1 May 2007. The mythical Hydra was virtually indestructible. Lop off one of its many heads, and another would take its place. Chopping off the first two N-terminal amino acids of Aβ1-40/42 exposes a glutamic acid residue that can be cyclized to pyroglutamate. That may not sound sinister, but evidence shows that pyroglutamate-derived variants are some of the most stable Aβ peptides and, with increased hydrophobicity, may seed growth of Aβ aggregates. Rather than destroying Aβ, truncating its head might have the opposite effect, unleashing a tenacious pathological cascade. For this reason, truncated Aβ, and pyroglutamate derivatives in particular, have come in for increased scrutiny in recent years, as exemplified by several presentations at last month’s 8th International Conference AD/PD in Salzburg, Austria.
Truncated Aβ as Early Markers of Disease
The existence of N-terminal variants of Aβ has been known for some time. Truncated, or “ragged,” N-termini were described as early as 1993 (see, e.g., Roher et al., 1993). “Now we know that at the N-terminus, Aβ can be truncated in many positions. 2, 3, 4, 5, 7, 8, 9, 10, 11, and almost all of the potential truncated species are found in the brain” said Luc Buee, of INSERM in Lille, France, in an interview. “This is, in part, thanks to work from other groups, including Takaomi Saido’s, who reported pyroglutamate variants, and Virginia Lee and John Trojanowski’s, who reported truncated versions,” he said. Buee and colleagues, including Andre Delacourte, also in Lille, and Kaj Blennow at Goteborg University, Sweden, proposed that these N-terminal truncated species of Aβ may represent good preclinical markers of AD. When they analyzed control brains, they detected various N-truncated forms of Aβ, in addition to a modicum of amyloid deposits and neurofibrillary tangles, in some cases (see Sergeant et al., 2003), suggesting that these species of Aβ are around during the very early stages of pathology. “When there are no neurofibrillary tangles or amyloid deposits, then we don’t see any truncated species, so their presence seems highly specific to preclinical dementia,” Buee said.
These researchers subsequently found N-truncated Aβ in the CSF and plasma of MCI patients. Importantly, their data suggest that truncated Aβ is primarily found in MCI patients who go on to develop full-blown AD (see Vanderstichele et al., 2005). If confirmed by larger studies, truncated Aβ could become the basis of a prognostic test for MCI patients. But Buee and colleagues are also interested in these truncated peptides as potential targets for immunotherapy. As Eugeen Vanmechelen, Innogenetics NV, Gent, Belgium, Buee, and coauthors reported in Salzburg, several truncated Aβ peptides, including N-truncated 4-42 and N-truncated 8-42 Aβ are present at the earliest stages for neuropathology and may represent the earliest pathological forms of the peptide.
Truncated Aβ Fate
Later on in the disease, these truncated species are less abundant than full-length Aβ. By then, the shorter peptides may have already wreaked havoc since they are more hydrophobic than full-length counterparts and may seed the growth of Aβ oligomers and aggregates, researchers believe. “I think truncated forms are very important for nucleation, but once the process is started then aggregation continues with full-length Aβ1-42,” suggested Buee.
Many species of truncated Aβ are quickly degraded. Takaomi Saido’s group at the RIKEN Brain Science Institute, Wako, Japan, has studied the fate of different Aβ derivatives and reported findings in Salzburg. The researchers tracked degradation of various radiolabeled Aβ species after injecting them into rat hippocampus, and found that species truncated at the 2, 3, and 4 N-terminal amino acid were degraded two to three times faster than full-length Aβ. Interestingly, while co-injection of a neprilysin inhibitor, thiorphan, completely blocked degradation of full-length Aβ, it had no such effect on the truncated species. “It appears that truncation makes Aβ susceptible to digestion by some other, as yet unknown, proteases,” suggested Saido in an interview with ARF.
Pyroglutamate-derived species are, however, the exception to this general trend. Saido and colleagues found that N3 pyroglutamate Aβ42 (N3pGlu42) was about three times more stable than Aβ1-42 and almost 10 times more stable than truncated forms without the cyclized glutamate. Given their stability and their hydrophobicity, these pyroglutamate peptides may be particularly dangerous in the brain. “This [N3pGlu42] is actually a major species of Aβ that you find in AD brain,” said Saido
(see Piccini et al., 2005). “However, as Steve Younkin’s and other groups have shown, in APP transgenic mice there is almost no pGlu-Aβ42, so the major difference between APP transgenic mice and human AD brains is in the structure of Aβ,” he said (see Kawarabayashi et al., 2001 and Guntert et al., 2006).
Does the lack of N3pGlu42 explain why transgenic mice do not fully recapitulate human AD pathology? To investigate this question, Saido and colleagues generated APP constructs that lack the aspartic acid and alanine residues in positions 1 and 2 of Aβ and ferried them into cortical neurons with viral vectors. Primary cultures of these neurons generate abundant N3pGlu42, indicating that cyclization can occur in the rodent cells (see Shirotani et al., 2002). There are indications that cyclization may exacerbate pathology in whole animals. Saido and colleagues used a slightly different approach to test this possibility. They recently reported that knocking out neprilysin exacerbates Aβ pathology, dampens synaptic plasticity, and compromises cognitive function in APP23 transgenic mice, which produce full-length Aβ42 (see Huang et al., 2006). Saido told ARF that when his group crossed APP transgenic mice with neprilysin knockout mice, they then were able to detect N3pGlu42 in the mouse brain. “This agrees with the notion that when neprilysin activity is lost, N-terminal truncation of Aβ and conversion to pyroglutamate 42 can occur,” said Saido. This idea is in keeping with the potential neprilysin proteolytic site between the third Aβ amino acid, glutamate, and the fourth Aβ amino acid, phenylalanine. Neprilysin should remove the first three amino acids, precluding formation of N3pGlu-Aβ.
The relationship between neprilysin and the formation of pyroglutamate derivatives of Aβ suggests a potential means to control Aβ toxicity. “One way to facilitate Aβ degradation is to upregulate neprilysin, but the other way may be to inhibit cyclization of glutamate,” Saido suggested. The latter is what a biotech company called Probiodrug AG, based in Sachsen-Anhalt, Germany, has in mind.
Prygoglutamate Aβ and Toxicity
Cyclization of glutamate was once thought to be a spontaneous chemical reaction, but that has since been discounted. “In vitro, to convert glutamic acid to pyroglutamine you need to heat the peptide to 135 degrees centigrade in the presence of equimolar water, so it must be an enzymatic process in vivo,” said Saido. In 2004, that is what Probiodrug’s Hans-Ulrich Demuth and colleagues found (see Schilling et al., 2004) .
The enzyme in question is glutaminyl cyclase (QC). QC normally catalyzes the cyclization of N-terminal glutamine, but Demuth and colleagues showed that it can also catalyze, albeit more slowly, the cyclization of N-terminal glutamate. In Salzburg, the scientists reported that, in HEK 293 cells, Aβ3-40 is converted to N3pGlu40 in cells co-transfected with a QC expression construct, and that a QC inhibitor blocked this conversion (see also Cynis et al., 2006). Similarly, the researchers showed that pyroglutamate-derived peptides formed after injecting Aβ3-40 into the cortex of rat brain but were also prevented by the QC inhibitor. And though the amount of N3pGlu40/42 peptides formed in Tg2576 mice is small, Demuth and colleagues demonstrated a dose-dependent reduction in the levels of these peptides upon administration of the cyclase inhibitor (see Bucholz et al., 2006.
Is targeting this cyclase a valid therapeutic strategy? Demuth and colleagues are banking on it. “If you look at the percentages, the plaques in human brain are almost entirely pyroglu-positive—50 to 70 percent if you do a whole plaque analysis. This fact has been ignored,” Demuth told ARF (see also Saido et al., 1995). “Furthermore, if you analyze the core plaques, which are more difficult to dissolve, then you find almost 100 percent of Aβ is pyroglu. So pyroglu forms seem to be the majority of these post-Aβ peptides. It looks like pyroglu peptides are formed early, because they are concentrated in the core and less so in the periphery of the plaques,” he said. This, he says, fits with the faster aggregation kinetics of N3pGluAβ than that of full-length Aβ1-40/42 (see Schilling et al., 2006). In Salzburg, Demuth and colleagues also presented preclinical data suggesting that Tg2576 animals treated with a QC inhibitor for 6 months improved their performance in a contextual fear condition model of learning and memory.
Pyroglutamate Aβ and Pathology
It is not yet clear how pyroglutamate derived Aβ species may be toxic, but they do have a greater propensity to aggregate (see, for example, He and Barrow, 1999; Schilling et al., 2006), which suggests that they might seed formation of larger oligomeric species. There are also indications that N3pGluAβ forms inside neurons and that it might disrupt axonal transport (see Wirths et al., 2006 and Alzforum Webinar by Thomas Bayer, Saarland University, Germany). Intraneuronal Aβ toxicity has become a subject of focused research in recent years (see ARF related news story). Extraneuronal N3pGluAβ is also neurotoxic, as reported by Thierry Pillot and colleagues from Lipidomix, Vandoeuvre-l`es-Nancy, France, in last week’s Neurobiology of Aging online.
If pyroglutamate-derived Aβ peptides turn out to be a major pathological species, then how they form could become central to AD pathology. N3pGluAβ could be derived from Aβ1-40/42 following removal of the first two amino acids. The enzyme dipeptidyl peptidase is a likely candidate for that reaction, suggested Saido, but there may be more to the story. “We believe that the pathway by which pyrogluAβ is formed in the human brain is different from the BACE1-mediated pathway,” said Demuth.
Heterogeneous β-cleavage of APP is not a new idea. Cleavage between the aspartic acid and alanine at positions 1 and 2 of Aβ yields Aβ2-42, for example (see Wiltfang et al., 2001). In Salzburg, Demuth and colleagues showed that HEK293 cells transfected with APP carrying the BACE-1-sensitive Swedish mutation (which substitutes an asparagine/leucine dipeptide for lysine/methionine at the β-secretase site), generated mostly Aβ1-40/42 and barely any N3pGluAβ. Cells transfected with both the Swedish and London mutations showed the same picture. In contrast, cells transfected with wild-type APP, or APP carrying only the London mutation (which substitutes an isoleucine for valine near the γ-secretase site far from the β-secretase site) generated high amounts of N3pGluAβ. The data suggest that there may be differences in APP processing that may be relevant to the treatment of sporadic and different familial forms of AD. In this regard, it is interesting that Pierluigi Gambetti and colleagues at Case Western Reserve University, Cleveland, Ohio, recently reported that N3- and N11-truncated forms of Aβ, which have cyclizable-N-terminal glutamate residues, are more abundant in AD patients carrying PS1 mutations (see Russo et al., 2000).
Demuth and colleagues believe that pyroglu forms of Aβ are generated in the secretory pathway. In Salzburg, they showed data to suggest that both QC and APP co-localize in secretory compartments in the brain. This localization of QC would be in keeping with what may be its primary physiological role, namely to cyclize glutamine residues at the N-terminal of thyrotrophin-releasing hormone (TRH) and other endocrine hormones that mature in secretory vesicles. The slightly acidic pH of the secretory pathway creates the right conditions for QC-catalyzed conversion of glutamate on Aβ, said Demuth. The pH optimum for glutamine cyclization, on the other hand, is slightly more alkaline. Demuth believes that these different pH optima can be exploited to develop a QC inhibitor that blocks formation of pyroglutamate Aβ while having less of an impact on glutamine cyclization. Another reason he is optimistic about this approach is grounded in enzyme kinetics. Because the affinity of QC for glutamine is two orders of magnitude higher than its affinity for glutamate, glutamine cyclization is less likely to be affected by QC inhibition.
On a final note, there may be an endogenous way of ridding the brain of pyroglutamate derivatives of Aβ that might warrant further exploration. In Salzburg, JM Martinez-Marto and colleagues from the University of Jaen, Spain, reported that the enzyme pyrrolidone carboxyl peptidase, which hydrolyzes N-terminal pyroglutamyl residues, can protect neuroblastoma cell lines against the toxicity of pyrogluAβ forms.—Tom Fagan.
Yussef I, Florent-Bechard S, Malaplate-Armand C, Koziel V, Bihain B, Olivier J-L, Leininger-Muller B, Kriem B, Oster T, Pillot T. N-truncated amyloid-beta oligomer induce learning impairment and neuronal apoptosis. Neurobiology of Aging. 2007, April. Advanced online publication.