβ formation. Thus, the pre-injection of the QC-inhibitor PBD150 on this side exerted a significant effect on the cyclization of Aβ(3–40), strongly implying that QC-catalysis is responsible for the generation of the N-terminal pyroglutamic acid residue.

Figure 5.

 Photomicroscopy of representative sections of four different rat brains 24 h after bilateral injection of either vehicle or PBD150 followed by Aβ(3–40). IHC-based detection of pGlu-Aβ(3–40), using diaminobenzidine as substrate, was performed. Panels a–d provides representative sections at the level of the injection site in all four animals.

ELISA analysis of pGlu-Aβ(3–40) formation after deep cortical injection of Aβ(3–40) and co-injection of QC-inhibitor PBD150

To quantify the reduction of pGlu-Aβ by the QC-inhibitor, the Aβ-peptides were sequentially extracted from the brains using 2% SDS and 70% formic acid consecutively. The vast majority of the Aβ peptides was extracted by 2% SDS (not shown). The extracted Aβ peptides were quantified applying ELISAs detecting either total Aβ40, i.e. N-truncated variants and also partially rodent Aβ, and specifically pGlu-Aβ(3–40). The total amounts of Aβ were determined and the ratio of pGlu-Aβ and total Aβ calculated. The ratio illustrates the effect of the QC-inhibitor specifically on the pGlu-Aβ formation and provides, in fact, the percentage of the Aβ, which was converted by QC.

In general, based on the ELISA analysis, only about 2–10% of the injected Aβ was recovered after extraction, implying a rapid clearance of the peptide after injection (not shown). Interestingly, about 2.5–5% of the recovered Aβ after injection of Aβ(3–40) was converted into pGlu-Aβ (Fig. 6). In comparison, the generation of pGlu-Aβ(3–40) from Aβ(1–40), however, was only marginal, substantiating the findings of the Aβ-immunohistochemistry (compare with Fig. 4). Thus, apparently the removal of the first two N-terminal amino acids from Aβ(1–40) occurs only very slowly in vivo compared with N-terminal cyclization. The pGlu-content after injection of Aβ(1–40) increased slightly after 48 h, indicating an initial aminopeptidase driven slow removal of the first two amino acids and the subsequent generation of the Aβ(3–40), which is prone to cyclization. Co-injection of the inhibitor did not affect the pGlu-Aβ generation, most likely because of the rapid clearance of the compound after injection.

Figure 6.

 Ratio of the N-terminally pGlu-modified and total Aβ40 determined using ELISAs after deep cortical injection and sequential extraction of the Aβ peptides. Injection of Aβ(3–40) resulted in significant formation of pGlu-Aβ(3–40) 24 h after injection. In contrast, pGlu-Aβ formation was marginal after injection of Aβ(1–40). The co-injection of the inhibitor PBD150 (5 μmol) reduced the pGlu-Aβ concentration significantly (hatched bars). Reduction of the compound to 1 μmol exerted a reduced effect, suggesting a dose dependency and thus inhibitory specificity (inset) (*p < 0.05, unpaired t-test).

In contrast, based on the results obtained with the inhibitor PBD150, the conversion of Aβ(3–40) into pGlu-Aβ(3–40) was substantially decreased. The injection of 5 μmol PBD150 resulted in a significant reduction in the pGlu-Aβ level after 24 and 48 h. In a second trial, the inhibitor was reduced to 1 μmol (Fig. 6, inset). Consequently, the reduction in the pGlu-Aβ content in relation to the total extracted Aβ lacked statistical significance. This, in turn, indicates a dose dependency of the inhibitory effect. Notably, none of the animal deceased after injection of PBD150 solution, suggesting that the QC-inhibition does not exert acute toxic effects.

Double-immunofluorescence stainings following injection of Aβ(3–40)

The ELISA analysis suggested that Aβ(3–40) was only partially converted into pGlu-Aβ. Therefore, it was a further aim to investigate, whether pGlu-Aβ and Aβ(x–40) form common aggregates after injection. Previous studies in vitro suggested a generation of such deposits (Schilling et al., 2006), and also studies in human AD brain support the abundance of different Aβ forms in amyloid plaques (Saido et al. 1995). In the present study, double-immunofluorescent stainings applying antibody 4G8 and the pGlu-specific antibody revealed a co-staining near the injection site (Fig. 7a–i). This, in turn, clearly substantiates that different N-terminal Aβ variants may form mixed aggregates also in vivo. According to the observation of pGlu-Aβ seeding in vitro, the QC-catalyzed formation of pGlu-Aβ could precede the deposition of the other peptides. As implied by the reduced staining for pGlu-Aβ and total Aβ following PBD150 co-injection (Fig. 7a and b, comparing the hemispheres), the pGlu-formation might trigger aggregation of other, more abundant species, thus functioning as a site for accumulation of Aβ in deposits. Although this requires further substantiation in transgenic animal models, the present approach in an acute model might help to clarify the role of seeding Aβ species for deposition of Aβin vivo.

Figure 7.

 Double-immunofluorescent stainings of total Aβ (a; 4G8, green) and pGlu-Aβ (b; N3pE, red) after cortical injection of Aβ(3–40). The overlay mode (c) suggests co-localization of the peptides, implying formation of mixed aggregates consisting of Aβ(x–40) and pGlu-Aβ(3–40) (inset d and e). The more intense staining for pGlu-Aβ and total Aβ in the hemisphere without injection of QC-inhibitor PBD150 implies abundance of Aβ aggregates and slower Aβ clearance, caused by seeding of pGlu-Aβ. (f–i) Staining of the injection sites of an additional injection approach as described before, however, in addition to 4G8 and N3pE, DAPI was introduced for labeling of nuclei (f–i). Corresponding images of left and right injections sides using triple-exitation-band-immunofluorescences (f and g) as well as merged single-excitation-immunofluorescence images (h and i) are given. Scale bars = 50 μm.

Discussion

Although substantial efforts were made as the discovery of N-terminally truncated Aβ peptides in amyloid deposits of AD patients, their role in neuropathophysiology as well as their formation in vivo is still ambiguous. Many reports suggested, that N-terminally truncated and pGlu-modified peptides are the major constituents in amyloid plaques (Iwatsubo et al. 1996; Saido et al. 1995, 1996; Hosoda et al. 1998; Harigaya et al. 2000; Shirotani et al. 2002). More recently, also evidence was provided that especially pGlu-amyloid peptides might play a decisive role in disease progression, because pGlu-Aβ peptides were present in brains of sporadic AD patients in significantly higher amounts compared with aged-matched controls displaying amyloid deposits (Piccini et al. 2005). Furthermore, the potential of pGlu-modified peptides to speed the disease process becomes also evident from several studies investigating the influence of the mutations in the presenilins and their impact on the amyloid plaque composition (Russo et al. 2000; Miravalle et al. 2005). Interestingly, the pGlu-modified peptides are over-represented in FAD cases, which are caused by mutations in PS1 or PS2, likely contributing to the early onset of the disease. Support for the potential disease-provoking role of N-truncated Aβ peptides is also provided by the analysis of transgenic murine animal models. A model, which has shown to generate N-terminally truncated Aβ peptides displays massive neuron loss (Casas et al. 2004). Intriguingly, the signs of neuronal dysfunction appear to be coupled with the occurrence of pGlu-Aβ(3–42). In contrast, the more commonly used animal models generating Aβ based on the over-expression of familial APP mutations, e.g. tg2576 mice, show less pGlu-modified Aβ peptides in the amyloid deposits and little or no neurodegeneration (Kawarabayashi et al. 2001; Kuo et al. 2001; Kalback et al. 2002; McGowan et al. 2006).

The molecular pathways leading to the generation of N-truncated amyloid peptides are only poorly understood. It has been suggested that especially aminopeptidases might cause the N-truncation of Aβ, thus generating Aβ(3–x) which is then prone to N-terminal cyclization. Originally, it was suggested that the N-terminal cyclization of glutamic acid is a spontaneous process similar to cyclization of glutamine (Hashimoto et al. 2002). However, also glutaminyl cyclization is enzyme catalyzed by QC in vivo. Furthermore, in contrast to spontaneous pGlu-formation from N-terminal glutamine, glutamic acid has been shown to cyclize extremely slow with cyclization half-lifes of months to years (Chelius et al. 2006; Yu et al. 2006; Dick et al. 2007). This, in turn, strongly implies an enzyme catalyzed pGlu-Aβ formation in vivo. Recently, it has been shown in vitro that QC catalyzes the cyclization of Aβ(3–x) at slow rates (Schilling et al. 2004; Cynis et al. 2006). Finally, the present study clearly substantiates that the pGlu-Aβ formation is a QC-catalyzed process in situ, as evidenced by the rapid conversion of Aβ(3–40) into pGlu-Aβ(3–40) and by its prevention co-applying the QC inhibitor PBD150 intracortically. In fact, several lines of evidence are supportive for a QC-mediated pGlu-Aβ formation in vivo: (i) as shown here for the first time by immunohistochemistry, QC is expressed in brain regions that are afflicted in AD, i.e. in cortex and hippocampus; (ii) QC and APP are transported and processed in secretory compartments, and QC is secreted from the expressing cells (Pohl et al. 1991); (iii) several recent studies including our own point to a negligible spontaneous pGlu-formation from N-terminal glutamic acid at physiological conditions, i.e. neutral to mildly acidic pH and 37°C (Schilling et al. 2004; Chelius et al. 2006; Yu et al. 2006; Dick et al. 2007). QC-mediated pGlu-formation has been shown, in turn, to be optimal under these conditions (Schilling et al. 2004).

Although the compelling evidence points to the QC-catalyzed pGlu-Aβ formation, the role of the pGlu-modified peptides for the disease is still not fully clear. However, because of the N-terminal lactam formation, conspicuous changes of the biophysical properties occur. Caused by the pGlu-residue, the Aβ-peptides are stabilized against N-terminal degradation by aminopeptidases and possibly also by endopeptidases (Hosoda et al. 1998; Saido 1998). The stabilization is based on the loss of the basic, under physiological conditions positively charged N-terminus of the peptide, which is required for substrate recognition by most unspecific aminopeptidases. The drastically increased proteolytic stability of the pGlu-modified Aβ might also cause the correlation of the pGlu-Aβ content with severity of the sporadic AD, which was suggested in a previous study (Guntert et al. 2006). The presence of pGlu-amyloid in vivo in relation to pathophysiological severity of neurodegeneration has been also studied recently in human and tg mouse brain tissue by imaging techniques. While in vivo positron emission tomography (PET) and ex vivo autoradiography analysis of ‘Pittsburgh Compound B’ (PIP) retention in amyloid-rich regions of human and mouse brain revealed strong correlation of PIP and pGlu-amyloid co-deposition, PIP-retention did not correlate to Aβ(1–40) and Aβ(1–42) deposition in the investigated tissue samples (Maeda et al. 2007).

The loss of N-terminal charged residues by truncation and cyclization, however, also causes a drastic increase in the hydrophobicity, likely provoking the enhanced tendency to aggregation that has been described for pGlu-Aβ(3–40/42) (He and Barrow 1999; Schilling et al., 2006). Hence, QC inhibition might lead to reduction of pGlu-amyloid formation in sporadic and familial AD as well as in the inherited amyloidoses familial Danish dementia (FDD) and familial British dementia (FBD). It appears, furthermore, that the pGlu-Aβ aggregates differ structurally from those of full-length Aβ (unpublished observations). The formation of mixed aggregates consisting of pGlu-Aβ(3–42) and Aβ(1–42) has been demonstrated in vitro recently (Schilling et al., 2006). The double-immunofluorescent stainings provided here (Fig. 7) indicate that also in situ mixed amyloid aggregates develop fast, further indicating that the pGlu-Aβ species might speed the aggregation of full-length Aβ, caused by their own insolubility. That pGlu-Aβ can serve as a nidus for and promoter of aggregation of Aβ1–40/42 was shown in vitro (Schilling et al., 2006). The data presented in Fig. 7 now support that an inhibition of pGlu-formation exerts secondary effects on the Aβ aggregation and accumulation in vivo. Considering the high amounts of Aβ(3–40), which were injected in the cortex, the 5% pGlu-Aβ formation detected by ELISA analysis might trigger accumulation and, in turn, reduced degradation as revealed by the differences in the total Aβ staining in both hemispheres (Fig. 7a).

Further support for the seeding theory of pGlu-modified amyloid is also provided by studies of FDD, which have been shown that ADan, amyloid peptides carrying also N-terminal pGlu, which is generated from N-terminal glutamic acid, is co-deposited with Aβ peptides in FDD, strongly suggesting that ADan provokes the deposition of Aβ. Moreover, Aβ aggregation by seeding of ADan could be shown in vitro (Schilling et al., 2006). Thus, the QC-catalyzed N-terminal pGlu-formation might represent a general concept leading to amyloid peptides displaying a massive seeding capacity and stability.

It is obvious that the immunohistological characterization of the injection sites (e.g. Fig. 3) shows a certain degree of variability in the staining intensity, levels of sectioning/staining in relation to the bilateral injection sites, and corresponding pGlu production – the latter being revealed by N3pE immunoreactivity. When considering this with regard to the potential usefulness of this assay as a pharmacological screening tool, it is conceivable that quantification is hardly to accomplish. Major obstacles are that the injected Aβ-species tend to flow out of the injection area, especially when the corpus callosum is close to this area and that the immunoreactivity of the primary antibody in the tissue being followed by a secondary detection system do not allow a linear (or any other) prediction of the corresponding absolute amount of a certain immunogene (in this case pGlu-Aβ). Focusing on the bilateral injection sites, a semi-quantitative scoring appears possible and – probably more important for the present approach – local reactivity, cellular effects (microglia activation, etc.) might well be studied using these techniques in the future. Thus, while this reasoning provides sufficient justification for the application of immunohistochemical studies at the sites of microinjection, a more reliable quantification of the pGlu-Aβ production will remain a domain of the ELISA-based approach established in the present work. Together, both techniques (immunohistochemistry and ELISA) provide a basis for the further establishment of this approach for medium throughput screening of compounds modulating pGlu-Aβ production. From this point of view, the present experimental approach provides an excellent tool to characterize the efficacy of QC-inhibitors to suppress the pGlu-Aβ formation in situ and thus enables an efficient profiling of QC inhibitors for future studies in vivo. The apparent conservation of the QC enzymatic properties between human and other mammalian species, as revealed here for rat QC for the first time, supports the reliability of rodent animal models for the studies aiming at the pGlu-peptide formation in general.

Acknowledgements

We thank S. Kuhlmann, S. Fassbender, and A. Stephan for their technical assistance. This work was supported by the German Federal Department of Education, Science and Technology, BMBF grant no. 3013185 to HUD.

Ancillary