Azetidine-2-carboxylic acid-induced oligodendrogliopathy in vitro and the pathogenesis of multiple sclerosis

Abstract

Azetidine 2-carboxylic acid (Aze) is consumed by humans and can be misincorporated in place of proline (Pro) in myelin basic protein (MBP). In systemically treated mice Aze induced distinct oligodendroglial (OL) alterations mimicking those in multiple sclerosis (MS) patient normal-appearing white matter. Here, Aze induced an unfolded protein response (UPR), cytoplasmic MBP aggregation, apoptosis and tumor necrosis factor secretion in the human OL lineage MO13.3 cell line. These alterations were counteracted by equimolar Pro suggesting that they are due to Aze substitution for Pro in OL proteins. Gene set enrichment analysis demonstrated extensive Aze-induced alterations of cell cycle, cytoskeletal, organelle, transport, developmental, inflammation-associated and myelination pathways that are altered in OL in MS patients and in toxin and inflammatory MS animal models. These data provide mechanistic support for the hypothesis that Aze protein misincorporation during early life myelinogenesis might over time result in a progressive UPR culminating in a pro-inflammatory/immunomodulatory phenotype, intracytoplasmic MBP aggregation, accelerated senescence and apoptosis in OL. This could occur prior to and independent of an external immune stimulus such as a viral infection. Aze-induced pathological alterations might enhance subsequent antiviral and autoimmune responses and contribute to MS susceptibility, lesion pathogenesis, remyelination failure, neurodegeneration and clinical progression.

INTRODUCTION

Edward Rubenstein hypothesized in 2008 that early life dietary exposure to the noncanonical amino acid azetidine 2-carboxylic acid (Aze), and its substitution for proline (Pro) in CNS myelin proteins, specifically the Pro-rich myelin basic protein (MBP), contribute to the pathogenesis of multiple sclerosis (MS)^.1 We previously investigated the effects of systemic administration of Aze in neonatal and adult mice. Behavioral effects were similar to those in MBP-deficient mouse mutants; histopathologic analyses demonstrated dose-dependent alterations of CNS white matter oligodendrocytes (OL) and microglial nodules in intact, normal-appearing white matter (NAWM).² Immunohistochemistry (IHC) demonstrated that Aze induced an endoplasmic reticulum (ER) stress/unfolded protein response (UPR), a pro-inflammatory/immunomodulatory phenotype and apoptosis of OL. The OL in neonatal mice exposed to Aze in utero were particularly susceptible to Aze-induced effects, consistent with the hypothesis that toxicity results from Aze misincorporation in OL proteins during developmental CNS myelinogenesis. This novel oligodendrogliopathy occurred in the absence of an external immune stimulus or leukocyte infiltration and replicates neuropathological findings in NAWM in MS patients.

To understand the cellular and molecular mechanisms of alterations induced directly by Aze on human OL lineage cells, we analyzed its effects on the MO13.3 oligodendroglial cell line using immunocytochemistry (ICC), assays of apoptosis and protein aggregation and ELISA of culture supernatants. The pathologic and pro-inflammatory effects on the Aze-treated cells in vitro recapitulated those in mice treated systemically with Aze. Moreover, these effects were counteracted in the presence of equimolar amounts of Pro but not alanine (Ala), consistent with specific misincorporation of Aze into OL proteins in place of Pro. Gene set enrichment analysis (GSEA) identified alterations of multiple metabolic pathways in Aze-treated MO13.3 cells that parallel OL alterations reported in inflammatory and toxin-induced MS models and other human leukoencephalopathies, as well as in MS patient CNS tissues. The identification and characterization of Aze-induced effects are relevant to an understanding of MS susceptibility, pathophysiology, and clinical progression and pathologic heterogeneity and for the development of appropriate therapies that may counteract effects of progressive OL protein misfolding and OL loss in MS patients.

METHODS

MO13.3 human oligodendroglial cell line

The human glial cell line MO3.13 (Creative-Biolabs, Shirley, NY) is an immortal human-human hybrid cell line that expresses phenotypic characteristics of primary OLs and is widely used as an in vitro OL model.^3–6 MO13.3 cells were grown in “H Oligo Medium” consisting of Dulbecco’s MEM Medium (Sigma, St. Louis, MO) containing 10% fetal bovine serum (American Type Culture Collection, Manassas, VA) and 1× Pen/Strep (Sigma). The medium was filtered, 0.2 µm and stored aseptically at 2 to 8 °C until use. One mL vials of frozen cells were thawed and washed with 19 mL H Oligo Medium and centrifuged at 11,000 RPM for 7 minute at 4 °C. Cell pellets were resuspended in 10 mL of Medium and inoculated on 100× 15 mL plates. All incubations were in a humidified chamber at 37 °C, 5% CO₂. When the cells reached confluence, trypsin/EDTA (2 mL/plate) (2 to 5 minute), was used to dislodge the cells; they were then centrifuged and the pellets resuspended in fresh medium and transferred to Permanox plastic 4 chamber slides (Lab-Tek, Thermo Fisher) (0.5 mL/well). Aliquots were also placed in cryovials (1 mL/vial) and stored at −80 °C for subsequent passages.

Initial characterization, routine staining and immunocytochemistry (ICC)

In pilot studies, MO13.3 cells were treated in the slide chambers with concentrations of L-Azetidine (SC-263441, Santa Cruz Biotechnology, Dallas, TX) from 0 to 5 mM Aze in H Oligo Medium for 24 hours. The treated cells were washed briefly 3× with 0.5 mL PBS, fixed with 0.5 mL/well of 10% neutral buffered formalin (NBF) and stored sealed at 2 to 8 °C. The growth chamber divisions of the 4 chamber slides were then removed and the slides were stained with hematoxylin. Because of reduced cell viability at concentrations of 1 mM Aze (or greater), subsequent experiments using higher Aze concentrations were not performed. A separate analysis confirmed significantly reduced viability at concentrations of 0.5 mM (Supplementary Figure S1).

For initial characterization, ICC experiments were conducted using rabbit antibodies (Abs) to MBP and oligodendrocyte transcription factor 2 (OLIG2) with 5 to 10% normal rabbit serum as a control and with mouse monoclonal antibodies (mAbs) to external and internal faces of myelin proteolipid protein (PLP),⁷ with PBS substituted for primary mAb as a control (Table 1). For ICC experiments, the cells were grown as described above and fixed cells were rinsed X2 for 5 minutes in PBS (0.5 mL/well), quenched first with 20 mM glycine in PBS 15 minute at RT, washed 5 minute with PBS and then quenched with 0.5 mL 0.3% H₂O₂ for 15 minutes, rinsed X2 in PBS, blocked with 0.5% normal goat or horse serum in PBS (for rabbit Abs or mouse mAbs, respectively) for 20 minutes at room temperature (RT). Antibodies were diluted in 0.1-0.5% normal goat serum in PBS (rabbit Abs, Table 1A) or PBS (mouse mAbs, Table 1B), 0.5 mL/well. The slides were incubated overnight at 2 to 8 °C. The wells were then washed 3× 5 minute with PBS. Biotinylated goat anti-rabbit Ig (or horse anti-mouse IgG) secondary antibodies (Vector Laboratories, Burlingame, CA) (30 µL/4 mL PBS) diluted in 0.1% normal serum were added to each well and incubated 30 minute at RT. After washing X3 with PBS, the slides were incubated with Vectastain Elite ABC reagent (Vector) for 45 minutes at RT and washed. The plastic growth chamber divisions were then removed from the slides. Reaction products were visualized with 3,3′ diaminobenzidine (0.025 mg/50 mL PBS, filtered and 0.5 mL 3% H₂O₂ added). Durations of exposure for optimal visualization of reaction product were determined by direct observation of the slides under microscopy. The slides were then washed 2× in PBS and reaction product was post-fixed in NBF. Slides were counterstained with hematoxylin, mounted in Aqua Polymount (Polysciences, Inc., Warrington, PA), and coverslipped. All cell ICC staining and analyses were performed at minimum in duplicate.

Table 1.

Antibodies.

Antigen/cell type	Antibody/clone name	Species/isotype	Concentrationa	Source
Commercial antibodies
ATF3 (C-19)	SC-188	Rb P	1:400	Santa Cruz Biotechnology, Dallas, TX
ATF4 (pSer245)	NB100-81802	Rb P IgG	1:400	Novus Biologicals, Littleton, CO
ATF6	LS-B2516	Rb P	1:800	LifeSpan Biosciences, Seattle, WA
Caspase-3, active (cleaved form)	AB3623	Rb P	1:500	Millipore, Burlington, MA
EIF2S1 (phosphor S51) [E90]	ab32157	Rb mAb IgG	1:100	Abcam
GADD 153 (R-20)	SC-793	Rb P	1:400	Santa Cruz Biotechnology
Myelin basic protein	Ab 5864	Rb P	1:800	Sigma
NFκB p65 (A)	SC-109	Rb P	1:250	Santa Cruz Biotechnology
Olig2	Ab 9610	Rb P	1:400	Sigma
PERK (H-300)b	SC-13073	Rb P	1:200	Santa Cruz Biotechnology
XBP1	LS-B188	Rb P	1:200	LifeSpan Biosciences

	PLP epitope	Isotype	Concentration (in PBS)
Mouse anti-PLP mAbsb
F4.4C2	50-69	M IgG1κ	3 µL/mL
F3.9E9	50-69	M IgG2aκ	3 µL/mL
2D2	100-123	M IgG1κ	7 µL/mL
P7.6A5	178-191	M IgG1κ	4 µL/mL (1:600)
F4.8A5	200-219	M IgG1κ	8 µL/mL

^aAntibodies diluted in 0.5% normal goat serum in PBS (blocking buffer).

^bAnti-PLP mAbs were kindly provided by V.K. Kuchroo and Nassim Kassam (Harvard Medical School, Boston, MA).⁷

Abbreviations: mAb, monoclonal antibody, M, mouse; P, polyclonal antibody; PBS, phosphate-buffered saline, pH 7.4; PLP, myelin proteolipid protein; Rb, rabbit.

To assess Aze-induced apoptosis, ICC using anti-caspase-3 (Table 1), and terminal deoxynucleotidyl transferase biotin-dUTP nick end labeling (TUNEL) staining were performed on slides prepared as described for ICC using ApopTag peroxidase in situ apoptosis detection kits (S7100, Sigma), according to the manufacturer’s instructions.

Aze, Pro and Ala treatments

Initial studies were performed on the four chamber slides to determine the effects of concentrations from 0 to 1 mM Pro and Ala (L-proline and L-alanine, both from Santa Cruz Biotechnology) on the MO13.3 cells. Effects of 24- or 48-hour exposure, (duration depending on viability at 24 hours), were determined for Aze, Pro and Ala and included assays of apoptosis, nuclear localization of specific markers, for example, NFκB, ATF3, and cytoplasmic aggregates of MBP immunoreactivity. These experiments were performed at minimum in paired comparison staining experiments. To determine the capacity for Pro or Ala to counteract UPR and apoptosis induction by Aze, the experimental conditions of cell exposure in different chambers were: Aze alone, Pro (or Ala) alone, equimolar Pro (or Ala) and Aze, or medium alone control. The Aze, Pro and Ala concentrations in these experiments were 0.2 mM.

Semiquantitative analyses of ICC

To ascertain the frequency of cells with nuclear immunostaining with the various UPR-associated Abs and cytoplasmic aggregation of MBP, total and immunostained cells were counted in 10 to 20 40× microscopic fields. Fields were selected in areas of the stained slides with uniform immunostaining and minimal cell overlap; the fields were acquired using Olympus cellSens Imaging Software (Olympus-Life Science Solutions, Center Valley, PA). To assess the frequency of apoptosis, fields were centered on TUNEL- or caspase-3-positive apoptotic cells; total cells/field and numbers of apoptotic nuclei in the fields were then determined.

Protein aggregation assay

MO3.13 cells were seeded in 6-well plates at 40% confluency. The following day, cells were treated with Pro or with Aze at varying concentrations for 48 hours. Protein aggregation in Aze- and Pro-treated and control cells was determined using a PROTEOSTAT^® Aggresome detection kit (Enzo Biochem, Farmingdale, NY), following the manufacturer’s protocol. This assay detects aggregates of all proteins in cell suspension samples.

Tumor necrosis factor (TNF) ELISA

For analyses of tissue culture supernatants, 60 mL cell suspensions of MO13.3 cells in H Oligo Medium were plated on 100 mm × 15 mm quadrant plates (3 mL/quadrant); 3 plates/condition were used for dose response treatments; 2 plates/condition were used for competitive treatments. Dose response experiments were performed at 0, 0.2 mM, 0.6 mM and 1 mM concentrations of Aze to determine effects of Aze on TNF secretion. To determine the capacity for Pro or Ala to counteract Aze-induced TNF secretion, the cells were exposed to: medium/PBS alone, Aze, Pro or Ala alone, and equimolar Pro or Ala and Aze.

After 24 hours, 1× Halt protease inhibitor (Thermo Fisher) and EDTA (5 mM/mL) were added to each plate; the harvested samples were centrifuged (15,000g, 10 minutes), the pellets discarded and the supernatants (400 µL/tube) frozen and stored at −80 °C. Analyses were performed using a TNF alpha Human ELISA Kit (KHC3011, Invitrogen), according to the manufacturer’s instructions. The ELISA plates were read within 30 minute to 2 hours at 450 nm. All standards, controls and samples were analyzed in duplicate.

Statistical analyses

Tissue culture and ELISA data were analyzed using t-tests and Brown-Forsythe or Welch ANOVA as appropriate using GraphPad Prism 5 (GraphPad, Inc., La Jolla, CA).

RNA extraction

MO13.3 cells were grown to near confluence as above in 6 individual sterile polystyrene Petri dishes (100 × 15 mm, Fisher). For harvesting, the medium was removed and the monolayers were briefly washed 3× with PBS at RT. Medium was replaced with 10 mL medium with 0.6 mM L Aze or with 10 mL medium alone (N = 3 each). After 2 days, the medium was removed and the monolayers were washed with sterile PBS X2, treated with 2 mL trypsin/EDTA to dislodge the cells and transferred to 6 50 mL centrifuge tubes. They were then washed with 18 mL H Oligo Medium. Cell suspensions were centrifuged 5 minute at 11,000 rpm. Supernatants were discarded and cell pellets were resuspended in 5 mL PBS, centrifuged, and the buffer discarded. Cell pellets were used for RNA extraction using an RNeasy Plus Mini Kit (Qiagen, Hilden, Germany) at RT according to the manufacturer’s protocols.

Gene set enrichment analysis (GSEA)

Microarray chip hybridization and data analysis were performed on Aze- and control treated MO13.3 cell cultures (N = 3 each) at Boston University Microarray and Sequencing Resource Core Facility, Boston, MA. Normalization and quality assessment, Principal Component Analyses (PCA), differential expression and microarray analyses and statistical methods are described in Supplementary Data S1 GSEA Methods. To identify major up- and down-regulated pathways, the fgsea package (https://bioconductor.org/packages/release/bioc/html/fgsea.html) was used. Pathways were filtered and joined with matched gene symbols for downstream interpretation. Analyses were performed in R programming and relevant Bioconductor packages.

RESULTS

Characterization of the MO13.3 cell line

The MO13.3 line cells demonstrated morphologic and IHC findings consistent with different stages of maturation/differentiation of OL lineage cells including nuclear expression of OLIG2 (Figure 1). They also exhibited the capacity to form major protein components of CNS myelin PLP and MBP. The extent of myelin protein immunopositivity in different experiments was greater than that of nuclear Olig2 and nuclear UPR markers, ie, generally more than half of the cells (Figure 1). The extent of detection of myelin protein epitopes varied, however, because specific experiments were terminated at different time points based on overall features of the cells. MBP showed a predominantly cytoplasmic and often perinuclear immunostaining pattern corresponding to the synthesis in the cytoplasm and localization of MBP on the cytoplasmic face of compact myelin (Figure 1). Although there was cell body staining with the anti-PLP mAbs attributable to their cytoplasmic site of synthesis, the mAbs to external surface PLP epitopes (ie, PLP 50-69 and 178-191) frequently showed membrane surface staining (Figure 1, Supplementary Figures S2 and S8). The staining with the mAbs to these PLP surface epitopes differed from that of the mAb to the internal epitope PLP 100-123, which was not on the cell surfaces (Supplementary Figure S2). Thus, expression of major myelin proteins and the demonstration of different subcellular distributions of MBP and PLP epitopes suggests that the MO13.3 cells could differentiate to a myelin protein production phenotype akin to that of human OL in developing CNS tissue.

Figure 1.

Oligodendroglial differentiation in MO13.3 cells. ICC demonstrates predominantly nuclear staining for OLIG2. Immunostaining for both PLP and MBP is present in the cytoplasm (where they are synthesized). There is more cell surface staining for the PLP 50 -69 (mAb F3.9E9) epitope (red arrows) and predominantly cytoplasmic and perinuclear (as well as surface) staining for MBP (green arrows). Staining is absent in the PBS control. Arrows in the upper panels indicate cells magnified in the fields in the lower panels. Bar = 100 µm, applies to the upper panels.

Effects of Aze and Pro on MO13.3 cells: morphology, apoptosis and OLIG2

Aze induced dose-dependent morphologic alterations of OL included cytoplasmic shrinkage, apoptosis and overall decline in viability when exposed to concentrations up to 1 mM (Figure 2).

Figure 2.

Dose-dependent effects of Aze on MO3.13 cell morphology and apoptosis. Left column: Hematoxylin counterstained slides demonstrate shrinkage of the cells with increasing Aze concentrations. The 0.6 and 1 mM Aze panels demonstrate overall depletion and cytoplasmic vacuolation of some remaining cells (arrows). Middle column demonstrates Aze concentration-dependent increases in numbers of apoptotic (TUNEL-positive) cells. Only cells with granular nuclear fragmentation are apoptotic; stained round nuclei seen in all panels do not fulfill criteria for apoptosis. Right column demonstrates Aze concentration-dependent increases in numbers of Caspase-3-positive apoptotic cells. Insets and adjacent panels are higher magnification of images of cells indicated with arrows. Bars = 100 µm and apply to panels in the corresponding columns.

Semiquantitative analyses of Aze dose-dependent induction of apoptosis were demonstrated using caspase-3 ICC (Supplementary Figure S3A) and TUNEL staining (Supplementary Figure S3B). The induction of apoptosis by Aze was counteracted in the presence of equimolar Pro (Figure 3A-D). By contrast, no comparable alterations were induced from 0 to 1 mM Pro (Supplementary Figure S4). There were also mild Aze dose-dependent reductions in proportions of cells with nuclear staining for OLIG2 (Figure 4).

Figure 3.

Aze-induced apoptosis of MO13.3 cells is counteracted in the presence of equimolar Pro. (A) ICC demonstrates more numerous TUNEL-positive apoptotic cells (arrows) in the presence of 0.2 mM Aze as compared to PBS control, 0.2 mM Pro alone and equimolar Aze + Pro. Scale Bar = 25 µM, applies to all panels. (B) *P < .0001, Brown-Forsythe ANOVA. (C) ICC demonstrates more numerous Caspase-3-positive apoptotic cells and apoptosomes (arrows) in the presence of 0.2 mM Aze as compared to the other conditions. Scale bar = 50 µM, applies to all panels. (D) P < .0001, Brown-Forsythe ANOVA and Welch ANOVA. Data are representative of 3 separate experiments for each apoptosis marker. Additional statistical analyses are in Supplementary Data S5.

Figure 4.

(A) Aze induces a dose-dependent reduction in proportions of MO3.13 cells with nuclear expression of OLIG2. Bars = 20 µm. (B) ***P = .0024, Welch ANOVA. Data are representative of 2 experiments. Additional statistical analyses are in Supplementary Data S5.

Nuclear translocation of ER stress/UPR-related molecules

Aze induced dose-dependent nuclear translocation of PERK, ATF4, NFκB and XBP1 (Supplementary Figure S5). Aze-treated cultures showed more numerous cells with nuclear staining for ATF3, ATF4, ATF6, NFκB, PERK and XBP1 than control, Pro-treated and Aze + Pro groups (Figures 5 and 6). Positively stained cells tended to have more round nuclei and less cytoplasm than adjacent cells without nuclear staining (Figures 5 and 6). This suggests that those cells are undergoing ER stress but are not yet apoptotic. Equimolar Pro counteracted the effects of Aze on nuclear translocation of ATF3, ATF4, ATF6, NFκB, PERK and XBP1 (Figures 5 and 6), whereas Pro alone did not induce nuclear translocation of PERK, XBP1 or NFκB (Supplementary Figure S6).

Figure 5.

Equimolar Pro counteracts Aze-induced nuclear localization of UPR activation components ATF3 (A, B), ATF4 (C, D), and ATF6 (E, F). There are more numerous immunostained nuclei in the Aze-treated panels vs the other conditions. There is some cytoplasmic staining for ATF3 (A) and ATF6 (E), whereas ATF4 nuclear immunostaining is dense and in a chromatin-like pattern (arrows, C). ***P < .0001, One-way ANOVA (B); P = .0007, One-way ANOVA (D), **P = .0003, Brown-Forsythe ANOVA (F). Data are representative of 2 experiments for each marker. Scale bars = 20 µm. Additional statistical analyses are in Supplementary Data S5.

Figure 6.

Equimolar Pro counteracts Aze-induced nuclear localization of UPR activation components NFκB (A, B), PERK (C, D), and XBP1 (E, F). There are more numerous immunostained nuclei in the Aze-treated panels vs the other conditions. ***P ≤ .0001 (B). P = .0003 (D). P ≤ .0001 (F). Brown-Forsythe and Welch ANOVA. Data are representative of 2 experiments for each marker. Scale bars = 20 µm. Additional statistical analyses are in Supplementary Data S5.

Total protein and MBP cytoplasmic aggregation assays

In preliminary analyses, Aze induced dose-dependent MBP cytoplasmic aggregation assessed by ICC in MO13.3 cells whereas Pro did not (Supplementary Figure S7A, B). A protein aggregation assay of lysed cell suspensions demonstrated a trend of Aze dose-dependent increases in protein aggregation. Only the maximal concentration tested (0.5 mM) resulted in significant measurable aggregation propensity factor (APF) in the cell suspensions whereas this concentration of Pro did not induce any protein aggregation (Figure 7A). This assay detects all protein aggregates in lysed cell suspensions and therefore does not distinguish MBP aggregates from all other protein aggregates. Specific MBP-immunopositive aggregates were detected by ICC in MO13.3 cells that had been exposed to the lower concentration of 0.2 mM Aze (Figure 7B). The Aze-induced increases in cytoplasmic MBP aggregates were counteracted by exposure to equimolar Pro (Figure 7B and C). In contrast to effects on MBP, Aze did not induce a similar effect on the pattern of immunostaining of PLP epitope 178-191 by ICC (Supplementary Figure S8). Together, both the biochemical and immunocytochemical assays demonstrate that Aze but not Pro induced protein aggregation and that this appeared to be (at least in part) a specific effect on MBP. The two assays provide consistent and complementary results but cannot be directly compared (eg, by dosages) because assay methods and measures differ.

Figure 7.

Aze induces protein aggregation in MO3.13 cells. A: Aze induced dose-dependent total protein aggregation in lysed cell suspensions whereas Pro does not. Cell cultures were untreated or treated with 0.5 mM Pro, or with varying concentrations of Aze up to 0.5 mM for 48 hous; protein aggregation in total cell suspensions was then determined. Error bars represent SD from 3 independent experiments. *P < .05, **P < .01, n.s., not significant, one-way ANOVA. APF, aggregation propensity factor. (B) MBP aggregation in vitro assessed by ICC is counteracted in the presence of equimolar Pro. Immunostaining for MBP is mostly diffuse through the cytoplasm (black arrows) in PBS-, Pro-, and Aze + Pro-treated MO13.3 cells whereas more numerous cells with Aze only exposure show coarse irregular MBP-positive cytoplasmic aggregates (red arrows). The arrows indicate cells magnified in the adjacent panels. Two sets of panels from Aze alone-exposed cultures highlight multiple examples of MBP aggregation in contrast to the three other conditions. Scale bars: 20 µm. Cell exposures in the ICC experiments were: medium control (PBS), 0. 2 mM Pro alone, 0.2 mM Aze alone or 0.2 mM Aze + 0.2 mM Pro. (C) Semiquantitative analysis of MBP aggregates/field. ****P < .0001, Brown-Forsythe ANOVA; P = .0046, Welch ANOVA. Additional statistical analyses are presented in Supplementary Data S5.

TNF ELISA

Combined results of TNF assays of supernatants from treated and co-treated MO13.3 cells are shown in Figure 8. Aze induced a dose-dependent increase in TNF in supernatants from MO13.3 cells (left portion of graph) whereas the levels of TNF induced by 0.2 mM Pro or 0.2 mM Ala were not different from the level induced by PBS alone (right portion of graph). Exposure of the cultures to equimolar Aze and Pro (0.2 mM) reduced the production of TNF in comparison to 0.2 mM Aze alone (++ comparisons), whereas the level produced by equimolar Aze and Ala was not different from that resulting from exposure to 0.2 mM Aze alone (right portion of graph) (n.s. comparisons).

Figure 8.

Composite graph from 4 TNF ELISA experiments demonstrates that TNF secretion from MO13.3 cells is counteracted in the presence of equimolar Pro whereas equimolar Ala does not counteract Aze-induced TNF secretion. Left side of graph: There are Aze dose-dependent increases of TNF secretion from MO13.3 cells (***P < .0001, Brown-Forsythe ANOVA). Right side of graph: Equimolar Pro counteracted Aze-induced TNF secretion whereas equimolar Ala did not 0.2 mM Aze alone vs 0.2 mM Pro + 0.2 mM Aze, (++), P = 0.0135. 0.2 mM Aze alone vs + 0.2 mM Ala + 0.2 mM Aze (n.s.). Exposure to 0.2 mM Pro or Ala alone also did not result in increased TNF secretion (not significant vs PBS for both). Paired comparisons by 2-tailed t-tests.

Gene set enrichment analysis (GSEA)

There were 3346/6226 gene sets upregulated in Aze-treated vs control cultures; 1333 gene sets were significant at FDR < 25%; 786 gene sets were significantly enriched at nominal P value < 1%; 1138 gene sets were significantly enriched at nominal P value < 5%. There were 2880/6226 gene sets downregulated in Aze vs control. 45 gene sets were significantly enriched at FDR < 25%; 126 gene sets were significantly enriched at nominal P value < 1%; 359 gene sets were significantly enriched at nominal P value < 5%. An Excel file containing the values of all principal components (PCs) is presented as Supplementary Data S2 Principal Component Results. An Excel file summarizing the GSEA results is presented as Supplementary Data S3 GSEA Results. Heat maps were made of the leading edge genes of all gene sets with FDR q < 0.25 (Supplementary Data S4). Representative heat maps showing examples of gene set upregulation and downregulation of particular relevance to immune mechanisms including potential antiviral responses are shown in Figures 9-12. Bar plots displaying the top 15 most upregulated and 15 most downregulated pathways based on Normalized Enrichment Scores (NES) for KEGG, Reactome, and other expert-curated pathway databases and gene ontology sets are shown in Figure 13. Major metabolic pathways and processes altered by Aze treatment and relevant to MS pathophysiology are listed in Table 2.

Figure 9.

Representative leading edge heat maps of GSEA results of Aze exposure to MO13.3 cells showing examples of upregulated gene set pathways (see Supplementary Data S1 GSEA Analysis). (A) GO regulation of sister chromatid segregation. Normalized enrichment score (NES) = – 2.19, P = .0000, FDR q = 0.002. Arrow highlights the key CCCTC-binding factor (CTCF) (upregulated in 42 gene sets) (see Ref. 77). (B) Reactome of antiviral mechanism by interferon-stimulated genes. NES = 2.15, P = .0000, FDR q = 0.0003. Black arrows highlight STAT1, a major transcription factor involved in immune gene regulation in oligodendrocyte precursor cells (OPC), and the IFN- γ-induced gene IRF1. Red arrows highlight nuclear pore proteins (NUP), suggesting involvement in transport of protein aggregates across the nuclear envelope and myelination.

Table 2.

Pathways altered by Aze vs control treatment of M03.13 cells identified by gene set enrichment analysis.

Processes predominantly upregulated include:

Cell Cycle (DNA replication, DNA damage and repair, DNA metabolic processes, mitosis, histone methylation and acetylation)

Chromatin organization, assembly and remodeling, binding, nuclear heterochromatin, silencing

Nuclear membrane/envelope, nuclear pore, nucleosome organization, actin, microtubule-related

RNA splicing, editing, metabolism

Intracellular transport of small molecules (amino acids, proteins, nucleocytoplasmic, transmembrane)

Protein metabolism (localization to chromosome, cytoskeleton organization, nucleus, acetylation, catabolism)

Amino acid (eg, glutamine) metabolism and degradation

Unfolded Protein Response

Protein autoubiquitination

Iron metabolism

Lipid biosynthesis and storage

OPC development, differentiation, myelination, for example, NOTCH binding, Wnt/β catenin signaling

Myelin assembly, compact myelin, myelin sheath, myelin maintenance, demyelination

Organelle biogenesis, assembly, morphology, fission, maintenance, localization

Mitochondrial organization, transcription, matrix, genome maintenance

Cellular responses to stimuli including ER stress, oxidative stress, nitric oxide, reactive nitrogen species, toxic substances

Heat shock proteins, for example, HSP90AB1, sacsin

Programmed cell death/apoptotic signaling

Type 1 interferon production and signaling

Hormone responses (estrogen, testosterone, steroid hormones)

Aginga

Cytokine and chemokine pathwaysa

Responses to virusesa

Antigen processing and presentationa

Processes predominantly down-regulated include:

Mast cell activation

Hypoxia/stress responses (including hypoxia-inducible factor-1)

TNF signaling via NFκB

Fibronectin binding

Leukocyte chemotaxis

MHC I

Insulin-like growth factor binding

Collagen formation

q < 0.05 or P < .05 for processes listed.

^aPro-inflammatory gene sets with specific upregulated principal components (PC). See Supplementary Data 1 for statistical methods, Supplementary data 2 for Excel file of Gene Sets, Supplementary Data 3 for Excel file containing values of all PC, and Supplementary Data 4 Leading Edge Heat Maps.

Figure 13.

Bar plots displaying the top 15 most upregulated and 15 most downregulated pathways identified based on Normalized Enrichment Score (NES) using the fgsea package. Significance was assessed using adjusted P-values (Padj < .05). Positive NES indicates upregulation; negative NES indicates downregulation in Aze vs control groups. (A) C2 (Curated Pathways) includes KEGG, Reactome, and other expert-curated pathway databases. (B) C5 (Gene Ontology Sets) captures functional biological processes.

DISCUSSION

Mechanisms of Aze toxicity: misincorporation and the UPR

The dramatic effects of Aze on the morphology, immunophenotype and induction of apoptosis of MO13.3 cells demonstrated in this study recapitulate those induced by systemic exposure to Aze in mice. Non-CNS effects were also observed in the mice, particularly in the liver, but the most prominent and widespread effects in the CNS were in OL.²

MBP is the second most abundant CNS myelin protein.⁸ It is an intrinsically disordered multifunctional protein that interacts with many cytoskeletal and extracellular signaling proteins.^9–11 Therefore, its misfolding could have major effects on fundamental OL functions. MBP has more Pro residues overall than the more abundant myelin PLP (18 vs 6). Compared to PLP and MBP, 2′,3′-cyclic-nucleotide 30-phosphodiesterase (CNP), myelin oligodendrocyte glycoprotein (MOG), and myelin-associated glycoprotein (MAG) are much less abundant in CNS myelin.⁸ Moreover, MBP is the only major myelin protein with a structurally critical triproline repeat and is, therefore, the myelin protein most likely to have functionally significant alterations induced by Aze substitution for Pro in OL.¹² However, misincorporation of Aze only into MBP has not been demonstrated in the present study; misincorporation into OL proteins other than MBP, such as HIF proteins,¹³ is likely and deserves further investigation. Moreover, other potential mechanisms of Aze toxicity¹⁴ have not been excluded.

Aze induces the UPR in vitro in many non-CNS and CNS cell types, including mammalian neurons, astrocytes and microglia.^14–16 The doses, cell types and durations of Aze exposure vary among these in vitro studies precluding direct comparisons; however, adverse effects in the MO13.3 cells were induced using concentrations lower than those that induce injury in other cell types.¹⁴ Moreover, no abnormalities of CNS cells other than OL were observed in mice treated systemically with Aze, suggesting that OL have a greater sensitivity to Aze.² As in mouse OL in vivo, Aze induced the nuclear translocation of the major ER stress/UPR activation molecules ATF3, ATF4, ATF6, NFκB, PERK and XP1 (Figures 5 and 6). Upregulation of numerous associated UPR pathways, (eg, Hallmark unfolded protein response and GO unfolded protein binding), was also identified in the GSEA (Table 2, Supplementary Data S4).

The UPR is essential for translation of myelin proteins, the viability and functions of mature OL and myelin maintenance.¹⁷ When the UPR is overwhelmed with misfolded proteins, apoptosis ensues.¹⁸ Aze induced MO13.3 cell apoptosis (Figures 2 and 3), and the GSEA demonstrated upregulation of 30 apoptosis-related pathways (Supplementary Data S4, Table 2). (These can be identified by searching “apopto” in the HeatMap Supplementary Figure S4). In Aze-treated mice, in addition to OL apoptosis, we also previously demonstrated ultrastructural evidence of mitochondrial damage characteristic of inflammation-associated OL stress.^2,19 In the present study, the GSEA demonstrated extensive perturbations of 15 mitochondrial metabolic pathways and pathways related to oxidative stress and response to reactive nitrogen species (Figure 13). These included GSTO1,²⁰ (which was upregulated in 15 pathways) (Supplementary Data S4, Table 2). Therefore, oxidative and nitrosative stress were also likely among the inter-related pathologic effects of Aze exposure on MO13.3 cells.

Pro counteracts Aze effects

The ability for equimolar Pro to counteract Aze-induced apoptosis, nuclear translocation of UPR molecules, MBP cytoplasmic aggregation, and TNF secretion in MO13.3 cells (Figures 3 and 5-8), suggests that the presence of additional Pro in the culture medium may have decreased the extent of Aze misincorporation, consequent protein misfolding and downstream effects. Aze can bind to both human prolyl- and alanyl-tRNA synthetases but was found to be preferentially substituted for Pro in a mammalian protein.²¹ Therefore, in contrast to equimolar Pro, the failure of equimolar Ala to affect Aze-induced TNF production (Figure 8), supports the hypothesized specificity of Aze misincorporation in place of Pro in MBP in MO13.3 cells. We did not determine the minimal amount of Pro sufficient to counteract Aze effects in this study but Piper et al recently reported that 50 µM Pro was sufficient to counteract toxic effects of 1000 µM Aze on BV2 murine microglial cells and on primary mouse microglia; moreover, they reported Aze misincorporation in proteins in the treated BV2 cells by triple/quadrupole mass spectrometry.²²

Immunomodulatory phenotype in the absence an exogenous immune stimulus

Induction of the UPR by misfolded proteins results in the production of pro-inflammatory cytokines.²³ The secretion of the pleotropic cytokine TNF by MO13.3 cells suggests an Aze-induced pro-inflammatory phenotype (Figure 8). A pro-inflammatory OL phenotype was also indicated by the upregulation of NFκB MHC II reactome (Figure 10), and mTORC1 signaling pathways (Supplementary Data S4), and by downregulation of type I interferons, chemokines and the immunosuppressive cytokine IL-10 (Figure 12). The anti-inflammatory and tissue protective heparin-binding EGF-like growth factor (HB-EGF) was also downregulated (Figure 11A). These alterations might promote antigen presentation to and activation of CD4+ and CD8+ T cells and contribute to opening of the blood-brain barrier, enhancement of T cell recruitment, immune-mediated neurotoxicity, and impaired recovery from autoimmune inflammation.^24–27

Figure 10.

Representative leading edge heat map of GSEA results of Aze exposure to MO13.3 cells showing gene set upregulation of reactome MHC class II antigen presentation pathway (see Supplementary Data S1 GSEA Analysis). NES = 1.33, P = .0371, FDR q = 0.2145. Arrow highlights MHC Class II HLA-DPA1.

Figure 12.

Representative leading edge heat map of GSEA results of Aze exposure to MO13.3 cells showing downregulation of GO leukocyte chemotaxis. NES = −1.80, P = .0016, FDR q = 0.1774 (see Supplementary Data S1 GSEA Analysis). Arrows highlight downregulated chemokine, cytokine and related genes, such as CXCL13, which recruits B cells, CCL2, which facilitates OPC migration, and the immunosuppressive cytokine IL-10.

Figure 11.

Representative leading edge heat maps of GSEA results of Aze exposure to MO13.3 cells showing examples downregulated gene set pathways (see Supplementary Data S1 GSEA Analysis). (A) Hallmark TNFA signaling via NFKB. Normalized enrichment score (NES) = −2.28, P = .0000, FDR q = 0.012. Black arrows highlight oligodendrogliogenesis- and OL differentiation-associated genes EGR1, KLF9, JAG1, and CXCL1, and the cytokine IL-6, which modulates microglial responses. FOS, JUN, CCL2 and JUNB are also downregulated. Blue arrow indicates the anti-inflammatory factor HBEGF. (B) Reactome interferon αβ signaling. NES = −1.80, P = .0018, FDR q = 0.1836. Black arrows highlight the downregulation of the immunomodulatory transcription factors STAT2, SOCS1, SOCS3 (also in panel A), and interferon-induced proteins with tetratricopeptide repeats (IFITs), which may participate in anti-viral immunity and PSMB8 (immunoproteosome subunit); green arrow indicates the IFNAR2 chain of the human type I IFN receptor; red arrows highlight downregulated MHC I molecules.

On the other hand, Aze effects on MO13.3 cells were not exclusively “pro-inflammatory”. The GSEA indicated that TNFA, SOCS1 and SOCS3,²⁸ MHC class I and MHC class II molecules (other than HLA-DP1, Figures 10 and 13B), and many leukocyte chemotaxis factors, for example, CXCL13 and CCL2, were downregulated by Aze exposure (Figures 11A, B, 12 and Table 2). By contrast, it should be noted that increased expression of MHC class I was observed on apoptotic OL but not on intact-appearing OL in situ in Aze-treated mice by IHC²; this may have been related to apoptotic DNA damage.²⁴ Taken together, the Aze-induced alterations in the present study might be better characterized as “pro-inflammatory/immunomodulatory”.^29–34 These alterations may in part represent strategies for protection from metabolic stresses.¹⁷ Such differences also highlight the complexities and discrepancies among in vitro and in vivo models.

In vitro differentiation, myelinogenesis and in vivo myelin maintenance

Aze exposure resulted in profound effects on MO13.3 cell chromosomes, organelles, and cytoskeleton (including microtubules) (Figures 9A, 13B and Table 2). For example, it induced upregulation of 234 heterogeneous nuclear ribonucleoproteins (hnrnps) (Supplementary Data S4). There was an Aze dose-dependent decrease in nuclear immunostaining for Olig2, the key lineage transcription factor for oligodendroglial differentiation, myelination, myelin repair and maintenance of axonal integrity^35–38 (Figure 4). Only cells with strong nuclear staining for Olig2 were considered positive (∼3% in the control condition in the experiments shown in Figure 4), but the proportions of Olig2-positive and -negative cells for each experimental condition were highly consistent within individual experiments. Because IHC may not detect low levels of Olig2 and does not provide information on mRNA levels, this might represent an underestimate of Olig2 gene expression. Our finding of reduced proportions of cells with Olig2 nuclear expression following exposure to increasing Aze concentrations (Figure 4), suggests effects of Aze on differentiation capacity of the MO13.3 cells, however. Downregulation of Olig2 is also associated with greater susceptibility to stress of oligodendrocyte precursor cells (OPC), including susceptibility to TNF.³⁹ The GSEA also demonstrated that a number of genes involved in oligodendrogliogenesis, OPC differentiation and myelination, including KLF9, NDRG1, JAG1, TGFβ2, BNIP3L and CXCL1,^40–46 were downregulated (Figures 11A, B, 12 and Supplementary Data S4), further indicating that Aze had some negative effects on essential OPC/OL differentiation and functions.

By contrast, some pathways involved in oligodendrogliogenesis, promotion of OL differentiation (including estrogen, androgen and steroid receptor pathways), initiation and maintenance of myelination, cellular organization of MBP, and OL maturation following demyelination were upregulated following Aze exposure (Figure 13). Pathways of myelin assembly, compaction and sheath formation were upregulated (Table 2). (These can be identified by searching “myelin” in the HeatMap Supplementary Data S4). In particular, both PLP1 and MBP genes were unexpectedly upregulated. We speculate that increased or more rapid myelin turnover (as a consequence of MBP aggregation, degradation and UPR-induced apoptosis), might explain these observations. Another possibility is that there is a greater likelihood of misincorporation of Aze vs Pro in myelin; how this might translate to in vivo conditions and possible instability of compact myelin at the single OL level is unclear. Other upregulated pathways included SIRT1 (110 pathways), individual ion channel genes, plp1, proteins associated with cell polarity, nuclear pore proteins (NUP), mTOR (48 pathways), BRCA1/2 (222 pathways), ZEB2 (11 pathways), SWAP70 (22 pathways), PAK1 (15 pathways), and Notch binding and autophagy pathways (Figures 9B, 13A and Supplementary Data S4, Table 2).^47–62 Thus, despite alterations that could have negative effects on OL differentiation and functions, the cellular machinery for OL development, myelin synthesis and myelination appeared to be at least partially preserved or possibly enhanced with the addition of Aze to the MO13.3 cells in this short-term experimental paradigm.

MO13.3 cells, non-MS human OL disorders and MS animal models

Multiple omics studies interrogating drug and other agent effects on OL and OPC have shown alterations of pathways and toxic effects overlapping with those of Aze in MO13.3 cells. These include DNA damage repair and ubiquitination pathways (Table 2).^63–65 The GSEA results also overlap with omics studies that have demonstrated UPR activation, metabolic stress and misfolding of myelin proteins over time, (eg, TRIB3 [upregulated in 30 pathways]),^66–69 and in human white matter pathologies with diverse etiologies (ie, other than primary demyelinating diseases).^70,71 Therefore, many of the Aze-induced alterations identified herein likely represent common pathways of OPC and OL metabolic injury and senescence.

In particular, Aze-induced alterations in MO13.3 cells also overlap with those induced by cuprizone, a widely studied in vivo model of OL injury, demyelination and remyelination.⁷² For example, Hist1h3d, (upregulated in 42 pathways), is also upregulated in cuprizone-treated mouse OL precursors.⁷³ However, effects and molecular mechanisms of Aze-induced oligodendrogliopathy are clearly distinct from those of cuprizone.⁷⁴ Cuprizone and other experimental OL toxins also do not as closely replicate the alterations of OL nucleomegaly and apoptosis found in MS NAWM.^2,75 Most importantly, Aze is consumed by humans and can be implicated in the pathogenesis of MS whereas exposures to cuprizone as well as to other myelin toxins have not been implicated in the pathogenesis of the human disease.

In the MS model experimental autoimmune encephalomyelitis (EAE) in mice, STAT1 and Irf1 genes, among others involved in immune regulation, show increased expression^76,77; they were upregulated in MO13.3 cells by Aze (Figure 9B). Meijer et al reported that OPC and OL exhibited primed chromatin accessibility in a large subset of genes prior to inflammatory insults and that the transcription factor CTCF, which binds enhancers and promoters of many immune genes was increased in IFN-γ-treated OPCs⁷⁷; CTCF was upregulated as part of 42 gene sets in the present study (Figure 9A, Supplementary Data S4). It should be noted, however, that Meijer et al found that FOS, JUN, JUN B and Stat2 were enriched in OL in EAE whereas these were downregulated in MO13.3 cells in the present study (Figures 11A and B).

Epigenomic priming, microglia and inflammation in MS patient CNS tissues

Meijer et al also found that immune genes exhibiting a primed chromatin state in healthy OL are transcriptionally activated in MS patient tissue through a series of epigenetic activations including histone modification, transcription factor binding, and chromatin reconfiguration pathways⁷⁷; these were also upregulated by Aze (Figures 9A, 13B and Table 2, Supplementary Data S4). Thus, exposure to Aze might similarly result in the induction of epigenomic priming of immune-related genes in OL thereby enhancing immune responses in the CNS of MS patients.

TNF and pro-inflammatory mediators have long been implicated in immunopathological mechanisms in MS.^78–82 As suggested by the induction of microglial nodules in NAWM in Aze-treated mice in situ,² and by Aze-induced TNF secretion by MO13.3 cells, an Aze-induced pro-inflammatory/immunomodulatory OL phenotype might contribute to the microglial activation and priming that are characteristic of MS,^83,84 as well as of other CNS inflammatory conditions.^81,85 Heat shock genes are also upregulated in MS tissue in parallel with immune system gene upregulation^86,87; Aze induced upregulated heat shock stress proteins, for example, hsph1, hspd1 and hspa6 (Supplementary Data S4). Similarly, IL-10 downregulation, (as induced by Aze, Figure 12), has been associated with MS worsening.^88,89

On the other hand, Kirby et al found that there was high expression of the immunoproteosome subunit PSMB8 in OL lineage cells in MS lesions with failed remyelination and not in NAWM; they suggested that PSMB8 might promote chronic inflammation in demyelinated lesions.⁹⁰ By contrast, we found reduced levels of PSMB8 in Aze-treated MO13.3 cells (Figure 11B). Thus, there are continuing challenges in defining the complex interrelationships between inflammatory responses, OL and MS regional pathology.^31,91

The initiation of MS and responses to viral infection

More than forty years ago, the neuropathologist Ellsworth (Buster) Alvord Jr, (1922-2010), wrote, “As a pathologist, … I still do not know what the first attack of MS looks like or whether the subsequent attacks begin the same way as the first one did … fundamentally we are stymied until we know whether the first abnormality requires leucocytes or not …”.⁹² The initial lesion and primary cause (or causes) of MS continue to remain controversial but alterations of MBP have long been considered to be central to MS pathogenesis.^9–11 There are also numerous lines of evidence indicating primary or central roles of ER stress/UPR and OL apoptosis resulting from structural instability of CNS myelin in MS pathogenesis and pathology.^{2,75,93–102} The present findings are consistent with this “inside-out hypothesis” and, indeed, the earlier proposal by Wilkin that autoimmune inflammation is a response to a primary lesion in the target tissue.¹⁰³

A number of studies have implicated Epstein-Barr virus (EBV) infection as an early event or trigger of MS onset.¹⁰⁴ In view of the high prevalence of EBV infection in both MS and non-MS populations, however, additional predisposing mechanisms intrinsic to the CNS may be critical determinants of MS susceptibility and subsequent progression.¹⁰⁵ Aze-induced alterations in MO13.3 cells are consistent with the potential for enhanced responses to viral infections (Supplementary Data S4, Figure 9B). Thus, developmental Aze misincorporation in myelin proteins could result in increased immune responses to EBV or other viruses in MS-susceptible populations. For example, downregulation of the IFITs and IFNAR2 in OL (Figure 11B) might lower antiviral immunity.¹⁰⁶ It is also possible that Aze could affect Pro-rich proteins in viruses that have also been implicated in MS pathogenesis.¹⁰⁷

Aze-induced OL alterations mimic MS pathology

The broad range of Aze-induced MO13.3 cell alterations parallel the many OL pathway alterations found in MS periplaques and NAWM.^108–111 For example, Nataf et al reported that irrespective of plaque activity, MS brain periplaques had increased HNRNPs and decreased expression of the OL gene NDRG1 and thus exhibited a TGF-β molecular signature¹⁰⁹; Jahan-Abad et al specifically implicated hnRNPA1 dysfunction.¹¹² Aze induced upregulated HNRNPs (including of hnRNPA1 in 7 pathways), and downregulated NDRG1 expression in MO13.3 cells (Supplementary Data S4); as noted above, however, TGFB2 was also downregulated (Table 2). Upregulation of SLC7A11 (40 pathways) and modulation of some KLF family genes (Supplementary Data S4) found in MS-affected tissues may represent protective responses to toxins and oxidative stress and prevent ferroptosis.^113,114 A considerable number of ion channels were also modified in MO13.3 cells by Aze, including KCNs, SCNs, RYR1 and CLCN5 (Supplementary Data S4, Table 2); these ion channel genes are altered in different types of WM lesions and have been associated with MS pathophysiology and progression.¹¹⁵

The intracytoplasmic MBP-positive aggregates in Aze-treated MO13.3 cells (Figure 7B), may represent the correlate of blebs or blistering of myelin identified in MS NAWM^102,116; they were also observed in Aze-treated mice.² These aggregates did not appear to incorporate a detectable PLP epitope (Supplementary Figure S8), consistent with a specific and possibly greater extent of Aze effects on MBP vs PLP misincorporation and misfolding relating to their differences in Pro composition. Such aggregates would not only be highly disruptive to multiple OL metabolism in vivo but could initiate or enhance anti-MBP autoimmunity.^10,117,118

Failure of repair mechanisms and remyelination of chronic MS lesions

In addition to promoting immune responses and demyelination, the present results suggest that Aze-induced downregulation of some components of OL maturation and myelination mechanisms could impede remyelination and repair.^45,119–122 For example, downregulation of CCL2 (Figure 12), would not only affect immune cells, but would also be detrimental to remyelination by impairing OPC migration to sites of demyelination and differentiation.^33,123 Intrinsic abnormalities in OL that prevent remyelination^124–126 were also upregulated by Aze; additional functions of OL and OPCs including axon remodeling and synapse engulfment, might also be affected in vivo.^127,128

MS progression and OL aging

CNS myelination is initiated and the bulk occurs in early life but the extent of turnover of myelin in adults and the effects of aging on OL in humans are incompletely understood.^129–131 In contrast to acute effects on proteins that turn over rapidly, developmental Aze misincorporation in long-lived myelin proteins might result in slowly progressive misfolding over decades. Random misincorporation at different critical neurodevelopmental timepoints and microanatomic sites might ultimately account for the highly variable localizations of MS lesions and heterogeneous OL depletion in NAWM and MS lesions.^{121,132–137} These in turn contribute to the profound heterogeneity of clinical manifestations and rates of progression in MS patients. Indeed, there is evidence for accelerated OL aging in MS patients.¹³⁸ Subclinical effects of Aze-induced proteotoxic aggregates in NAWM leading to OL death, microglial activation and axonal dysfunction might explain early neurodegeneration and MS progression despite the absence of overt inflammation and clinical relapses in patients receiving current disease-modifying therapies.^139–145

Identifying a process of random spontaneous apoptosis in NAWM that occurs at a very slow rate over decades prior to death is challenging. Although there may be a residual focal inflammatory response (microglial nodules), because of the transient nature of individual cell apoptosis, OL that have undergone complete apoptosis cannot be identified. In our study of CNS tissues of Aze-exposed mice, identification of randomly distributed apoptotic cells in myelinated CNS tissues (ie, without leukocyte infiltration or demyelination), required extensive screening of high magnification fields over the entire CNS of each (Aze-treated and control) mouse.² Because diagnostic biopsies and autopsy sampling of humans with demyelinating diseases appropriately target lesions (detected either by neuroimaging or direct visualization), a comparable unbiased screening of large areas of NAWM in autopsy tissue at high magnification and an intensive search for cells that are undergoing apoptosis at the time of the patient’s death would be required to quantify OL apoptosis in human NAWM. Moreover, terminal events could also result in OL apoptosis adding further to the difficulties in performing a comparable thorough analysis of MS NAWM. A slowly evolving UPR that results in apoptosis of individual OL might be a nidus of a focal lesion, for example, by secreting TNF and activating microglia. We speculate that over many years this could contribute to the diffuse white matter degeneration that occurs in chronic MS NAWM remote from overt plaques.

Potential clinical and therapeutic implications

The GSEA indicated that Aze induced modulation of genes in MO13.3 cells that may represent plasma or CSF biomarkers of relapses or clinical progression in MS patients. These included upregulation of CHEK1, PIK3R1 and FGF9,¹⁴⁶ NCAM (Figure 10B, 13 pathways),¹⁴⁷ EPHA4,¹⁴⁸ RTN4,¹⁴⁹ and ANXA1,¹⁵⁰ and downregulation of CXCL1⁴⁶ (eg, Figure 11A). Potential therapeutic targets for MS patients that were modulated by Aze included components of the UPR/stress responses,^151–153 EGFR (upregulated in 50 pathways) (Figure 13A),^154,155 Piezo1,¹⁵⁶ potassium channels,¹⁵⁷ MAPK3,¹⁵⁸ gpr37 (upregulated in 9 pathways),¹⁵⁹ HDAC2 (upregulated in 36 pathways),¹⁶⁰ clusterin (upregulated in 44 pathways),¹⁶¹ and others.^162–165 Because MBP (unlike PLP and MOG) is a major component of peripheral nerve myelin as well as CNS myelin, Aze misincorporation in Schwann cell MBP might also account for peripheral nerve dysfunction in some MS patients.¹⁶⁶ Evolving technologies that enable editing of individual residues in endogenous proteins in live mammalian cells,¹⁶⁷ suggest the potential for future therapeutic applications in diseases with misincorporated noncanonical amino acids.

Human exposure to Aze

To our knowledge there has been no direct demonstration of Aze misincorporation in human CNS tissues to date. The Rubenstein hypothesis postulated that sugar beets and beet byproducts are the major dietary sources of human exposure to Aze through the food chain.¹ However, the dietary sources and amounts of possible Aze exposure in human populations remain largely unknown and investigating the extent to which dietary Aze might contribute to MS pathogenesis and progression remains particularly challenging due to the inherent difficulties of investigating temporal and geographic factors that contribute to intra-uterine and early life dietary exposure.

Aze also occurs in natural bacterial and other micro-organismal metabolic pathways,^14,168 and may, therefore, represent an additional source of exposure. The concept of a dietary noncanonical amino acid inducing human neurological diseases has precedence, however. For example in the 1980s, the eosinophilia-myalgia syndrome was attributed to ingestion of altered L-tryptophan,^169,170 and the microbial neurotoxin ß-methylamino-L-alanine (BMAA) has been implicated in neurodegenerative conditions.¹⁷¹ Indeed, how the gut microbiome influences human neurologic diseases is currently of considerable interest,¹⁷² particularly in MS.^173–182 Moreover, there are many naturally-occurring microbial Pro analogues and Aze derivatives that also might impact human health^183–185; in the context of the Aze hypothesis, the maternal and neonatal microbiomes¹⁸⁶ would be particularly pertinent. Addressing possible relationships of Aze to MS is further complicated by the fact that exposure to Aze would also have effects on other non-CNS and CNS cells and tissues.^2,14,15,22 Whether and the extent to which Aze is present in human microbiomes, whether it could affect CNS immunity and function, and how it might affect human CNS myelin both in early life and once developmental myelination has occurred remain completely unknown.

Limitations

In addition to the caveats discussed above, there are other specific limitations of the present study that prevent direct extrapolations to conditions in humans. Importantly, mRNA levels assessed by the GSEA may not reflect protein levels, which need to be validated.¹⁸⁷ Furthermore, the use of the MO13.3 cell line, an artificial in vitro system with cells in different stages of maturation and metabolic stress¹⁸⁸ and without in situ context, may be suboptimal for analyzing OL and OPC maturation, myelinogenesis/myelination and response to injury in vivo.⁶ In particular, although the presence of MBP-positive aggregates indicates that the Aze-treated cells had the capacity for myelin protein synthesis, it was not possible to correlate the injuries induced in individual cells with their maturation stages. Moreover, the GSEA experiment compared 0 to 0.6 mM concentrations of Aze; the high concentrations are likely non-physiological and may not reflect in vivo pathologic processes.

As in inflammatory and toxin MS models, (and in ²) the goals of in vitro and in vivo (ie, experimental animal) experiments are to induce severe acute measurable injury effects over relatively short time periods for the purpose of identifying mechanisms and potential therapeutic targets. Longer durations of lower Aze concentrations in vitro and in vivo might result in more protein misincorporation and have less dramatic and possibly different effects, for example, slower myelin degeneration. Thus, specific extrapolations from this in vitro model to conditions in humans may not be possible and caution is necessary regarding specific Aze effects and means of counteracting them. In particular, although the results suggest the possibility of beneficial effects of exogenous Pro treatment of MS patients by dietary modification or Pro supplementation, caution is indicated because of the ubiquity of Pro in the human proteome,¹⁸³ and the potential adverse effects of hyperprolinemia.¹⁸⁹

Summary and Conclusions

Using the human oligodendroglial lineage MO13.3 cells as a model of myelin-producing OL, we found that exposure of these cells to Aze recapitulated the OL alterations and apoptosis observed following systemic treatment of adult and neonatal mice with Aze. These included nuclear translocation of immunoreactivity of ER stress/UPR molecules culminating in apoptosis, MBP (but not PLP), cytoplasmic aggregation, and TNF production. These in vitro effects were counteracted by equimolar Pro. GSEA demonstrated alterations in multiple pathways involving OL metabolic stress, differentiation and a pro-inflammatory/immunomodulatory state as a consequence of Aze exposure.

The present data corroborate and extend our previous observations in the in vivo model of Aze oligodendrogliopathy and support the hypothesis of specific Aze misincorporation in MBP contributing to the initial pathogenesis of MS. Even if Aze is not the only agent causing progressive ER stress in OL, the Aze hypothesis presents a novel paradigm for investigating MS pathogenesis that accounts for global epidemiology and other clinical and pathological mysteries surrounding it that are not addressed by other MS models. The pro-inflammatory/immunomodulatory OL phenotype evidenced in our studies recapitulate alterations also observed in EAE and MS suggesting that intrinsic OL abnormalities could precede an autoimmune-mediated event in the absence of an external immune stimulus. Because developmental myelination largely occurs in early life, Aze misincorporation could be a predisposing factor that enhances subsequent immune reactions and is therefore potentially relevant to both inflammatory and degenerative processes in the pathophysiology and progression in MS patients.

FUNDING

This study was supported by Stanford University Department of Pathology Gift Fund, Palo Alto Veterans Institute for Research and Boston University Microarray and Sequencing Resource Core Facility.

CONFLICTS OF INTEREST

None declared.