Nonvesicular lipid transfer drives myelin growth in the central nervous system

Abstract

Oligodendrocytes extend numerous cellular processes that wrap multiple times around axons to generate lipid-rich myelin sheaths. Myelin biogenesis requires an enormously productive biosynthetic machinery for generating and delivering these large amounts of newly synthesized lipids. Yet, a complete understanding of this process remains elusive. Utilizing volume electron microscopy, we demonstrate that the oligodendroglial endoplasmic reticulum (ER) is enriched in developing myelin, extending into and making contact with the innermost myelin layer where growth occurs. We explore the possibility of transfer of lipids from the ER to myelin, and find that the glycolipid transfer protein (GLTP), implicated in nonvesicular lipid transport, is highly enriched in the growing myelin sheath. Mice with a specific knockout of Gltp in oligodendrocytes exhibit ER pathology, hypomyelination and a decrease in myelin glycolipid content. In summary, our results demonstrate a role for nonvesicular lipid transport in CNS myelin growth, revealing a cellular pathway in developmental myelination.

Fig. 1. Tubular ER is enriched in developing myelin and contacts the myelin membrane.

a Volume EM (ATUM-SEM) analysis of developing myelin from postnatal day 14 (P14) mouse optic nerve. Different compartments are marked on the bottom left image: axon (A), myelin (M) and inner tongue (I). ER membrane and axon membrane are segmented, and labeled in magenta and yellow, respectively. 3D reconstruction of ER network is shown together with (upper) or without (bottom) axon membrane, from an 11.25 µm-thick stack (226 slices with 50 nm interval, scale cube 1x1x1 µm). b Representative image of the TEM dataset for quantification of membrane-bound organelles in the inner tongue. Arrow heads: ER. c Examples of other membrane-bound organelles (arrows). d Ratio of inner tongues containing specified organelles, showing mean $\pm$SD, ER: 82.36 $\pm$ 1.51, others: 20.13 $\pm$ 3.80 (n = 3 wildtype P14 mice, two-tailed unpaired t-test, t = 26.37, df = 4, ****p < 0.0001). e Immunohistochemistry of developing myelin from P14 mouse spinal cord shows tubular ER markers (REEP5 and RTN4) appear as puncta and overlap with MBP marked myelin. f Ratio of myelin (MBP⁺ ring) overlapping with specified tubular ER marker. Quantification of (e) showing mean $\pm$SD, REEP5: 81.80 $\pm$3.97, RTN4: 72.79 $\pm$ 3.91 (n = 3 wild-type P14 mice). g An example with zoom-in views showing short distance between the ER and myelin membrane. h Distribution of inner tongue and axonal ER’s distance from the plasma membrane. For inner tongue ER, 62% ≤ 10 nm, 83% ≤ 20 nm, 90% ≤ 30 nm (n = 175 ER from three wild-type P14 mice). For axonal ER, 5% ≤ 10 nm, 16% ≤ 20 nm, 29% ≤ 30 nm (n = 337 ER from three wild-type P14 mice). Scale bars: 0.5 µm (a), 0.5 µm (b), 0.1 µm (c), 1 µm (e), 100 nm (g left), 50 nm (g zoom-in). Source data are provided as a Source Data file.

Fig. 2. Tubular ER is associated with active myelination.

a, b Immunohistochemistry of P14 mouse cortex. Arrow heads: pre-myelinating oligodendrocytes (MAG^-BCAS1⁺); arrows: myelinating oligodendrocytes (MAG⁺BCAS1⁺). c Quantification of (a) and (b) showing mean $\pm$SD. 1.65$\pm$0.55% or 3.54$\pm$2.60% premyelinating cells are REEP5⁺ or RTN4⁺, 93.04$\pm$0.25% or 82.34$\pm$4.16% myelinating cells are REEP5⁺ or RTN4⁺ (n = 3 wild-type P14 mice, two-way ANOVA followed by Sidak’s multiple comparison test, ****P < 0.0001). d, f Immunohistochemistry of P14 and six-month-old (6mo) adult mouse thoracic spinal cord, focusing on dorsal white matter e, g quantification of tubular ER (REEP5 or RTN4) fluorescence signal normalized by myelin (MBP) signal, in the dorsal white matter of thoracic spinal cord, showing mean $\pm$SD, (e) P14: 0.95$\pm$0.27, 6-month-old: 0.27$\pm$0.06 (n = 3 wildtype mice for each time point, two-tailed unpaired t-test, t = 4.302, df=4, *p = 0.0126) (g) P14: 1.35$\pm$0.30, 6-month-old: 0.46$\pm$0.13 (n = 3 wildtype mice for each time point, two-tailed unpaired t-test, t = 4.677, df=4, **p = 0.0095) h Left: Immunocytochemistry of primary oligodendrocyte culture for tubular ER markers (RTN4 and RTN1), rough ER marker SEC61B, ER sheet marker KDEL and oligodendrocyte marker O1 and MBP for outlining the cells. Right: The ratio of fluorescence signals from different ER subtypes at the cellular processes to that at the cell body, showing mean ± SD, KDEL: 0.58$\pm$0.10, RTN1: 0.79$\pm$0.11, SEC61B: 0.56$\pm$0.10, RTN4: 0.75$\pm$0.09 (n = 24 cells for KDEL/RTN1, 25 cells for SEC61B/RTN4, one-way ANOVA, F (3, 94) = 36.09, Tukey’s post-hoc test: **** p < 0.0001) Scale bars: 50 µm (a, b), 100 µm (d, f), 10 µm (h). Source data are provided as a Source Data file.

Fig. 3. Glycolipid transfer protein (GLTP) is associated with active myelination.

a Immunohistochemistry of P14 mouse cortex. Arrow heads: pre-myelinating oligodendrocytes (MAG^-BCAS1⁺); arrows: myelinating oligodendrocytes (MAG⁺BCAS1⁺). b Quantification of (a) showing mean $\pm$SD, 3.33$\pm$1.36% premyelinating cells are GLTP⁺, 92.62$\pm$3.08% myelinating cells are GLTP⁺ (n = 3 wild-type P14 mice, two-tailed paired t-test, t = 87.45, df=2, ***p = 0.0001) c Immunohistochemistry of P14 and six-month-old (6mo) adult mouse thoracic spinal cord, focusing on dorsal white matter. d quantification of GLTP fluorescence signal normalized by myelin (MBP) signal, in dorsal white matter of thoracic spinal cord, showing mean $\pm$SD, P14: 0.76$\pm$0.19, 6-month-old: 0.39$\pm$0.03 (n = 3 wildtype mice for each time point, Two-tailed unpaired t-test, t = 3.403, df=4, *p = 0.0272). e zoom-in view of P14 spinal cord f Ratio of myelin (MBP⁺ ring) overlapping with GLTP. Quantification of (e) showing mean $\pm$SD, 99.51$\pm$0.44% myelin overlaps with GLTP (n = 3 wild-type P14 mice). g Immunocytochemistry of primary oligodendrocyte culture showing GLTP subcellular localization. h The ratio of fluorescence signals from GLTP and KDEL at the cellular processes to that at the cell body, showing mean ± SD, GLTP: 0.54 ± 0.20, KDEL: 0.32 ± 0.09 (n = 24 cells, two-tailed paired t-test, t = 5.834, df=23, ****p < 0.0001). Scale bars: 50 µm (a), 100 µm (c), 1 µm (e), 10 µm (g). Source data are provided as a Source Data file.

Fig. 4. Mice lacking GLTP in oligodendrocytes exhibit ER pathology in myelin.

a Membrane rings in the inner tongue at the spinal cord of P14 Gltp cKO (Cnp-Cre, Gtlp^flox/flox). Arrows: thin rings; arrowheads: thick rings; Star: a comet tail-like structure associated with the ring; double arrowheads: rings in the outer tongue b Ratio (mean ± SD) of inner tongue containing rings: 0.08$\pm$0.13% of inner tongues in the optic nerve of Cnp-Cre (Cre on), 19.34$\pm$3.65% in the optic nerve of Cnp-Cre, Gtlp^flox/flox (cKO on), 33.90$\pm$6.87% in the spinal cord of Cnp-Cre, Gtlp^flox/flox (cKO sc) (n = 3 mice for each condition, from 436 + 537 + 543 inner tongues of “Cre on”, 508 + 525 + 531 of “cKO on”, 589 + 585 + 664 of “cKO sc”, One-way ANOVA: F (2, 6) = 42.77, Tukey’s post-hoc test: Cre on vs. cKO on **p = 0.0046, cKO on vs. cKO sc *p = 0.0173) c Volume EM of the optic nerve from Gltp cKO. Ring membrane is segmented in green, and 3D reconstruction of ring from a 1.5 µm-thick stack (31 slices with 50 nm interval), from dark green to light green across the stack. Scale cube: 1 × 1 × 1 µm. d Amount of ER decrease in the presence of rings, suggested by the ratio (mean ± SD) of inner tongue containing the ER: 90.34$\pm$3.33% of ring- vs. 44.69$\pm$6.01% of ring+ inner tongues in optic nerve of Cnp-Cre, Gtlp^flox/flox (cKO on), 90.67$\pm$5.47% of ring- vs. 51.31$\pm$2.78% of ring+ inner tongues in the spinal cord of Cnp-Cre, Gtlp^flox/flox (cKO sc) (n = 3 mice for each condition, one-way ANOVA: F (3, 8) = 86.07, Sidak’s multiple comparisons test: ****p < 0.0001) Scale bars: 1 µm (a), 150 nm (a zoom-in view) 200 nm (c). Source data are provided as a Source Data file.

Fig. 5. Mice lacking GLTP in oligodendrocytes exhibit hypomyelination and dysmyelination.

a Optic nerve of Gltp cKO and Cre control mice at P28. b, c g-ratio (a measurement of myelin thickness defined as inner diameter of myelin divided by outer diameter), quantification from three mice for each condition. b shows g-ratio of individual myelin relative to the enwrapped axon’s diameter. c shows the mean of myelin g-ratio from each mouse, showing mean $\pm$SD, Cre: 0.754$\pm$0.015, cKO: 0.783$\pm$0.003 (n = 3 mice for each condition, Two-tailed unpaired t-test, t = 3.282, df=4, *p = 0.0304) d mean of inner tongue diameter from three mice for each condition, showing mean $\pm$SD, Cre: 0.080$\pm$0.004 µm, cKO: 0.107$\pm$0.016 µm. (n = 3 mice for each condition, Two-tailed unpaired t-test, t = 2.943, df=4, *p = 0.0422) e examples of myelin whorls f Ratio of myelin whorls, showing mean $\pm$SD, Cre: 0.424$\pm$0.170%, cKO: 2.852$\pm$0.425% (n = 3 mice per condition, Two-tailed unpaired t-test, t = 9.203, df=4, ***p = 0.0008). g Percentage of axons that are myelinated, showing mean $\pm$SD, Cre: 83.720$\pm$6.148%, cKO: 76.780$\pm$11.43% (n = 3 mice per condition, two-tailed unpaired t-test). Scale bars: 1 µm (a and e). Source data are provided as a Source Data file.

Fig. 6. Delivery of glycolipid to myelin is impaired in Gltp mutants.

a–f Lipidomics analysis of myelin purified from Gltp cKO (Cnp-Cre Gltp^flox/flox) and Cre control (Cnp-Cre) mice (n = 4 mice for each genotype/time point). Source data are provided as a Source Data file. a Volcano plot shows differentiated regulated lipid species in cKO vs. Cre control. Vertical lines mark -/+ 1.5 fold change. Data points above horizontal line have p value < 0.05. Data were grouped based on genotype (Cre and cKO), and transformed to log2 for two-tailed Welch’s t-test. b and c Top 10 decreased and increased lipid species sorted by fold changes. P values are available in the source data. d Relative amount (mole %) of HexCer containing normal fatty acid (NFA) or 2-hydroxy fatty acid (HFA), mean $\pm$SD: P14-Cre-NFA: 2.15$\pm$0.39, P14-cKO-NFA: 1.25$\pm$0.08, P28-Cre-NFA: 2.44$\pm$0.19, P28-cKO-NFA: 1.78$\pm$0.17, P14-Cre-HFA: 2.65$\pm$0.44, P14-cKO-HFA: 2.42$\pm$0.11, P28-Cre-HFA: 4.61$\pm$0.3, P28-cKO-HFA: 4.62$\pm$0.2. (Two-way ANOVA followed by Tukey’s post-hoc tests, P14-Cre-NFA vs. P14-cKO-NFA ***p = 0.0003, P28-Cre-NFA vs. P28-cKO-NFA **p = 0.0088). e Relative amount (mole %) of various lipid classes. Color code is the same as in d. (Two-way ANOVA followed by Tukey’s post-hoc test. Stars indicate significant changes compared to the control of same age. P28-Cre-Chol vs. P28-cKO-Chol ****p < 0.0001, P14-Cre-HexCer vs. P14-cKO-HexCer *p = 0.0278, P14-Cre-PE O- vs. P14-cKO-PE O- ****p < 0.0001, P14-Cre-PS vs P14-cKO-PS ****p < 0.0001, P28-Cre-PS vs P28-cKO-PS ****p < 0.0001). f Principal Component Analysis (PCA) of lipid species from the samples. Each sample is myelin from one animal. g Mouse primary oligodendrocyte’s extracellular(magenta) and intracellular(green) GalCer signal obtained before and after Triton X-100. h Quantification of WT and KO cells from three mice for each condition, showing mean $\pm$SD, WT (Gltp ^fl/fl): 4.32$\pm$3.18, cKO (Cnp-Cre, Gltp ^fl/fl): 2.50$\pm$2.78. (n = 286 for WT, n = 265 for cKO, two-tailed unpaired t-test, t = 7.127, df=549, ****p < 0.0001). P value is calculated by Student’s t-test. i Quantification of non-targeting and Gltp siRNA, showing mean $\pm$SD, non-targeting siRNA: 1.90$\pm$1.43, Gltp siRNA: 1.38$\pm$1.13. (n = 47 for nt, n = 50 for siRNA, Two-tailed unpaired t-test, t = 1.986, df=95, *p = 0.0500) Scale bars: 50 µm g. Source data are provided as a Source Data file.

Fig. 7. Graphical illustration of model showing transport of GalCer in developing myelin.

a In wild-type myelin, GLTP transfers GalCer (red) from the ER to myelin membrane. Double-head arrow indicates GalCer reach an equilibrium in ER. Single-head arrow indicates that a yet-to-be-identified machinery transports GalCer to the extracellular leaflet of myelin membrane, keeping GalCer low in the intracellular leaflet of myelin membrane, creating a gradient and establishing the transfer direction as ER-to-myelin membrane. b Ring formation in Gltp cKO occurs in two steps. Step 1: GalCer attract each other and zip up the ER from the lumen. Step 2: GalCer accumulating in the cytoplasmic leaflet of ER make the ER to roll up. Created with BioRender.com.