In reply to Letter to the Editor from Paudel: Comment on “Changes in iPSC-Astrocyte Morphology Reflect Alzheimer’s Disease Patient Clinical Markers”

Paudel comments on our recent publication, “Changes in iPSC-Astrocyte Morphology Reflect Alzheimer’s Disease Patient Clinical Markers,”¹ and highlights several aspects of our study that represent a major advancement in Alzheimer’s disease (AD) research. We are grateful for Paudel’s comments and recognition of the importance of our study. Here we build on Paudel’s commentary.

It is widely recognized that induced pluripotent stem cells (iPSC)-derived cells can model general aspects of AD pathogenesis, and many patient versus control studies have shown AD patient-specific differences in cellular and molecular phenotype. To date, most studies have understandably focused on neuronal phenotypes. The large number of studies in neurons indicates numerous phenotypes, including elevated levels and ratios of Aβ42, Aβ40, and Aβ38.^2-4 Other phenotypes include increases in total and phosphorylated tau,⁵ GSK3β,⁶ and synaptic,⁷ oxidative stress, and mitochondrial changes.^8,9 These studies include iPSC neurons derived from familial AD patients bearing pathogenic mutations in APP, PSEN1, PSEN2, but have been demonstrated in sporadic AD patients, with notably more variation in phenotype,³ reflecting diverse etiology leading to disease pathogenesis.¹⁰

In recent years, more research has been published on astrocytic phenotypes deriving astrocytes from familial and sporadic lines, notably focusing on AD risk genes such as APOE. These studies have also demonstrated wide-ranging phenotypes associated with AD, including key processes such as neurotrophic support,¹¹ morphology changes,¹² Aβ production and clearance, inflammatory response,^13,14 cholesterol and lipid metabolism,^7,15 oxidative stress, and calcium regulation.^16,17 These studies uncover cell-type specific insight into the mechanisms behind AD pathology, but cannot address the complexity and diversity of disease phenotypes across patients, with most of these studies typically using anywhere from 1 to 4 patient lines versus control comparison. However, these studies do not attempt to correlate the cellular and molecular phenotype with the clinical presentation of the patients from whom the iPSCs were derived.

Perhaps, the most significant finding of our study is the demonstration that iPSC-derived astrocytes derived from a panel of AD patients not only provide a general cellular model of AD but also exhibit phenotypes that correlate with patient clinical data. Specifically, we show that astrocytes from AD patients with high and low levels of the AD biomarker, CSF YKL-40, can be stratified based on how they undergo structural changes in response to Aβ. This builds upon an earlier study where we showed cognitive vulnerability correlated with individual cellular vulnerability to extrinsic amyloid-β in vitro, as measured by synapse loss and function from the same patient cohorts.¹⁸ Correlations between iPSC-neuron and CSF Aβ were also observed, where both of these findings have been echoed in a similar study revealing significant associations between specific Aβ and tau species in iPSC neurons and the levels of Aβ plaque and neurofibrillary tangle deposition in the brain and the trajectory of cognitive decline.¹⁹

These studies are important for 2 principle reasons. First, they allow patient stratification by providing cellular models of disease that reflect the heterogeneity of complex disorders, which can, in turn, lead to the discovery of early biomarkers for disease progression. Second, identification of patient-specific cellular phenotypes can lead to the development of novel therapeutics that target specific subsets of patients, potentially leading to tailored treatment strategies based on cellular phenotypes.

Moreover, in our study, we add another dimension to this by demonstrating the utility of AI to correlate astrocyte morphology with patient clinical data. AI is increasingly being deployed to interrogate cellular, molecular, and clinical data sets, aiding earlier detection and diagnosis (such as exampled by our study, brain imaging,²⁰ and biomarker discovery²¹), understanding of the underlying disease mechanisms, and drug discovery. Integration of multi-modal data by AI offers the ability to gain a comprehensive understanding of disease complexity. Moreover, AI has been used to impute gene expression profiles from cellular morphology in iPSC models of amyotrophic lateral sclerosis and expedite correlations between cellular and patient phenotype.²² This potential of AI to predict changes in molecular phenotype from cellular phenotype is an exciting prospect of these approaches, with huge implications for novel drug screening platforms.

In order to fully harness the potential of iPSCs to model disease heterogeneity, several challenges remain to be overcome. Scale is important; robust correlations require a large sample size to distinguish signal from noise to uncover subtle phenotypes. Large patient cohorts with attendant longitudinal data, coupled with an ability to measure cellular phenotype in multiple iPSC lines, are required. AI is revolutionizing the analysis of cellular phenotype and will change the face of drug discovery, and as shown by our study, provides a valuable tool in stratifying disease phenotype. We anticipate a surge of reports that use and extend these approaches to use iPSCs to interrogate the relationship between cellular and patient phenotype in multiple diseases, which, in turn, will greatly enhance patient stratification and personalized medicine.

Conflicts of interest

None declared.