Changes in iPSC-astrocyte morphology reflect Alzheimer's disease patient clinical markers

Abstract

Human induced pluripotent stem cells (iPSCs) provide powerful cellular models of Alzheimer's disease (AD) and offer many advantages over non-human models, including the potential to reflect variation in individual-specific pathophysiology and clinical symptoms. Previous studies have demonstrated that iPSC-neurons from individuals with Alzheimer's disease (AD) reflect clinical markers, including β-amyloid (Aβ) levels and synaptic vulnerability. However, despite neuronal loss being a key hallmark of AD pathology, many risk genes are predominantly expressed in glia, highlighting them as potential therapeutic targets. In this work iPSC-derived astrocytes were generated from a cohort of individuals with high versus low levels of the inflammatory marker YKL-40, in their cerebrospinal fluid (CSF). iPSC-derived astrocytes were treated with exogenous Aβ oligomers and high content imaging demonstrated a correlation between astrocytes that underwent the greatest morphology change from patients with low levels of CSF-YKL-40 and more protective APOE genotypes. This finding was subsequently verified using similarity learning as an unbiased approach. This study shows that iPSC-derived astrocytes from AD patients reflect key aspects of the pathophysiological phenotype of those same patients, thereby offering a novel means of modelling AD, stratifying AD patients and conducting therapeutic screens.

Keywords: YKL-40; astrocyte; deep learning; morphology; stem cell.

Graphical Abstract

Figure 1.

Characterization of iPSC-derived astrocytes from patients with high and low CSF YKL-40 values. (A) Astrocytes were matured to over 150 days, and then treated for 24 hours in either control conditions (blue) or with 10 μM Aβ25-35 (green). Astrocytes were subsequently stained for Hoechst (blue), YKL-40 (yellow), NF-κB (red), s100B (green). (B) Expression of s100B was calculated for each patient line with and without Aβ treatment as determined by mean intensity expression. (C) YKL-40 expression. (D) NF-κB expression. Scale bar represents 50 μm. Lines indicate patients 1-4 have “high” levels of CSF YKL-40 (shown in red) and patients 5-8 have ‘low’ levels of CSF YKL-40 (shown in green).

Figure 2.

(A) Cell width to length of control iPSC-astrocytes (blue) or treated with 10μM Aβ25-35 (green). Lines indicate patients 1-4 have “high” levels of CSF YKL-40 (shown in red) and patients 5-8 have “low” levels of CSF YKL-40 (shown in green). (B) Correlation of CSF YKL-40 levels to the change in cell width to length between control and Aβ25-35 treated*. (C) Correlation of CSF YKL-40 to the APOE value of the patient where higher AD risk genotypes equates to a higher APOE value. (D) Correlation of APOE value to change in cell width to length between control and Aβ25-35 treated. Red represents patients with high levels of CSF YKL-40, with green representing low levels. *Data remain significant with the removal of patient 7 representing the greatest change in cell width to length (R² = 0.6944, P < .0001).

Figure 3.

Use of deep learning to compare effect of Aβ treatments in an unbiased approach. (A) Schematic demonstrating the network setup using images from the s100B channel. A model was developed to determine how similar images in the s100B channel were from the untreated control of each patient were to its training dataset, and how images of astrocytes treated with Aβ were to their respective controls, Similarity learning was applied where 0 represents no similarity, and 1 an identical image (B) Similarity of the images of the test dataset to that of each patient’s control training dataset. The control test dataset is shown in blue, the Aβ treated test dataset is shown in green. Lines indicate patients 1-4 have “high” levels of CSF YKL-40 (shown in red) and patients 5-8 have “low” levels of CSF YKL-40 (shown in green). (C) Correlation of the change in width to length ratio of control and Aβ-treated astrocytes to the similarity of control images minus the similarity of Aβ treated astrocytes (where higher values represent more divergence). (D) Correlation of the change in similarity of images to patient YKL-40 CSF levels. Red represents patients with high YKL-40 CSF levels, green represents patients with low YKL-40 CSF levels. (E) Cell width to length measured in isogenic iPSC-astrocytes from four individuals edited to carry either APOE ε3/ε3 or ε4/ε4. Cell width to length was only significantly different in the ε3/ε3 astrocytes after Aβ treatment.