16 April 2009. Pyroglutamate, or pyroGlu (pGlu) to the initiated, is an aromatic ring form of glutamate that has been discovered in some fragments of amyloid-β, namely, those missing the first two (Aβ3-x) or 10 (Aβ11-x) amino acids. Growing evidence suggests that pGlu forms of Aβ play a role in the pathology of Alzheimer disease (AD), given that they are more stable and more likely to seed aggregation (see ARF related news story) than the unmodified peptides. But research into pGluAβ has so far found traction among only a few AD researchers. That seems poised to change. “There are now quite a few labs beginning to work on this,” Cynthia Lemere, Brigham and Women’s Hospital, Boston, told ARF. Lemere, who heads one of those labs, recently chaired a pyroGluAβ session at the 9th International Conference AD/PD, held 11-15 March in Prague, Czech Republic. Among the reports that stood out at that session were those showing that other proteins linked to the pathology of dementia undergo the same cyclization process, and that in addition to BACE, there may be other secretases that cleave APP, suggested Lemere. Those mystery secretases might cleave between the second and third amino acids of the Aβ sequence of APP, leaving an N-terminus glutamate that would be vulnerable to cyclization. For her part, Lemere is keeping some hot pyroGlu data under her hat until the International Conference on Alzheimer’s Disease and Related Disorders next July in Vienna, Austria.
The Prague pyroGlu session was sponsored by Probiodrug AG, Halle, Germany, a biotechnology company that has been driving much of the research into this variant of Aβ. Probiodrug’s Stephan Schilling described some biophysical properties of pyroGlu Aβ, showing that at pH 5-7, pGluAβ3-x is less soluble than the full-length peptide. “This reduced solubility is the driving force for aggregation propensity at physiological pH,” said Probiodrug’s Uli Demuth in a post-meeting interview with this reporter. Schilling also reported that ABri and ADan, the amyloidogenic peptides responsible for British and Danish dementia, respectively, undergo a similar N-terminal pyroglutamate cyclization and that this renders these peptides more insoluble and more hydrophobic, as well. “I thought this was amazing, as it suggests that the same mechanism is consistent among these Aβ-like peptides,” said Lemere. In fact, pGluABri and pGluADan are even more insoluble than pGluAβ3-42. “This reflects their relative toxicities and is in keeping with the fact that they [ABri and ADan] are primarily found in the vasculature and are not readily transported,” said Demuth.
Holger Cynis, also from Probiodrug, offered up yet another dementia-relevant substrate for glutaminyl cyclase (QC), the enzyme that catalyzes pGlu formation. Monocyte chemoattractant protein 1 (MCP-1, also known as CCL2) is a major monocyte/glial chemokine that is elevated in AD (see Galimberti et al., 2006) and may mediate the chronic inflammation associated with the disease (see Sokolova et al., 2008). Blocking MCP-1 signalling in the brain suppresses gliosis, Aβ accumulation, and learning impairment in APP/PS1 mice (see Kiyota et al., 2009). Cynis showed that the N-terminus glutamine of MCP-1 can be cyclized and that the resulting pGlu form of the chemokine was the most potent form in a cellular assay of chemotaxis. pGluMCP-1 seemed relevant in an in-vivo model of peritonitis based on injecting thioglycollate to stimulate monocytes, because administering a QC inhibitor 30 to 60 minutes before immune stimulation dramatically reduced monocyte infiltration.
But the story is more complex than that. MCP-1 is not only susceptible to QC but also to proteases that remove peptides from its N-terminus. One of these proteases is dipeptidyl dipeptidase 4 (DP4), which nibbles away the first two amino acids of proteins when proline is in the second position. Cynis showed that DP4 cleavage of the glutamine-proline N-terminus of MCP-1 turns the chemokine from an agonist to an antagonist. Because DP4 does not cleave pGluMCP-1, elevated QC could therefore cause a double whammy, Demuth suggested. By simultaneously stabilizing Aβ and MCP-1, QC might exacerbate both the pathology and neuroinflammatory consequences of Aβ toxicity. “We think pGluAβ may be increasing with age, initiating the death of neurons, causing release of MCP-1, which in turn activates resting microglia that upregulate QC, leading to even more MCP-1,” said Demuth. “QC inhibition would, therefore, hit both sides of that coin.” Support for that idea came from Makoto Higuchi, National Institute of Radiological Sciences, Molecular Imaging Center, Chiba, Japan. In a different session he reported that in AD mouse models, reactive microglia are characterized by massive overexpression of MCP-1.
New animal data support this scenario. Stephan von Horsten, also from Probiodrug, presented data showing that QC is expressed in several nuclei of the rat brain. The enzyme is found in neurons, microglia, and astrocytes in many regions of the brain including the hippocampus, various cortical structures, the striatum, the thalamus, and hypothalamus. Van Horsten also reported that injecting Aβ into the brain caused an increase in QC and in the number of QC-positive microglia (see Schilling et al., 2008). In primary human brain cell cultures, QC is expressed in astrocytes and microglia. In the poster session, Anca Alexandru from Probiodrug’s daughter company Ingenium Pharmaceuticals in Munich showed that an animal model of pGluAβ toxicity fits with the pGluAβ/activated microglia double hit hypothesis. In mice expressing Aβ3-42 directly from the thy1 promoter, highly sensitive ELISAs detect pGluAβ3-42 in the hippocampus after just four weeks of age. At two months, the animals exhibit different severe behavioral and locomotive phenotypes, at three months there is neuron loss accompanied by pGluAβ immunohistochemical staining, microgliosis, and astrogliosis, and by five months no more invading glia are seen “because there is nothing left to clean up,” said Demuth.
Thomas Bayer, University of Goettingen, Germany, also reported on a pGluAβ model, wherein Aβ3-42, having either a glutamate or a glutamine at the 3 position, is fused to mouse pre-pro-thyrotropin releasing hormone. These models generate large amounts of intraneuronal pGluAβ3-42 and by eight weeks of age have extensive neurological impairment that seems to correlate with Purkinje cell loss. Bayer also reported on an APP/presenilin-1 knock-in mouse, which develops neuron loss, axonopathy, and synaptic and learning deficits that all seem to correlate with intraneuronal aggregation of Aβx-42 (see ARF related news story). Interestingly, pGluAβ is elevated about 40-fold in the brains of these animals compared to controls. Those findings support the idea that intraneuronal Aβ, and pGluAβ in particular, may be a major player in AD pathology.
Probiodrug hopes that inhibition of QC will be a strategy for treating AD. In Prague, Steffen Rossner, University of Leipzig, Germany, reviewed some preclinical data published toward that goal (see ARF related news story). Briefly, in this study inhibiting QC in two different mouse models reduced pGluAβ3-42, total Aβ, and overall plaque burden, and improved performance in learning and memory tasks. QC inhibition reduced gliosis, which could be due, in part, to reining in MCP-1. Demuth told ARF that QC inhibition can prevent atherosclerotic plaques, which are driven by inflammatory responses, in mice fed a high cholesterol/fat diet. The Austrian biotech company AFFiRiS, betting on pGluAβ being a player in AD, has started a vaccine program to target these peptides.
There are still questions lingering over the pGluAβ peptides. Above all, it’s unclear how the N-terminus gets truncated to begin with and how these pGlu peptides might be expunged from the brain. Aβ3-42 could be formed in two ways. Either the responsible protease truncates full-length Aβ peptide N-terminally, or it cleaves APP between position 2 and 3. That implies that some non-BACE secretase is at work. “My feeling is that Aβ3-x is generated by an unknown protease that removes the two N-terminal amino acids of Aβ after the latter is formed, so that the formation of Aβ3-x depends on prior BACE1 cleavage,” wrote Bob Vassar, Northwestern University, Evanston, Illinois, in an e-mail to ARF. Vassar did not attend the pyroGluAβ session in Prague, but his lab has done seminal work on BACE. “I think pyroGluAβ is potentially important in AD because it's very stable, but I don't think it's BACE1 independent,” he wrote.
Demuth has been coy about identifying the mystery protease. At AD/PD he presented indirect evidence to suggest that secretases other than BACE may indeed be involved in pGluAβ formation. He reported that in BACE1/2 knockout fibroblasts expressing wild-type APP, the same amount of pGluAβ is formed as in BACE-expressing fibroblasts. “We believe there are many different β site activities,” Demuth said. “We have studied that in different cell systems, including primary cortical neurons, and we see about one-half of Aβ is formed via the BACE pathway. The other half is formed by different protease(s), among them being the proteolytic pathway that generates pGluAβ,” said Demuth.
“This is a very hot topic,” suggested Lemere. “There is definitely a lot of work going on in multiple labs right now investigating whether there are other secretases involved in pre-cleavages or cleavage of Aβ at the -1, -2, +2, +3 positions, etc., and it may be that enzymes are responsible for those alternative cleavage sites,” she said.
Last but not least, what happens to pGluAβ once it is formed? In his AD/PD presentation, Takaomi Saido of the RIKEN, Wako, Japan, reported that pGluAβ is very much resistant to in vivo degradation. The half-life of pGluAβ3-42 in the brain was approximately five times longer than full-length Aβ42. In contrast, truncated Aβ species other than pGluAβ3-42, such as Aβ3-42 and Aβ4-42, were more susceptible to in vivo degradation than Aβ42. This may account for the selective deposition of pGluAβ in aging human brains, suggested Saido. He also demonstrated, by immunohistochemistry and mass spectrometry, that neprilysin deficiency resulted in elevated deposition of pGluAβ in APP-Tg mice, indicating a connection between the aging-associated decline of neprilysin activity and accumulation of pGluAβ in the brain. The underlying mechanism remains yet to be identified. Strikingly, Saido and colleagues found that QC is upregulated about fivefold in the APP/NEP-/- animals. This is relevant to earlier work by Demuth’s group showing upregulation of QC in brain tissue of AD patients but not of age-matched controls (see Schilling et al., 2008). Whether the same driving force is responsible for elevated QC in aging humans and NEP-negative mice remains to be seen.—Tom Fagan.