Over and over in Alzheimer’s clinical trials, compounds have failed to live up to their preclinical prowess. In the October 6 JAMA Neurology, researchers led by Shauna Yuan at the University of California, San Diego, suggest that one problem may be differences between mouse and human neurons. In the latter, a common γ-secretase inhibitor is less potent, while a modulator turns out a different suite of Aβ fragments, according to the study. The researchers generated the human neurons from induced pluripotent stem cells (iPSCs) made from Alzheimer’s patients. “These cells may be more physiologically comparable to the human neurons that are most affected in Alzheimer’s disease, and may be more accurate for drug testing,” Yuan told Alzforum.

“In the future, neurons derived from human iPSCs will be a promising addition to secondary screenings and optimization of drugs that target neurons,” wrote Fan Liao and David Holtzman at Washington University School of Medicine, St. Louis, in an accompanying commentary.

Pharmaceutical companies have long been interested in curtailing the toxic Aβ42 peptide by targeting γ-secretase, the enzyme that makes the final cut snipping Aβ from the amyloid precursor protein. However, γ-secretase cleaves other crucial substrates as well, and adverse effects of γ-secretase inhibitors (GSIs) such as semagacestat scuttled human trials (see Aug 2011 conference storyNov 2012 news storyDec 2012 news story).

Companies have now turned to γ-secretase modulators (GSMs), which do not block enzyme activity but shift the cleavage site in favor of smaller Aβ fragments. Because these compounds allow the enzyme to cleave other substrates, they are believed to be safer than GSIs (see Apr 2011 conference storyFeb 2012 news story). Early GSMs shared structural properties with non-steroidal anti-inflammatory drugs (NSAIDs), but researchers have since developed second-generation, highly potent GSMs with a distinct, non-NSAID structure. These compounds enter the brain better than their precursors and effectively suppress Aβ42 production in rodents. One example, GSM-4, was developed by co-author Steven Wagner, also at UCSD, when he worked at the now-defunct TorreyPines Therapeutics.

Yuan and colleagues wondered how these drugs would perform in human neurons. First author Qing Liu generated iPSC lines from four AD patients and two age-matched controls. Each patient carried one of three different pathogenic mutations (A246E, H163R, or M146L) in presenilin 1, the catalytic component of the γ-secretase complex. Liu and colleagues differentiated the cells into neurons, then treated them with either semagacestat or GSM-4 for three days. They measured the concentration of secreted Aβ42, Aβ40, and Aβ38 using ELISAs. As expected, treatment lowered Aβ42, Aβ40, and the Aβ42/Aβ40 ratio in both control and patient cells.

The data also turned up surprises. One, human neurons required fivefold higher doses of semagacestat than transgenic mouse neurons to achieve the same suppression of Aβ42. Because clinical trials did not administer doses that high, semagacestat may never have reached effective concentrations in trial populations, Yuan speculated. Two, GSM-4 gave rise to an unexpected profile of Aβfragments in the induced neurons. In cell-free assays using purified mammalian γ-secretase and in animal studies, GSMs pumped up Aβ38 levels while slashing Aβ40 and Aβ42 (see Apr 2011 conference storyAug 2013 news storyJul 2014 news story). However, in the induced neurons from both patients and controls, the compound squelched Aβ38 along with the other peptides. Since total Aβlevels did not change, other fragments must have increased, the authors reasoned. By western blot, they think they made out an uptick in the Aβ37 and Aβ39 bands, but they detected no other fragments. “We suspect there are other, smaller fragments that went up as well,” Yuan told Alzforum, adding that Wagner will use mass spectrometry to clarify the issue.

While the results hint at differences between human and animal neurons, the iPSC system still needs to be correlated against human tissues, Yuan cautioned. Other researchers agreed. “Any cellular system, including iPS cells, needs to be validated by observing the same effect in humans,” said Samantha Budd at AstraZeneca in Cambridge, Massachusetts. AstraZeneca is currently investigating the potential of iPS cells for drug discovery, she added.

Others expressed skepticism, noting that the jury is still out on whether iPS cells make better models than transgenic mice. Todd Golde at the University of Florida, Gainesville, noted some drawbacks of induced neurons. They are time-consuming and expensive to make, and results from iPS cells do not always reproduce well. To interpret the data in the current study, researchers need to find out why GSM-4 shifted Aβ production differently in induced human neurons than in mice, Golde said. Was this due to a direct action on γ-secretase, or an effect of some unrelated factor such as toxicity? “It remains to be seen whether iPS cells really are a better intermediary step for drug testing,” Golde told Alzforum. Christian Haass at the Ludwig-Maximilians-Universität in Munich noted that mouse data are often sufficient to make good judgments. “The problems of semagacestat were fully predictable based on work on animal models. Here I do not see a major advantage of iPS cells,” he wrote to Alzforum. 

If these cellular systems pan out, they might advance the practice of personalized medicine, Holtzman and Liao suggested in their editorial. “Patient-specific neurons from iPSCs could provide unique models to study individual response to drugs,” they wrote. Yuan will test induced neurons made from patients who carry other familial AD mutations to determine how these variations affect drug efficacy. For example, in the current study, GSM-4 was twice as effective in suppressing Aβ42 in neurons that carried the A246E mutation as it was in control neurons.—Madolyn Bowman Rogers


  1. This paper brings up an important topic, namely the uncertainty of current approaches in the preclinical evaluation of novel drug candidates for Alzheimer’s disease. As stated by the authors, many studies have been performed on transfected cell lines and transgenic animal models that have a marked overexpression of the protein of interest, which may not correspond to how drug candidates act in human neuronal cells, or more so in the human brain.

    In the present study, the effect of γ-secretase modulators (GSMs) and γ-secretase inhibitors (GSIs) on Aβ metabolism was tested on human induced pluripotent stem cells (iPSCs). Following application of a GSM, a biomarker pattern characterized by a reduction in Aβ42, Aβ40, and Aβ38 without any change in total Aβ was found. To get this equation balanced, an increase in other Aβ species is to be expected, and using western blot instead of ELISA, an increase of smaller peptides including Aβ39 and Aβ37 was found.

    The results, which differ from those obtained by other cellular models and transgenic animals, may, as the authors suggest, be due to the huge overexpression of APP in these models. This suggestion is also supported by data from a previous study in which a GMS was tested in healthy beagle dogs. In ventricular CSF samples, a modest reduction in Aβ42 and Aβ40 was accompanied by a marked increase in Aβ37 (Portelius et al., 2010).

    The final question will be: Which model will most accurately reflect the biomarker signature, i.e., the effect on Aβ metabolism by drug candidates, in the human brain? Further translational studies, comparing experiments performed in different models (either cellular or animal) with results from human studies, will be necessary to answer this question. Encouraging data suggesting that small and short-term Phase 1 studies in healthy volunteers will provide solid proof of target engagement was recently reported in a study of a BACE1 inhibitor. A dose-dependent decrease in both plasma Aβ, CSF Aβ, and sAPPβ was found, corresponding to the change found in plasma and CSF in beagle dogs, and in brain tissue in APP transgenic mice (May et al., 2011). 


    . Acute effect on the Aβ isoform pattern in CSF in response to γ-secretase modulator and inhibitor treatment in dogs. J Alzheimers Dis. 2010;21(3):1005-12. PubMed.

    . Robust central reduction of amyloid-β in humans with an orally available, non-peptidic β-secretase inhibitor. J Neurosci. 2011 Nov 16;31(46):16507-16. PubMed.

    View all comments by Kaj Blennow

Make a Comment

To make a comment you must login or register.


News Citations

  1. Paris: Semagacestat Autopsy and Other News of Trial Tribulations
  2. Déjà Vu? AD Patients Again Look Worse on γ-Secretase Inhibitor
  3. Drug Company Halts Development of γ-Secretase Inhibitor Avagacestat
  4. Barcelona: Live and Learn—γ-Secretase Inhibitors Fade, Modulators Rise
  5. Paper Alert: γ-Secretase Modulators Trump Inhibitors
  6. Barcelona: Allosteric γ Modulation Moves Toward Clinic
  7. More Evidence that γ-Secretase Modulators Spare Essential Substrates
  8. Scientists Pinpoint γ-Secretase Modulator Binding Spot

Mutations Citations

  1. PSEN1 A246E
  2. PSEN1 H163R

External Citations

  1. M146L

Further Reading

Primary Papers

  1. . Effect of potent γ-secretase modulator in human neurons derived from multiple presenilin 1-induced pluripotent stem cell mutant carriers. JAMA Neurol. 2014 Dec;71(12):1481-9. PubMed.
  2. . Human neurons derived from induced pluripotent stem cells as a new platform for preclinical drug screening and development. JAMA Neurol. 2014 Dec;71(12):1475-6. PubMed.