PNAS PLUS

Rapid and efficient generation of oligodendrocytes

from human induced pluripotent stem cells using
transcription factors
Marc Ehrlicha,b, Sabah Mozafaric,d,e,f, Michael Glatzab, Laura Starosta,b, Sergiy Velychkob, Anna-Lena Hallmanna,b,
Qiao-Ling Cuig, Axel Schambachh, Kee-Pyo Kimb, Corinne Bachelinc,d,e,f, Antoine Marteync,d,e,f, Gunnar Hargusa,b,
Radia Marie Johnsoni, Jack Antelg, Jared Sterneckertj, Holm Zaehresb,k, Hans R. Schölerb,l, Anne Baron-Van
Evercoorenc,d,e,f, and Tanja Kuhlmanna,1
a
Institute of Neuropathology, University Hospital Münster, 48149 Muenster, Germany; bDepartment of Cell and Developmental Biology, Max Planck
Institute for Molecular Biomedicine, 48149 Muenster, Germany; cINSERM, U1127, F-75013 Paris, France; dCNRS, UMR 7225, F-75013 Paris, France; eSorbonne
Universités, Université Pierre et Marie Curie Paris 06, UM-75, F-75005 Paris, France; fInstitut du Cerveau et de la Moelle epinière-Groupe Hospitalier
Pitié-Salpêtrière, F-75013 Paris, France; gMontreal Neurological Institute, McGill University, Montreal, QC, Canada H3A 2B4; hInstitute of Experimental
Hematology, Hannover Medical School, 30625 Hannover, Germany; iDepartment of Physiology, McGill University, Montreal, QC, Canada H3A 2B4; jDFG
Research Center for Regenerative Therapies, Technische Universität Dresden, 01307 Dresden, Germany; kMedical Faculty, Department of Anatomy and
Molecular Embryology, Ruhr-University Bochum, 44801 Bochum, Germany; and lMedicial Faculty, Westphalian Wilhelms-University of Muenster, 48149
Muenster, Germany

Edited by Brigid L. M. Hogan, Duke University Medical Center, Durham, NC, and approved February 1, 2017 (received for review August 30, 2016)

Rapid and efficient protocols to generate oligodendrocytes (OL) more, these protocols require long culture periods (70–150 d) to
from human induced pluripotent stem cells (iPSC) are currently obtain O4+ OL and show limited efficiency (9–12).

NEUROSCIENCE
lacking, but may be a key technology to understand the biology of Here, we describe an efficient strategy that facilitates and op-
myelin diseases and to develop treatments for such disorders. Here, timizes the generation of human O4+ OL from human iPSC-
we demonstrate that the induction of three transcription factors derived neural progenitor cells (NPC) (13). This approach yields
(SOX10, OLIG2, NKX6.2) in iPSC-derived neural progenitor cells is up to 70% O4+ OL within 28 d of differentiation, using a com-
sufficient to rapidly generate O4+ OL with an efficiency of up to bination of three transcription factors (TF), namely SOX10,
70% in 28 d and a global gene-expression profile comparable to OLIG2, and NKX6.2. Furthermore, 30% of the O4+ iPSC-derived
primary human OL. We further demonstrate that iPSC-derived OL OL (iOL) differentiate into mature myelin basic protein-positive
disperse and myelinate the CNS of Mbpshi/shi Rag−/− mice during (MBP+) OL within 7 additional days. The global gene-expression
development and after demyelination, are suitable for in vitro mye- pattern of O4+ OL is comparable to that of human primary OL
lination assays, disease modeling, and screening of pharmacological (pOL). The induced human iOL produce myelin-like structures
compounds potentially promoting oligodendroglial differentiation. around nanofibers or iPSC-derived neurons. After transplantation
Thus, the strategy presented here to generate OL from iPSC may into MBP-deficient shiverer mice (Shi/Shi Rag2−/−) iOL disperse
facilitate the studying of human myelin diseases and the develop- widely and myelinate host axons in the developing central nervous
ment of high-throughput screening platforms for drug discovery. system (CNS), as well as the adult demyelinated spinal cord.

|
human iPSC oligodendrocytes | forward patterning | myelination | Significance
disease modeling
Understanding of myelin diseases and development of new

O ligodendrocytes (OL) play a key role in myelin-related dis-

eases, including multiple sclerosis (MS), leukodystrophies, as
well as periventricular leukomalacia, and there is an increasing
treatment options are at least partly hampered by the limited
availability of human oligodendrocytes. Induced pluripotent stem
cells (iPSC) may be an ideal tool to circumvent this problem;
awareness of their potential role in neurodegenerative diseases (e.g., however, rapid and efficient protocols to generate oligodendro-
multiple system atrophy and amyotrophic lateral sclerosis) or trau- cytes from human iPSC are currently lacking. The induction of the
matic spinal cord injury (1–6). OL form and maintain the myelin transcription factors SOX10, OLIG2, and NKX6.2 in iPSC-derived
sheaths that insulate axons and organize the distribution of axonal neural progenitors accelerates oligodendroglial differentiation
voltage-gated ion channels, a prerequisite for conduction of action significantly resulting in up to 70% of O4+ oligodendrocytes
potentials and trophic support of axons. Demyelination in MS within 28 d. These oligodendrocytes myelinate the CNS during
contributes to axonal damage and disease progression (7). Immu- development and after demyelination, and are suitable for
nosuppressive or immunomodulatory therapies, including complete pharmacological screens and disease modeling. The strategy
ablation of the immune system by radiation and chemotherapy, presented herein will markedly facilitate the studying of hu-
prevent new inflammatory lesions that underlie clinical relapses but man myelin diseases and the development of screening plat-
do not arrest disease (8). Therapies promoting remyelination rep- forms for drug discovery.
resent a promising new treatment strategy to protect and restore
axonal integrity and neurologic function (4). The development of Author contributions: M.E. and T.K. designed research; M.E., S.M., and L.S. performed research;
S.V., A.-L.H., Q.-L.C., A.S., K.-P.K., C.B., A.M., G.H., R.M.J., J.A., J.S., H.Z., H.R.S., and A.B.-V.E.
such therapeutics is hampered, at least in part, by the limited contributed new reagents/analytic tools; A.B.-V.E. designed animal experiments; M.E., S.M.,
availability of human OL. Stem cell technologies are a promising and M.G. analyzed data; and M.E., A.B.-V.E., and T.K. wrote the paper.
tool to circumvent this problem. Availability of OL derived from Conflict of interest statement: M.E. and T.K. have a pending patent application for the
human induced pluripotent stem cells (iPSC) would permit studies oligodendroglial differentiation protocol.
to delineate mechanisms regulating repair by endogenous myelin This article is a PNAS Direct Submission.
lineage cells and provide a source of autologous cells for re- Data deposition: The data reported in this paper have been deposited in the Gene Ex-
placement therapy. Such cells would also provide new opportunities pression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE79914).
to identify pathological mechanisms underlying de- or dysmyelinat- 1
To whom correspondence should be addressed. Email: [email protected].
ing diseases. However, to date only a few protocols have resulted in This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
the successful generation of human stem cell-derived OL. Further- 1073/pnas.1614412114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1614412114 PNAS Early Edition | 1 of 10

Furthermore, iOL can be exploited for disease modeling and to dendroglial progenitor cells (OPC) and OL, after 14 d of exposure
test the potential of pharmacological compounds in promoting (Fig. 1B). Controls, including NPC either uninfected or infected
oligodendroglial differentiation. with RFP only, did not yield any O4+ cells (Fig. 1A). We sub-
sequently determined the oligodendroglial induction capacity of
Results SOX10 in combination with any of the remaining six TF. We
Identification of OL Lineage Inducing TF in Human NPC. Human PSC identified OLIG2 as a factor that substantially increased the
present a valuable source for the generation of myelinogenic OL SOX10-mediated oligodendroglial lineage commitment (Fig. 1 C
for research and autologous cell-replacement therapies (9–12). and G), whereas ASCL1 and MYT1 significantly decreased the
Currently, available protocols are technically challenging and number of O4+ cells (Fig. 1 D and G). The combination of SOX10
time-consuming, limiting the application of these cells for research and OLIG2 with NKX6.2, a TF associated with oligodendroglial
and regenerative medicine. In these protocols, NPC are rapidly maturation, significantly increased further the portion of O4+ cells
and efficiently derived from human PSC, which is followed by the (Fig. 1 E, F, and H). Thus, we concluded that the ectopic ex-
rate-limiting step of oligodendroglial specification and differenti- pression of SOX10, OLIG2, and NKX6.2 (subsequently referred
ation. Therefore, we first aimed to identify TF accelerating the to as SON) was the most efficient combination of TF tested here
oligodendroglial specification and differentiation from human to induce OL from iPSC-derived NPC.
iPSC-derived NPC. We performed literature data mining and
selected a set of seven TF, which are enriched in OL compared Ectopic Expression of TF Accelerates Oligodendrogenesis from iPSC–NPC.
with other neural lineages (2, 14), and are required for oligo- To further enhance the generation of human iOL, we generated a
dendroglial specification (15, 16): ASCL1, MYT1, NKX2.2, polycistronic lentiviral expression vector containing SON and RFP
NKX6.1, NKX6.2, OLIG2, and SOX10. Coding sequences for as a reporter gene under control of the retroviral spleen focus-
these proteins, as well as red fluorescent protein (RFP), were forming virus (SFFV) promoter (Fig. 2A). iPSC-derived NPC (13)
individually cloned into a tetracycline (tet)-inducible lentiviral were infected with SON/RFP-expressing lentivirus (Fig. 2B). After
vector. Human iPSC-derived NPC, which can be frozen and cost- induction of SON, a two-step differentiation protocol was sufficient
efficiently expanded, as previously described (17), were transduced to derive increasing numbers of iOL over 28 d (Fig. 2B). One day
with a combination of lentiviruses expressing one of the TF can- after viral transduction, NPC medium was replaced by glial induction
didates and the reverse tet-controlled transactivator (rtTA). medium (GIM) containing different supplements, including PDGF,
Among all TF candidates, only SOX10 was capable of inducing SAG (smoothened agonist), and thyroid hormone (Triiodo-L-
O4 (9.99 ± 0.81%), a highly specific marker of late-stage oligo- Thyronine, T3; 10 ng/μL). After 4 d, this medium was replaced by

A B C

D E F

Fig. 1. Screening for oligodendroglial lineage in-

ducing TF in human NPC. Human iPSC-derived NPC
were infected with individual OL-specific TFs or RFP
control virus. (A–F) OL-lineage commitment of in-
fected iPSC-derived NPC was analyzed 14 d after
G H transgene induction by immunostaining using the
OL-specific antibody O4 (green). Nuclei were coun-
terstained with Hoechst (blue). (A) Control cultures
did not express the O4 epitope. (B) SOX10 was the
only tested TF inducing O4+ OL. (C) Addition of
OLIG2 enhanced the OL-lineage commitment,
whereas (D) ASCL1 led to a decreased number of
O4+ iOL. (E) Coexpression of SOX10, OLIG2, and
NKX6.2 increased the number of O4+ cells, (F) ac-
companied by the appearance of iOL with a more
mature oligodendroglial morphology. (Scale bars,
50 μm in A–E and 25 μm in F.) (G and H) Quantifi-
cation of O4+ iOL over all cells with indicated TF
combinations 2 wk after transgene induction. Data
are presented as mean of replicates from three in-
dependent experiments + SD. One-way ANOVA with
Bonferroni’s multiple comparisons test was used as
statistical test (*P < 0.05, **P < 0.01, ***P < 0.001).

2 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1614412114 Ehrlich et al.

 PNAS PLUS
Fig. 2. SON induce a rapid and efficient oligoden-
droglial lineage commitment. (A) Schematic pre-
A B sentation of the lentiviral expression vector used for
the polycistronic expression of SON. (B) Schematic
summary of the differentiation protocol using NPCM,
GIM, and DM. (C–F) Representative immunofluores-
cence images of different NPC and OL markers during
differentiation. Nuclei were counterstained with
C D E F Hoechst (blue). (C) iPSC-derived NPC homogenously
expressed the neural progenitor marker NESTIN
(green) and SOX1 (red). (D) Seven days after trans-
gene induction, NG2+ and O4+ oligodendroglial line-
age cells were detected. (E) By day 28, iOL expressed
the O4-epitope, the more mature OL marker GALC,
and presented with a branched morphology. (F) Fur-
ther maturation led to the emergence of MBP+ ma-
ture iOL forming myelin sheaths. (Scale bars, 50 μm.)

differentiation medium (DM) containing higher concentrations of T3 gene expression during differentiation, we generated a tet-
(60 ng/μL) and lacking PDGF and SAG. To ensure the reproducibility inducible lentiviral expression vector containing SON (Fig. S3A).
of our protocol, all experiments were performed with four in- Quantification of O4+ cells at day 28 of differentiation revealed
dependent NPC lines derived from three different iPSC lines and one that induced expression of SON for 7 d is already sufficient to
embryonic stem cell (ESC) line. All NPC lines homogeneously obtain a stable and transgene independent oligodendroglial
expressed the neural stem cell marker SOX1 and NESTIN (Fig. 2C). lineage commitment (Fig. S3B). Interestingly, longer expression
Seven days after SON induction, OL cultures comprised mainly NG2+ of SON (14 and 21 d) did not significantly enhance the differ-

NEUROSCIENCE
glial progenitor cells (82.26 ± 2.36%) together with O4+ cells (8.7 ± entiation efficiency. Constitutive tet-controlled expression of
3.0%) (Fig. 2D). At day 28 after SON induction, the majority of cells SON for 28 d demonstrated the highest differentiation efficiency
expressed the OL marker O4 (65.5 ± 11.1%) and presented with a (Fig. S3B), suggesting that a subset of O4+ cells was still de-
more complex morphology identified by an increased number of pendent on ectopic expression of SON.
primary and secondary processes (Fig. S1). Additionally, cells star- Next, we determined the influence of SON overexpression on
ted to express more mature oligodendroglial markers, like GALC the lineage commitment of NPC. ICC analysis of SON infected
(19.19 ± 1.46%) and MBP (30.37 ± 7.87% of O4+ cells) by days 28 NPC cultures compared with RFP-infected control cultures at
and 35, respectively (Fig. 2 E and F; see also Fig. 5 D and E). day 28 revealed a decreased number of SOX1+ NPC (Fig. 3I and
To assess the kinetics, efficiency, and yield of SON-mediated Fig. S4A) and a significant switch from neuronal to oligoden-
oligodendroglial lineage specification, we conducted weekly flow droglial cell fate (Fig. 3J and Fig. S4B). In contrast, the astroglial
cytometry analyses of the O4 epitope expression during differ- lineage commitment was not affected (Fig. S4 C and D).
entiation (Fig. 3A). As a control, NPC were infected with RFP
expressing lentivirus and cultured under the same differentiation Global Gene-Expression Profiling Demonstrates That iOL Resemble
conditions. All NPC lines tested were found to perform similarly Primary Human Adult OL. To further characterize the cellular iden-
with respect to iOL generation, starting from 8.7 ± 3.0% O4+ tity of iOL, we compared the global gene-expression profiles of O4+
cells at day 7 to 65.5 ± 11.1% O4+ cells by day 28 (Fig. 3B). In iOL purified at day 28 of differentiation with human MBP+ pOL-
contrast, only 1.4 ± 0.5% O4+ cells were identified in RFP- derived from surgically resected brain samples from adult patients
transduced cell cultures (Fig. 3B). The protocol was highly effi- (Fig. S5), as well as with iPSC-derived NPC before induction of SON.
cient and reproducible among all cell lines, illustrated by the As a negative control, we used gene-expression values of un-
quantification of O4+ cells at day 28, ranging from 62.1 ± 9.5% differentiated iPSC. The unbiased hierarchical clustering clearly
(ESC–NPC) to 79.0 ± 14.8% (iPSC–NPC-3) (Fig. 3C). demonstrated that iOL and pOL exhibit highly comparable gene-
Yields of O4+ iOL (total O4+ cell number/starting NPC cell expression signatures and form a distinct cluster significantly segre-
number) ranged from 133.70 ± 24.83% at day 14 to 241.20 ± gating from NPC and iPSC (Fig. 4A). When we compared neural
19.07% at day 28 (Fig. 3D). These data suggest an expansion of lineage-specific gene sets, we identified a strong up-regulation of OL-
SON-expressing cells during our differentiation protocol. Identi- specific genes, such as OLIG1, MOG, and MBP in iOL compared
fication of proliferative cells using Ki-67 revealed a proliferation with NPC, whereas NPC-related genes, including SOX1, PAX6, and
rate of 35% among RFP+ cells at day 14, which declined to 10% PAX7, were down-regulated in iOL (Fig. 4 B and C). Because we
by day 28, confirming that the transgene-expressing cell pop- compared O4-sorted iOL with MBP+ pOL, it is not surprising that
ulation further expanded during differentiation (Fig. 3E). In- iOL expressed some OPC-specific genes, such as PDGFRA and
terestingly, the proliferation capability was retained in 20% of O4+ ST8SIA1, to a higher extent than pOL, indicating a more immature
iOL at day 14 and diminished to 5% by day 28 (Fig. 3 F and G). cell identity of iOL (Fig. S6A). Coherently, pOL exhibit a stronger
Next, we tested whether iOL cultures can be expanded by a pro- expression of some mature OL-specific genes, such as MAG and
longed exposure to a glial expansion medium (GEM) postlentiviral MOBP, compared with O4+ iOL. To further analyze the influence of
transduction. Compared with GIM, GEM additionally contained ectopic SON expression on the oligodendroglial lineage commitment
FGF2 (5 ng/mL) and completely lacked IGF-1. SON-transduced of NPC, we determined differentially expressed genes in iOL com-
NPC were cultured in GEM for either 4, 8, or 12 d and subsequently pared with the original NPC population. This analysis revealed 755
transferred to DM for an additional 28 d. FACS analyses of the iOL commonly up- and 955 commonly down-regulated genes among all
cultures after 0, 14, and 28 d in DM revealed that although expo- iOL cell lines (Fig. S6 B and C). Gene Ontology (GO) terms asso-
sure to GEM for 12 d reduced the differentiation efficiency (Fig. ciated with up-regulated genes in iOL include categories such as “cell
S2A), yields of O4+ iOL were significantly increased (Fig. S2B). adhesion,” “myelin sheath,” “axon ensheathment,” “myelin,” and
Furthermore, flow cytometry analyses exhibited the presence “regulation of action potential.” Conversely, GO terms associated
of an O4+/RFP− cell population in SON-transduced cultures with down-regulated genes include categories such as “cell cycle,”
(Fig. 3A), which comprised up to 50% of the O4+ cell population “DNA replication,” “mitosis,” and “nucleoplasm” (Table S1).
and which could be confirmed by immunocytochemical (ICC) These results indicate that ectopic expression of SON induces
analysis (Fig. 3H), suggesting transgene silencing in a subset of an oligodendroglial gene-expression profile comparable to native
iOL. To analyze whether iOL become independent from trans- human adult OL.

Ehrlich et al. PNAS Early Edition | 3 of 10

A B

Fig. 3. Quantification of oligodendroglial lineage

cells after SON induction. (A) Representative flow
cytometry analyses for the expression of O4 and RFP in
control and SON cultures 7 and 28 d after transgene
C induction. (B) Quantification of O4+ cells in control
and SON cultures 1 to 4r wk after transgene induc-
tion. Data are presented as mean of replicates from
four independent experiments, each using NPC de-
rived from an independent human PSC line + SD.
(C) Quantification of O4+ iOL at day 28 derived from
one human ESC and three independent iPSC lines.
Data are presented as mean of replicates from three
to five independent differentiation experiments per
cell line + SD. (D) Quantification of O4+ iOL yields at
days 14 and 28 after transduction. Data are presented
D E F G as mean of replicates from three independent differ-
entiation experiments, each using an independent
iPSC-derived NPC line + SD. (E) Quantification of Ki-
67+ transgene-expressing cells (RFP) and of (F) Ki-67+
/O4+ iOL at days 14 and 28 after induction. Data are
presented as mean of replicates from three in-
dependent differentiation experiments + SD (**P <
0.01, ***P < 0.001). (G) Immunostaining of iOL for O4
(purple) and the proliferation marker Ki-67 (green) at
day 14 after transduction. (H) Representative immu-
H I J nofluorescence image of O4+ iOL (green) 28 d after
transduction either expressing (filled arrowhead) or
silencing (empty arrowhead) the transgenes. (Scale
bars, 40 μm.) (I) Quantification of SOX1+ iPSC-derived
NPC and (J) TUJ1+ neurons in control and SON cultures
at day 28. Data are presented as mean of replicates
from three independent differentiation experiments
+ SD. Student’s t test was performed for statistical
analysis (***P < 0.001).

iOL Differentiate into Mature MBP-Expressing OL in Vitro and Produce MACS-purified O4+ iOL in the immune- and MBP-deficient Shi/
Myelin-Like Sheaths. Next, we assessed the terminal differentia- Shi Rag2−/− mouse CNS. To facilitate the identification of trans-
tion potential of iOL in vitro. At day 35, iOL cultures contained planted cells, iOL coexpressed RFP as a reporter. These mice carry
many highly branched O4+ cells (Fig. 5A), as well as mature OL- a deletion of seven exons of the MBP gene and completely lack
expressing CNP (2′,3′-cyclic nucleotide 3′-phosphodiesterase) (32.82 ± MBP protein expression. Because of the lack of endogenous MBP
2.80%) (Fig. 5B). Additionally, 30.37 ± 7.87% of O4+ cells differen- expression in Shi/Shi Rag2−/− mice, myelin generated by trans-
tiated into mature MBP+ iOL with myelin-like sheaths (Fig. 5 D and planted cells can be easily identified by MBP immunohistochemistry
E) of which 36.62 ± 8.93% coexpressed MAG (Fig. 5C). To evaluate and by detection of myelin compaction, as well as the presence of
the myelinogenic capability of iOL in vitro, we purified O4+ iOL using the major dense line by electron microscopy (EM) (18, 19). To
magnetic cell separation (MACS) at day 21 and cultured them for address developmental myelination, cells were grafted bilaterally
14 d on 3D cell culture surfaces with aligned nanofibers. ICC analysis and rostrally to the corpus callosum of newborn mice. Analysis of
of mature MBP+ iOL in these cultures revealed the extension of sagittal sections 16 wk postgrafting (wpg) indicated the presence of
multiple processes along the nanofibers, with some of these extensions numerous MBP+ myelin profiles associated with RFP+ and human
wrapping around the nanofibers (Fig. 5F). Evidence for ensheathment nuclei-positive (STEM101+) cells (Fig. 6 A and B). Higher magni-
of axons in vitro was evaluated in cocultures of O4+ iOL with iPSC- fication using confocal microscopy showed that iOL, identified by
derived neurons. After 3 wk, the cultures exhibited myelin-like combined human nuclei (STEM101+) and cytoplasmic (STEM121+)
sheaths surrounding the axons, identified by confocal analysis of MBP antigens to highlight their cellular contours, extended processes
and TUJ1 expression (Fig. S7A). Three-dimensional reconstruction frequently connected to MBP+ myelin, thus further validating the
of confocal optical sections in high magnification showed colabeling donor origin of the myelin (Fig. 6 C and G). MBP+ myelin gener-
of neuronal processes (TUJ1) with MBP (Fig. 5 G and H), which was ated by grafted iOL wrapped around SM312+ host axons (Fig. 6D).
further evaluated by orthogonal projections clearly displaying the The presence of human-derived myelin surrounding SM312+ axons
formation of MBP+ structures around neuronal processes (Fig. S7B). was further validated combining MBP detection with human
Control cultures completely lacked these MBP+ structures. These NOGO-A, which identifies late stages of mature OL (Fig. 6 E and
data clearly illustrate the capability of iOL to mature into MBP+ OL F). The human-derived myelin internodes were often associated
and to ensheath neuronal processes in vitro. with the paranodal marker CASPR flanking Ankyrin-G nodal
structures, demonstrating functionality of the human cell-derived
iOL Myelinate the Developing Brain and Remyelinate the Demyelinated myelin (Fig. 6H). Some of the animals were also used for ultra-
Spinal Cord of Dysmyelinating Mice. The differentiation of iOL structural analysis of myelin compaction. In control Shi/Shi Rag2−/−
into myelin-forming OL was further validated by grafting day 14 mice, myelin sheaths were thin and noncompacted (Fig. S8). In

4 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1614412114 Ehrlich et al.

 PNAS PLUS
Fig. 4. Global transcriptional profiling of iOL. (A) Hierarchical clustering of whole-genome expression profiles of iPSC (black), iPSC-derived NPC (green), iOL (28
d after transduction) (red), and human adult OL pOL (blue) revealed a strong correlation between iOL and pOL. (B and C) Pairwise scatterplot analysis of log2-
adjusted global gene-expression values of iPSC-derived NPC and their corresponding iOL (n = 10). Genes presenting with a <twofold difference in gene expression

NEUROSCIENCE
are illustrated in gray. (B) Characteristic OL-enriched genes were up-regulated in iOL, whereas (C) characteristic NPC-enriched genes were down-regulated.

mice that received iOL, numerous normal compacted myelin grafted cells differentiated into mature OL was further con-
sheaths with alternating major dense lines and intermediate lines were firmed by widespread coexpression of human cytoplasmic
observed (Fig. 6 I–K), validating unambiguously that iOL have the STEM121 with human NOGO-A (Fig. S11). The extent of hu-
capacity to differentiate into functional myelin-forming cells in vivo. man-derived myelination in the spinal cord of Shi/Shi Rag2−/−
We quantified the percentage of myelinated axons over axons with a mice was evaluated by immunolabeling of MBP and MOG, and
diameter >1 μm in grafted (n = 3) vs. nongrafted mice (n = 3) by EM further confirmed the widespread, integrated, and high amount
analyses. We did not detect a difference in the percentage of mye- of human cell-derived myelin (Fig. S12). Analysis of the MBP+
linated axons (grafted mice: 69.7 ± 3.28% vs. nongrafted mice 73.9 ± area expressed by the human cells (exogenous myelin) per
5.1%; P value: 0.2892). This result is most likely because of the MOG+ area, identifying endogenous as well as exogenous myelin
competition between endogenous and grafted cells and the relatively at the lesion site (dorsal funiculus), showed that at 12 wpg,
early time point after transplantation of human iOL, which are known 12.95% ± 2.02% of the total myelin was derived from the grafted
to differentiate more slowly than rodent cells. However, in the corpus cells. MBP was detected over 18.14 ± 6.39 mm in the grafted
callosum of transplanted mice, 12% of the axons were myelinated by mice (Fig. S12). MBP expression was specific for grafted cells, as
grafted cells, as determined by the percentage of axons with com- no MBP expression was found in sections remote from the lesion
pacted myelin over the total numbers of myelinated axons. Moreover, (Fig. S12). Higher magnification showed that processes extended by
the g-ratio in grafted mice was significantly smaller compared with iOL were frequently connected to MBP+ donut-shaped myelin in-
nongrafted mice (76.8 ± 1.8 vs. 89.5 ± 1.36; P value: 0.0003). ternodes, thus validating the exogenous source of the myelin (Fig.
To address the ability of iOL to remyelinate demyelinated axons, S9 E–G). Although most NF165+ axons were surrounded by RFP+
RFP+ iOL were grafted into the dorsal funiculus of adult Shi/Shi processes, fewer of them coexpressed MBP, indicating that myeli-
Rag2−/− mice spinal cord. Although Shi/Shi Rag2−/− mice are nation was still ongoing (Fig. S9H). MBP+ myelin structures were
hypomyelinated, numerous axons are surrounded by MBP− non- often colabeled for the paranodal protein CASPR, as viewed on
compacted myelin (20) (Fig. S8). To optimize the number of longitudinal sections (Fig. S9I), indicating the formation of nodes of
demyelinated axons accessible for remyelination, mice were injected Ranvier and suggesting that the iOL-derived newly formed myelin
with lysophosphatidyl-choline (LPC) 48 h prior grafting to induce was also functional in the adult demyelinated spinal cord.
demyelination (21, 22). Twelve weeks postgrafting, immunolabeling
of serial cross-sections for nucleic STEM101 and cytoplasmic RFP, Human iOL Respond to Compounds Promoting Oligodendroglial
together with MBP, revealed MBP+ donut-like myelin structures, Differentiation. Identification of drugs inducing remyelination via
indicating that grafted iOL not only colonized and remyelinated the promotion of oligodendroglial differentiation presents a promising
lesion site, but also myelinated the entire neuraxis, including ventral approach for the treatment of demyelinating disorders like MS.
and dorsal white and gray matter (Fig. S9 A, C, and D). Thus, we assessed whether iOL can be used to identify compounds
The differentiation of human grafted cells into various neural promoting oligodendroglial differentiation. We selected six drug
cell types was assessed by colabeling of human nucleic STEM101 candidates (miconazole, clobetasol, benztropine, indometacin,
and RFP with cell-specific markers of oligodendroglial cells clemastine, and oxybutynin), which have been previously described
(OLIG2, CC1), neurons (NeuN), and astrocytes (GFAP) (Figs. to promote differentiation or myelination of rodent OL (1, 23–25).
S9B and S10). Quantification of the various populations at the iOL were cultured in a minimum growth medium lacking differ-
lesion site indicated that about 81% of the grafted human cells entiation-promoting factors and were treated with either vehicle
generated OLIG2+ oligodendroglial cells with 25.64% ± 4.65% [0.01% (vol/vol) DMSO] as a negative control, T3 as a positive
OPCs (OLIG2+/CC1−) and 56.06% ± 3.11% mature OL control, or the drug candidate dissolved in DMSO at three different
(OLIG2+/CC1+) (Fig. S10G). The remaining population (19%) concentrations (0.5, 1, 5 μM) (Fig. 7 and Fig. S13). In DMSO-
differentiated into NeuN+ neurons (2.66% ± 0.66%) and treated control cultures, 14.01 ± 2.89% O4+ iOL were observed in
GFAP+ astrocytes which could not be quantified because of minimum DM after 21 d of culture, whereas addition of T3 resulted
highly interdigitated processes, but which by deduction represent in the doubling of O4+ cells (28.25 ± 3.47%). Several drug candi-
most likely 17% of the grafted cells. That the majority of the dates performed as well as T3 and demonstrated a dose-dependent

Ehrlich et al. PNAS Early Edition | 5 of 10

A B C D E Fig. 5. iOL differentiate into mature OL and
DNA O4 DNA CNP DNA MAG DNA MBP ensheath iPSC-derived neurons in vitro. (A) Thirty-five
days after transgene induction, O4+ iOL presented a
branched morphology typical for mature OL and (B–D)
expressed the mature oligodendroglial markers CNP,
MAG, and MBP. (E) Quantification of mature MBP+ iOL
over all O4+ iOL. Data are presented as mean of repli-
cates from four independent differentiation experi-
ments + SD. (F) Immunostaining of iOL 14 d after
F G H replating on 3D nanofiber scaffolds illustrating the
DNA MBP PC MBP DNA MBP TUJ1 Colocalization formation of myelin sheaths around nanofibers. Nuclei
are counterstained with Hoechst. (G and H) Human in
vitro myelination assay: coculture of O4+ iOL purified at
day 21 by MACS with iPSC-derived neurons for 3 wk.
(G) Three-dimensional reconstruction of confocal im-
ages for MBP (green) and the neuronal marker TUJ1
(red), suggesting wrapping of axons. Nuclei were
counterstained with Hoechst (blue). (H) Three-di-
mensional illustration of MBP and TUJ1 colocalization
(white) from the same detail. [Scale bars, 100 μm (A),
20 μm (B and C), 50 μm (D), and 10 μm (F and H).]

increase of O4+ cells (Fig. 7A). Interestingly, clemastine and oxy- Discussion
butynin failed to promote the formation of O4+ iOL. Furthermore, Previously established protocols using in vitro patterning to derive
we observed a toxic influence of miconazole, benztropine, and OL from murine iPSC require fewer than 30 d (33). Additionally,
clemastine on iOL at higher concentrations. Quantification of ma- it has been shown that mouse fibroblasts can be directly converted
ture MBP+ iOL revealed a fourfold increase in the presence of T3 into OPC by the forced expression of Sox10, Olig2, and Zfp536 or
compared with DMSO control (Fig. 7B). The effect was most Nkx6.2, resulting in ∼15% O4+cells after 20 d (2, 14). However,
striking with 1 μM miconazole, inducing an almost 10-fold increase the generation of human OL from iPSC or ESC is more challenging
of MBP+ cells compared with DMSO control cultures. Clobetasol and characterized by much longer culture periods (70–150 d),
and benztropine enhanced the formation of MBP+ mature iOL limited efficiencies, and variable reproducibility. Although the
comparable to T3-treated cultures. These data demonstrate that protocols to generate human OL have been further optimized to
iOL can be used to identify compounds that promote differentiation reduce culture times and increase efficiency, they still require at
into O4+ as well as maturation into MBP+ OL. least between 60 and 130 d of culture to generate OPC from iPSC/
ESC-derived NPC and only a small percentage of cells become
iOL from Patients Carrying the N279K Microtubule-Associated Protein MBP+ mature OL (9–11). One of the first studies demonstrating
Tau Mutation Have a Higher Susceptibility to Oxidative Stress- the successful derivation of human OL from iPSC was published
Induced Cell Death. Next, we wanted to determine whether iOL in 2013 (11) using a modified in vitro-patterning strategy from
can be used for in vitro disease modeling. The microtubule- ESC-based protocols (34, 35). However, the transition from pre-
associated protein tau (MAPT) is developmentally expressed in OPC to OPC was hampered; by day 150, 20–40% of the cells
OL (26, 27) and mutations in MAPT have been associated with expressed the OPC marker CD140a, whereas the late OPC
frontotemporal dementia with Parkinsonism linked to chromo- marker O4 was only present in 5–10%. Interestingly, the appli-
some 17 (FTDP-17), a disease also characterized by pathological cation of low levels of oxygen (3%) during the differentiation of
changes in white matter (28–30). Therefore, we generated iOL PSC to OL could greatly enhance oligodendroglial lineage com-
from two iPSC clones from one patient carrying the N279K mitment, resulting in around 40% O4+ cells at day 135 (10). The
MAPT mutation associated with FTDP-17 (17) and compared use of retinoic acid from the beginning of differentiation, followed
these to their isogenic controls (31). Additionally, we included by the application of sonic hedgehog agonist SAG could further
another independent control iPSC line. increase the oligodendroglial lineage commitment, resulting in up
After 28 d of differentiation, O4+ iOL harboring the N279K to 70% of O4+ OPCs at day 75 (9). To accelerate oligodendroglial
differentiation that is orchestrated by TF, we tested individual and
mutation (MAPT-OL) were morphologically indistinguishable
combinations of TF previously shown to be involved in oligo-
from their gene-corrected control cell lines (MAPT-GC-OL) (Fig.
dendroglial differentiation in rodents (15, 36–40). Our results have
8A) and featured similar differentiation efficiencies among all cell defined a combination of three different TF that efficiently in-
lines included (Fig. 8B). We next set out to investigate whether the duces iOL and indeed overcomes the rate-limiting steps of the
N279K MAPT mutation induces an altered expression of tau transition from a neurogenic to a glial progenitor and accelerates
isoforms in iOL. We purified O4+ iOL using FACS before RNA oligodendroglial differentiation. In line with an earlier study, in
sample preparation. Analysis of MAPT expression revealed mu- which overexpression of SOX10 alone was sufficient to induce
tation-specific significantly higher levels of 4R tau compared with oligodendroglial commitment (∼22% O4+ cells 14 d after trans-
MAPT-GC-OL (Fig. 8C), which is in line with observations in duction) in neural progenitors derived from the human fetal brain
iPSC-derived neurons and brains of FTDP-17 patients harboring (16), in our experiments SOX10 was the only TF that induced
this mutation (17, 32). FTDP-17 patients display widespread expression of O4 in iPSC-derived NPC, demonstrating that
neurodegeneration because of increased cellular vulnerability. SOX10 is one key TF to induce oligodendroglial lineage com-
Therefore, we investigated whether MAPT-OL are more suscep- mitment. However, combination of SOX10 with OLIG2 and
tible to oxidative stress induced by rotenone, an inhibitor of the NKX6.2 further enhanced the commitment into the oligoden-
mitochondrial complex I. Exposure of MAPT and MAPT-GC-OL droglial lineage, resulting in a significantly higher percentage of
for 48 h to rotenone revealed an enhanced vulnerability of MAPT- O4+ cells 14 d after induction. Our protocol was highly efficient
OL to oxidative stress identified by an increased number of and reproducible, resulting in 50–70% O4+ cells and yields of 220–
cleaved CASPASE-3+ iOL in MAPT-cultures (Fig. 8D). This ef- 260% after 28 d using three different iPSC-derived NPC lines and
fect was obvious in all tested concentrations of rotenone (100, 250, a single ESC-derived NPC line.
and 500 nM), leading to an average increase of cell death of 48.9 ± The myelinating capacity of iOL was tested in vitro and in vivo.
18.7% in MAPT-OL (Fig. 8E). In vitro iOL ensheathed the neuronal processes of iPSC-derived

6 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1614412114 Ehrlich et al.

 PNAS PLUS
NEUROSCIENCE
Fig. 6. iOL give rise to functional myelin following engraftment in brains of newborn mice. (A) Transplantation of iOL into the corpus callosum of newborn Shi/Shi
Rag2−/− mice resulted in extensive generation of MBP+ myelin (green) by human cells expressing RFP and staining positive for the human nuclei marker STEM101 (red)
16 wpg. (B) Higher magnification of the boxed area in A. (C) Human OL revealed by combined human nuclei STEM101 and cytoplasmic STEM121 (red) markers send
multiple processes connected with MBP+ myelin. Orthogonal view of the boxed area illustrating the colocaliziation of cytoplasmic STEM121 with MBP is depicted in
G. (D) Colabeling of MBP (green) and axonal neurofilament (SM312, red), highlights wrapping of host axons by donor-derived myelin. (E and F) Mature human
NOGO-A+ oligodendrocyte (red, yellow arrow) connected to MBP+ myelin (green) wrapped around host axons (blue). The small panels in F illustrate the merged and
single fluorochromes images represented in the boxed area. Note that unlike MBP, NOGO-A is expressed in the cell cytoplasm (cell body, processes, and paranodal
loops) but not in compact myelin. (H) Human-derived (STEM121+, white)/MBP+ myelin (green) integrate into axo-glial elements expressing Ankyrin-G nodal protein
(blue, yellow arrow) and CASPR paranodal proteins (red). The boxed area is enlarged to illustrate a typical node defined by a STEM121+ grafted cell and its MBP+
myelin internode, with Ankyrin-G+ aggregate flanked by two CASPR+ domains. (I–K) EM images demonstrate that human-derived myelin undergoes final maturation
via compaction and formation of the major dense line. Axons surrounded by compact myelin are indicated by yellow stars. J and K are higher magnifications of boxed
axon in I. n = 4 for immunostaining, n = 3 for EM. [Scale bars, 100 μm (A), 50 μm (B), 20 μm (C), 5 μm (D), 10 μm (E and F), 5 μm (H), 2 μm (I), and 200 nm (J).]

neurons, as well as nanofibers confirming that physical properties the connection of donor cell processes to myelin internodes sur-
of axons are sufficient to induce wrapping of axons, as it has been rounding host axons, and the integration of these myelin inter-
described for rodent OL (41, 42). The coculture of iOL with nodes into nodal and paranodal elements. We provided additional
nanofibers facilitates the identification of compounds that exclu- proof of the donor origin of myelin by the detection at the ultra-
sively promotes axon ensheathment without potentially modulating structural level of thick, compact myelin with the major dense line,
molecular axonal signaling. We did not detect CASPR accumula- a hallmark of wild-type myelin compared with thin, noncompacted,
tions, indicative for paranode formation in iPSC-derived neuron/ and dense line-null shiverer myelin (18, 19). It is known that de-
iOL cocultures, suggesting that distinct axonal signaling cascades spite the lack of MBP, the number of axons ensheathed by non-
required for the formation of paranodes and nodes were not ac- compacted myelin increases progressively in shiverer mice over
tivated in iPSC-derived neurons. However, CASPR+ paranodes time. In the gracile fasciculus in the dorsal spinal cord, the number
were easily detectable in vivo after transplantation into Shi/Shi of myelinated axons increases from 56% 2 wk after birth to up to
Rag2−/− mice. The transplanted cells not only efficiently myelinated 77% at 20 wk (20). Injection of lysolecithin in the dorsal funiculus
the forebrain in newborn Shi/Shi Rag2−/− mice, but also remyeli- results in complete demyelination of these axons (22). (Re)myeli-
nated the adult demyelinated spinal cord. Proof of the donor origin nation, although more efficient in the lesion, was not limited to the
of the myelin was provided by immunohistochemistry using human- dorsal funiculus because MBP+ exogenous myelin was found
specific nuclei and cytoplasmic markers, allowing us to highlight throughout the spinal cord (over 22 mm) with numerous myelin-

Ehrlich et al. PNAS Early Edition | 7 of 10

 pharmacological screens aimed at the identification of compounds
promoting remyelination. In large compound screens using pri-
mary or iPSC-derived OL from rodents, a number of Food and
Drug Administration-approved drugs has been identified that
were able to promote oligodendroglial differentiation in vitro and
remyelination in vivo (1, 24, 25). To determine whether iOL may
be suitable for pharmacological screens, we cultured iOL in the
presence of compounds identified in earlier rodent studies (1, 23–
25). In contrast to these earlier studies, some but not all of these
compounds increased the number of O4+ iOL in a dose-de-
pendent manner and were at least as effective as T3, a known
promoter of oligodendroglial differentiation. Furthermore, only a
subset of these drugs enhanced the maturation of O4+ iOL into
MBP+ mature OL, suggesting that the compounds affect different
stages of oligodendroglial differentiation. Miconazole demon-
strated the strongest effect on iOL; this is in line with an earlier
publication by Najm et al., in which they reported a strong dif-
ferentiation promoting effect of miconazole on OL (1). However,
we observed a toxic effect with a fivefold-higher concentration,
suggesting that miconazole might have a narrow range of efficacy.
Our observations thus suggest that there are species-specific dif-
ferences between rodent and human OL that could be relevant for
drug screens aiming at identifying compounds that promote
oligodendroglial differentiation.
To determine whether iOL are suitable for disease modeling,
Fig. 7. iOL are suitable to test the differentiation promoting effects of se-
lected compounds. Quantification of (A) O4+ and (B) MBP+ iOL after treatment
we characterized in a proof-of-concept study the phenotypes of
with either vehicle, T3, or the drug candidate dissolved in DMSO at three
iOL derived from a patient diagnosed with an inherited form of
different concentrations (0.5, 1, 5 μM) for 21 d in minimum DM. Data are FTD. FTD is characterized by cortical degeneration of the frontal
presented as mean of replicates from three independent experiments + SD. and temporal lobe that in 15–20% of patients with an inherited
One-way ANOVA with Dunnett’s multiple-comparisons test was performed for form of FTD is a result of mutations in the MAPT gene that en-
statistical analysis comparing the mean of each sample with DMSO control (*P < codes the MAPT located on chromosome 17q21. The neuropa-
0.05, **P < 0.01, ***P < 0.001). 0* = Toxic culture condition. thology of FTDP-17 patients with mutations in the MAPT gene is
characterized by tau+ inclusions in neurons and glia, including OL
(for review, see ref. 43). Furthermore, extensive myelin pathology
ated axons in gray matter and ventral white matter, suggesting that can be observed in patients with FTD (28–30). In OL tau regulates
grafted cells dispersed widely in their adult host environment. and stabilizes the microtubule network that is also involved in the
Similar observations were reported for transplanted human fetal transport of RNA granules, for example, those containing MBP
neural progenitors (21), indicating that O4+ iOL are as migratory mRNA. Knockdown of tau or mutated tau in rodent OL impairs
as human fetal NPC in adult Shi/Shi Rag2−/− mice. Compared with process outgrowth and the differentiation into MBP+ myelinating
the developing CNS, the adult CNS represents an environment mature OL (44, 45). Therefore, we assessed whether changes in
with impaired tissue plasticity and trophic support. Whereas the OL may directly contribute to the white matter pathology ob-
colonization and remyelination potential of murine iPSC-derived served in FTDP-17 patients. In iOL from patients with a N279K
NPC has already been described in this environment (22), this mutation in the MAPT gene, we observed as expected, significantly
study demonstrates that human iPSC-derived oligodendroglial increased expression levels of the 4R tau isoform. Furthermore,
lineage cells have similar capabilities. However, one has to keep in we observed an increased susceptibility to cell death induced by
mind that the endogenous cells are compromised because of the respiratory stress compared with gene-corrected control cell lines,
lack of MBP expression in contrast to the grafted cells. Therefore, similar to that reported in iPSC-derived neurons from the same
the grafted cells may have a competitive advantage and not only patient (17). These data suggest that MAPT mutations in OL may
remyelinate demyelinated axons but also replace the endogenous directly contribute to myelin pathology and thus to disease pro-
pool of myelinating cells. In the adult spinal cord, the majority of gression in patients with FTDP-17.
iOL differentiated into OL (81%), rarely into neurons (2%), and In summary, our studies demonstrate that a combination of three
few into astrocytes (estimated to be 17%), confirming previous TFs, namely SOX10, OLIG2, and NKX6.2, greatly accelerates the
data obtained with iPS-derived O4+ cells (9) or O4+ human fetal generation of OL from iPSC-derived NPC and that these cells are
cells (11) grafted in the newborn shiverer mouse brain. In- suitable for disease modeling and pharmacological screens. Thus,
terestingly iOL-derived neurons were observed only in gray matter our protocol should significantly facilitate the development of high-
surrounding the lesion, and iOL-derived astrocytes were detected throughput screening platforms and the study of human myelin
only in the demyelinated area but not in the ventral or lateral white diseases and repair strategies using patient-derived iPSC.
matter tracks, suggesting that their differentiation is highly regu-
Methods
lated by cell-specific environmental cues.
The iPSC technology is an emerging tool for drug development. Culturing of Human PSC. The iPSC included in this study have previously been
generated and characterized (13, 17). iPSC colonies were maintained on a
Promotion of remyelination represents until now an unmet
layer of mitotically inactivated mouse embryonic fibroblasts in human ESC
treatment strategy for patients with MS. Without SON, less than medium consisting of knockout DMEM (Invitrogen) with 20% (vol/vol)
2% of NPC differentiated into O4+ OL. Importantly, SON Knockout Serum Replacement (Invitrogen), 1 mM β-mercaptoethanol (Invi-
transduced NPC cultured in minimal medium lacking T3 and trogen), 1% nonessential amino acids (NEAA; Invitrogen), 1% penicillin/
growth factors (except dbcAMP), resulted in ∼14% of O4+ OL. streptomycin/glutamine (PAA), freshly supplemented with 5 ng/mL FGF2
Addition of T3 increased the efficiency to 28% and culturing in (Peprotech). PSC were split at ratios of 1:6–1:8 every 7 d by mechanic dis-
DM to 40% after 21 d (as shown in Figs. 3B and 7, respectively). aggregation with 1 mg/mL collagenase IV (Invitrogen).
These data demonstrate that SON is sufficient to induce an oli-
godendroglial cell fate; however, growth factors are necessary to Generation and Culturing of Human NPC. NPC were derived from human PSC
further increase maturation of the cells. The increase in differ- by treatment with small molecules as previously described (13, 17). Meth-
entiation by exogenous factors is prerequisite to use iOL for odological details are provided in SI Methods.

8 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1614412114 Ehrlich et al.

 PNAS PLUS
A B

C D E

NEUROSCIENCE
Fig. 8. MAPT-OL exhibit mutation related phenotypes. (A) Immunostaining for O4 (green) demonstrating differentiation of iPSC carrying the N279K MAPT
mutation (MAPT1, MAPT2) and genetic corrected controls (MAPT1 GC, MAPT2 GC) into iOL. Nuclei were counterstained with Hoechst (blue). (Scale bar, 25 μm.)
(B) Flow cytometry-based quantification of O4+ iOL after 28 d of differentiation in MAPT mutation cultures, genetic corrected cultures, and an independent
healthy control culture. Data are presented as mean of replicates from three independent experiments + SD. (C) Quantitative RT-PCR analysis on control, MAPT
gene-corrected, and MAPT mutated iOL cultures for 4R tau isoforms containing exon 10. Expression levels were normalized to total tau expression and control
lines. Data are presented as mean of replicates from three independent experiments + SD. One-way ANOVA with post hoc Tukey test was performed for statistical
analyses (**P < 0.01, ***P < 0.001). (D) Quantification of cleaved CASPASE 3+ iOL in control and MAPT cultures after 48 h of either vehicle [0.01% (vol/vol) DMSO]
or rotenone treatment. Data are presented as mean of replicates from three independent experiments + SD. One-way ANOVA with post hoc Tukey test was
performed for statistical analysis (*P < 0.05, **P < 0.01). (E) All results combined after normalization by setting all control cultures to 100%, show that MAPT
N279K causes a higher sensitivity to oxidative stress. Error bars present SD. Student’s t test was performed for statistical analysis (***P < 0.001).

Lentiviral Vector Construction and Production of Lentiviral Particles. Details Neuronal Differentiation, in Vitro Myelination Assay, and 3D Culture. Human
including the cloning strategies are provided in SI Methods and Fig. S14. iPSC-derived NPC were differentiated into neurons as previously described (17).
To assess the in vitro myelination capacity of iOLs, O4+ cells were purified at
Transduction of NPC for TF Screening. Human NPC were seeded with a density differentiation day 21 by magnetic cell sorting using anti-O4 MicroBeads (Mil-
of 1 × 105 cells per well in 12-well plates, allowed to attach overnight, and tenyi Biotec) following the manufacturer’s protocol. Purified iOLs were added to
transduced with equal volumes of concentrated Lenti-rtTA and 1-TF virus 21-d-old neuronal cultures derived from iPSC-derived NPC populations or
particle supplemented with 5 μg/mL protamine sulfate (Sigma) in fresh NPC nanofibers. Details are provided in SI Methods.
expansion medium (NPCM). The 2-TF infections were done by mixing
equivalent volumes of Lenti-rtTA, pLV-TetO-SOX10 and 1-TF virus particle Isolation of Adult Human pOL. Brain tissue was obtained from adults un-
for infection. For 3-TF infections, the volume of each virus was reduced by dergoing surgical resections as treatment for nontumor-related intractable
one quarter and equivalent volumes of Lenti-rtTA, pLV-TetO-SOX10, pLV- epilepsy in accordance with the guidelines set by the Biomedical Ethics Unit of
McGill University. As described previously (46), tissue specimens were enzy-
TetO-OLIG2, and 1-TF were mixed for NPC transduction. Further details are
matically digested and placed on a linear 30% Percoll density gradient
provided in SI Methods.
(Pharmacia Biotech). Further details are provided in SI Methods.
Oligodendroglial Differentiation. For oligodendroglial differentiation, human
Whole-Genome Expression Analysis. Details including the microarray data
NPC were seeded with a density of 1 × 105 cells per well in 12-well plates,
processing are provided in SI Methods.
allowed to attach overnight, and transduced with concentrated SON lenti-
viral particle and 5 μg/mL protamine sulfate in fresh NPCM. Viral medium
Cell Transplantation. Details of cell transplantation into Shi/Shi Rag2−/− dys-
was removed after 24 h and replaced with GIM. After 4 d, GIM was replaced
myelinating immunodeficient mice are provided in SI Methods. Animal ex-
by DM. Details are provided in SI Methods.
periments were performed according to European Community regulations
and INSERM ethical committee (authorization 75-348; 20/04/2005) and were
ICC. Details of ICC are provided in SI Methods and Table S2. approved by the local Darwin ethical committee.

Flow Cytometry Analysis. Cells were enzymatically detached and singularized Immunohistochemistry and EM. The methodology for immunohistochemistry
by accutase treatment for 10 min at 37 °C. Following washing with PBS, and EM is described in detail in SI Methods.
singularized cells were resuspended in PBS/0.5% BSA buffer and filtered
through a 70-μm cell strainer (BD Falcon). After determination of cell Compound Screen. To assess the sensitivity of iOL toward differentiation promoting
number, cells were incubated with mouse IgM anti–O4-APC antibody (Mil- drugs, iPSC-derived NPC were transduced with concentrated SON virus particle, as
tenyi Biotec) following the manufacturer’s protocol. Stained cells were described above. Viral medium was removed after 24 h and replaced with GIM
washed, resuspended in PBS/0.5% BSA buffer (5 × 106 cells/mL), and imme- lacking T3. The end of the virus infection period was termed day 0. Further details
diately sorted using a FACSAria cell sorter (BD Biosciences). Debris, doublets, including the differentiation with drug candidates are provided in SI Methods.
and aggregates were excluded by appropriate gating using forward and
side scatter. Additionally, DAPI was used for dead cell exclusion. Unstained Generation and Characterization of N297K MAPT NPC and Isogenic Controls.
cells were used to set background fluorescence and undifferentiated human The N279K MAPT iPSC-derived NPC included in this study have previously
NPC were used as negative controls. been generated and characterized (17). Frozen NPC, termed FTDP-17-1-I and

Ehrlich et al. PNAS Early Edition | 9 of 10

FDTP-17-1-II in the aforementioned publication were thawed at passages ACKNOWLEDGMENTS. We thank Jacqueline Trotter (Molecular Cell Biology,
10–14 and designated as MAPT-1 and MAPT-2 in this study. Department of Biology, Johannes Gutenberg University Mainz) for critical
reading and helpful discussion of the manuscript, and Martina Sinn and
Quantitative RT-PCR. Details of RT-PCR including a list of primers are provided Ingrid Gelker for excellent technical support. This study was supported by
in SI Methods and Tables S3. German Research Foundation Grant SFB-TR128-B7; Ku1477/6-1 (to T.K.);
Interdisciplinary Clinical Research Center, Münster Grant IZKF; KuT3/012/15
(to T.K.); an operating grant from the Multiple Sclerosis Society of Canada
Stress-Induced Cell Death. Details are described in SI Methods.
(to J.A.); Cluster of Excellence REBIRTH (A.S.); Multiple Sclerosis Society of the
United States Grant RGPA-1501-02553; and INSERM, the program “Inves-
Quantifications. Quantifications were performed as described in detail in SI Methods. tissements d’Avenir” ANR-10-IAIHU-06 and ANR-11-INBS-0011–NeurATRIS.
S.M. is supported by a Du Pré grant from the Multiple Sclerosis International
Statistics. Data of at least three independent differentiation experiments are Foundation, Fondation des Treilles, Ecole de Neurosciences de Paris, and
presented as mean + SD. Statistical significance was determined by Student’s Electricité de France Foundation. A.M. is supported by the European Leuko-
t test and with one-way ANOVA, respectively. dystrophy Foundation.

1. Najm FJ, et al. (2015) Drug-based modulation of endogenous stem cells promotes 26. LoPresti P (2002) Regulation and differential expression of tau mRNA isoforms as
functional remyelination in vivo. Nature 522(7555):216–220. oligodendrocytes mature in vivo: Implications for myelination. Glia 37(3):250–257.
2. Yang N, et al. (2013) Generation of oligodendroglial cells by direct lineage conversion. 27. Gorath M, Stahnke T, Mronga T, Goldbaum O, Richter-Landsberg C (2001) De-
Nat Biotechnol 31(5):434–439. velopmental changes of tau protein and mRNA in cultured rat brain oligodendro-
3. Goldman SA, Nedergaard M, Windrem MS (2012) Glial progenitor cell-based treat- cytes. Glia 36(1):89–101.
ment and modeling of neurological disease. Science 338(6106):491–495. 28. Lam BY, Halliday GM, Irish M, Hodges JR, Piguet O (2014) Longitudinal white matter
4. Franklin RJ, ffrench-Constant C, Edgar JM, Smith KJ (2012) Neuroprotection and re- changes in frontotemporal dementia subtypes. Hum Brain Mapp 35(7):3547–3557.
pair in multiple sclerosis. Nat Rev Neurol 8(11):624–634. 29. Kovacs GG, et al. (2008) White matter tauopathy with globular glial inclusions: A
5. Sreedharan J, Brown RH, Jr (2013) Amyotrophic lateral sclerosis: Problems and pros- distinct sporadic frontotemporal lobar degeneration. J Neuropathol Exp Neurol
pects. Ann Neurol 74(3):309–316. 67(10):963–975.
6. Kim JB, et al. (2015) Oct4-induced oligodendrocyte progenitor cells enhance func- 30. Lu PH, et al. (2014) Regional differences in white matter breakdown between fron-
tional recovery in spinal cord injury model. EMBO J 34(23):2971–2983.
totemporal dementia and early-onset Alzheimer’s disease. J Alzheimers Dis 39(2):
7. Tallantyre EC, et al. (2010) Clinico-pathological evidence that axonal loss underlies
261–269.
disability in progressive multiple sclerosis. Mult Scler 16(4):406–411.
31. Hallmann AL, et al. (2017) Astrocyte pathology in a human neural stem cell model of
8. Atkins H, Freedman M (2009) Immune ablation followed by autologous hematopoi-
frontotemporal dementia caused by mutant TAU protein. Sci Rep 7:42991.
etic stem cell transplantation for the treatment of poor prognosis multiple sclerosis.
32. Goedert M, Spillantini MG (2000) Tau mutations in frontotemporal dementia FTDP-17
Methods Mol Biol 549:231–246.
and their relevance for Alzheimer’s disease. Biochim Biophys Acta 1502(1):110–121.
9. Douvaras P, et al. (2014) Efficient generation of myelinating oligodendrocytes from
33. Czepiel M, et al. (2011) Differentiation of induced pluripotent stem cells into func-
primary progressive multiple sclerosis patients by induced pluripotent stem cells. Stem
tional oligodendrocytes. Glia 59(6):882–892.
Cell Rep 3(2):250–259.
34. Hu BY, Du ZW, Zhang SC (2009) Differentiation of human oligodendrocytes from
10. Stacpoole SR, et al. (2013) High yields of oligodendrocyte lineage cells from human
embryonic stem cells at physiological oxygen tensions for evaluation of translational pluripotent stem cells. Nat Protoc 4(11):1614–1622.
biology. Stem Cell Rep 1(5):437–450. 35. Izrael M, et al. (2007) Human oligodendrocytes derived from embryonic stem cells:
11. Wang S, et al. (2013) Human iPSC-derived oligodendrocyte progenitor cells can Effect of noggin on phenotypic differentiation in vitro and on myelination in vivo.
myelinate and rescue a mouse model of congenital hypomyelination. Cell Stem Cell Mol Cell Neurosci 34(3):310–323.
12(2):252–264. 36. Sugimori M, et al. (2008) Ascl1 is required for oligodendrocyte development in the
12. Piao J, et al. (2015) Human embryonic stem cell-derived oligodendrocyte progenitors spinal cord. Development 135(7):1271–1281.
remyelinate the brain and rescue behavioral deficits following radiation. Cell Stem 37. Pozniak CD, et al. (2010) Sox10 directs neural stem cells toward the oligodendrocyte
Cell 16(2):198–210. lineage by decreasing Suppressor of Fused expression. Proc Natl Acad Sci USA 107(50):
13. Reinhardt P, et al. (2013) Derivation and expansion using only small molecules of 21795–21800.
human neural progenitors for neurodegenerative disease modeling. PLoS One 8(3): 38. Lu QR, et al. (2000) Sonic hedgehog-regulated oligodendrocyte lineage genes en-
e59252. coding bHLH proteins in the mammalian central nervous system. Neuron 25(2):
14. Najm FJ, et al. (2013) Transcription factor-mediated reprogramming of fibroblasts to 317–329.
expandable, myelinogenic oligodendrocyte progenitor cells. Nat Biotechnol 31(5): 39. Jessberger S, Toni N, Clemenon GD, Jr, Ray J, Gage FH (2008) Directed differentiation
426–433. of hippocampal stem/progenitor cells in the adult brain. Nat Neurosci 11(8):888–893.
15. Liu Z, et al. (2007) Induction of oligodendrocyte differentiation by Olig2 and Sox10: 40. Maire CL, et al. (2009) Directing human neural stem/precursor cells into oligoden-
Evidence for reciprocal interactions and dosage-dependent mechanisms. Dev Biol drocytes by overexpression of Olig2 transcription factor. J Neurosci Res 87(15):
302(2):683–693. 3438–3446.
16. Wang J, et al. (2014) Transcription factor induction of human oligodendrocyte pro- 41. Lee S, et al. (2012) A culture system to study oligodendrocyte myelination processes
genitor fate and differentiation. Proc Natl Acad Sci USA 111(28):E2885–E2894. using engineered nanofibers. Nat Methods 9(9):917–922.
17. Ehrlich M, et al. (2015) Distinct neurodegenerative changes in an induced pluripotent 42. Bechler ME, Byrne L, Ffrench-Constant C (2015) CNS myelin sheath lengths are an
stem cell model of frontotemporal dementia linked to mutant TAU protein. Stem Cell intrinsic property of oligodendrocytes. Curr Biol 25(18):2411–2416.
Rep 5(1):83–96. 43. Ghetti B, et al. (2015) Invited review: Frontotemporal dementia caused by microtu-
18. Lachapelle F, et al. (1983-1984) Transplantation of CNS fragments into the brain of
bule-associated protein tau gene (MAPT) mutations: A chameleon for neuropathol-
shiverer mutant mice: Extensive myelination by implanted oligodendrocytes. I. Im-
ogy and neuroimaging. Neuropathol Appl Neurobiol 41(1):24–46.
munohistochemical studies. Dev Neurosci 6(6):325–334.
44. Seiberlich V, et al. (2015) Downregulation of the microtubule associated protein tau
19. Gansmuller A, et al. (1986) Transplantations of newborn CNS fragments into the brain
impairs process outgrowth and myelin basic protein mRNA transport in oligoden-
of shiverer mutant mice: Extensive myelination by transplanted oligodendrocytes. II.
drocytes. Glia 63(9):1621–1635.
Electron microscopic study. Dev Neurosci 8(4):197–207.
45. Klein C, et al. (2002) Process outgrowth of oligodendrocytes is promoted by in-
20. Inoue Y, Inoue K, Terashima T, Mikoshiba K, Tsukada Y (1983) Developmental
teraction of fyn kinase with the cytoskeletal protein tau. J Neurosci 22(3):698–707.
changes of oligodendroglia in the posterior funiculus of “Shiverer” mutant mouse
46. Cui QL, et al. (2010) Response of human oligodendrocyte progenitors to growth
spinal cord, with special reference to myelin formation. Anat Embryol (Berl) 168(2):
159–171. factors and axon signals. J Neuropathol Exp Neurol 69(9):930–944.
21. Buchet D, Garcia C, Deboux C, Nait-Oumesmar B, Baron-Van Evercooren A (2011) 47. Ahfeldt T, et al. (2012) Programming human pluripotent stem cells into white and
Human neural progenitors from different foetal forebrain regions remyelinate the brown adipocytes. Nat Cell Biol 14(2):209–219.
adult mouse spinal cord. Brain 134(Pt 4):1168–1183. 48. Warlich E, et al. (2011) Lentiviral vector design and imaging approaches to visualize
22. Mozafari S, et al. (2015) Skin-derived neural precursors competitively generate the early stages of cellular reprogramming. Mol Ther 19(4):782–789.
functional myelin in adult demyelinated mice. J Clin Invest 125(9):3642–3656. 49. Sommer CA, et al. (2009) Induced pluripotent stem cell generation using a single
23. Preisner A, et al. (2015) Non-steroidal anti-inflammatory drug indometacin enhances lentiviral stem cell cassette. Stem Cells 27(3):543–549.
endogenous remyelination. Acta Neuropathol 130(2):247–261. 50. Irizarry RA, et al. (2003) Exploration, normalization, and summaries of high density
24. Mei F, et al. (2014) Micropillar arrays as a high-throughput screening platform for oligonucleotide array probe level data. Biostatistics 4(2):249–264.
therapeutics in multiple sclerosis. Nat Med 20(8):954–960. 51. Marteyn A, et al. (2016) Modulation of the innate immune response by human neural
25. Deshmukh VA, et al. (2013) A regenerative approach to the treatment of multiple precursors prevails over oligodendrocyte progenitor remyelination to rescue a severe
sclerosis. Nature 502(7471):327–332. model of Pelizaeus-Merzbacher disease. Stem Cells 34(4):984–996.

10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1614412114 Ehrlich et al.

Supporting Information
Ehrlich et al. 10.1073/pnas.1614412114
SI Methods Transduction of NPC for TF Screening. Viral medium was removed
Generation and Culturing of Human NPC. Human PSC colonies from from human NPC after 24 h and replaced by N2B27 medium
passages 10–15 were mechanically sectioned and enzymatically supplemented with 1 μM SAG, 10 ng/mL PDGF (Peprotech),
detached from mouse embryonic fibroblasts. Pieces of PSC col- 10 ng/mL NT3 (Peprotech), 10 ng/mL IGF-I (Peprotech), 200
onies were collected by sedimentation, resuspended in ESC μM AA, 1:1,000 Trace Elements B (Corning), and 60 ng/mL T3
medium (without FGF) supplemented with 10 μM SB-431542 (Sigma). The end of the virus infection period was termed day
(Ascent Scientific), 1 μM dorsomorphin (Tocris), 3 μM 0 and transgene expression was induced with 2 μg/mL doxycy-
CHIR99021 (CHIR; Axon Medchem), and 0.5 μM purmorph- cline (Clontech) for 14 d. Medium was replaced every other day
amine (Alexis), and subsequently cultured as embryoid bodies in and cells were fixed in 4% paraformaldehyde (PFA; Sigma) in
Petri dishes. The medium was changed after 2 d to N2B27 me- PBS (Invitrogen) for ICC analysis at day 14 of differentiation.
dium consisting of equal parts DMEM-F12 (Invitrogen) and Investigators were blinded for ICC analyses.
Neurobasal (Invitrogen) with 1:200 N2 supplement (Invitrogen),
1:100 B27 supplement lacking vitamin A (Invitrogen), 1% pen- Oligodendroglial Differentiation. GIM consisted of DMEM-F12
icillin/streptomycin/glutamine, and with the same small-molecule with 1:200 N2 supplement, 1:100 B27 supplement lacking vitamin
supplements as mentioned before. On day 4, SB-431542 and A, 1% penicillin/streptomycin/glutamine, 1 μM SAG, 10 ng/mL
dorsomorphin were withdrawn and 150 μM ascorbic acid (AA; PDGF, 10 ng/mL NT3, 10 ng/mL IGF-I, 200 μM AA, 1:1,000
Sigma) was added to the medium. On day 6, embryoid bodies Trace Elements B, 10 ng/mL T3. GIM was changed every other
were disintegrated into smaller pieces and plated on Matrigel- day and replaced at day 4 posttransduction by DM consisting of
coated (Matrigel, growth factor reduced, high concentration; BD DMEM-F12 with 1:200 N2 supplement, 1:100 B27 supplement
Biosciences) 12-well plates (Nunc) in NPCM consisting of lacking vitamin A, 1% penicillin/streptomycin/glutamine, 60 ng/mL
N2B27 medium supplemented with 3 μM CHIR, 0.5 μM SAG T3, 10 ng/mL NT3, 10 ng/mL IGF-I, 200 μM AA, 1:1,000 Trace
(Cayman Chemical), and 150 μM AA. Cells were split once a week at Elements B, and 100 μM dbcAMP (Sigma). After 7–10 d of dif-
ratios of 1:15–1:20 by treatment with accutase (Sigma). Regular tests ferentiation, cells were detached and singularized by treatment with
for mycoplasma contamination using the MycoAlert mycoplasma accutase and reseeded at densities of 2.5 × 105 cells in 12-well plates
detection kit (Lonza) were negative. and 1.5 × 105 cells in 24-well plates containing glass coverslips. DM
was changed every other day.
Lentiviral Vector Construction and Production of Lentiviral Particles.
The coding regions of SOX10, OLIG2, ASCL1, NKX2.2, ICC. For ICC analysis, an equal volume of 4% PFA in PBS was
NKX6.1, NKX6.2, MYT1, and RFP were amplified by PCR, added to the culture medium and cells were prefixed for 10 min at
validated by sequencing, cloned into pCR8/GW/TOPO (Invi- room temperature. After removing the supernatant, cells were
trogen) according to the manufacturer’s instructions, and fixed for an additional 15 min with 4% PFA in PBS and washed
recombined into pLV-tetO-attR1/R2 (47) by LR clonase II three times with PBS. Fixed cells were permeabilized by adding
(Invitrogen). pLV-tetO-attR1/R2 and Lenti-rtTA were kindly 0.2% Triton X-100 (Sigma) in PBS for 15 min at room temperature
provided by Chad Cowan, Harvard University, Cambridge, MA. (this step was omitted for NG2, O4, and GalC staining). Sub-
To construct the polycistronic lentiviral SON vector, we used a sequently, blocking was performed by incubating the cells with 5%
third-generation lentiviral vector (kindly provided by L. Naldini, normal goat serum (NGS; Gibco) and 5% FCS (Gibco) for 30 min
Ospedale San Raffaele, Milan, Italy), which we further equipped at room temperature. Primary antibodies were applied overnight at
with the retroviral SFFV U3 promoter and the reprogramming 4 °C in blocking solution. Following three washing steps with PBS,
cassette (48). To be optionally able to excise the reprogramming cells were incubated with Alexa Fluor-conjugated secondary an-
cassette later, we incorporated a Flp recognition target site in the tibodies diluted in PBS for 1 h at room temperature. Cells were
3′ U3 region. The human cDNAs encoding SON were inserted subsequently washed three times with PBS, including a Hoechst
to create a three-in-one vector, in which the transcription factor counterstaining for nuclei in the second washing step. Cells on
genes are coexpressed and linked by 2A self-cleavage sites (P2A, glass coverslips were mounted in Dako Fluorescent Mounting
T2A). Furthermore, we introduced an internal ribosome entry Medium (Dako) and visualized on a Zeiss LSM700 confocal mi-
site (IRES)-RFP cassette (encoding the dimeric RFP variant croscope. Cells on plastic cell culture plates were visualized on a
dTomato) for visualization of vector expression, as previously Leica DMI6000 B inverted microscope. Primary antibodies used
described (48) (Fig. S14). Tet-inducible SON was generated by in this study are listed in Table S2.
cloning SON cassette into pHAGE, a third-generation lentiviral
vector previously described (49). The construct was also equip- Neuronal Differentiation. For the derivation of human neurons,
ped by IRES-pac cassette allowing puromycin selection (Fig. iPSC-derived NPC were cultured with N2B27 medium supple-
S3A). The constructs were validated by sequencing. Cloning mented with 1 μM SAG (Cayman Chemical), 2 ng/mL BDNF
details are available upon request. Lentiviral particles were (Peprotech), 2 ng/mL GDNF (Peprotech), and 100 μM AA
produced by cotransfection of 293T cells with individual ex- (Sigma) for 6 d and afterward with N2B27 medium supplemented
pression vectors in combination with the packaging plasmids with 2 ng/mL BDNF (Peprotech), 2 ng/mL GDNF (Peprotech),
psPAX2 (Addgene #12260) and pMD2.G (Addgene #12259). 0.5 ng/mL TGF-β3 (Peprotech), 100 μM dbcAMP, and 100 μM
Additional reverse tetracycline transactivator lentivirus (FUW- AA. Additionally, 5 ng/mL Activin A (Sigma) was added to the
M2rtTA, Addgene #20342) was generated to coinfect with tet- medium from days 7–9. After 9 d of neuronal differentiation, cells
inducible lentiviruses. Virus-containing supernatants were har- were detached, singularized by treatment with accutase, and
vested at 48 and 72 h posttransfection and filtered with 0.45-μm reseeded at densities of 2 × 105 per well in 24-well plates
PVDF membrane (Millipore). Viral particles were subsequently containing glass coverslips. After 21 d of differentiation, cells
concentrated by ultracentrifugation, resuspended in N2B27 were used for coculture experiments. Transgene expression
medium, and stored at −80 °C. for the oligodendroglial differentiation experiments using the

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 1 of 12

tet-inducible SON construct was induced with 1 μg/mL doxy- formed using a robust multiarray average method implemented in
cycline (Clontech). the “oligo” package (50). Variance stabilization was performed
using the log2 scaling. Differentially expressed genes among iOL
In Vitro Myelination Assay and 3D Culture. To assess the in vitro and iPSC-derived NPC samples were identified through an un-
myelination capacity of iOL, O4+ cells were purified at differ- paired one-way between-subject ANOVA. P values were cor-
entiation day 21 by magnetic cell sorting using anti-O4 rected for multiplicity according to the Benjamini–Hochberg
MicroBeads (Miltenyi Biotec) following the manufacturer’s procedure with a threshold of 0.05 (false-discovery rate). Results
protocol. Purified iOL were added to 21-d-old neuronal cultures were further filtered by fold-change magnitude (jfold changej ≥ 2).
derived from iPSC-derived NPC populations with the afore- Graphics were obtained using ComplexHeatmap and VennDia-
mentioned protocol at densities of 1 × 105 cells per well in gram R-packages. Hierarchical cluster of samples was performed
Matrigel-coated 24-well plates containing glass coverslips. Co- with pvclust R-package using the one minus the sample correlation
cultures were maintained in DM supplemented with 2 ng/mL metric and complete-linkage clustering method. Probe mapping to
BDNF and 2 ng/mL GDNF. After 14–28 d of coculture, cells the corresponding gene information was performed using the Bio-
were fixed in 4% PFA for ICC analysis. conductor package Annotationdbi.
For the 3D culture experiments, nanofiber chamber slides
(Nanofiber Solutions) containing aligned nanofiber polymers Cell Transplantation. Shiverer mice were crossed to Rag2-null
were precoated with 10 μg/mL laminin (Sigma) and incubated immunodeficient mice to generate a line of Shi/Shi Rag2−/−
with DM supplemented with 2 ng/mL BDNF and 2 ng/mL dysmyelinating immunodeficient mice. Mice were housed under
GDNF at 37 °C overnight. O4+ iOLs were purified after 21 d of standard conditions of 12-h light/ 12-h dark cycles with ad libi-
differentiation using MACS and were reseeded at a density of tum access to dry food and water cycle at the Institut du Cerveau
5 × 104 cells per chamber in DM supplemented with 2 ng/mL et de la Moelle épinière animal facility (PHENOPARC). Experi-
BDNF and 2 ng/mL GDNF. Half of the medium was changed ments were performed according to European Community regula-
every other day and cells were fixed in 4% PFA after 14 d for tions and INSERM ethical committee (authorization 75-348; 20/04/
ICC analysis. All experiments have been repeated with at least 2005) and were approved by the local Darwin ethical committee. To
three biological independent iOL populations. assay iOL contributions to developmental myelination, newborn pups
(n = 7) were cryoanesthetized before bilateral transplantation of
Isolation of Adult Human pOL. Microglia were separated and re- RFP+ iOL, rostral to the corpus callosum. Injections (105 cellsμL
moved from the specimens by an initial adhesion step in which the each) were performed 1-mm caudally, 1-mm laterally from the
total cell fraction was cultured for 24 h in noncoated flasks. The bregma, and to a depth of 1 mm (51). Animals were killed at 16 wpg
floating cell fraction was subjected to immunomagnetic bead for immunohistological studies (n = 4) and EM (n = 3). To assay iOL
selection with the A2B5 antibody (IgM) to select out progenitor involvement in remyelination, mice (n = 4) of 8–9 wk of age, were
cells (48). The nonselected fraction (referred to as pOL) was anesthetized by intraperitoneal injection of a mixture of 100 mg/kg
plated on poly-L-lysine–coated glass-chamber slides in defined ketamine (Alcyon) and 10 mg/kg xylazine (Alcyon). Focal de-
medium (DFM) consisting of DMEM-F12 supplemented with myelination was performed as previously described (51) by stereotaxic
N1 (Sigma), 0.01% BSA, 1% penicillin-streptomycin and B27 injection of 1 μL of 1% LPC (Sigma-Aldrich) in 1% PBS into the
supplement lacking Vitamin A, 10 ng/mL PDGF, 10 ng/mL dorsal funiculus of the spinal cord at the level of the 13th thoracic
bFGF (Sigma), and 2 nM T3. After 6 d in DFM, cells were either vertebrae. Forty-eight hours after demyelination, mice received a
lysed with Isol-RNA Lysis Reagent (Thermo Fisher) or fixed in single injection (1 μL, 105 cells/μL) of RFP+ iOL at the site of de-
4% PFA for ICC analysis, demonstrating that ∼90% of cells myelination. All injections (LPC or cells) were performed at low
were O4+ and MBP+ (Fig. S5). Expression of O1 (galactocer- speed (1 μL/2 min) using a stereotaxic frame equipped with a mi-
broside) has been documented on these cells by flow cytometry. cromanipulator and a Hamilton syringe. Animals were killed at
12 wpg for immunohistological studies.
Whole-Genome Expression Analysis. Total RNA was quantified
using a NanoDrop Spectrophotometer ND-1000 (NanoDrop Immunohistochemistry. For immunohistochemistry, mice were
Technologies) and its integrity was assessed using a 2100 Bio- killed by transcardiac perfusion-fixation with 4% PFA in PBS and
analyzer (Agilent Technologies). Five nanograms of total RNA processed for freezing. After dissection, tissues were postfixed in
was used as input for cRNA synthesis with the Genechip WT Pico the same fixative for 1 h, and then processed in 20% sucrose in 1×
Reagent Kit (Affymetrix) according to the manufacturer’s pro- PBS overnight. Sagittal brain and serial cross spinal cord sections
tocol. The processed human iPSC-derived samples were hy- of 12-μm thickness were performed with a cryostat (CM3050S;
bridized onto GeneChip Human Transcriptome 2.0 Arrays Leica). Transplanted cells were identified by RFP expression and
(Affymetrix) and the primary human samples were hybridized with anti-human cytoplasm (STEM121, 1:500; SC Proven) and
onto Human Gene 2.0 ST Arrays (Affymetrix) for 16 h following anti-human nuclei (STEM101, 1:100; SC Proven) antigens. In
the manufacturer’s protocol. Next, GeneChips were washed and vivo characterization of grafted cells was performed by immu-
stained using the GeneChip Hybridization, Wash and Stain Kit nostaining using the following antibodies: anti-MBP [(Abcam,
(Affymetrix) and the GeneChip Fluidics Station 450 (Affyme- ab7349) Rat, 1:100; (Millipore, MAB387) IgG2b, 1:100], anti-
trix). The Arrays were scanned by the GeneChip Scanner 3000 MOG (mouse IgG1 hybridoma, clone C18C5; 1:20; provided by
7G (Affymetrix) and first data processing was performed by the C. Linnington, University of Glasgow, Glasgow, United King-
GeneChip Command Console Viewer 3.2 (Affymetrix). iPSC- dom), anti-hGFAP (Covance; SMI21, IgG1, 1:1,000), anti-
derived NPC and iOL RNA samples were processed at Uni- OLIG2 (Millipore AB 9610; rabbit, 1:400), anti-CC1(Millipore,
versity Hospital Muenster, and pOL samples were processed by OP80; IgG2b, 1:100), anti-NeuN (Abcam, ab177487; rabbit,
McGill University and the Genome Quebec Innovation Centre. 1:100). Identification of neurofilaments was performed by anti-
NF200 (SMI 312, BioLegend, 837904; mixture IgG1-IgM, 1:200)
Microarray Data Processing. Gene-expression data for iPSC sam- or anti-NF165 (mouse IgG1 hybridoma 2H3, 1:20). Nodes of
ples were obtained from Gene Expression Omnibus (GSE61358) Ranvier were detected by anti-CASPR (1:1,000, rabbit poly-
and used as a negative control. All microarray data from iPSC, clonal antibody, a generous gift from E. Peles, Weizmann In-
pOL, iPSC-derived NPC, and iOL samples were processed using stitute, Rehovot, Israel; from D. R. Colman, McGill University,
Bioconductor package “oligo.” Background subtraction, quantile Montreal, Canada, 1:200) and anti-AnkyrinG (Santa Cruz Bio-
gene-expression normalization, and summarization were per- technology, sc-166602; IgG2b, 1:75). For MOG and MBP staining,

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 2 of 12

sections were pretreated with ethanol (10 min, room temperature). tions (0.5 μM, 1 μM, 5 μM) in minimum DM. The drug candidates
For CASPR staining, slices were incubated in methanol (10 min, comprised miconazole (Sigma), clobetasol (TCI), benztropine
20 °C) and saturated in the presence of 0.1% glycine (Research (Sigma), indometacin (Sigma), clemastine (Sigma), and oxybutynin
Organics). Secondary antibodies conjugated with FITC, TRITC (Sigma). Medium was changed every other day and cells were fixed
(SouthernBiotech), or Alexa Fluor 647 (Life Technologies) were in 4% PFA for ICC analysis after 21 d of treatment.
used, respectively, at 1:100 and 1:1,000. Nuclei were stained with
Dapi (1 μg/mL; Sigma-Aldrich) (1:1,000). Tissue scanning, cell visu- Stress-Induced Cell Death. To examine the effect of rotenone
alization and imaging were performed with a Carl Zeiss microscope (Sigma) on the viability of iOL derived from either N279K MAPT
equipped with ApoTome. 2. or control cells, O4+ iOL were purified using the MACS tech-
nology after 21 d of differentiation and replated at a density of
Electron Microscopy. For EM, grafted (n = 3) and nongrafted Shi/ 8 × 103 cells per well into Matrigel-coated 96-well plates. After
Shi Rag2−/− mice (n = 3) were perfused with 1% PBS followed by another 6 d in DM, cells were treated with either vehicle [0.01%
a mixture of 4% paraformaldehyde/5% glutaraldehyde (Electron
(vol/vol) DMSO] or rotenone dissolved in DMSO at three dif-
Microscopy Science) in 1% PBS. After 2 h postfixation in the
ferent concentrations (100, 250, 500 nM). After 24 h, cells were
same solution, brains were cut sagitally in 100-μm-thick sections
and fixed in 2% osmium tetroxide (Sigma-Aldrich) overnight. fixed and cell toxicity was determined by ICC double-staining
After dehydration, samples were embedded in Epon. Ultrathin using antibodies against cleaved CASPASE 3 and O4.
sections (80 nm) of the medial part of the corpus callosum were
Quantifications.
examined and imaged with a HITACHI 120kV HT-7700 elec-
Cell differentiation in vitro. Quantification of cell populations using
tron microscope.
ICC or FACS were obtained from at least three independent
Quantitative RT-PCR. Total RNA was isolated from cell lysates using differentiation experiments. For ICC quantification, at least four
the RNeasy mini kit (Qiagen) according to the manufacturer’s randomly chosen fields were analyzed and investigators were
protocol and including an on-column DNase digest (RNase free blinded for quantification.
+
DNase Set; Qiagen). Quantification of total RNA was performed Cell differentiation in vivo. Quantification of STEM101/RFP grafted
with Nanodrop ND1000 (Peqlab). cDNA was generated from cells expressing cell-specific markers in adult demyelinated spinal
isolated RNAs by reverse transcription using the High Capacity cord, was performed in the lesion core of the dorsal funiculus
cDNA reverse Transcription Kit (Applied Biosystems). Quanti- according to our previously published method (40). The lesion was
tative RT-PCR was performed on an Applied Biosystems StepOne defined by MOG and cell density by Hoechst staining of all cell
Plus real time cycler (Applied Biosystems) with the Power SYBR nuclei on adjacent sections. For each animal (n = 3), eight trans-
Green PCR master mix (Applied Biosystems). Specificity of the verse serial sections at a 120-μm interval per animal were analyzed.
primers used for RT-PCR reactions was determined beforehand by Cell counts were expressed as the percentage of total STEM101/
agarose gelelectrophoresis. The melting curve of each sample was RFP+ cells in the lesion.
determined to ensure the specificity of the products. The quantitative MBP extension within the dorsal funiculus was performed on
RT-PCR conditions were 2 min at 50 °C, 10 min at 95 °C, 40 cycles of serial cross-sections of the grafted spinal cords (n = 3) scanned with
15 s, at 95 °C, and 1 min at 60 °C. Relative expression levels were an automated slide scanner equipped with fluorescence (Zeiss,
calculated using the 2−ΔΔct method and normalized to biological Axioscan). MBP extension corresponds to the distance spanning
reference samples and using GAPDH as the housekeeping gene, between the most rostral and most caudal sections containing
unless otherwise noted. The primer sequences used in this study are MBP in each animal × distance (120 μm) between two sections.
listed in Table S3. EM analyses. The g-ratio in grafted versus nongrafted mice was
determined as previously described (22). Briefly, the maximum and
Compound Screen. At day 5 of differentiation, cells were detached,
singularized, and reseeded at densities of 5 × 104 cells in Matrigel- minimum diameters of a given axon and the maximum and min-
coated 48-well plates or 1.5 × 105 cells in 24-well plates containing imum axon plus myelin sheath diameter were measured with the
Matrigel-coated glass coverslips. Cells were allowed to recover in ImageJ software at a magnification of 62,000× for at least 70
minimum DM consisting of DMEM-F12 with 1:200 N2 supple- myelinated axons per animal. Data were expressed as the mean of
ment, 1:100 B27 supplement lacking vitamin A, 1% penicillin/ the maximal and minimal values for each axon for grafted (n = 3
streptomycin/glutamine, 200 μM AA, and 100 μM dbcAMP. After mice) or nongrafted Shi/Shi Rag2−/− mice (n = 3). The percentage
24 h cells were treated with vehicle alone [0.01% (vol/vol) DMSO] of myelinated axons in the corpus callosum of the adult grafted or
as a negative control, 60 ng/mL T3 as a positive control, or with a nongrafted mice was evaluated for the axons larger than 1-μm
drug candidate dissolved in DMSO at three different concentra- diameter on 40 visual fields (magnification 62,000×) per animal.

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 3 of 12

Fig. S1. iOL develop increased numbers of primary and secondary processes during differentiation. To analyze morphological changes of O4+ iOL during
differentiation, mean numbers of primary and secondary processes per cell body were quantified at day 7 and day 28 of differentiation. Mean numbers of
primary and secondary processes significantly increased during differentiation, demonstrating a more complex morphology at day 28. Data are presented as
mean of replicates from three independent differentiation experiments, each using an independent iPSC-derived NPC line + SD. Process numbers of at least
100 O4+ cells were quantified per differentiation. Student’s t test was performed for statistical analysis (**P < 0.01).

Fig. S2. Impact of prolonged expansion of SON transduced NPC in GEM on oligodendroglial differentiation efficiencies and yields. SON-transduced NPC were
expanded in GEM for either 4 (blue), 8 (green), or 12 d (red) and subsequently transferred to DM for an additional 28 d. FACS analyses at days 0, 14, and 28 of
differentiation revealed a trend to decreased differentiation efficiency after 12 d of expansion (red curve) (A), whereas yields of O4+ cells (red columns) were
significantly increased (B). Data are presented as mean of replicates from three independent experiments, each using an independent iPSC-derived NPC line +
SD. Two-way ANOVA with Tukey’s multiple comparisons test was performed for statistical analysis (**P < 0.01, ***P < 0.001).

Fig. S3. Analysis of iOL transgene dependency using a tet-inducible lentiviral expression construct. (A) Tet-inducible SON was generated by cloning a poly-
cistronic SON cassette into pHAGE, a third-generation lentiviral expression vector. An IRES-PAC cassette was introduced following the SON expression cassette
allowing optional puromycin selection. (B) NPC were cotransduced with lentiviruses expressing tet-inducible SON and lentiviruses expressing rtTA. Expression
of SON was induced by doxycycline for different time periods including 0 d as a negative control and 28 d as a positive control. Oligodendroglial lineage
commitment was assessed by FACS analyses at day 28. Transduced NPC populations without doxycycline exposure (0) demonstrated comparable differentiation
efficiencies as nontransduced NPC populations (Ctr), confirming the tight control of SON expression. Constitutive expression of SON (28 d) resulted in an
increased differentiation efficiency compared with transient SON expression (7, 14, 21 d), indicating that a subpopulation of O4+ iOL was dependent on the
ectopic expression of SON.

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 4 of 12

Fig. S4. ICC analysis of the lineage commitment of SON-transduced and control NPC at day 28. The neural lineage commitment of control and SON-transduced
NPC was assessed by SOX1 (A), TUJ1 (B), and GFAP (C) immunostaining at differentiation day 28. (Scale bars, 50 μm.) (D) Quantification of GFAP+ cells revealed
no differences in the differentiation efficiencies in control and SON-transduced NPC populations. Data are presented as mean of replicates from three in-
dependent differentiation experiments + SD. Student’s t test was performed for statistical analysis.

Fig. S5. ICC analysis of pOL used for whole-genome expression analysis. Representative immunofluorescence image of pOL obtained from adults undergoing
surgical resections as treatment for nontumor-related intractable epilepsy after 6 d in vitro. The vast majority of cells were O4+ (red) and MBP+ (green). Nuclei
were counterstained with Hoechst (blue). (Scale bar, 50 μm.)

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 5 of 12

Fig. S6. Global transcriptional profiling of iOL. (A) Heatmap illustrating gene expression for cell-type–enriched genes comparing iPSC-derived NPC, iOL and
pOL. Each biological replicate of NPC and iOL presents the mean of two to three independent experiments. (B and C) Venn diagram showing the overlap of
genes significantly up-regulated (B) or down-regulated (C) in four biological independent iOL cell lines compared with their corresponding iPSC-derived NPC
population. Each iOL cell line presents the mean of replicates from two to three independent experiments.

Fig. S7. Confocal analysis of in vitro myelination assays. (A) Confocal immunofluorescence image of iOL cocultured with iPSC-derived neurons for 21 d. The
image illustrates the colocalization of MBP (green) with neuronal processes visualized by TUJ1 (red). Nuclei were counterstained with Hoechst. No MBP ex-
pression was detectable in control cultures. (B) Orthogonal projection illustrates the formation of MBP+ (green) sheaths around TUJ1+ (red) neuronal processes.
(Scale bars, 25 μm in A and 1 μm in B.)

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 6 of 12

Fig. S8. EM analysis of the corpus callosum of nongrafted shiverer mice. In the corpus callosum of adult shiverer mice, 73% of myelin-competent axons (>1 μm
diameter) are ensheathed by thin and noncompacted myelin. Inset shows a 3.5-fold higher magnification of noncompacted myelin. (Scale bar, 1 μm.)

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 7 of 12

Fig. S9. Functional differentiation of iOL into bona fide mature re/myelinating OL following transplantation in adult demyelinated mice. (A) Coronal section
illustrating widespread distribution of iOL (red) derived MBP+ myelin (green) after engraftment into the dorsal funiculus (highlighted by dotted line) of the
adult demyelinated Shi/Shi Rag2−/− spinal cord 12 wpg. (C and D) Higher magnification illustrating that grafted RFP+/ STEM101+ (red) human cells not only
remyelinated the lesion site (C) by producing high amounts of MPB+ myelin (green), but also myelinated axons throughout the spinal cord white and gray
matters, including the ventral column (D). (B) The majority of RFP+/ STEM101+ (red) human cells found in the dorsal funiculus of adult mice were CC1+ (green)/
OLIG2+ (white) mature OL (yellow arrowheads highlight some of them). (E–G) Colabeling for human cytoplasmic/human nuclei (STEM121/STEM101 in green)
revealed that many human cells (E and G) were connected to multiple MBP+ donut-shaped myelin-like structures (red, E and F) indicated by yellow arrowheads
in F and G. (H) MBP+ myelin sheaths (green) surrounding NF165+ host axons (blue) in the dorsal funiculus. It also shows that in demyelinated spinal cord many
axons are wrapped by human cells (RFP+) not yet expressing MBP, indicating a prospective larger remyelination potential of the grafted iOL. The boxed area is
illustrated at higher magnification in the small panels below. (I) Longitudinal view of functional human-derived MBP+ myelin (green) integrated into nodes of
Ranvier revealed by paranodal marker CASPR (red) in adult spinal cord. n = 4 mice for all staining. (Scale bars, 200 μm in A, 20 μm in B–H, and 5 μm in I.)

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 8 of 12

Fig. S10. The majority of human cells grafted in dorsal funiculus were mature OL. (A and B) Illustration that most of huNuclei+ grafted cells (red) were
negative for the neuronal marker NeuN. (C and D) The majority of the huNuclei+ grafted cells differentiated into CC1+/Olig2+ mature OL. Nuclei are stained by
Dapi in A–D. (E and F) Donor-derived MOG+/MBP+ myelin (green) integrated among MOG+/MBP− endogenous myelin (red) in the demyelinated dorsal fu-
niculus. (G) Percentage of the different cell types generated by the huNuclei+ grafted cells. n = 3 mice. (Scale bars, 200 μm in A, C, and E and 50 μm in B, D, and
F.) B, D, and F are higher magnifications of dorsal funiculus of A, C, and E.

Fig. S11. Human iOL grafted in dorsal funiculus expressed human-specific mature oligodendrocyte marker NOGO-A. Triple-labeling of the human cells with
human antinuclei STEM101, anticytoplasmic STEM 121, and anti-human NOGO-A revealed that most of the grafted cells expressed mature oligodendroglial
marker NOGO-A (red). Note that unlike MBP, NOGO-A is expressed in the cell cytoplasm (cell body, processes, and paranodal loops) but not in compact myelin.
Merged image in boxed area is detailed in small panels as NOGO-A (Upper) and STEM121/STEM101 (Lower). n = 3 mice. (Scale bar, 20 μm.)

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 9 of 12

Fig. S12. Human myelin disperses widely and integrates very well among endogenous myelin in adult mice. (A) Serial coronal sections of the grafted adult Shi/
Shi Rag2−/− mouse spinal cord stained for MOG (red) and MBP (green) reveal MOG+/MBP− endogenous and MOG+/MBP+ exogenous myelin, respectively, and
illustrates the widespread distribution of human-derived MBP+ myelin along the rostro-caudal axis (between 10 and 22 mm in length) 12 wpg. (B) Illustration of
more caudal sections expressing MOG only (internal negative controls). n = 3 mice. (Scale bar, 1,000 μm.)

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 10 of 12

Fig. S13. ICC analysis of iOL in DM containing selected compounds. (A and B) Representative immunofluorescence images of iOL cultures treated with either
vehicle [0.01% (vol/vol) DMSO], T3 as a positive control, or 1 μM of the drug candidate miconazole for 21 d in minimum DM. Oligodendroglial lineage
commitment was assessed by (A) O4 (green) and (B) MBP (green) immunostaining; nuclei were counterstained with Hoechst (blue). (Scale bars, 50 μm.)

Fig. S14. Schematic presentation of the polycistronic all-in-one SON lentiviral vector. The human cDNAs encoding SOX10, OLIG2, and NKX6.2 were linked by
2A self-cleavage sites and were inserted into a third-generation lentiviral expression vector equipped with the retroviral SFFV U3 promoter. For the visuali-
zation of transgene expression, an IRES-dTomato cassette was introduced following the SON expression cassette.

Table S1. Gene ontology analysis performed for up- and down-
regulated genes in iOL compared with iPSC-derived NPC
GO term Genes P value

Up-regulated
Cell adhesion 66 4.1E-12
Regulation of action potential 13 1.1E-6
Lipoprotein 41 1.3E-5
Myelin sheath 6 7.9E-5
Axon ensheathment 8 3.0E-4
Ensheathment of neurons 8 3.1E-4
Myelination 6 6.3E-3
Down-regulated
DNA replication 62 2.9E-32
Nucleoplasm 124 2.2E-27
Cell cycle 113 1.2E-25
DNA repair 60 1.6E-20
Mitosis 51 2.7E-19

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 11 of 12

 Table S2. Primary antibodies for ICC
Antibody Dilution Source

Mouse anti-NESTIN 1:300 R&D (MAB1259)

Goat anti-SOX1 1:150 R&D (AF3369)
Rabbit anti-NG2 1:200 Millipore (AB5320)
Mouse anti-O4 1:100 R&D (MAB1326)
Mouse anti-GALC 1:100 Millipore (MAB342)
Rat anti-MBP 1:50 Abcam (AB7349)
Rabbit anti–Ki-67 (SP6) 1:250 Abcam (AB16667)
Mouse anti-CNP 1:250 BioLegend (836401)
Mouse anti-MAG 1:400 Abcam (AB89780)
Mouse anti-TUBBIII (TUJ1) 1:750 Covance (MMS-435P)
Mouse anti-AT8 1:150 Innogenetics (90206)

Table S3. Primer for quantitative RT-PCR

Primer Sequence

hGAPDH_for CTG GTA AAG TGG ATA TTG TTG CCA T

hGAPDH_rev TGG AAT CAT ATT GGA ACA TGT AAA CC
hMAPT_total_for CTC GCA TGG TCA GTA AAA GCA A
hMAPT_total_rev GGG TTT TTG CTG GAA TCC TGG T
hMAPT_Exon10_for CCA AGT GTG GCT CAA AGG AT
hMAPT_Exon12_rev CCC AAT CTT CGA CTG GAC TC

Ehrlich et al. www.pnas.org/cgi/content/short/1614412114 12 of 12