Are N- and C-terminally truncated Aβ species key pathological triggers in Alzheimer's disease?

The histopathology of Alzheimer's disease (AD) is characterized by neuronal loss, neurofibrillary tangles, and senile plaque formation. The latter results from an exacerbated production (familial AD cases) or altered degradation (sporadic cases) of 40/42-amino acid–long β-amyloid peptides (Aβ peptides) that are produced by sequential cleavages of Aβ precursor protein (βAPP) by β- and γ-secretases. The amyloid cascade hypothesis proposes a key role for the full-length Aβ42 and the Aβ40/42 ratio in AD etiology, in which soluble Aβ oligomers lead to neurotoxicity, tau hyperphosphorylation, aggregation, and, ultimately, cognitive defects. However, following this postulate, during the last decade, several clinical approaches aimed at decreasing full-length Aβ42 production or neutralizing it by immunotherapy have failed to reduce or even stabilize AD-related decline. Thus, the Aβ peptide (Aβ40/42)-centric hypothesis is probably a simplified view of a much more complex situation involving a multiplicity of APP fragments and Aβ catabolites. Indeed, biochemical analyses of AD brain deposits and fluids have unraveled an Aβ peptidome consisting of additional Aβ-related species. Such Aβ catabolites could be due to either primary enzymatic cleavages of βAPP or secondary processing of Aβ itself by exopeptidases. Here, we review the diversity of N- and C-terminally truncated Aβ peptides and their biosynthesis and outline their potential function/toxicity. We also highlight their potential as new pharmaceutical targets and biomarkers.

A defining characteristic of Alzheimer's disease pathology is the presence of extraneuronal plaques composed of aggregated ␤-amyloid peptides (A␤). 2 A␤ terminology usually refers mainly to a mix of canonical 40/42 amino acid peptides excised by endoproteolysis of a type I transmembrane protein called ␤-amyloid precursor protein (␤APP) through the sequential action of two enzymes: ␤-site APP-cleaving enzyme (BACE1) and ␥-secretase (Fig. 1A) (1). Processing of ␤APP by these two enzymatic activities also generates an intracellular fragment, APP intracellular domain (AICD), that behaves as a transcription factor (2,3). Once produced, A␤ peptides are secreted and, upon various triggers that could be of genetic or environmental natures, accumulate and yield oligomeric aggregates. These oligomeric structures can transiently remain soluble or ultimately fibrillize and seed to form one of the main histological stigmata of AD pathology, senile plaques. However, the number of plaques is not clearly associated with disease progression and severity. Recent studies suggest that soluble, oligomeric forms of A␤ have an important role in neurotoxicity and memory loss (4). Thus, these oligomeric species of A␤ have been shown to cause synaptic dysfunction, to disrupt long-term potentiation (LTP) (5,6), and to affect behavior in transgenic mice (7,8).
The amyloid cascade hypothesis is strongly supported by genetic facts. Thus, all mutations responsible for early onset and aggressive forms of AD share as common denominators a modulation of total A␤ load, modification of A␤42 over the A␤40 ratio, or generate A␤ species prone to aggregation (9,10). Thus, the amyloid cascade hypothesis is at the center of gravity of AD pathology. However, the vast majority of clinical trials centered on A␤ by either blocking their production or neutralizing them once produced have failed. It is thus a challenge to reconcile genetic grounds with clinical failures. It could be envisioned that the physiological function of full-length A␤ could have been underestimated. Furthermore, the contribution of A␤-derived fragments generated by secondary cleavages or additional APP-derived fragments distinct from A␤ could have been underestimated. Thus, other fragments issued from ␤APP processing, such as the membrane-tethered fragment, C99, or ␤CTF, could very well contribute to pathological dysfunctions (11)(12)(13). Also, one can consider the fact that several aspects of physiological ␤APP processing have still to be delineated. This is exemplified by recent data showing that besides classical secretases, novel proteolytic actors recently came on stage. For example, recent works unmasked a novel cleavage site on ␤APP (14,15), named Eta () cleavage site (14), which gives rise to a subset of new fragments (A␣, A␤, and CTF) (Fig. 2). The enzyme responsible for ␤APP cleavage at the site has been identified as the matrix metalloproteinase MT5-MMP (15,16).  In this Minireview, we will focus on the N-and C-terminally truncated ␤-amyloid peptides that are produced either by primary cleavages taking place on ␤APP or yielded by secondary cleavages occurring on A␤ full-length A␤40 or A␤42. We will describe here several N-and C-terminally truncated A␤ peptides, their production, biophysical properties in term of aggregation, neuronal toxicity, as well as their putative impact on Alzheimer's disease progression. We will also describe the A, amyloidogenic pathway involves a first cleavage of ␤APP by ␤-secretase generating a soluble fragment, sAPP␤, and a membrane-anchored fragment, ␤CTF. Secondary cleavage of ␤CTF by ␥-secretase releases an intracellular fragment, AICD, and full-length A␤ (A␤40 and A␤42). A nonamyloidogenic pathway involves a first cleavage by ␣-secretase releasing a soluble fragment, sAPP␣, and a membrane-anchored fragment, ␣CTF. The latter is then processed by ␥-secretase to produce an intracellular domain, AICD, and a shorter peptide, A␤(17-X). B, processing of ␤APP at ␤Ј or ␣ cleavage sites combined with ␥-secretase-mediated hydrolysis gives rise to A␤(11-X) and A␤(17-X) forms of A␤. ␤APP processing by -secretase. Full-length ␤APP can be processed by an alternative pathway involving a first cleavage by -secretase (MT5-MMP) releasing a soluble fragment sAPP and a membrane-anchored fragment CTF. CTF is then cleaved either by ␣-secretase or by ␤-secretase, producing soluble fragments named A␣ and A␤, respectively. A␣ can undergo subsequent cleavage by ␤-secretase. Membrane-anchored CTFs (␣CTF and ␤CTF) are processed by ␥-secretase activity to give rise to AICD, full-length A␤ peptides, and A␤(17-X) peptide. potential of N-and C-terminally truncated A␤ peptides as biomarkers as well as pharmaceutical targets.
It is interesting to note that FAD-linked mutations occurring on ␤APP or on the ␥-secretase-associated proteins presenilin-1 and presenilin-2 (19) usually result in either an increase in total A␤ load or, alternatively, a selective increase in A␤42 and thereby A␤42/A␤40, and consistently lead to early-onset AD (20). Interestingly, Ancolio et al. (21) showed that cells overexpressing a ␤APP bearing the V715M mutation indeed behave differently because they secrete less A␤40, whereas levels of A␤42 remained unchanged. More strikingly, they described a 7-fold increase in the recovery of A␤ (11/17-42) with no change in A␤ (11/17-40) (21). These data suggest the potential toxic effect of 11/17-X fragments in AD pathology. This conclusion is supported by both anatomical, biophysical, and clinical grounds. Thus, these shorter fragments have been described in senile plaques as well as in AD and Down syndrome brains. These fragments are increased by FAD mutations on PS1 as well (22). Finally, both cleavage fragments could indicate a pathophysiological condition because expressions of the enzymes responsible for their production are regulated during AD. Moreover, A␤(11-X) and A␤(17-X), have been both detected in cerebrospinal fluids in patients suffering from mild cognitive impairment (MCI), i.e. at very early stages of the disease (23). Although levels of A␤(11-X) in CSF of MCI patients seem to be lowered, A␤(17-X) tends to be expressed at higher levels than in controls. Moreover, A␤40 appeared to be higher in the cerebrospinal fluid of patients with Alzheimer's disease than in patients suffering from other dementias (24).
de Strooper and co-workers (25) documented the fact that slight alterations of the A␤40/42 ratio could lead to drastic modifications in toxic potential. Ancolio et al. (21) support this view and further indicate that the ratio of A␤-truncated fragments could also account for early onset cases of AD even when the total A␤ load is reduced.

Truncated fragments resulting from secondary cleavages on A␤ peptides
Truncated A␤ species could just represent by-products of A␤ degradation. However, these could also harbor their own spectrum of physiological and/or (according to their concentration) toxic functions. In the latter case, they could be seen as biotransformation derivatives of A␤. Thus, this questions the potential weight of truncated A␤ species in AD pathology. As a corollary, they could represent a new set of early diagnostic markers, and thus, enzymatic activities implicated in their formation could be seen as potential therapeutic targets.

A␤38, A␤37, and A␤39
Although a plethora of articles have addressed the biology of A␤40 and A␤42, the C-terminally truncated A␤ peptides ( Fig.  3) have received much less consideration. However, in addition to A␤40 and A␤42, several A␤ shorter species truncated in their C-terminal moiety, among them A␤37, A␤38, and A␤39, have been identified in blood plasma samples and human cerebrospinal fluids of patients suffering from AD (26). A␤38 was even found to be the second prominent A␤ form after A␤40 in quantity. These peptides may be produced according to different pathways. The stepwise release of the GVV and VIA tripeptides generates A␤37 and A␤39 from A␤40 and A␤42, respectively. Moreover, A␤37 can be produced by another pathway after release of the GVVIA peptide from the A␤42 sequence (27). Production of A␤38 is influenced by nonsteroidal anti-inflammatory drugs (NSAIDs). Although NSAID treatments tend to globally decrease A␤42 production, these compounds induce a slight shift of ␤APP cleavage leading to an increase of A␤38 formation (28), and this appears to occur independently of their primary target, Cox2. A␤38 production has been shown to be increased by a subset of ␥-secretase modulators (29).
Immunohistological studies of A␤37 and A␤39 C-terminally truncated peptides in AD brains and transgenic mouse models have revealed that both peptides were found to accumulate in meningeal and parenchymal vessels in the brains of familial AD cases as well as in sporadic AD (30). The pattern of deposition differs between AD brains and transgenic mouse models. In sporadic or familial AD brains, the C-terminally truncated peptides appeared to be aggregated in plaques, but in transgenic mouse models, the presence of truncated peptides in plaques was more variable.
Interestingly, expression of shorter species of A␤, from A␤37 to A␤40, does not elicit toxicity in Drosophila and even appeared to attenuate A␤42 toxicity (31). Such results suggest that treatments regulating ␤APP processing by favoring an increase in A␤37, A␤38, or A␤39 production could be somehow beneficial and seen as an A␤42-related inactivating pathway.

C-terminally truncated A␤34 variant
It has been demonstrated that A␤34 derived from hydrolysis of A␤ by BACE1 (32,33). A␤34 is increased in cells overex- pressing both human ␤APP and human BACE1, and it is interesting to note that a ␥-secretase inhibitor treatment impairs A␤34 levels suggesting that A␤34 production results from a secondary cleavage that occurs after A␤ release by ␥-secretase (34).
BACE1 inhibition decreases A␤34 in CSF (35). However, in vitro experiments have also pointed out that A␤34 could be produced by a secondary cleavage step on A␤40 due to matrix metalloproteases (MMP2 and MMP9) (36). These MMPs could also degrade A␤34 into shorter species (A␤30 and A␤16).
A␤34 physiology and the impact on AD progression have not been deeply assessed. This may be due to a lack of tools to directly monitor the peptide and assess its pathophysiological influence. Caillava et al. (37) characterized and developed an A␤X-34 -directed specific polyclonal antibody recognizing the C-terminal part of A␤X-34. They demonstrated that A␤X-34 peptides are present in 3ϫTgAD mice brains as well as in AD patient's brains. More recently, it has been shown by immunofluorescence and immunohistological studies in three different mouse models that A␤34-like immunoreactivity appears as a punctate diffuse pattern and does not label the center of senile plaques (38). Moreover, a similar histological pattern was found in the brains of AD patients, and several studies have shown that A␤34 is an abundant species in CSF of AD patients (26,38,39). Thus, BACE1 contributes to A␤ production but is also involved in its secondary cleavage as well. Whether a disruption in the balance governing and the equilibrium between A␤ and A␤34 underlies part of AD pathology or whether BACE1 could be seen as a beneficial enzyme contributing to A␤ catabolism remains to be established. At first sight, the age-related increase in BACE1 activity and the associated elevation of A␤34 argue in favor of the first hypothesis. However, Caillava et al. (37) showed that A␤34 could display a protective phenotype in HEK293 cells. It is possible that the A␤34-associated protective phenotype could be abolished when its aggregation occurs as has been documented for A␤ (40). In this context, it is noteworthy that, as is the case for A␤ (2,41), A␤34 undergoes degradation by neprilysin, a peptidase, the activity and expression of which are reduced during aging. Thus, both age-related augmentation of BACE1 (42) and reduction of neprilysin expressions and activities could very well account for augmentation of A␤34 levels, aggregation, and pathogenic phenotype.

A␤24
C-terminally truncated A␤24 (A␤ (1-24)) is a peptide produced upon secondary cleavage of full-length A␤ after activation of glial cells. Such a peptide has not yet been described in the brains of AD patients. However, it has been shown that intracranial injection of synthetic A␤24 in WT mice impairs full-length A␤42 clearance through the blood-brain barrier and promotes A␤42 aggregation via its seeding properties. Moreover, the synthetic A␤24 peptide tends to promote A␤42 aggregation, whereas the A␤24 peptide itself presents a lowaggregation propensity (43). The exact nature of the enzyme responsible for A␤24 production still remains unknown. However, it is interesting to notice that a cleavage between a valine and a glycine, as is the case for A␤24 generation, is not a usual signature of endopeptidases or exopeptidases. Noteworthy, some matrix metalloproteinases (MMP-9 or MMP-2), which are regulated by microglial activation, appeared to be involved in A␤ degradation, generating C-terminally truncated fragments such as A␤23, A␤30, or A␤34 (36). Their involvement in A␤24 production still awaits firm demonstration.

A␤2-X
The amino-truncated ␤-amyloid peptide A␤(2-42) has been detected in a detergent-soluble fraction of AD brains (44) as well as in the CSF of sporadic and familial AD patients. Does this A␤ peptide have a physiological or a pathological function or is it only an intermediate form for another cleavage giving rise to a shorter peptide? Even if A␤(2-42) has been described in CSF and brains of AD patients, very little is known about this A␤ species and its biological properties. However, an important aspect concerned the fact that, unlike the case for canonical A␤ species (45), A␤(2-42) production remains poorly affected by presenilin-1 deficiency in neurons (44), although ␥-secretase cleavage is obviously necessary for releasing this fragment. This could be explained by a selective involvement of presenilin-2 in A␤42 formation. This would imply that PS1 and PS2 occur in distinct cellular compartments, one of which is permissive for A␤(2-42) generation. Indeed, recent works show that PS1 and PS2 occur in distinct cellular compartments (46,47). Alternatively, one cannot rule out the possibility that a PS1-independent activity(ies) (48 -52) could account for at least part of A␤(2-42) production.

A␤3-X and pE3-XA␤
Another pyroglutamate-modified A␤ that begins with the glutamate in position 3 of A␤ (pE3-XA␤) has been described. Several lines of anatomical clues suggest a potential key role of this A␤ species in AD pathology. Thus, pE3-XA␤ has been shown to be present in quantities similar to full-length A␤ in senile plaques (53) but also in diffuse plaques (54) and in the vascular wall (55). Noticeably, pE3-XA␤ is also present in Down's syndrome-affected brains (56). pE3-42A␤ appeared to be a dominant isoform in the hippocampus and cortex in patients with AD (57).
pE3-XA␤ fragments can trigger hippocampal neuronal loss, microglial activation, and astrogliosis and impair long-term potentiation in transgenic animals expressing human pE3-XA␤ (58).This toxicity appears to be accounted for by the ability of pE3-XA␤ to seed A␤ and promote its deposition (59).
pE3-XA␤ peptide formation is a two-step process involving an N-terminal truncation releasing the first two A␤ residues Asp-Ala followed by an enzymatic cyclization of the glutamyl in position 3 (Fig. 4). Studies aimed at deciphering the enzymes responsible for A␤3-X formation and pyroglutamate conversion have highlighted several candidates. Sevalle et al. (60) delineated the contribution of aminopeptidase A (APA) to the truncation of full-length A␤ leading to the 3-X species. Thus, by means of two distinct selective inhibitors, APA overexpression and APA-expressing membranes, it was demonstrated that the initial step consisting of the release of the aspartyl 1 residue was elicited by APA. This was consistent with the APA's affinity for acidic residues (61). APA seems not to be the only exopeptidase responsible for N-terminal truncations of A␤ and exposition of glutamate at position 3. Implication of a member of the dipeptidyl peptidase family enzymes (DPP) has been previously suggested (62). Recently, in vitro experiments using MALDI-TOF MS applications have pointed out the possible implication of DPPIV in A␤3-X formation (63). Interestingly, A␤40 peptide appeared to be the more prone to DPPIV truncation compared with A␤42.
After removal of the two first residues by APA and/or the dipeptidyl aminopeptidase activities, the glutamate residue at position 3 is converted into pyroglutamate forming a peptide more resistant to exopeptidasic attack. Several anatomical, pharmacological, and genetic evidences indicated that the enzyme responsible for A␤3-X cyclization was an acyltransferase named glutaminyl cyclase (QC) (64 -66). First, QC is unevenly distributed in brain and is up-regulated in AD-affected brains (67). Second, QC protein and mRNA expressions colocalize with pE3-XA␤ in human temporal and entorhinal cortices and, more importantly, correlate better with cognitive alterations assessed by mini-mental state examination than the unmodified A␤ peptides (40). Third, in animal models, QC overexpression triggers behavioral deficits, whereas conversely, QC depletion rescues defects observed in an AD-transgenic model (68). Fourth, pharmacological blockade of endogenous QC by selective inhibitors reduces the pE3-XA␤ load in mouse and Drosophila AD models and reduces plaques, astrogliosis, and cognitive alterations in AD mouse models (64). It should be added that calcium homeostasis dysregulation, which is commonly observed in cellular AD models, increases QC mRNA expression and activity in neuron-like differentiated SK-N-SH (69). It should be noted that ␤Ј cleavage mediated by BACE1 (see above and Fig. 1B) generates a glutamyl residue that can undergo cyclization. Whether GC is involved in pE11-XA␤ remains to be established.

N-truncated A␤4-X
N-truncated A␤(4 -42) was one of the first A␤-truncated species being reported (70). This truncated form, which starts with a phenylalanine at position 4, was found to be highly abundant in AD brains, aged controls, and vascular dementia (57,71). It has been demonstrated that this peptide rapidly formed stable aggregates (72). A␤4-X species concentrate in the core of the plaques in several AD mouse models (38). In vitro toxicity assays showed that A␤(4 -42) is as toxic as pE3A␤ and A␤42 (72), but A␤(4 -40) was less toxic compared with A␤ (4 -42). In terms of the kinetics of appearance, A␤4-X variants seemed to precede pE3-XA␤ accumulation in the 5ϫFAD transgenic mouse models (73). In vivo studies have indicated that intraventricular injections of A␤ (4 -42) in WT mice tend to affect working memory as assessed with a Y maze test (72). Little is known about the catalytic events responsible for A␤4-X production. One can envision an exopeptidasic release of either glutamate at position 3 that would occur before its cyclization or removal of the pE3 residues once formed. Based on theoretical grounds, two types of peptidases could indeed perform these cleavages. On the one hand, free glutamate residues could be released by acidic peptidase such as aminopeptidase A that is already involved in Asp-1 removal (see above and Ref. 60). On the other hand, there exists two types of pyroglutamyl peptidases I and II that are specialized in the attack of pE residues (61). In vitro studies with fluorimetric substrates show that the enzyme cleaves tripeptides where the third Xaa could be indeed interchanged but only TRH natural substrate fulfills the requirement and is cleaved (74). Pyroglutamyl peptidase I only requires a pE residue without clear requirement for residues in the P2 or P3 position. Assessment of its involvement in A␤4-X genesis is still pending.

Biophysical properties of truncated peptides
Amyloid peptide aggregation is a hallmark of AD pathology. However, precise mechanisms leading to seed formation and accelerated aggregation during the disease progression are still under investigation. Infusion of brain extracts derived from AD transgenic mouse model in WT rodent brains is not per se sufficient to trigger A␤ aggregation (75). Also, aged synthetic A␤40 and A␤42 did not trigger deposits in APP23 transgenic mice (75). This suggests first that in physiological conditions, an efficient clearance mechanism occurs that keeps the A␤ concentration below its threshold of aggregation. This also indicates that cellular, genetic, or environmental cofactors may govern seeding formation and its acceleration in pathological conditions (75). Interestingly, this increased capability to induce full-length A␤ seeding has been described for several truncated A␤ variants such as A␤24 (43) and pE3A␤ (76).
Glutamate cyclization results in the loss of a negative charge that contributes to the hydrophobicity, resistance to catabolism (77), and aggregation propensity of N-terminally truncated A␤ species pE11-A␤ and pE3-A␤. pE3-42A␤ influences misfolding of full-length unmodified A␤ (76). Moreover, pEA␤ peptides are known to seed the aggregation of other A␤ fragments. It has been suggested that the C-terminal part of A␤42 was the locus of interaction with pE3-A␤ (78). Aside pE-A␤ variants, A␤ (4 -42) has also a propensity to form very stable aggregates (72). Fast aggregation properties of such fragments have been described in vitro (59,79) as well as in vivo in a Drosophila model (80).
Increased aggregation propensity and related toxicity have been described also for pyroglutaminylated forms of ADan and ABri, two peptides signing Familial Danish dementia and Familial British dementia, respectively (79,81,82).
Neuronal loss and synaptic loss in the hippocampus have been related to the early stages of AD. Several studies have pointed out the effects of truncated A␤ peptides on cellular toxicity. Some A␤ species protect against A␤42 toxicity, whereas others appeared to be more aggressive than full-length A␤ peptides. As described above, A␤34 protects cells overexpressing ␤APP WT or bearing Swedish mutation from caspase-3-related cell death (37), and therefore, in physiological conditions it could be considered as a beneficial isoform. This protective phenotype could be hampered by A␤34 aggregation. On the contrary, pE3-A␤ behaves like a noxious peptide. It has a high capacity to induce lipid peroxidation and to influence membrane permeabilization in primary cultured neurons (83). Toxicity of pE3-42A␤, assayed on rat cultured hippocampal neurons, appeared to be increased compared with the toxicity of full-length A␤ (77). More recently, development of transgenic mice expressing pE3A␤ peptides has helped to demonstrate that such truncated forms are implicated in neuronal loss (58). A Drosophila model has been characterized where pE3-42A␤ peptide is expressed in neurons. Interestingly, the life span of transgenic flies was affected by pE3-42A␤. Moreover, expression of the pyroglutamylated peptide induced behavioral dysfunctions, and toxicity was observed by a disorganization of the eye structure (80). Finally, pE3-42A␤ has recently been shown to induce synaptic plasticity impairment by different mechanisms than A␤42 and independently of a co-oligomerization process (84). Precise mechanisms involved in pE3-42A␤ toxicity have still to be addressed.

Truncated A␤ species in animal models
Several truncated and modified A␤ species have been found in AD murine models. C-terminally truncated A␤37 and A␤39 are widely expressed in the vasculature of human sporadic and familial AD patients (30). The same study also addressed A␤37 and A␤39 expressions in several transgenic AD mouse models (APP/PS1⌬E9, 5ϫFAD, PDAPP, APP23, 3ϫTgAD, and APP/ PS1KI). However, expression patterns appeared to be drastically distinct. In mice, both C-terminally truncated ␤-amyloid peptides were found in plaques, but vascular expression was almost absent in all the mouse models tested. Aside from A␤40 and A␤42, several other A␤ peptides were found in the 5ϫFAD mouse model. The more abundant truncated peptide detected by MS was A␤38, followed by A␤(4 -42), pE3-42A␤, and A␤39. A␤ (4 -40), A␤ , and A␤37 were also present but to a lesser extent (30).
A transgenic mouse model expressing the N-truncated A␤ (4 -42) peptide in the brain has been engineered (Tg (4 -42)) to investigate the effect of a chronic exposure of this toxic peptide that appears particularly abundant in human brain (57,70). This transgenic mouse model has been shown to express A␤ (4 -42) in the CA1 region of the hippocampus. However, such expression decreased with aging because of a massive neuronal loss in the region (72) associated with working memory dysfunction. Moreover, neurodegeneration was supported by early astrogliosis and microglial activation at only 2 months of age. The Tg(4 -42) mouse model showed spatial memory deficits starting at 5 months of age and being severely impaired at 6 months of age (85). In this model, A␤(4 -42) hippocampal expression correlates with a significant neuron loss in the CA1 layer of the hippocampus (85).
Another transgenic mouse model, with a glutamine instead of the glutamate at position 3 of the A␤ peptide, has been generated to examine the effect of cyclization on pathology development (86). The TBA42 mice that do not express ␤APP present a very rapid onset of symptoms, accumulate pE3-A␤, and harbor microglial activation and impaired LTP (58). These mice showed an age-dependent neuronal loss in the CA1 region of the hippocampus. TBA42 mice were then crossed to 5ϫFAD to engineer the FAD42 mouse model, which at 6 months of age showed an aggravated phenotype compared with the 5ϫFAD mouse model (87). This set of data corroborates the view of a pathological influence of pE3-A␤ species.

A␤ truncated species as biomarkers
Almost all truncated A␤ species, yielded by primary cleavages of ␤APP or secondary cleavages directly on A␤, have been detected in the CSFs and therefore could represent an interest as putative early biomarkers. Thirteen C-terminally truncated species have been detected in CSF (88,89). Quantitation of several A␤ species reveals, for example, that A␤(1-38) decreases in AD fluids compared with controls CSF (39). It appears that the pattern of truncated A␤ could help differentiate between some neurological disorders. For example, A␤(2-42) levels in cerebrospinal fluids are decreased in AD, although they are unchanged in fronto-temporal dementia (90). A␤(11-X) and A␤(17-X) peptides are of interest as new biomarkers for MCI detection because they have been identified in cerebrospinal fluids in patients with very early MCI (23).

Pharmaceutical strategies
As discussed in this Minireview, truncated amyloid species and pyroglutamate A␤ are very toxic forms of A␤, and thus, they correspond to potential therapeutic targets. One of the advantages of a therapeutic strategy aimed at abolishing truncated A␤ species-related pathology relies on the fact that such an approach will not interfere with any physiological function of soluble full-length A␤ (40,60,91) Accordingly, two main classical strategies could be envisioned that concern immunotherapy aimed at neutralizing truncated or modified A␤ or inhibitors of enzymes implicated in truncation and cyclization.
A specific antibody, targeting A␤4-X variants (NT4X-167), has been developed and characterized in transgenic mouse models as well as sporadic and FAD patients (73). Although the NT4X-167 antibody reacts with senile plaques in 5ϫFAD mouse models, a different pattern was found in AD brains where staining was observed in blood vessels as in Down's syndrome cases (92). Considering also A␤(4 -42) as a target for a therapeutic strategy appeared relevant because passive immunization with NT4X antibody decreases neuronal loss in CA1, rescues spatial memory deficits in Tg(4 -42) mouse model, and reduced amyloid plaques in 5ϫFAD mice (85). A chronic passive immunization against pE3 in the APPswe/PS1⌬E9 transgenic mouse model also triggers beneficial effects on plaque deposition, cerebral amyloid angiopathy, as well as gliosis (93). Another pharmaceutical strategy relies on the effect of QC inhibition to prevent formation of pyroglutaminylated A␤ species. Several QC inhibitors are actually in development (64,94). Obviously, a pre-requirement of these pharmacological approaches remains the firm identification of enzymes responsible for fragment formation. As stated above, in some cases, this identification is still awaited. Additional problems to overcome could be related to the lack of exclusive specificity displayed by these peptidases (61). Thus, close examination of the potential side-effects linked to proteolysis of additional substrates, a key feature that was initially underestimated when ␥-secretase inhibitors were designed and envisioned as therapeutic probes, will be a prior and redhibitory requirement.

Conclusion
The monitoring of full-length A␤ peptides (A␤(1-40) and A␤(1-42)) as biomarkers of AD pathology has to be deeply reconsidered with respect to the fact that many additional A␤-related species are generated and recovered in biological fluids.
Secondary cleavages onto a canonical A␤ peptide sequence have only been considered for a while as a clearance paradigm, aimed at depleting A␤(1-40) and A␤(1-42) and generating biologically inert bypass products. More likely, it appears that it gives rise to new players with potential pathological properties. Many A␤ truncated peptides that aggregate or favor seed aggregation yield variable oligomer profiles. Whether this represents a pathogenic signature accounting for specific differences observed in variable settings and the progression of AD in patients has to be envisioned.