iPSC-derived myelinoids to study myelin biology of humans

Abstract

Myelination is essential for central nervous system (CNS) formation, health, and function. Emerging evidence of oligodendrocyte heterogeneity in health and disease and divergent CNS gene expression profiles between mice and humans supports the development of experimentally tractable human myelination systems. Here, we developed human iPSC-derived myelinating organoids ("myelinoids") and quantitative tools to study myelination from oligodendrogenesis through to compact myelin formation and myelinated axon organization. Using patient-derived cells, we modeled a monogenetic disease of myelinated axons (Nfasc155 deficiency), recapitulating impaired paranodal axo-glial junction formation. We also validated the use of myelinoids for pharmacological assessment of myelination-both at the level of individual oligodendrocytes and globally across whole myelinoids-and demonstrated reduced myelination in response to suppressed synaptic vesicle release. Our study provides a platform to investigate human myelin development, disease, and adaptive myelination.

Keywords: adaptive myelination; disease modelling; human myelination; human stem cell; iPSC; myelin; myelinoid; oligodendrocyte; organoid; paranode.

Graphical abstract

Figure 1

Generation and characterization of iPSC myelinoids (A) Schematic of protocol for generating spinal cord-patterned organoids. (B) Representative images of CNP⁺ oligodendrocytes, NF-H⁺ axons, SOX10⁺ oligodendroglial cells and GFAP⁺ astrocytes at MI-0 (scale bar, 250 μm). (C) Representative images of CNP⁺ oligodendrocytes, NF-H⁺ axons, SOX10⁺ oligodendroglial cells and GFAP⁺ astrocytes at MI-12 (scale bar, 250 μm). (D) Heatmap of qRT-PCR-derived assessment of rostral and caudal gene expression in MI-0 organoids. Forebrain, LHX2; midbrain, OTX2; hindbrain, HOXB4; cervical, HOXB5, cervical, HOXB8. 1/ ΔCt values are normalized to 18S rRNA expression levels, n = 3 batch-conversions per cell-line. (E) Immunostaining of neuronal dendrites, axons, and cell bodies by MAP2, NF-H, and NEUN, respectively (scale bar, 25 μm). (F) Immunostaining of ChAT⁺ motor neurons (scale bar, 10 μm). (G) Immunostaining of ChAT and parvalbumin (PV) shows differentiation of distinct neuronal subtypes (scale bar, 25 μm). See also Figure S1.

Figure 2

Temporal development of myelin formation in iPSC myelinoids (A) Representative images of myelin development between MI-0 and MI-12 (scale bar, 250 μm). (B) Representative images of oligodendrocyte morphology between MI-0 to MI-4 (Bi and Bii) and MI-8 to MI-12 (Biii and Biv; scale bar, 10 μm). (C) Representative image of SOX10⁺CNP⁺MBP⁺ oligodendrocytes in MI-12 myelinoids (scale bar, 25 μm). (D) The proportion of mature oligodendroglial cells (MBP⁺CNP⁺/SOX10⁺) was stable between MI-4 and MI-12 at 18.1% (linear mixed effects regression, no change). Two images were taken per myelinoid from n = 22 (MI-4), 24 (MI-8), and 29 (MI-12) myelinoids across 3 cell lines (3–5 batch-conversions each). (E) The proportion of mature MBP⁺CNP⁺ oligodendrocytes engaged in myelination increased by 4-fold between MI-4 and MI-12 (95% CI: 3.42- to 4.89-fold; p < 0.001); generalized linear mixed model (GLMM) with time point as fixed effect and unique myelinoid ID as random effect. Two images were taken per myelinoid from n = 22 (MI-4), 24 (MI-8), and 29 (MI-12) myelinoids across 3 cell lines (3–5 batch-conversions each). (F) Immunostaining of MBP, NF-H, and MAP2 shows colocalization of myelin only on NF-H⁺ axons (scale bar, 25 μm). (G) Frequency distribution of MBP⁺ myelin sheath lengths between MI-4 and MI-12. Myelin sheath length increased by 52% between MI-4 and MI-12 (95% CI: 44% to 60%; p < 0.001); GLMM with time point as fixed effect and unique myelinoid ID as random effect, n = 1,061 sheaths from 6 myelinoids (MI-4), 754 sheaths from 5 myelinoids (MI-8), 1,083 sheaths from 5 myelinoids (MI-12). (H) Representative images of myelinated ChAT⁺ axons (white arrows), scale bar, 10 μm. (I) Representative images of myelinated PV⁺ axons (white arrows), scale bar, 10 μm. Boxplots show the medians, interquartile ranges and Tukey-style whiskers that extend to 1.5 times the interquartile range. n.s. = not significant, ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. See also Figure S2.

Figure 3

Structural organization of myelinated axons and myelin compaction (A) Representative image of MBP and CASPR immunostaining at MI-12 (scale bar, 25 μm). (B) Paranodal and nodal assembly in MI-12 myelinoids: CASPR and CLAUDIN-11 highlight PNJ formation and ANKYRIN-G highlights node formation. Pan-Nfasc identifies both Nfasc186 at the node and Nfasc155 at the paranode. White dotted lines mark these boundaries. Each pair of figures are from different myelinated axons. Scale bars, 5 μm (CASPR and CLDN-11) and 2 μm (ANK-G and pan-Nfasc). (C) Toluidine blue staining of MI-12 myelinoids reveals presence of compact myelin (scale bar, 5 μm). (D) Representative TEM images of compactly myelinated axons at MI-12 (scale bars for Di, Dii, Div, and Dv, 1 μm, scale bars for Diii and Dvi, 500 nm). (E) Scatterplot of axon diameter against g-ratio (axon diameter divided by fiber diameter) with a logarithmic regression curve (124 axons from 6 myelinoids across distinct conversions from MI-8 onward).

Figure 4

Nfasc155^−/− patient-derived myelinoids recapitulate disease pathology of disordered myelinated axon organization (A) Immuno-staining of neuronal dendrites, axons, and cell bodies in Nfasc155^−/− MI-12 myelinoids by MAP2, NF-H, and NEUN respectively (scale bar, 25 μm). (B) Immunostaining of ChAT⁺ and PV⁺ neurons in Nfasc155^−/− MI-12 myelinoids (scale bar, 25 μm). (C) Immunostaining of DAPI and GFAP⁺ astrocytes in Nfasc155^−/− MI-12 myelinoids (scale bar, 25 μm). (D) Representative image of cryosectioned Nfasc155^−/− MI-12 myelinoid stained with DAPI, MBP, CNP, and SOX10 (scale bar, 25 μm). (E) Analysis of the proportion of SOX10⁺ oligodendroglial cells showed no difference between Ctrl and Nfasc155^−/− MI-12 myelinoid cryosections (GLMM with cell-line as fixed effect and unique myelinoid ID as random effect); n = 3 myelinoids each. (F) Analysis of the proportion of mature MBP⁺ oligodendroglial cells showed no difference between Ctrl and Nfasc155^−/− MI-12 myelinoid cryosections (GLMM with cell-line as fixed effect and unique myelinoid ID as random effect, n = 3 myelinoids each. (G) Representative image of Nfasc155^−/− patient-derived myelinoid at MI-12 (scale bar, 250 μm). (H) Representative images of pan-Nfasc, CASPR and MBP across two Ctrl cell lines and two independently generated Nfasc155^−/− cell-lines (clone 1 and clone 2) at MI-12. Nfasc155^−/− myelinoids lack paranodal neurofascin expression and demonstrate disrupted PNJ formation. Axonally expressed neurofascin at the node remains intact (5–10 nodes analyzed per condition from < 3 myelinoids each; scale bar, 2 μm). (I) Representative images of ankyrin-G, CASPR and MBP across two Ctrl cell-lines and two independently generated Nfasc155^−/− cell-lines at MI-12. ANK-G expression further demonstrates that nodal assembly is preserved in Nfasc155^−/− myelinoids (5–10 nodes analyzed per condition from < 3 myelinoids each; scale bar, 2 μm). Boxplots show the medians, interquartile ranges and Tukey-style whiskers that extend to 1.5 times the interquartile range. n.s. = not significant. See also Figure S3.

Figure 5

Myelinoids predictably respond to pharmacological cues at both individual cell and whole-myelinoid levels (A) Representative image of tiled area for tracing individual myelin sheaths per cell (scale bar, 100 μm). (B) Manual tracing of CNP⁺ myelin sheath lengths. CASPR⁺ PNJs (red arrows) precisely overlap the distal ends of manually traced myelin sheath lengths (magenta) (scale bar, 100 μm). (C) Analysis of sheath number per cell (hPSC1: mean 12.5 ± 8.5 sheaths, hPSC2: 11.3 ± 5.1 sheaths, hPSC3: 12.3 ± 7.3 sheaths). n = 211 cells across 17 myelinoids (hPSC1), 110 cells across 9 myelinoids (hPSC2), and 90 cells across 9 myelinoids (hPSC3) from 2–5 conversions each. (D) Analysis of mean sheath length per cell (hPSC1: mean 93.0 μm ± 29.8 μm, hPSC2: 85.6 μm ± 20.4 μm, hPSC3: 67.5 μm ± 28.5 μm). n = 211 cells across 17 myelinoids (hPSC1), 110 cells across 9 myelinoids (hPSC2), and 90 cells across 9 myelinoids (hPSC3) from 2–5 conversions each. (E) Percentage frequency distribution of myelin sheath lengths at MI-12. n = 211 cells across 17 myelinoids (hPSC1), 110 cells across 9 myelinoids (hPSC2), and 90 cells across 9 myelinoids (hPSC3) from 2–5 conversions each. (F) Skeletonized CNP⁺ myelin sheaths of individual cells from Ctrl and blebbistatin-treated cultures (scale bar, 25 μm). (G) Sheath number per cell was increased in blebbistatin-treated cultures by 44.5% compared to Ctrl (95% CI: 10% to 89%; p = 0.007); GLMM with treatment as fixed effect and unique myelinoid ID and batch-conversion as random effects; n = 94 cells from 7 myelinoids (Ctrl) and 60 cells from 6 myelinoids (blebb) across 2 conversions. (H) Mean sheath length per cell was decreased in blebbistatin-treated cultures by 24% compared to Ctrl (95% CI: 10% to 35%; p < 0.001); GLMM with treatment as fixed effect and unique myelinoid ID and batch-conversion as random effects; n = 94 cells from 7 myelinoids (Ctrl) and 60 cells from 6 myelinoids (blebb) across 2 conversions. (I) Representative images demonstrating automated segmentation of MBP⁺ myelin sheaths at a particular Z-step. MBP expression is shown in greyscale and segmented objects overlaid in color. (J) Automated analysis of myelin volume normalized to NF-H intensity over time revealed an overall 8-fold increase between MI-4 and MI-8 (95% CI: 3- to 24-fold; p = 0.006) and a 15-fold increase between MI-4 and MI-12 (95% CI: 6- to 40-fold; p < 0.0001); GLMM with timepoints as fixed effects and batch-conversion and cell-lines as random effects. n = 19 (MI-4), 13 (MI-8) and 34 (MI-12) myelinoids across three cell-lines (2–7 conversions each). (K) Representative images of MBP expression in Ctrl and BDNF-treated whole-mounted myelinoids (scale bar, 250 μm). (L) Automated analysis of myelin volume normalized to NF-H intensity revealed a 2.09-fold increase in BDNF-treated myelinoids (95% CI: 1.13-fold to 2.74-fold; p = 0.0024); GLMM with treatment as fixed effect and batch-conversion as random effect). n = 13 (Ctrl) and 13 (BDNF) myelinoids from 3 conversions each. (M) Automated analysis of NF-H intensity showed no change between Ctrl and BDNF-treated myelinoids (GLMM with treatment as fixed effect and batch-conversion as random effect). n = 13 (Ctrl) and 13 (BDNF) myelinoids from 3 conversions each. Boxplots show the medians, interquartile ranges and Tukey-style whiskers that extend to 1.5 times the interquartile range. n.s. = not significant, ^∗∗p < 0.01, ^∗∗∗p < 0.001. See also Figure S4.

Figure 6

TeNT suppresses human myelination (A) Skeletonised CNP⁺ myelin sheaths of individual cells from Ctrl and TeNT-treated cultures (scale bar, 25 μm). (B) Sheath number per cell in Ctrl and TeNT-treated myelinoids. There was an overall 20% reduction in sheath number per cell in TeNT-treated myelinoids (95% CI: 6% to 31%; p = 0.006); GLMM with treatment as fixed effect and unique myelinoid ID and cell-line as random effects. hPSC1: 19% reduction (95% CI: 3% to 33%; p = 0.02); hPSC2: 23% reduction (95% CI: 0.4% to 40%; p = 0.04); n = 314 cells from 23 myelinoids (Ctrl) and 290 cells from 20 myelinoids (TeNT) across two cell-lines and 5 conversions. (C) Mean sheath length per cell in Ctrl and TeNT-treated myelinoids. There was an overall 14% increase in mean sheath length per cell in TeNT-treated myelinoids (95% CI: 4% to 21%; p < 0.01); GLMM with treatment as fixed effect and unique myelinoid ID and cell-line as random effects). hPSC1: 12.5% increase (95% CI: 2% to 23%; p = 0.018); hPSC2: no statistical difference; n = 314 cells from 23 myelinoids (Ctrl) and 290 cells from 20 myelinoids (TeNT) across two cell-lines and 5 conversions. (D) Nearest neighbor analysis shows no change in oligodendrocyte density between Ctrl and TeNT-treated myelinoids (GLMM with treatment as fixed effect and unique myelinoid ID and cell-line as random effects). n = 49 cells from 12 myelinoids (Ctrl) and 51 cells from 9 myelinoids (TeNT) across two cell-lines and 4 conversions. (E) Representative images of Ctrl and TeNT-treated myelinoids (scale bar, 250 μm). (F) Automated analysis of myelin volume normalized to NF-H intensity demonstrated an overall reduction of 38% in TeNT-treated myelinoids (95% CI: 4% to 52%; p = 0.017); GLMM with treatment as fixed effect and individual batch-conversions and cell-lines as random effects). hPSC1: 35% reduction (95% CI: 1.4% to 57% p = 0.043); hPSC2: 65% reduction (95% CI: 4.1% to 87.3%; p = 0.041); n = 29 (Ctrl) and 22 (TeNT) myelinoids across two cell lines and 5 conversions. (G) Automated analysis of NF-H intensity (GLMM with treatment as fixed effect and individual batch-conversions and cell-lines as random effects, no change. n = 29 (Ctrl) and 22 (TeNT) myelinoids across two cell lines and 5 conversions. Boxplots show the medians, interquartile ranges and Tukey-style whiskers that extend to 1.5 times the interquartile range. n.s. = not significant, ^∗p < 0.05. See also Figure S5.