At the Keystone conference on Alzheimer disease held last month in Keystone, Colorado, Takaomi Saido of the RIKEN in Wako, Japan, and Hans-Ulrich Demuth from the German biotech company, Probiodrug AG, based in Halle, presented new data to argue that pyroglutamate (pGlu) derivatives of Aβ may be critically important in the underlying pathology of Alzheimer disease. Far from being some marginal byproduct of Aβ degradation, pGlu forms of Aβ may be the very seeds that start oligomerization, according to Demuth. In a telephone-based rerun of his talk with this reporter, Demuth likened pGluAβ to a snowball perched on a pristine slope of normal Aβ, saying that when that ball starts to roll, it gathers the peptides that would otherwise melt in the glare of proteases and metabolic clearance.

Pyroglutamate derivatives of Aβ have been known since the early 1990s, but these highly stable and amyloidogenic peptides have come in for closer scrutiny only in the last few years. The peptides form when the first two amino acids of Aβ1-x are removed to expose a glutamate residue. Aβ3-x can then undergo a cyclization reaction at the N-terminus to form pyroglutamate species (a similar fate can befall Aβ11-x). Though researchers led by Hiroshi Mori, then at the University of Tokyo, were the first to report the existence of these peptides (see Mori et al. 1992), their presence, amount, and importance was largely underestimated, said Saido. At Keystone, Saido showed why. PyroGluAβ peptides bind avidly to reverse phase HPLC columns, which must be heated to 50 degrees Celsius using a basic solvent for complete elution. pGlu peptides are also underestimated in mass spectroscopic analysis of Aβ mixtures, because their altered charge-to-mass ratio resulting from truncation and cyclization makes them harder to ionize (e.g., in MALDI TOF MS). On top of that, it only recently became apparent that these species need not arise by the spontaneous and inexorably slow chemical cyclization of glutamate. Instead, cyclization is catalyzed by glutaminyl cyclase, an enzyme present in the brain. (For an introduction to pGluAβ, see ARF related news story).

These factors conspired to relegate pGluAβ to minor importance, suggested Saido. Over the years this has begun to change, as use of proper purification techniques shows that pGluAβ represents as much as 50 percent of total Aβ in human tissue, Saido said. Saido and colleagues previously showed that pyroglutamate-derived Aβ species are dominant in senile plaques and actually occur prior to plaques containing normal Aβ in Down syndrome patients (see Saido et al., 1995).

This observation jibes with what is now known about pGluAβ stability, namely, that pGlu-modified peptides have a much longer half-life than full-length Aβ. For its part, however, full-length has a longer half-life than non-cyclized, N-terminally truncated Aβ1, 2, 3, 4 or 5-x. So how do the modified peptides end up as a major constituent of plaques? Saido suggested this could happen if the degradation machinery somehow failed to remove the truncated and cyclized peptides. At Keystone, he showed that just this scenario can occur in transgenic mice.

One thing that sets existing transgenic mice apart from humans is that the rodents normally make very little pGluAβ. This changes when the protease neprilysin is taken out of the picture, Saido said. His group crossed neprilysin knockout mice with APP23 transgenic animals, which express human APP with the Swedish mutation. Mass spectroscopy shows that these crosses produce Aβ3-x and also pGluAβ species. Saido suggested that failure to form pGluAβ in neprilysin-competent animals might be the major limitation of mice as models of AD. He also noted that pGluAβ correlates to the PET signals detected using the Aβ ligand PIB. Researchers led by Makoto Higuchi at the National Institute of Radiological Sciences, Chiba, Japan, recently developed a PIB radiotracer that is suitable for use in small rodents and found that in both human and mouse brain, PIB retention correlates with levels of an N-terminal, pyroglutamate derivative of Aβ. These researchers proposed that one difference between mouse models of AD and the human condition is that the accelerated production of Aβ in the mice does not allow sufficient time for pGluAβ—and consequently “AD-like” plaques enriched in the pyroglutamate derivative—to form (see ARF related news story).

In humans, neprilysin levels inversely correlate with Aβ deposition in both demented and non-demented individuals (see, e.g., Russo et al., 2005 and Hellström-Lindahl et al., 2006), though it is not clear if that correlation results from increased production of pGluAβ.

These forms of Aβ are not only more stable than the unmodified peptide, but also more fibrillogenic, making them potentially more dangerous to have lurking in one’s brain. This was a point Demuth emphasized. He showed that while pGluAβ3-40 undergoes fibrillization faster than Aβ40, even further acceleration occurs when pGluAβ3-40 is used to seed solutions of Aβ40. This hints that in the AD brain, even a small amount of pGluAβ3-40, or of the even more amyloidogenic pGluAβ3-42, may be sufficient for such a snowball effect. A similar dynamic was seen pGluAβ3-38 in comparison to Aβ38, raising the question of whether γ-secretase modulation to increase Aβ38 production at the expense of Aβ42 production will by itself have the desired effect on amyloid pathology.

To test the relevance of this idea, Demuth and colleagues, together with Thomas Bayer and colleagues at the University of Göttingen, Germany, have created a transgenic mouse model where the third position of Aβ, the glutamate (E), is changed to a glutamine (Q). Glutamine gets cyclized more than three orders of magnitude faster by endogenous glutaminyl cyclase than is glutamate, so if N-terminal pyroglutamate is important for neurodegeneration, one might expect these animals to have accelerated pathology. This is exactly what Demuth showed at Keystone. These transgenic animals, called TBA2, show extremely rapid onset of symptoms, having behavioral problems and motor activity deficits (they cannot find or reach food) by the age of two months. At the same age, the animals’ hippocampus and cortex reacts with antiserum that recognizes both monomeric and oligomeric Aβ.

The Aβ in these animals is produced as part of a prothyrotropin-releasing hormone (proTRH)-Aβ chimera driven by a brain-specific promoter. The N3QAβ is cleaved from proTRH-Aβ by prohormone convertases in the secretory pathway. A similar construct (release of Aβ by furin from a BRI-Aβ sequence) was used by Eileen McGowan and colleagues at the Mayo Clinic, Jacksonville, Florida, to circumvent the complications of γ-secretase cleavage at both the 40 and 42 positions. McGowan and colleagues found that Aβ40 produced from this construct is actually neuroprotective (see ARF related news story). That N3QAβ driven from a similar system is destructive is indicative of the toxicity of pGluAβs. “This is the first Alzheimer’s-related model that shows a tremendous behavioral phenotype after only two months of existence,” said Demuth. He further demonstrated the toxicity of N3QAβs in hippocampal rat brain slices. Conditioned medium from HEK293 cells transfected with APP-NLQ (a construct having asparagine, leucine, and glutamine in the first three positions of Aβ) more strongly depressed hippocampal LTP than did medium from cells expressing normal APP or APP-NLE, which would yield full-length Aβ or Glu3Aβ with slower N-terminal cyclization rates. To Demuth, the in-vivo and tissue culture experiments together confirm the toxicity of pGluAβ.

Human Glutaminyl Cyclase—A Therapeutic Target? If pGluAβ drives AD pathology, then what is driving cyclization of the N-terminal glutamate in the diseased state that is not at work in normal brain? Possible answers here include enhanced Aβ production, enhanced removal of the N-terminal dipeptide, or more rapid cyclization.

Glutaminyl cyclase (QC), the enzyme that cyclizes N3Aβ glutamine in the above mouse model, is normally found in the brain, but at Keystone, Demuth presented data to suggest that this enzyme is massively overexpressed in AD brain. Enzyme and pGluAβ levels show the same pattern, with QC levels being increased up to 40 times depending on the severity of the investigated stages. The enzyme is expressed in brain areas affected in AD, including the CA3 layer of the hippocampus and the cortex.

Such data hint that targeting QC might be one approach to tackling AD pathology, and that’s just what Probiodrug has in mind. Demuth presented data on one experimental inhibitor, PBD150, suggesting efficacy in animal models. Used as a prophylactic in Tg2576 mice beginning at four or six months, PBD150 reduced both pGluAβ and total Aβ, and it improved memory in a contextual fear paradigm. Its efficacy was similar when used as a treatment in older (10 months) animals that have formed plaques. PBD150 reduced pGluAβ and improved memory. Prophylaxis worked in the TASD41 mouse, which produces very high amounts of human APP carrying both Swedish and London mutations (see Rockenstein et al., 2001). Demuth said that Probiodrug has profiled better compounds for passage of the blood-brain barrier.

One potential bonus of QC inhibition is that it may also reduce reactive gliosis. In characterizing the effect of PBD150, Demuth’s collaborators Steffen Rossner at the Brain Research Institute at University of Leipzig, and Manfred Windisch of the contract research organization JSW Research in Graz, Austria, noticed reduction in the number of reactive glia surrounding plaques in treated Tg2576 and TASD41 mice, respectively. This may simply be due to the reduction of plaque load or the reduction of seed peptides, but it may also be related to a direct inhibition of glial responses. In this regard, Demuth noted that other pyroglutamate peptides, in particular pGlu derivatives of monocyte chemoattractant protein (MCP-1), which has an N-terminal glutamine, trigger glial cell migration. Therefore, blocking QC may also reduce inflammatory responses in the brain and even elsewhere. Probiodrug is screening QC inhibitors for inflammatory diseases such as atherosclerosis, as well.

Could this anti-inflammatory aspect be a downside to the strategy, leading to side effects? “The binding constants for the cyclization of glutamine peptides are much lower than for glutamate peptides, so you don’t have to fully inhibit QC’s negative side function of glutamate cyclization within a certain time frame to get an effect on Aβ but still have sufficient glutamine cyclization for maturation of proteins, for example, peptide hormones,” said Demuth.

There may be other surprises in store from this pyroglutamate story. Demuth briefly described experiments that lead him to believe BACE may not be the only important enzyme that leads to Aβ production. This possibility has been entertained before but currently represents a fringe view (see ARF related news story). Cultured HEK293 cells produce much less pGluAβ when transfected with APP containing the Swedish mutation, which promotes cleavage at the BACE site. This suggests that there may be another protease that cleaves Aβ in such a way that the glutamate is unmasked at the N-terminus and can be cyclized. One possibility is that an alternate protease cleaves APP at the three position of Aβ.

Whether another protease is important in vivo remains to be seen. Robert Vassar of Northwestern University, Chicago, attended the meeting but is not involved in this work. He told ARF that the PDAPP mouse used by Elan is wild-type at the β-secretase cleavage site and, when crossed with BACE knockouts, produces no Aβ or amyloid plaques, suggesting that BACE is the major protease for Aβ even in wild-type APP (McConlogue et al., 2007). “There are other enzymes that cut [at the β site], but they cut rarely and the amounts of their Aβ are very low, so I don’t think they play a significant role in AD. They do start to cut at higher levels when you overexpress APP. It is possible that under condition of overexpression, and when you inhibit BACE so you have more full-length APP, that some of these aberrant cleavages appear to increase,” Vassar suggested. Perhaps any BACE doppelganger will reveal itself at the next Keystone meeting.—Tom Fagan.


  1. In a separate discussion, Dr. Vassar cites McConlogue et al. (2008), which shows that knocking out BACE1 in mice expressing human APP containing the wild-type β-secretase site (PDAPP) significantly reduces Aβ and amyloid plaques in support of his conclusion that BACE1 is the major protease for cleavage of wild-type APP and that other enzymes cannot play a significant role.

    However, the PDAPP mouse model massively overexpresses human APP. In contrast, Hirata-Fukae et al. (2008) recently reported on the effects of BACE1 in mice expressing physiological levels of mouse APP, which contains the human wild-type β-secretase site sequence. They found that adult BACE1-overproducing mice and age-matched control mice made the same amount of endogenous brain Aβ. They concluded that, under normal physiological levels of APP, BACE1 protein level has a minimal effect on endogenous Aβ level. Moreover, they speculated that factors other than BACE1 must be involved in the modulation of Aβ in the adult brain.

    It appears that the animal model used dramatically affects the conclusions drawn regarding the importance of BACE1 on Aβ production. In animals in which human APP is greatly overexpressed, reducing BACE1 reduces Aβ and suggests that BACE1 is the primary producer of Aβ. However, in animals expressing physiological levels of endogenous APP, BACE1 levels have a minimal effect on Aβ, suggesting that factors other than BACE1 must be involved.


    . Partial reduction of BACE1 has dramatic effects on Alzheimer plaque and synaptic pathology in APP Transgenic Mice. J Biol Chem. 2007 Sep 7;282(36):26326-34. PubMed.

    . Beta-site amyloid precursor protein-cleaving enzyme-1 (BACE1)-mediated changes of endogenous amyloid beta in wild-type and transgenic mice in vivo. Neurosci Lett. 2008 Apr 25;435(3):186-9. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Salzburg: Aβ’s N-terminal Shenanigans
  2. Hot Stuff—PIB News From the Pacific Rim
  3. Aβ40—The Anti-Amyloid?
  4. SfN: The Old and the New—BACE and Cathepsin B Share the β-secretase Stage

Paper Citations

  1. . Mass spectrometry of purified amyloid beta protein in Alzheimer's disease. J Biol Chem. 1992 Aug 25;267(24):17082-6. PubMed.
  2. . Dominant and differential deposition of distinct beta-amyloid peptide species, A beta N3(pE), in senile plaques. Neuron. 1995 Feb;14(2):457-66. PubMed.
  3. . Neprylisin decreases uniformly in Alzheimer's disease and in normal aging. FEBS Lett. 2005 Nov 7;579(27):6027-30. PubMed.
  4. . Age-dependent decline of neprilysin in Alzheimer's disease and normal brain: inverse correlation with A beta levels. Neurobiol Aging. 2008 Feb;29(2):210-21. PubMed.
  5. . Early formation of mature amyloid-beta protein deposits in a mutant APP transgenic model depends on levels of Abeta(1-42). J Neurosci Res. 2001 Nov 15;66(4):573-82. PubMed.
  6. . Partial reduction of BACE1 has dramatic effects on Alzheimer plaque and synaptic pathology in APP Transgenic Mice. J Biol Chem. 2007 Sep 7;282(36):26326-34. PubMed.

Other Citations

  1. APP23

Further Reading