03 Jun 2021

In recent years, researchers have honed the craft of transforming skin cells directly into neurons. Now it turns out that these neurons have an advantage over their better-known stem-cell-derived counterparts, according to a study led by Fred Gage at the Salk Institute in La Jolla, California. Directly induced neurons reflect the age of their donors, and even whether they came from a person who has Alzheimer's disease. Though the skin fibroblasts were spared the degenerative disease that was raging in the brain, neurons induced from them adopted an AD signature. Their neuronal identity was shaky, their synaptic function weak, and they showed signs of stress. This signature may have been passed on via epigenetic modifications. In contrast, neurons derived from induced pluripotent stem cells (iPSCs) showed a much milder disease signature. The findings, published April 27 in Cell Stem Cell, suggest that induced neurons may make more accurate models of AD than iPSC-derived ones.

Transcriptome of induced neurons matches that seen in AD.
Induced neurons from people with AD are immature, have few synapses.
These iNs reflect the age of their donors; iPSC-derived neurons do not.

IPSC-derived neurons have become the workhorses for investigating disease mechanisms at the cellular level. However, because they make a pit stop in the embryonic state before differentiating into neurons, iPSC-derived cells are imbued with the characteristics of young cells. Alternatives are neurons induced directly from another differentiated cell type. By using the just-right mix of transcription factors, researchers can transform fibroblasts straight into neurons, bypassing the pluripotent step (Jan 2010 news; Herdy et al., 2019).

Previously, Gage and colleagues had reported that unlike iPSC-derived neurons, which possess attributes of newborn neurons, induced neurons express a transcriptome that reflects the age of the donors (Oct 2015 news). Might they also bear the stamp of neurodegenerative disease?

Finally, LOAD Neurons? Starting with skin biopsies from AD patients and controls, fibroblasts are transformed into neurons. Compared to control iNs, AD iNs have fewer synapses but more stress and cell cycle gene expression. They seem de-differentiated, with open chromatin. iPSC-derived neurons show few signs of disease. [Courtesy of Mertens et al., Cell Stem Cell, 2021.]

To address this question, first author Jerome Mertens and colleagues generated induced neurons (iNs) from fibroblasts of 16 people with AD and of 19 age-matched, non-demented controls per neuropsychological testing and MRI. Then they sequenced both the fibroblast and the iN transcriptomes. The fibroblasts that came from people with AD and controls had no significant differences. However, after conversion to neurons, 778 genes were differentially expressed between the two groups. What's more, this iN AD signature overlapped extensively with gene-expression profiles of neurons plucked directly from postmortem brain samples of people with AD. Genes involved in synaptic transmission were turned down, while stress response and cell cycle re-entry genes were cranked up. “These data indicate that known characteristics of the disease, including functional neuronal failure, stress response, and cell cycle re-entry, are reflected in AD iNs,” the authors wrote.

The donor's disease state was also borne out by functional deficits seen in the iNs. Compared to iNs from controls, those from people with AD had fewer synapses, fired them less frequently, pumped out more reactive oxygen species, and had signs of DNA damage. Curiously, they appeared to take on a hypo-mature state in that their gene-expression pattern suggested a de-differentiation back toward a stem-cell fate. The latter finding jibes with prior studies suggesting the same happens to neurons in the AD brain (Arendt, 2012).

How was this AD signature imparted to fibroblasts, then awakened in neurons? Suspecting epigenetic modifications, the researchers investigated chromatin. They discovered vast stretches of open chromatin in AD iNs, as if transcriptional control was loosening globally. They identified 7,188 stretches of the genome that were differentially accessible in AD iNs compared to control iNs. Of these, 97 percent were more open in AD neurons. The authors believe this epigenetic erosion could explain the fading neuronal identity of AD iNs relative to control iNs. What causes the erosion is unclear, but tau pathology in the AD brain has been tied to a loosening of chromatin (Apr 2017 conference news).

Do iPSC-derived neurons share these characteristics? To investigate, the scientists prepared iPSC-derived neurons from a subset of 20 of their iN donors. They found no significant gene expression differences between iPSC-derived neurons from people with AD versus those from controls. Some gene expression differences nearly reached significance; they largely overlapped with the AD genes identified in directly induced neurons, suggesting a dampened signature.

Since age is the strongest non-genetic risk factor for AD, the researchers asked if the iNs' AD signature reflected a state of accelerated aging. They used several gauges of age, including DNA methylation patterns and expression of age-related genes. As expected, iNs closely aligned with adult neurons, while iPSC-derived neurons resembled fetal cells. Interestingly, the researchers detected no difference in aging markers between AD and control iNs, suggesting that AD is not merely a state of accelerated aging. Rather, AD signatures are distinct from normal aging, but dependent on age.

“The work provides a great model to study the link between aging and neurodegeneration,” commented Bart De Strooper of KU Leuven, Belgium. He thinks the age-related erosion of epigenetic control might be important in AD. “It turns the reasoning in the field upside down, suggesting that the pathological phenotype in AD is a consequence of pathological aging manifested as epigenetic alteration," he wrote to Alzforum (comment below).

De Strooper wondered how the defining neuropathological signature of AD relates to the pathological transcriptional signatures that cropped up in iNs. “The fact that the latter are captured in neurons derived from fibroblasts which were not exposed to amyloid or tangle pathology suggests a cell-autonomous, upstream effect,” he wrote. “How these then affect mechanisms that result in the classical neuropathology remains to be seen. “

The findings raise the question of whether directly induced neurons are better suited to model AD and other diseases of aging than their iPSC-derived counterparts. IPSCs are becoming widely available via a growing repository of isogenic cell lines that can be transformed into different neural cell types. The human induced pluripotent stem cell Neurodegenerative Disease Initiative (iNDI), led by Michael Ward at the National Institute of Neurological Disorders and Stroke and Mark Cookson at the National Institute on Aging, both in Bethesda, Maryland, will introduce 134 disease associated mutations into iPSC lines (May 2021 news). Because fibroblasts are recalcitrant to genome engineering and grow slowly, iNs are less well-suited for this purpose. In a comment to Alzforum, Cookson, Ward, and Andy Singleton of the NIA noted that while iNs may be uniquely positioned to detect signatures of sporadic AD, iPSC-derived neurons harboring familial AD mutations do have transcriptional abnormalities (Aug 2019 news; Feb 2021 news). “For these reasons, our strategy in iNDI complements experiments performed in fibroblast-derived neurons,” they wrote (comment below).—Jessica Shugart

Comments

- Mark Cookson
  National Institute on Aging
- Michael Ward
  National Institute of Health
- Andrew Singleton
  National Institutes of Health
- Posted: 03 Jun 2021
The paper by Mertens et al. identifies an important approach to using human cells in the context of both sporadic and familial AD. One especially intriguing possibility would be to identify specific age-related pathways that the fibroblast-derived neurons possess and use those pathways to age iPSCs. As demonstrated by Mertens et al., iPSC-derived neurons more closely resemble fetal neurons than adult neurons, so an ability to identify specific aging pathways that interact with genetic mutations would be a big step forward for the field. It should be noted that there are other aging mimics that have been used to try to address this problem (Miller et al., 2013; Vera et al., 2016), although there are some caveats about how well they may work (Pandya et al., 2021).

It is also important to note that other groups have indeed seen significant transcriptional abnormalities in iPSC-derived neurons harboring familial ADRD mutations in well-controlled studies (Kwart et al., 2019, and Guttikonda et al., 2021, were covered by Alzforum previously). There appear to be overlapping transcriptional signatures between these different studies, which likely implies that there are similarities between inherited and sporadic AD cases. Furthermore, for highly penetrant variants, we may be able to readily identify molecular signatures related to disease process, even in relatively immature cells. The choice of platform will likely be informed by whether one wishes to examine single monogenic causes of disease or try to capture more subtle polygenic risk, for which the relative effect of aging may be more substantial.

While the direct reprogramming of aged fibroblast into derived neurons is both technically and conceptually important, there are currently some limitations. They include 1) lack of an ability to efficiently genome-engineer fibroblasts, making generation of isogenic controls difficult, and 2) challenges in scaling up experiments due to slow growth and limited outgrowth potential. This results in significant hurdles to disseminating cell reagents to the community and prohibits certain types of experiments that require large cell numbers. For these reasons, our strategy in the IPSC Neurodegenerative Disease Initiative complements experiments performed in fibroblast-derived neurons (Ramos et al., 2021).

An isogenic series of mutant lines may be an appropriate strategy for creating cell models of relatively highly penetrant ADRD alleles that can be disseminated to other researchers and will likely enable us to identify molecular convergences across gene mutations that would otherwise be masked by intergenic variability across lines from different patients. Per George Edward Pelham Box’s aphorism, all models are wrong, but it remains our aspiration that some models might be useful.

References:

Miller JD, Ganat YM, Kishinevsky S, Bowman RL, Liu B, Tu EY, Mandal PK, Vera E, Shim JW, Kriks S, Taldone T, Fusaki N, Tomishima MJ, Krainc D, Milner TA, Rossi DJ, Studer L. Human iPSC-based modeling of late-onset disease via progerin-induced aging. Cell Stem Cell. 2013 Dec 5;13(6):691-705. PubMed.

Vera E, Bosco N, Studer L. Generating Late-Onset Human iPSC-Based Disease Models by Inducing Neuronal Age-Related Phenotypes through Telomerase Manipulation. Cell Rep. 2016 Oct 18;17(4):1184-1192. PubMed.

Pandya VA, Crerar H, Mitchell JS, Patani R. A Non-Toxic Concentration of Telomerase Inhibitor BIBR1532 Fails to Reduce TERT Expression in a Feeder-Free Induced Pluripotent Stem Cell Model of Human Motor Neurogenesis. Int J Mol Sci. 2021 Mar 23;22(6) PubMed.

Kwart D, Gregg A, Scheckel C, Murphy EA, Paquet D, Duffield M, Fak J, Olsen O, Darnell RB, Tessier-Lavigne M. A Large Panel of Isogenic APP and PSEN1 Mutant Human iPSC Neurons Reveals Shared Endosomal Abnormalities Mediated by APP β-CTFs, Not Aβ. Neuron. 2019 Oct 23;104(2):256-270.e5. Epub 2019 Aug 12 PubMed.

Guttikonda SR, Sikkema L, Tchieu J, Saurat N, Walsh RM, Harschnitz O, Ciceri G, Sneeboer M, Mazutis L, Setty M, Zumbo P, Betel D, de Witte LD, Pe'er D, Studer L. Fully defined human pluripotent stem cell-derived microglia and tri-culture system model C3 production in Alzheimer's disease. Nat Neurosci. 2021 Mar;24(3):343-354. Epub 2021 Feb 8 PubMed.

Ramos DM, Skarnes WC, Singleton AB, Cookson MR, Ward ME. Tackling neurodegenerative diseases with genomic engineering: A new stem cell initiative from the NIH. Neuron. 2021 Apr 7;109(7):1080-1083. PubMed.

View all comments by Mark Cookson View all comments by Michael Ward View all comments by Andrew Singleton

- Bart De Strooper
  UK Dementia Research Institute@UCL and VIB@KuLeuven
- Posted: 03 Jun 2021
In this groundbreaking work, Mertens and colleagues compare the transcription profiles in AD and control neurons derived either directly from fibroblasts or from iPSC cells from the same donors. The directly induced neurons maintain epigenetic changes associated with aging, while the iPSC cells are reprogrammed and lose this signature. The differences in gene expression between control fibroblast- and AD fibroblast-derived neurons are striking and indicate an important role for age-related “epigenetic erosion” in the development of AD. This is corroborated by demonstrating that the AD expression profile is lost when iPSC cells are used to generate neurons.

The work provides a great model to study the link between aging and neurodegeneration. It turns reasoning in the field upside down, suggesting that the pathological phenotype in AD is a consequence of pathological aging manifested as epigenetic alteration. The question is whether anything specific (e.g., polygenic risk) is upstream to this pathological profile.

One wonders what the secret is to maintaining a healthy epigenetic profile in the brain in very old age, as seen in the centenarian cohorts of Henne Holstege. More down to earth is how the defining neuropathological signature of AD relates to the pathological transcriptional signatures identified here. The fact that the latter are captured in neurons derived from fibroblasts that were not exposed to amyloid or tangle pathology suggests cell-autonomous, upstream effects. How these then affect mechanisms that result in the classical neuropathology remains to be seen.

A further development of the approach would be to integrate these neurons into more complex multicellular models of disease to learn how the epigenetic erosion in neurons (or other cells) affects vascular-glia-neuron interactions.

View all comments by Bart De Strooper