Two recent papers tap an old hand and a relative newcomer as potential Alzheimer disease therapies—both are peptides but they have very different resumes. In the February 2 PNAS online, researchers led by William Klein at Northwestern University in Evanston, Illinois, offer increasing support for the idea that insulin, and insulin signaling in particular, protects against the ravages of toxic amyloid-β and could be used to ward off AD. In the January 28 Journal of Neuroscience, researchers led by Eliezer Masliah at the University of California, San Diego, report that fragments of the peptide neurotransmitter neuropeptide Y (NPY) also protect against Aβ in both cell culture and in transgenic mice. Interestingly, the NPY fragments are generated by none other than neprilysin, which is better known for degrading Aβ. The work suggests that neprilysin may be doing double duty in protecting against AD pathology. “This is the first time this type of dual activity has been demonstrated for neprilysin,” Masliah told ARF.

Growing evidence indicates that insulin signaling may be disrupted in AD. In addition to epidemiology linking diabetes with AD (see ARF related news story), researchers led by Konrad Talbot at the University of Pennsylvania, Philadelphia, have shown that insulin receptor (IR) activation gets lost as patients progress from a diagnosis of MCI to one of AD (see ARF related news story). More directly, work from Dennis Selkoe’s group at Brigham and Women’s Hospital, Boston, and previous work from Klein’s group has tied Aβ oligomers to IR dynamics. Selkoe’s group showed that Aβ inhibits autocatalytic activation of the receptors, while Klein’s group showed that the receptors are lost from neuronal dendrites under an Aβ onslaught. Now, Klein and colleagues extend those observations by demonstrating a tit-for-tat: while small, soluble Aβ oligomers (which they call ADDLs, or Aβ-derived diffusible ligands) drive uptake of insulin receptors, insulin signaling drives loss of Aβ oligomer binding sites. What those binding sites are is still not clear, but the evidence suggests that they are not the insulin receptors themselves.

“To us, it was a total surprise that insulin was doing to the ADDL binding sites what ADDLs were doing to the insulin binding sites,” Klein told ARF in an interview. Using hippocampal neuron cultures, first author Fernanda De Felice and colleagues found that not only did insulin completely block the ADDL-induced uptake of insulin receptors from dendrites, but it also prevented the oligomers from binding to neurons. In so doing, insulin reduced Aβ oligomer-induced loss of dendritic spines and oligomer-associated oxidative stress. The simplest explanation for these protective effects might be that insulin and the Aβ oligomers compete for the same binding site, namely the insulin receptors. “But of course the simplest explanation was too simple,” said Klein. “If we blocked insulin receptor activity, then insulin was totally without effect.”

The findings suggest that insulin signaling somehow downregulates surface ADDL binding sites. In support of this, rosiglitazone, a PPAR-γ (peroxysome proliferator-activating receptor-γ) agonist and activator of IR kinase activity, potentiated insulin’s effects, whereas AG1024, an IR protein tyrosine kinase inhibitor, eliminated the ability of insulin to reduce Aβ binding. Klein said it is not clear what the surface ADDL binding sites are, “but this work gives us a handle on that, potentially, because if the insulin is causing them to go down, we can use that to look for the missing smoking gun.”

Interestingly, the researchers found that the oligomer-induced loss of IRs on the neurons’ surface did not occur in the presence of inhibitors of casein kinase 2 (CK2) and calcium/calmodulin-dependent kinase II (CaMKII). Both enzymes are involved in activity-dependent uptake of NMDA receptors. Blocking CK2 completely prevented ADDL-induced loss of both IRs and NMDARs from hippocampal neuron dendrites, suggesting that the mechanism for receptor loss may be the same in both cases. Insulin has been shown to affect synaptic plasticity in an NMDAR-dependent manner (see van der Heide et al., 2005), but whether the NMDARs represent the Aβ binding sites that are sensitive to insulin signaling is not clear.

“Results strongly support the hypothesis that insulin signaling plays a role in defending CNS neurons against AD and provides a disease-specific basis for treatments based on stimulating CNS insulin pathways,” the authors write. Interestingly, Talbot’s work showing increased activation of IR downstream kinases, despite loss of IR in AD (see ARF related news story), might suggest that some kind of compensation is already going on in the brain. Insulin-based strategies, including intranasal administration of the hormone itself, and PPAR-γ agonists, have shown some efficacy in clinical trials (see ARF related news story), and GlaxoSmithKline is currently testing rosiglitazone in several Phase 3 clinical trials (see REFLECT-1, REFLECT 3, REFLECT-4, and REFLECT-5).

Intranasal administration to the brain is being tested for various therapeutic agents; it is one potential route for the neuropeptide Y fragments that Masliah’s group has identified. Consisting of a 36-amino acid peptide, NPY is a neurotransmitter released by inhibitory interneurons. It is highly expressed in the hippocampus, a site of early AD pathology. The peptide draws researchers’ curiosity mostly for its role in controlling appetite and circadian rhythms, but according to Masliah also has proliferative and neuroprotective activity and has been linked to memory function. A variety of proteases besides neprilysin degrade NPY, but the breakdown products they generate were not considered active. “Most people thought of this degradation as a terminal event,” said Masliah.

First author John Rose and colleagues tested the effects of increased neprilysin activity in the mouse brain. Boosting the protease is one potential mechanism of clearing Aβ from the brain (see ARF related news story), and the researchers wanted to see if doing so might have untoward consequences. They analyzed a number of neuropeptides (other than Aβ) in the hippocampus of mice expressing ~1.5-fold normal levels of neprilysin and found that only levels of full-length NPY were lower than normal. Using an antibody to the C-terminal of NPY and also mass spectrometry, the researchers found two major NPY fragments were generated, a piece corresponding to amino acids 21-36, and a smaller 31-36-amino acid fragment.

To see if these fragments are biologically active, the researchers infused them into the brains of normal and APP transgenic mice. After four weeks, sham-infused mice had lost dendrites (as judged by levels of neuronal marker MAP-2), while the dendritic arbor in mice infused with the peptides was not statistically different from that of non-transgenic animals. The findings suggested that the NPY fragments were neuroprotective, which the authors confirmed using human primary neuronal cultures. In those cells, pretreatment with the NPY fragments prevented synapse loss (as per synaptophysin staining) in response to Aβ42.

How these fragments protect against Aβ is not entirely clear, but they may work by binding to NPY receptors. There are six different NPY receptors (Y1 to Y6) with different affinities for full-length NPY and its fragments. That neprilysin can generate C-terminal fragments of NPY in vitro has been documented previously (see Medeiros Mdos et al., 1996), leading to the suggestion that they may be active. In support of that idea, Rose and colleagues showed that the protective effects of the C-terminal fragments of NPY are blocked by a Y2 inhibitor—a Y1 inhibitor had no effect.

These findings add to a growing number of studies that cast NPY in the umbra of AD research. The Y2 receptor has recently been implicated in learning and memory (see Redrobe et al., 2004) and NPY upregulation has been seen in mouse models of AD (see Diez et al., 2003) and may be related to epileptic seizures seen in animal models (see ARF related news story).

Whether these fragments could become useful therapeutically is another question. Their in vivo half-lives are unknown at present, and they would require viruses or intranasal delivery to get them past the blood-brain barrier.—Tom Fagan


  1. Insulin as a Potential Therapeutic Agent in Alzheimer Disease
    The belief that insulin has little if any role in brain function is outdated. There is now abundant evidence that insulin plays important roles in the brain, including synaptic diverse roles affecting cognition (van der Heide, 2006). Definitive proof for insulin generation by adult neurons is still lacking, but it is well established that pancreatic insulin is actively taken up across the blood-brain barrier and that relatively high levels of insulin receptors occur in various forebrain areas, including the hippocampal formation.

    The new paper by De Felice et al. in PNAS suggests that insulin may serve a protective function in Alzheimer disease. The authors demonstrate that insulin can protect cultured hippocampal neurons from deleterious effects of amyloid-β-derived diffusible ligands or ADDLs. They specifically show that insulin and, where tested, the insulin sensitizing drug rosiglitazone can impair or even prevent neuronal binding of ADDLs, dendritic spine loss, dendritic insulin receptor (IR) internalization, and oxidative stress. This raises the prospect of a new therapeutic strategy for retarding the progression of Alzheimer disease.

    Several questions are nevertheless raised by this report. The most basic is whether the effects are exclusively the result of effects on the IR. The effects reported were achieved with 100 nM and 1 uM insulin, which is far higher than the 1-5 nM concentrations at which insulin selectively activates the IR (Li et al., 2005). Even at 10 nM, insulin can activate insulin-like growth factor 1 (IGF-1) receptors (Li et al., 2005; Denley et al., 2007). Since IGF-1 receptors are abundant in the hippocampal formation (Doré et al., 1997) and in cultured hippocampal neurons (Doré et al., 1997), it remains possible that both IRs and/or IGF-1Rs mediate the neuroprotective effects reported by De Felice et al.

    Another question is whether the ADDL binding sites are IRs. This difficult question has been addressed previously by the same investigators (Zhao et al., 2008). Arguing against the idea are colocalization studies showing that all neurons in hippocampal cultures have abundant IRs, but only a lesser number of neurons binding ADDLs. This is evident in Fig. 1F of the De Felice paper, which shows at least one neuron with IRs but no ADDL binding in contrast to adjacent neurons. More importantly, co-immunoprecipitation studies of Zhao et al. (2008) established only that IRs are closely associated with ADDL binding sites. While their work also showed that ADDLs can bind isolated IRs, this occurred only under insulin stimulation. Similarly, De Felice et al. found that ADDL binding can be abolished simply by inhibiting the protein kinase activity of the IR’s intracellular segment with AG1024, which should not affect the extracellular alpha chain's ability to bind ADDL. So there is good reason to think that ADDLs do not directly bind IRs even though they may bind membrane receptors very close to them.

    What, then, are the ADDL binding sites? Could they be NMDA receptors? Despite some seemingly supportive evidence, this also proves unlikely. ADDLs promote rapid decreases in cell surface expression of not only IRs, but also NMDARs (Lacor et al., 2007). Moreover, the Zhao et al. (2008) showed that the effect of ADDLs on IRs was blocked by NMDAR antagonists. This raised the possibility that NMDARs may be binding sites of ADDLs, consistent with the finding that pretreatment of hippocampal cultures with antibodies to the extracellular domain of NMDARs reduces both ADDL binding and generation of ADDL-induced reactive oxygen species (Klein et al., chapter in Memories: Molecules and Circuits [B. Bontempi et al., eds.], 2007). Yet direct evidence for NMDARs as ADDL binding sites has not been obtained, and Zhao et al. (2008) note that after ADDL treatment dendritic loss of IRs occurs more quickly than that of NMDARs. Consequently, the results presented by De Felice et al. cannot be attributed entirely to an effect of ADDLs on NMDARs.

    How does ADDL binding promote IR internalization? Insulin binding of IRs is well known to promote internalization of those receptors, a process heavily dependent on phosphorylation of tyrosines 953 and 960 in the intracellular portion of the IR (Backer et al., 1992). ADDLs are known to inhibit phosphorylation of tyrosines 1162 and 1163 of the (Zhao et al., 2008), the sites important for transmitting insulin signals from the receptor to targets within cells. Though speculative, it is possible that decreased phosphorylation of tyrosines 1162 and 1163 promotes phosphorylation of tyrosines 953 and 960 and thereby triggers receptor internalization.

    While the findings of De Felice et al. do suggest that using insulin and/or an insulin sensitizer such as rosiglitazone may be a viable therapeutic approach for treating Alzheimer disease, it must be recognized that much work remains to be done before its actual potential can be assessed. There are two reasons for this. The findings are based on in vitro work using rat embryonic hippocampal neurons and, as noted above, 100 nM-1 uM insulin (with or without rosiglitazone), which is far beyond the normal levels of insulin in CSF (i.e., about 360 pM = 0.360 nM; see Craft et al., 1998). (CSF provides our best estimate of insulin in the extracellular spaces in vivo.) It thus remains to be shown if the effects reported by De Felice occur in adults in vivo without unwanted side effects. This can be tested in animal models of AD, in which it would be worth exploring whether the necessary amount of insulin and perhaps rosiglitazone can be administered intranasally given reports that intranasal insulin administration improves memory in Alzheimer cases (Reger et al., 2008).


    . Insulin signaling in the central nervous system: learning to survive. Prog Neurobiol. 2006 Jul;79(4):205-21. PubMed.

    . Insulin at physiological concentrations selectively activates insulin but not insulin-like growth factor I (IGF-I) or insulin/IGF-I hybrid receptors in endothelial cells. Endocrinology. 2005 Nov;146(11):4690-6. PubMed.

    . Differential activation of insulin receptor substrates 1 and 2 by insulin-like growth factor-activated insulin receptors. Mol Cell Biol. 2007 May;27(10):3569-77. PubMed.

    . Distribution and levels of [125I]IGF-I, [125I]IGF-II and [125I]insulin receptor binding sites in the hippocampus of aged memory-unimpaired and -impaired rats. Neuroscience. 1997 Oct;80(4):1033-40. PubMed.

    . Presence and differential internalization of two distinct insulin-like growth factor receptors in rat hippocampal neurons. Neuroscience. 1997 May;78(2):373-83. PubMed.

    . Amyloid beta oligomers induce impairment of neuronal insulin receptors. FASEB J. 2008 Jan;22(1):246-60. PubMed.

    . Abeta oligomer-induced aberrations in synapse composition, shape, and density provide a molecular basis for loss of connectivity in Alzheimer's disease. J Neurosci. 2007 Jan 24;27(4):796-807. PubMed.

    . The insulin receptor juxtamembrane region contains two independent tyrosine/beta-turn internalization signals. J Cell Biol. 1992 Aug;118(4):831-9. PubMed.

    . Cerebrospinal fluid and plasma insulin levels in Alzheimer's disease: relationship to severity of dementia and apolipoprotein E genotype. Neurology. 1998 Jan;50(1):164-8. PubMed.

    . Intranasal insulin administration dose-dependently modulates verbal memory and plasma amyloid-beta in memory-impaired older adults. J Alzheimers Dis. 2008 Apr;13(3):323-31. PubMed.

    View all comments by Konrad Talbot
  2. The comment that the cleavage of neuropeptide Y to generate a biologically active fragment by neprilysin (Neutral EndoPeptidase-24.11) is the first such example for the enzyme is incorrect. At least one example has previously been reported in the metabolism of calcitonin gene-related peptide (CGRP) (Davies et al., 1992).


    . Endopeptidase-24.11 cleaves a chemotactic factor from alpha-calcitonin gene-related peptide. Biochem Pharmacol. 1992 Apr 15;43(8):1753-6. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Type 2 Diabetes and Neurodegeneration—The Plot Caramelizes
  2. Diabetes-Insulin Roundup: Dementia Connection Grows Stronger, Part 2
  3. Madrid: Highs and Lows of The Insulin Connection
  4. DC: New Neprilysin Methods Reduce Brain Aβ
  5. Do "Silent" Seizures Cause Network Dysfunction in AD?

Paper Citations

  1. . Insulin modulates hippocampal activity-dependent synaptic plasticity in a N-methyl-d-aspartate receptor and phosphatidyl-inositol-3-kinase-dependent manner. J Neurochem. 2005 Aug;94(4):1158-66. PubMed.
  2. . Metabolism and functions of neuropeptide Y. Neurochem Res. 1996 Sep;21(9):1125-32. PubMed.
  3. . Characterization of neuropeptide Y, Y(2) receptor knockout mice in two animal models of learning and memory processing. J Mol Neurosci. 2004;22(3):159-66. PubMed.
  4. . Neuropeptide alterations in the hippocampal formation and cortex of transgenic mice overexpressing beta-amyloid precursor protein (APP) with the Swedish double mutation (APP23). Neurobiol Dis. 2003 Dec;14(3):579-94. PubMed.

External Citations

  1. REFLECT-1
  2. REFLECT 3
  3. REFLECT-4
  4. REFLECT-5

Further Reading


  1. . The insulin/Akt signaling pathway is targeted by intracellular beta-amyloid. Mol Biol Cell. 2009 Mar;20(5):1533-44. PubMed.

Primary Papers

  1. . Protection of synapses against Alzheimer's-linked toxins: insulin signaling prevents the pathogenic binding of Abeta oligomers. Proc Natl Acad Sci U S A. 2009 Feb 10;106(6):1971-6. PubMed.
  2. . Neuropeptide Y fragments derived from neprilysin processing are neuroprotective in a transgenic model of Alzheimer's disease. J Neurosci. 2009 Jan 28;29(4):1115-25. PubMed.