Human neural stem cells differentiate and promote
locomotor recovery in spinal cord-injured mice
Brian J. Cummings*†, Nobuko Uchida‡§, Stanley J. Tamaki‡§, Desire ´e L. Salazar¶?, Mitra Hooshmand¶?,
Robert Summers**, Fred H. Gage**, and Aileen J. Anderson*¶?
Departments of *Physical Medicine and Rehabilitation and¶Anatomy and Neurobiology and?Reeve-Irvine Research Center, University of California, Irvine,
CA 92697;‡StemCells, Inc., 3155 Porter Drive, Palo Alto, CA 94304; and **The Salk Institute for Biological Studies, 10010 North Torrey Pines Road,
La Jolla, CA 92037
Communicated by Irving L. Weissman, Stanford University School of Medicine, Stanford, CA, August 16, 2005 (received for review December 15, 2004)
as neurospheres (hCNS-SCns) survive, migrate, and express differ-
entiation markers for neurons and oligodendrocytes after long-
term engraftment in spinal cord-injured NOD-scid mice. hCNS-SCns
engraftment was associated with locomotor recovery, an obser-
vation that was abolished by selective ablation of engrafted cells
by diphtheria toxin. Remyelination by hCNS-SCns was found in
both the spinal cord injury NOD-scid model and myelin-deficient
shiverer mice. Moreover, electron microscopic evidence consistent
with synapse formation between hCNS-SCns and mouse host
neurons was observed. Glial fibrillary acidic protein-positive astro-
cytic differentiation was rare, and hCNS-SCns did not appear to
contribute to the scar. These data suggest that hCNS-SCns may
possess therapeutic potential for CNS injury and disease.
behavioral assessment ? differentiation ? stem cell transplantation
into rodent models of spinal cord injury. Many of these studies
focused on cell survival and did not address differentiation, func-
tional recovery, or the causal relationship between successful
engraftment and observed behavioral improvements. When differ-
entiation was investigated, embryonic and adult neural stem cells
were reported to principally assume glial fibrillary acidic protein
(GFAP)-positive astrocytic phenotypes after grafting into nonneu-
(3–5). Furthermore, although in vitro predifferentiation paradigms
designed to generate neural lineage restricted precursors success-
fully generated ?-tubulin III (Tuj-1)-positive neuronal phenotypes
either in vitro or after transplantation into uninjured spinal cord,
this commitment was overridden by environmental cues in the
injured spinal cord (6).
Transplants of human brain-derived stem cells or human spinal
cord tissue into injured rat spinal cord have been described (7–9).
Moreover, several human cell transplantation paradigms recently
have been reported to promote locomotor recovery: human um-
bilical cell infusion in a rat spinal cord injury model, although only
within 3 weeks or less postgrafting (10); neurons differentiated
cells and transplanted into a rat spinal cord injury model (11);
human ES cells differentiated in vitro to oligoprogenitors and
transplanted into a rat spinal cord injury model (12); and human
neural stem?progenitor cells transplanted into a monkey spinal
cord injury model (13). In general, these studies lack some or all of
the following: definitive identification of transplanted cells, long-
term survival and engraftment data, evidence of differentiation,
and?or direct evidence of functional integration of human cells in
the injured spinal cord. The current study addresses three previ-
ously unexplored issues in stem cell transplantation research for
injury experiments using prospectively isolated, human CNS stem
cells grown as neurospheres (hCNS-SCns) derived from fetal brain
(14, 15). First, we report that hCNS-SCns survive, engraft, differ-
ecent studies have used a variety of immortalized, engineered,
or isolated rodent-derived precursor?stem cells transplanted
entiate, and are associated with locomotor improvements after
traumatic spinal cord injury in NOD-scid mice. Second, we report
that selective ablation of engrafted hCNS-SCns by diphtheria toxin
(DT) results in loss of locomotor recovery. Third, we show that
mutants, remyelinate axons in traumatically injured NOD-scid
spinal cord, and differentiate into neurons exhibiting electron
microscopic criteria consistent with synapse formation with mouse
Further experimental details are given in Supporting Text, which is
published as supporting information on the PNAS web site.
NOD-scid mice received a laminectomy at vertebral level T9. One
cohort received a 50-kilodyne (kd) (1 dyne ? 10 ?N) (n ? 38)
contusion spinal cord injury, and the second cohort received a
60-kd (n ? 30) contusion spinal cord injury using an Infinite
Horizon Impactor (Precision Systems and Instrumentation, Lex-
ington, KY). Seven days after spinal cord injury, mice in the 50-kd
cohort were tested by using the Basso, Beattie, and Bresnahan
or vehicle (n ? 19). Five hCNS-SCns-transplanted animals and 4
vehicle-injected animals were excluded by using prehoc criteria,
resulting in 11 hCNS-SCns-grafted animals and 15 vehicle control
animals. In the 60-kd cohort, mice were randomized to receive
hCNS-SCns, human fibroblasts (hfibroblasts), or vehicle as above.
Before analysis, 2 hCNS-SCns-transplanted animals and 2 hfibro-
blast-injected animals were excluded by using prehoc criteria,
resulting in 8 hCNS-SCns, 8 hfibroblast, and 10 vehicle controls.
Human Cell Culture, Maintenance, and Injections. Long-term hCNS-
SCns cultures from fetal brain are described in ref. 15. Neuro-
spheres were concentrated to a density of 75,000 cells per ?l in
50% X-vivo medium. For cohort 2, hfibroblasts from fetal liver
were grown to confluence in Iscove’s modified Dulbecco’s medi-
um?10% FBS, dissociated with trypsin, washed, and concentrated
to 75,000 cells per ?l. Nine days after spinal cord injury, mice
anterior aspect of T10 and the posterior aspect of T8. Each site
received 250 nl of cells or vehicle. Functional recovery was assessed
by using a modified (18 point) BBB locomotor rating scale weekly
for 1 month and then biweekly (16, 17) by observers blind to
Freely available online through the PNAS open access option.
Abbreviations: SCI, spinal cord-injured; DT, diphtheria toxin; hCNS-SCns, human CNS stem
kd, kilodyne; BBB, Basso, Beattie, and Bresnahan; hfibroblast, human fibroblast.
†To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
§N.U. and S.J.T. are employees of StemCells, Inc., which stands to profit from this research.
© 2005 by The National Academy of Sciences of the USA
September 27, 2005 ?
vol. 102 ?
no. 39 ?
treatment. At 16 weeks, a subset of animals was also videotaped on
a horizontal ladder beam task in a series of three trials and scored
blind for step errors. After BBB assessment at 16 weeks, mice
received two i.p. injections of DT (50 ?g?kg) 24 h apart, were
reassessed on the BBB 1 week later, and were then euthanized
according to the Institutional Animal Care and Use Committee
guidelines at University of California, Irvine.
Perfusion and Tissue Collection. Mice were anesthetized and tran-
scardially perfused with 30 ml of PBS followed by 100 ml of 4%
paraformaldehyde. Spinal cords were dissected, and the segments
corresponding to T2–T5, T6–T12, and T13–L3 were blocked.
Blocks were postfixed overnight in 4% paraformaldehyde. For
immunocytochemistry, some blocks were equilibrated in 30%
sucrose?PBS for 12 h, embedded in OCT compound, and frozen at
?65°C in isopentane for sectioning. For EM, 1-mm-thick coronal
sections were postfixed in 4% glutaraldehyde immediately after
in ref. 18. SC101, a monoclonal antibody for human nuclei; SC121,
a monoclonal antibody for human cytoplasm; and SC123, a mono-
clonal antibody specific for human GFAP, were produced by
StemCells, Inc. Other primary antibodies included: ?-tubulin III
(Covance, Berkeley, CA); glutamate decarboxylase (GAD)-67,
NeuN, NG2, and Neurofilament-150 (Chemicon); pan-GFAP
(DAKO); Vimentin (Cell Marque, Hot Springs, AR); CC-1?APC
(Oncogene); and myelin basic protein (MBP) 1-188 (Biogenesis,
Bournemouth, U.K.). Secondary antibodies in double-labeling ex-
periments were Alexa Fluor 488 and 555 from Molecular Probes.
system using LASERSHARP 2000 software with lambda strobing to
reduce bleed-through associated with simultaneous scanning.
EM. Tissue was either cut at 50 ?m on a Vibratome and immuno-
stained with SC121 and diaminobenzidine or washed in 0.1 M
sodium cacodylate buffer overnight, fixed with osmic acid, and
embedded in Spurr resin.
NOD-scid?shi Crosses. The MBPshimutation was backcrossed for
four generations onto the NOD-scid background. NOD-scid?shi
mice received either intracerebellar grafting of hCNS-SCns at
postnatal day 1–2 or a contusion injury to T9 at 6 weeks of age
combined with hCNS-SCns or vehicle; survival was for 4 weeks.
hCNS-SCns Survive and Engraft Within the Injured Spinal Cord of
NOD-scid Mice. Mice receiving hCNS-SCns grafts were euthanized
at 24 h, 48 h, 4 weeks, and 17 weeks posttransplantation. Immu-
nocytochemistry for either a human nuclear antigen (SC101) or a
human cytoplasmic antigen (SC121) revealed extensive human cell
survival and engraftment within the injured mouse spinal cord. At
24 and 48 h posttransplantation, numerous human cells had mi-
grated out from the injection site. By 17 weeks, many human cells
had migrated away from the lesion epicenter, such that a clear
injection site was no longer visible. SC121-immunopositive cells
were found in both gray and white matter, and human cells were
morphologically distinct depending on whether they were found
within white matter or gray matter (Fig. 1). All hCNS-SCns- or
hfibroblast-transplanted mice exhibited successful engraftment
upon euthanasia. In sagittal sections, progeny of hCNS-SCns were
found ?1 cm from the lesion epicenter. Migration was greater
rostral to the injection site, and migration into the site of injury was
rare; a rim of spared tissue in the contused cord, even at the
epicenter, was frequently observed (Fig. 6, which is published as
supporting information on the PNAS web site). Myelination or
sprouting?regeneration of spared axons in this region to form
bridge circuits has been hypothesized to promote recovery of
function after spinal cord injury.
hCNS-SCns Promote Locomotor Recovery. At 16 weeks postengraft-
ment, BBB scoring suggested a recovery of coordinated forelimb–
hindlimb locomotor function in 50-kd spinal cord-injured (SCI)
controls (n ? 15) (Fig. 2 A and C). ?2analysis of coordination
recovery revealed a higher observed frequency of coordination
(BBB score ? 12) in hCNS-SCns vs. vehicle controls (P ? 0.05; ?2
? 3.94). Repeated-measures ANOVA yielded a significant main
effect for hCNS-SCns vs. vehicle control BBB performance, F (1,
24) ? 2.28, P ? 0.01, with a Cohen’s effect size d ? ? 0.73,
indicating the effect was ‘‘moderately large’’ (Fig. 2A). Individual
t tests showed significant differences between hCNS-SCns and
vehicle control groups beginning at 6 weeks posttransplantation
(asterisks in Fig. 2A; see also bold values in Table 1, which is
published as supporting information on the PNAS web site).
Additionally, a subset of mice tested on a linear quantitative
horizontal ladder beam task also showed that hCNS-SCns-grafted
mice (n ? 9) exhibited significantly fewer errors than vehicle
controls (n ? 12), averaging 4.2 (SE ? 1.2) vs. 13.5 (SE ? 4.1)
errors, respectively (Fig. 2C; Student’s t test, P ? 0.05).
To address the potential for any transplanted cell population to
produce locomotor recovery in this paradigm and the reproduc-
a hfibroblast control and increased the injury severity to 60 kd (Fig.
2 B and D) in cohort 2. At 16 weeks postengraftment, BBB scoring
suggested a recovery of stepping ability in 60-kd SCI hCNS-SCns-
transplanted mice (n ? 8) in comparison with vehicle (n ? 10) or
hfibroblast (n ? 8) controls. ?2analysis revealed a higher observed
frequency of stepping (BBB score ? 11) in hCNS-SCns vs. vehicle
(P ? 0.02; ?2? 5.45) or hfibroblast (P ? 0.05; ?2? 4.00) controls.
hfibroblast and vehicle control groups were not significantly dif-
ferent from each other (P ? 0.80; ?2? 0.06). Repeated-measures
ANOVA did not yield a significant main effect for the groups;
however, individual t tests showed significant differences between
hCNS-SCns and either vehicle or hfibroblast-transplanted controls
beginning at 4 weeks postgrafting (asterisks in Fig. 2B and bold
values in Table 1). Further, hCNS-SCns-grafted mice (n ? 8)
exhibited significantly fewer errors on the quantitative and more
averaging 9.9 (SE ? 1.2) vs. 17.4 (SE ? 3.6) errors (Fig. 2D;
in gray vs. white matter. Shown are human immunopositive cells (SC121,
elongated, oligodendrocyte-like morphology (arrowheads). (B) In gray mat-
ter, some human cells exhibit neuronal morphologies (arrowheads); lesion
perimeter is above and left of the dashed line. Insets in A and B show higher
magnification of the morphologies adopted by human cells within white and
gray matter, respectively. (Scale bars: 50 ?m; Insets, 10 ?m.)
The progeny of hCNS-SCns migrate and are morphologically distinct
www.pnas.org?cgi?doi?10.1073?pnas.0507063102Cummings et al.
from vehicle controls, averaging 12.5 (SE ? 2.1) errors (Student’s
t test, P ? 0.87), but approached significance in comparison with
hCNS-SCns-grafted mice (Student’s t test, P ? 0.17). To our
knowledge, this is the longest time that mice receiving stem cell
grafts of any type have been tracked behaviorally.
To investigate the effect of selective ablation of engrafted
hCNS-SCns on locomotor recovery, we treated animals with DT
after behavioral recovery stabilized and reassessed locomotor
performance on the BBB and ladder beam 1 week later. Murine
cells are 100,000 times less sensitive to DT than are human cells
(19), and DT has been used extensively as a tool for targeted cell
ablation in transgenic rodent models. These models either condi-
tionally drive DT A-chain expression in specific subpopulations of
cells or express the human DT receptor under a promoter that is
specific to a selective cell population. DT-targeted cells die by
neighboring tissues (20–22). Grafted mice that received vehicle
retained an extensive number of hCNS-SCns (Fig. 2E) or hfibro-
blasts (Fig. 2F), whereas grafted mice that received DT had few or
no remaining human cells (Fig. 2 G and H; see also Fig. 7, which
is published as supporting information on the PNAS web site).
Vehicle-treated hCNS-SCns-transplanted mice were unaffected or
improved slightly, whereas DT-treated hCNS-SCns-transplanted
mice showed significant decrements on both the BBB and ladder
beam (Fig. 2 A–D and Table 1). Comparable results were observed
in both the 50-kd (Fig. 2 A and C) and 60-kd SCI cohorts (Fig. 2 B
and D). No significant differences in behavioral performance were
as a cellular control in the 60-kd SCI groups (Fig. 2 B and D),
suggesting that a general toxic effect of cell ablation on host
locomotor function is unlikely.
hCNS-SCns Do Not Contribute to the Glial Scar. Tissue from mice 4
and 17 weeks posttransplantation was examined for differentiation
markers for astrocyte, neuron, and oligodendrocyte lineages. Sur-
prisingly, few GFAP-immunopositive astrocytes colocalized with
either human specific nuclear (SC101) or cytoplasmic (SC121)
markers (Fig. 3A), and only 2.9% (SD ? 4.1) of SC121-
immunopositive cells exhibited double labeling for GFAP in con-
focal images 17 weeks posttransplantation. Many animals, in fact,
contained no detectable human cells colocalized with GFAP.
Further, using an antibody specific to human-GFAP (SC123), we
rarely detected human cells that had differentiated into SC123-
positive astrocytes (Fig. 8D, which is published as supporting
information on the PNAS web site). We did not observe that
hCNS-SCns formed a border surrounding the lesion area or ap-
peared to contribute to a glial scar. These data suggest that few
into the immunodeficient NOD-scid mouse.
hCNS-SCns Differentiate into Neurons and Form Synapses. On aver-
age, 26.4% (SD ? 22.0) of SC121-immunopositive human cells
exhibited double labeling for ?-tubulin III. Infrequent glutamate
decarboxylase (GAD)-67-immunopositive neurons were colocal-
ized with human nuclei (SC101) (Fig. 3B). We also detected
NeuN-positive nuclei colocalized with human cytoplasm (Fig. 3C).
GAD-67- and NeuN-positive cells were observed only in gray
(B) in individual t tests of recovery times. Asterisks indicate t test (P ? 0.05). DT was administered after 16-week BBB scores were obtained. (C and D) Errors on
a ladder beam task are significantly decreased in 50-kd SCI hCNS-SCns-grafted mice compared with vehicle controls (P ? 0.05) (C) and in 60-kd SCI
mice were reversed 1 week after treatment with DT in A–D. (E–H) hCNS-SCns (E) or human hfibroblasts (F) 17 weeks posttransplantation after treatment with
vehicle in comparison with DT (G and H), which ablated the majority of human cells. No evidence of toxicity to surrounding host tissues was apparent. See Fig.
7 for additional photos.
Surviving hCNS-SCns promote locomotor recovery. (A and B) BBB locomotor performance is significantly improved in 50-kd SCI hCNS-SCns-grafted mice
Cummings et al.
September 27, 2005 ?
vol. 102 ?
no. 39 ?
differentiation along a neural lineage (see also Fig. 9, which is
published as supporting information on the PNAS web site). To
verify terminal differentiation of hCNS-SCns to neurons, we used
immuno-EM in tissue labeled for SC121. SC121-positive neuronal
structures were detected at the EM level, and evidence of newly
differentiated human neurons sending processes into the mouse
parenchyma (Fig. 4 A and B) and forming putative synapses with
all of the ultrastructural criteria for synapses between grafted cells
and host, including membrane thickening and apposition, pre- and
postsynaptic densities, a synaptic cleft, and synaptic vesicles. The
close association of human cells exhibiting neuronal morphologies
with mouse axons is supported by immunocytochemical evidence
that hCNS-SCns are capable of terminal neuronal differentiation.
hCNS-SCns Differentiate into Oligodendrocytes and Myelinate. At 4
weeks posttransplantation, we detected numerous hCNS-SCns
colocalized with anti-NG2, an early marker for oligodendroglial
predominantly in white matter. At 17 weeks, both NG2- and
CC-1?APC-positive cells were colocalized with human nuclei (Fig.
3 D and E), suggesting differentiation of engrafted cells along an
oligodendroglial lineage. Of the SC121-immunopositive human
cells, 64.1% (SD ? 23.1) were double-labeled for CC-1?APC,
suggesting robust terminal differentiation to oligodendrocytes.
Using Hoechst counterstaining, we never observed human immu-
nopositive cells with two nuclei, suggesting that cell fusion, if it
occurred at all, was a very rare event.
Because of the number of hCNS-SCns observed to express
oligodendroglial lineage markers, we used three approaches to test
the hypothesis that hCNS-SCns could myelinate axons in vivo. We
first crossed NOD-scid with shiverer (shi) mice, which have a partial
deletion in the gene for MBP. hCNS-SCns were transplanted into
rons, and oligodendrocytes at 17 weeks. (A) Few human cy-
toplasm-positive (SC121, red), GFAP-positive (green) astro-
cytes (arrowheads) could be detected. Most human cells were
negative for GFAP (arrows). (B) Glutamate decarboxylase
sionally colocalized with human nuclei (SC101, green), indi-
cating GABAergic neurons. (C) Human cytoplasm-positive cell
(red) colocalized (arrowheads) with the neural marker NeuN
(green). (D and E) Several APC-positive oligodendrocytes
(red), one (arrow) colocalized with a human nuclei (SC101,
green). Other APC-positive cells (arrowhead) were not colocalized. (F) Injection of hCNS-SCns into neonatal NOD-scid?shi mice demonstrated MBP (green,
arrowheads) wrapping neurofilament-positive mouse axons (red) in the cerebellum. Other axons remained dysmyelinated (arrows). Blue nuclei are Hoechst-
stained. (Scale bars: A, D, and E, 20 ?m; B, C, and F, 10 ?m.)
Differentiation of hCNS-SCns into astrocytes, neu-
into putative neurons. Immuno-EM for the
human cytoplasm marker SC121 reveals
many engrafted cells within the spinal neu-
ropil. (A) Nucleus of a human cytoplasm-
labeled cell adjacent to the nucleus of an
endogenous mouse cell (mN). Human posi-
tive perikaryal cytoplasm surrounds the nu-
cleus (hN), which exhibits patchy chromatin
aggregates (Cr) typical of neurons. A single
apical process (arrowheads), together with
the paucity of cytoplasm, suggests that this
cytoplasm-labeled process (Hp). The upper
axon (Ax1) is myelinated; the lower axon
(Ax2) is not. Ax2exhibits a bouton-like swell-
ing containing synaptic vesicles (Sv) in appo-
sition with structure Hp; the presence of mi-
tochondria (Mt) and smooth endoplasmic
reticulum (SR) within Hp suggests that this
process may be a dendrite. (C) Human cyto-
plasm-labeled process (Hp) cradled by a
mouse axon terminal (At). Synaptic vesicles
(Sv) and a single mitochondrion (M) are visi-
ble in the mouse axon terminal in close ap-
position with structure Hp; there are no in-
terposing membranes between Hp and At,
but no cleft is visible. (D) Human cytoplasm-
metric synapse with a mouse axon terminal.
There is thickening of both the presynaptic
(black arrowheads) and postsynaptic (white
arrowheads) membranes, with widening of
membrane apposition and greater electron
density within the cleft. Synaptic vesicles (Sv)
Inset shows the ‘‘synapse region’’ without
EM of differentiation of hCNS-SCns
www.pnas.org?cgi?doi?10.1073?pnas.0507063102 Cummings et al.
the cerebella of NOD-scid?shi mice at postnatal day 1. At this early
stage of development, myelination is just beginning; therefore,
hCNS-SCns would encounter the fewest obstacles to integrating
with host axons. In ungrafted NOD-scid?shi mice, an antibody
against the deleted portion of MBP did not detect mouse myelin,
as expected. In contrast, in the cerebella of NOD-scid?shi mice
transplanted with hCNS-SCns, MBP staining was detected sur-
rounding many neurofilament-positive axons, suggesting that en-
grafted human cells differentiated into mature oligodendrocytes
capable of expressing normal MBP and integrating with mouse
axons (Fig. 3F).
We next transplanted hCNS-SCns into contusion-injured spinal
heterozygous uninjured NOD-scid?shi mice have normal myelina-
tion and frequently exhibit ?25 wraps of myelin per axon (Fig. 5A).
Ungrafted, contusion-injured NOD-scid?shi mice never exhibited
more than nine lamellae per axon (Fig. 5B), consistent with
previous characterizations of homozygous shi mice (23, 24). In
contrast, 6-week-old, contusion-injured NOD-scid?shi mice trans-
planted with hCNS-SCns (n ? 5) exhibited axons ensheathed in
D), suggesting terminal oligodendrocyte differentiation by human
Finally, to demonstrate more conclusively that human cells are
before injury, we examined contusion-injured NOD-scid mice by
using preembedding immuno-EM. Ungrafted NOD-scid mice did
not exhibit staining for human cells at the light or EM level. In
examined at the EM level exhibited human cytoplasm-positive
staining (SC121) within oligodendrocytes closely associated with
axons (Fig. 5E). Additionally, we found human-labeled oligoden-
drocytic tongue processes extending to ensheath unlabeled mouse
axons (Fig. 5F), strongly suggesting active remyelination of mouse
axons by engrafted hCNS-SCns.
Role of hCNS-SCns in Spinal Cord Injury Repair. Hypothetically,
cell-based therapeutics for spinal cord injury could affect histolog-
ical and?or functional outcome in a number of ways: (i) differen-
tiation and functional integration of new cells into spared spinal
decreasing host glial scaring, and (iii) increasing host-mediated
regeneration or remyelination. The possibility that hCNS-SCns
affect host locomotor recovery by means of multiple mechanisms
must also be considered.
hCNS-SCns-engrafted, SCI NOD-scid mice suggests that hCNS-
SCns survival in the host plays a role in the maintenance of
to have caused extraneous damage to host cells?tissue, affecting
behavioral performance nonspecifically. The extensive character-
ization of DT cell ablation as apoptotic and nondamaging to
surrounding tissues in previous studies (20–22), lack of observed
damage to the host neuropil, and lack of a DT treatment effect on
behavioral performance in hfibroblast-grafted animals suggests
that nonspecific toxicity is unlikely. Additionally, the probable
source of such a nonspecific effect is activation of the host immune
response, which would be mitigated in NOD-scid mice. Acute
host cell function and lead to decrements in behavioral perfor-
mance, but this possibility is mitigated by the long-term behavioral
plateau before DT treatment in this study. Further studies to
identify additional or alternative mechanisms of recovery are
needed, and the post-DT period should be lengthened by using
younger NOD-scid animals. However, taken together, these data
suggest that hCNS-SCns differentiation to myelinating oligoden-
drocytes and neurons with EM criteria for host synaptic connec-
hCNS-SCns-Mediated Remyelination. We report that transplanted
hCNS-SCns can remyelinate axons, both in naı ¨ve NOD-scid?shi
CNS and in injured NOD-scid?shi and NOD-scid spinal cord. To
our knowledge, these data are the first to demonstrate remyelina-
stem cells to myelinate in atraumatic models of CNS demyelination
and disease. Transplanted adult rat neurospheres in an experimen-
tal autoimmune encephalomyelitis model are associated with re-
myelination and functional recovery (25). Similarly, rat CNS pre-
cursors identified as Schwann cells and some oligodendrocytes
spinal cord. Six-week-old NOD-scid?shi mice
injury and examined at 10 weeks of age. (A)
Normal myelination of axons (Ax) within the
dorsal funiculus of heterozygous shiverer lit-
scid?shi reveals hypomyelination (Inset shows
eight lamellae), loops of myelin, and a lack of
NOD-scid?shi mouse demonstrating thicker,
dense myelination. (D) Boxed area in C show-
ing compact myelin, ?20 lamellae, and pres-
ence of the major dense line. (E) In spinal cord
of contusion-injured NOD-scid mice, im-
muno-EM for human cytoplasm (SC121) re-
veals grafted human stem cells 17 weeks after
spinal cord injury. Two human oligodendro-
cytes (Oligo) appear associated with neighbor-
ing axons. (F) Cross section of a human immu-
nopositive oligodendrocytic tongue process
(Op) and residual immunopositive cytoplasm
within the outermost wrap of myelin (arrow-
heads). (Scale bars: A–C and F, 200 nm; D, 100
nm; E, 1 ?m.)
Cummings et al.
September 27, 2005 ?
vol. 102 ?
no. 39 ?
remyelinated rat spinal cord axons in focal chemical demyelination
models (26). Additionally, predifferentiation of cultured mouse or
human ES cells to generate ‘‘oligospheres’’ before transplantation
has been reported to yield engrafted cells expressing mature
oligodendrocyte differentiation markers in association with in vivo
remyelination (12, 27, 28). Finally, direct or i.v. injection of bone
marrow-derived stem cells has been reported to result in remyeli-
nation in focal demyelination models (29, 30); however, the origin
of the remyelinating cells is unclear, and the stimulation of a host
remyelination response has not been ruled out.
hCNS-SCns-Mediated Human–Mouse Synapse Formation. We present
that both naı ¨ve and neuronal differentiated rodent ES cells can
these cells are functional based on electrophysiological recording
(31, 32). Additionally, mouse and rat CNS neural precursors can
form synapses with same-species host cells after transplantation,
based on electrophysiology (33) or EM analysis (34). Similarly, it
has been reported recently that human neural stem cells form
synapses at the EM level after transplantation into a gerbil brain
ischemia model (35). At present, whether sufficient human–mouse
synaptic connections are established by hCNS-SCns in our para-
digm to account for the locomotor recovery observed is not clear,
and electrophysiological evidence for functional circuit integration
graft survival is an issue that affects all neurotransplantation
research, particularly xenograft studies. Although immunosuppres-
sants are routinely used to prevent clinical allograft rejection, these
drugs may fail to provide long-term protection from xenograft
rejection (36). To minimize xenotransplantation barriers for long-
term engraftment after spinal cord injury, we chose the NOD-scid
mouse (37). Stem cells, including neurospheres, can respond to
chemokine signaling (38, 39); consequently, both constitutive im-
munodeficiency and pharmacological immunosuppression could
affect the engraftment, migration, and differentiation of trans-
planted cells. Additionally, the lack of a fully functional immune
system in NOD-scid mice could affect the evolution of spinal cord
injury; however, the gross histology of the lesion in Nod-scid mice
does not differ dramatically from other mouse strains, and NOD-
scid mice exhibit clear and sustained locomotor impairments after
spinal cord injury. Further, the NOD-scid mouse model avoids the
confound of cellular toxicity associated with pharmacological
These results suggest that hCNS-SCns are capable of surviving and
differentiating in a traumatically injured environment without
contributing to glial scarring. Selective ablation of these cells
suggests that hCNS-SCns play a role in locomotor recovery. Re-
cently, these hCNS-SCns have also been shown to engraft in a
model of ischemia?reperfusion?stroke (38). Together, these find-
ings suggest that hCNS-SCns could have potential benefits for
multiple CNS diseases and injuries. However, this study represents
an initial step toward defining potential clinical applications; addi-
tional animal studies are necessary both to establish the mecha-
nism(s) of recovery and to evaluate the potential of these cells for
possible therapeutic use.
We thank Edwin Apilado, Monika Dohse, Dongping He, Ilse Kraxberger,
Hong-Li Lui, Rebecca Nishi, Richard Silva, Yumin Tang, Tanya Thamk-
ruphat, and Robert Tushinski for technical assistance. This work was
supported by National Institutes of Health Grants R43NS046975 and
R01NS049885 and the Christopher Reeve Foundation.
1. Taupin, P. & Gage, F. H. (2002) J. Neurosci. Res. 69, 745–749.
2. Cao, Q., Benton, R. L. & Whittemore, S. R. (2002) J. Neurosci. Res. 68, 501–510.
3. Cao, Q. L., Zhang, Y. P., Howard, R. M., Walters, W. M., Tsoulfas, P. &
Whittemore, S. R. (2001) Exp. Neurol. 167, 48–58.
4. Wu, S., Suzuki, Y., Kitada, M., Kitaura, M., Kataoka, K., Takahashi, J., Ide, C.
& Nishimura, Y. (2001) Neurosci. Lett. 312, 173–176.
5. Vroemen, M., Aigner, L., Winkler, J. & Weidner, N. (2003) Eur. J. Neurosci. 18,
6. Cao, Q. L., Howard, R. M., Dennison, J. B. & Whittemore, S. R. (2002) Exp.
Neurol. 177, 349–359.
7. Wictorin, K. & Bjorklund, A. (1992) NeuroReport 3, 1045–1048.
8. Stepanov, G. A., Karpenko, D. O., Aleksandrova, M. A., Podgornyi, O. V.,
Poltavtseva, R. A., Pevishchin, A. V., Marey, M. V. & Sukhikh, G. T. (2003) Bull.
Exp. Biol. Med. 135, 397–400.
9. Akesson, E., Holmberg, L., Jonhagen, M. E., Kjaeldgaard, A., Falci, S., Sund-
strom, E. & Seiger, A. (2001) Exp. Neurol. 170, 305–316.
10. Saporta, S., Kim, J. J., Willing, A. E., Fu, E. S., Davis, C. D. & Sanberg, P. R.
(2003) J. Hematother. Stem Cell Res. 12, 271–278.
11. Saporta, S., Makoui, A. S., Willing, A. E., Daadi, M., Cahill, D. W. & Sanberg,
P. R. (2002) J. Neurosurg. 97, 63–68.
12. Keirstead, H. S., Nistor, G., Bernal, G., Totoiu, M., Cloutier, F., Sharp, K. &
Steward, O. (2005) J. Neurosci. 25, 4694–4705.
13. Iwanami, A., Kaneko, S., Nakamura, M., Kanemura, Y., Mori, H., Kobayashi, S.,
Yamasaki, M., Momoshima, S., Ishii, H., Ando, K., et al. (2005) J. Neurosci. Res.
14. Tamaki, S., Eckert, K., He, D., Sutton, R., Doshe, M., Jain, G., Tushinski, R.,
Reitsma, M., Harris, B., Tsukamoto, A., et al. (2002) J. Neurosci. Res. 69, 976–986.
15. Uchida, N., Buck, D. W., He, D., Reitsma, M. J., Masek, M., Phan, T. V.,
Tsukamoto, A. S., Gage, F. H. & Weissman, I. L. (2000) Proc. Natl. Acad. Sci.
USA 97, 14720–14725.
16. Basso, D. M., Beattie, M. S. & Bresnahan, J. C. (1995) J. Neurotrauma 12, 1–21.
17. Dergham, P., Ellezam, B., Essagian, C., Avedissian, H., Lubell, W. D. &
McKerracher, L. (2002) J. Neurosci. 22, 6570–6577.
18. Anderson, A. J., Cummings, B. J. & Cotman, C. W. (1994) Exp. Neurol. 125,
19. Pappenheimer, A. M., Jr., Harper, A. A., Moynihan, M. & Brockes, J. P. (1982)
J. Infect. Dis. 145, 94–102.
20. Saito, M., Iwawaki, T., Taya, C., Yonekawa, H., Noda, M., Inui, Y., Mekada, E.,
Kimata, Y., Tsuru, A. & Kohno, K. (2001) Nat. Biotechnol. 19, 746–750.
21. Drago, J., Padungchaichot, P., Wong, J. Y., Lawrence, A. J., McManus, J. F.,
Sumarsono, S. H., Natoli, A. L., Lakso, M., Wreford, N., Westphal, H., et al.
(1998) J. Neurosci. 18, 9845–9857.
22. Cailhier, J. F., Partolina, M., Vuthoori, S., Wu, S., Ko, K., Watson, S., Savill, J.,
Hughes, J. & Lang, R. A. (2005) J. Immunol. 174, 2336–2342.
Lett. 12, 107–112.
24. Inoue, Y., Inoue, K., Terashima, T., Mikoshiba, K. & Tsukada, Y. (1983) Anat.
Embryol. 168, 159–171.
25. Pluchino, S., Quattrini, A., Brambilla, E., Gritti, A., Salani, G., Dina, G., Galli,
R., Del Carro, U., Amadio, S., Bergami, A., et al. (2003) Nature 422, 688–694.
26. Keirstead, H. S., Ben-Hur, T., Rogister, B., O’Leary, M. T., Dubois-Dalcq, M. &
Blakemore, W. F. (1999) J. Neurosci. 19, 7529–7536.
27. Liu, S., Qu, Y., Stewart, T. J., Howard, M. J., Chakrabortty, S., Holekamp, T. F.
& McDonald, J. W. (2000) Proc. Natl. Acad. Sci. USA 97, 6126–6131.
28. Nistor, G. I., Totoiu, M. O., Haque, N., Carpenter, M. K. & Keirstead, H. S.
(2004) Glia 49, 385–396.
29. Akiyama, Y., Radtke, C. & Kocsis, J. D. (2002) J. Neurosci. 22, 6623–6630.
30. Sasaki, M., Honmou, O., Akiyama, Y., Uede, T., Hashi, K. & Kocsis, J. D. (2001)
Glia 35, 26–34.
31. Harkany, T., Andang, M., Kingma, H. J., Gorcs, T. J., Holmgren, C. D., Zilberter,
Y. & Ernfors, P. (2004) J. Neurochem. 88, 1229–1239.
32. Wernig, M., Benninger, F., Schmandt, T., Rade, M., Tucker, K. L., Bussow, H.,
Beck, H. & Brustle, O. (2004) J. Neurosci. 24, 5258–5268.
33. Auerbach, J. M., Eiden, M. V. & McKay, R. D. (2000) Eur. J. Neurosci. 12,
34. Snyder, E. Y., Yoon, C., Flax, J. D. & Macklis, J. D. (1997) Proc. Natl. Acad. Sci.
USA 94, 11663–11668.
35. Ishibashi, S., Sakaguchi, M., Kuroiwa, T., Yamasaki, M., Kanemura, Y., Shizuko,
36. Larsson, L. C., Frielingsdorf, H., Mirza, B., Hansson, S. J., Anderson, P., Czech,
K. A., Strandberg, M. & Widner, H. (2001) Exp. Neurol. 172, 100–114.
37. Greiner, D. L., Hesselton, R. A. & Shultz, L. D. (1998) Stem Cells 16, 166–177.
38. Kelly, S., Bliss, T. M., Shah, A. K., Sun, G. H., Ma, M., Foo, W. C., Masel, J.,
39. Ji, J. F., He, B. P., Dheen, S. T. & Tay, S. S. (2004) Neurosci. Lett. 355, 236–240.
www.pnas.org?cgi?doi?10.1073?pnas.0507063102Cummings et al.