The Journal of Neuroscience, June 1, 1999, 19(11):4370–4387
Transplants of Fibroblasts Genetically Modified to Express BDNF
Promote Regeneration of Adult Rat Rubrospinal Axons and
Recovery of Forelimb Function
Yi Liu,1 Duckhyun Kim,1 B. Timothy Himes,1,2 Stella Y. Chow,1 Timothy Schallert,3 Marion Murray,1
Alan Tessler,1,2 and Itzhak Fischer1
Department of Neurobiology and Anatomy, Medical College of Pennsylvania/Hahnemann University, Philadelphia,
Pennsylvania 19129, 2Philadelphia Veterans Administration Hospital, Philadelphia, Pennsylvania 19104, and 3Department
of Psychology and Institute for Neuroscience, University of Texas at Austin, Austin, Texas 78712
1
Adult mammalian CNS neurons do not normally regenerate
their severed axons. This failure has been attributed to scar
tissue and inhibitory molecules at the injury site that block the
regenerating axons, a lack of trophic support for the axotomized neurons, and intrinsic neuronal changes that follow axotomy, including cell atrophy and death. We studied whether
transplants of fibroblasts genetically engineered to produce
brain-derived neurotrophic factor (BDNF) would promote rubrospinal tract (RST) regeneration in adult rats. Primary fibroblasts
were modified by retroviral-mediated transfer of a DNA construct encoding the human BDNF gene, an internal ribosomal
entry site, and a fusion gene of lacZ and neomycin resistance
genes. The modified fibroblasts produce biologically active
BDNF in vitro. These cells were grafted into a partial cervical
hemisection cavity that completely interrupted one RST. One
and two months after lesion and transplantation, RST regeneration was demonstrated with retrograde and anterograde trac-
ing techniques. Retrograde tracing with fluorogold showed that
;7% of RST neurons regenerated axons at least three to four
segments caudal to the transplants. Anterograde tracing with
biotinylated dextran amine revealed that the RST axons regenerated through and around the transplants, grew for long distances within white matter caudal to the transplant, and terminated in spinal cord gray matter regions that are the normal
targets of RST axons. Transplants of unmodified primary fibroblasts or Gelfoam alone did not elicit regeneration. Behavioral
tests demonstrated that recipients of BDNF-producing fibroblasts showed significant recovery of forelimb usage, which
was abolished by a second lesion that transected the regenerated axons.
Most of the functional deficits after spinal cord injury result from
the interruption of descending and ascending axons and the lack
of successful regeneration. The failure of axons to regenerate is
now generally attributed to the nonpermissive environment of the
adult mammalian CNS, the lack of trophic/tropic support for
axotomized neurons, and changes intrinsic to the neurons after
axotomy (Tetzlaff et al., 1994; Schwab and Bartholdi, 1996;
Joosten, 1997; Tessler et al., 1997; Stichel and Muller, 1998).
The strategies that have been used to promote regeneration of
injured mammalian CNS axons are designed, in general, to provide a growth-permissive environment or to enhance the regenerative effort of axotomized CNS neurons (for review, see
Schwab and Bartholdi, 1996; Stichel and Muller, 1998). E xamples
of the first approach are to graft peripheral nerves (David and
Aguayo, 1981; Richardson et al., 1982), fetal CNS tissue
(Bernstein-Goral and Bregman, 1993; Himes et al., 1994; Iwashita
et al., 1994; Miya et al., 1997; Mori et al., 1997; Diener and
Bregman, 1998), or non-neuronal cells (Xu et al., 1995a,b; Chen
et al., 1996; Honmou et al., 1996; Li et al., 1997), or to neutralize
CNS inhibitory molecules (C aroni and Schwab, 1988; Schnell and
Schwab, 1990, 1993; Bregman et al., 1995; Z’Graggen et al., 1998)
(for review, see Schwab and Bartholdi, 1996). In the second
category, examples are application of neurotrophic factors (Diener and Bregman, 1994; Tetzlaff et al., 1994; Oudega and Hagg,
1996; Kobayashi et al., 1997; Shibayama et al., 1998) or overexpression of growth-associated genes (e.g., GAP-43, c-Jun, and
Bcl-2). None of these strategies alone has been sufficient in the
adult CNS, but in combination they have elicited regeneration
from several descending pathways (Schnell et al., 1994; Xu et al.,
1995a; Cheng et al., 1996; Bregman et al., 1997; Kobayashi et al.,
1997; Ye and Houle, 1997).
Ex vivo gene therapy is an especially promising approach
because the CNS environment can be modified, and neurotrophic
factors can be delivered by one manipulation. In this strategy,
cultured cells are genetically modified to express therapeutic gene
products, such as neurotrophins, and then grafted into a CNS
lesion site to deliver the therapeutic products and to reestablish
tissue continuity (Gage et al., 1987; Whittemore and Snyder,
1996; Snyder and Senut, 1997). Genetically engineered fibroblasts
have been shown to promote axon regeneration in brain (Rosenberg et al., 1988; Kawaja et al., 1992), and intraspinal grafts of
Received Dec. 17, 1998; revised March 15, 1999; accepted March 22, 1999.
This work was supported by National Institutes of Health Grant NS24707 and
Training Grants NS10090 and HD07467, the Eastern Paralyzed Veterans of America, the International Spinal Cord Research Trust, a Center of Excellence Grant
from Medical College of Pennsylvania /Hahnemann University, and the Research
Service of the Veterans Administration. We thank Dr. L. Lillien for the generous gift
of the LIG retrovirus vectors, Dr. L. Reichardt for the human BDNF cDNA, and Dr.
J. Solowska-Baird for constructing LIG/ BDNF. We thank Maryla Obrocka, Theresa
Connors, and Kathy Bozek for their technical help.
Correspondence should be addressed to Dr. Itzhak Fischer, Department of
Neurobiology and Anatomy, Medical College of Pennsylvania /Hahnemann University, 3200 Henry Avenue, Philadelphia, PA 19129.
Copyright © 1999 Society for Neuroscience 0270-6474/99/194370-18$05.00/0
Key words: spinal cord injury; cell transplantation; retrovirus;
axon regeneration; anterograde tracing; retrograde tracing;
neurotrophin; recovery of function
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
J. Neurosci., June 1, 1999, 19(11):4370–4387 4371
Figure 1. The LIG/BDNF retrovirus. The virus encodes
the full-length human BDNF cDNA and GEO, which is a
fusion gene of b-gal and neomycin resistance genes. The
entire BDNF-IRES-GEO sequence is driven by the RSV
LTR promoter and is transcribed into a polycistron
mRNA. The EMC V IRES located between BDNF and
GEO allows cap-independent initiation of translation of
the polycistron mRNA.
neurotrophin-3 (NT-3)-expressing fibroblasts have enhanced corticospinal tract (C ST) regeneration (Grill et al., 1997). However,
the regenerating CST axons failed to grow into the graft or host
white matter, and the length of growth was very limited (Grill et
al., 1997). N T-3 and brain-derived neurotrophic factor (BDN F)producing fibroblasts have also been shown to induce oligodendrocyte proliferation and axon remyelination after spinal cord
contusion (McTigue et al., 1998).
In the present study, we have used a gene therapy strategy to
elicit regeneration of the rubrospinal tract (RST). The rubrospinal system is readily identified by retrograde or anterograde
tracers, and rubrospinal neurons are known to express the fulllength Trk-B receptor, which accounts for a regenerative response to the application of the neurotrophins BDNF and N T-4/5
(Xu et al., 1995a; Kobayashi et al., 1997; Ye and Houle, 1997). In
addition, the almost complete (.99%) contralateral trajectory of
RST allows an unambiguous interpretation of the anatomical tracing results (Brown, 1974), in contrast to other descending pathways
in which spared axons and collateral sprouting may complicate the
interpretation (Waldron and Gwyn, 1969; Tracey, 1995).
We tested regeneration in a partial cervical hemisection model
in which the lateral funiculus (containing the entire RST) and
part of the ventral white matter were ablated, whereas the ipsilateral gray matter was partially preserved and the dorsal columns
and CST were left intact (see Fig. 2). This model resembles the
lesion paradigms in which RST regeneration into a peripheral
nerve graft has been observed previously (Richardson et al., 1984;
Kobayashi et al., 1997) but preserves the host gray matter, which
is a potential growth substrate for regenerating axons (Cheng et
al., 1996; Grill et al., 1997). We demonstrate that intraspinal
grafts of primary fibroblasts genetically engineered to express
BDNF promote RST regeneration and functional recovery in
adult rats with high cervical spinal cord injury.
MATERIALS AND METHODS
Preparation of the retrovirus vector, the pack ag ing cell line, and BDNFproducing fibroblasts. A retroviral construct (Fig. 1) encoding the BDNF.IRES.GEO sequence was prepared using the LIG vector (provided by
Dr. L. Lillien, University of Pittsburgh) and a 850 bp fragment of the
human BDN F cDNA containing the coding region (provided by Dr. L.
Reichardt, University of C alifornia at San Francisco). The BDN F fragment was isolated by digestion with NotI–EcoRV restriction enzymes and
subcloned into the NotI–SnaBI sites of LIG as shown in Figure 1. The
resulting vector, which was named LIG/ BDN F, contained the human
BDN F gene, an encephalomyocarditis virus (EMC V) internal ribosomal
entry site (IRES), and a GEO gene that is a f usion gene of the lacZ
(encoding Escherichia coli b-galactosidase) and neomycin resistance
(neo) genes. The entire BDN F.IRES.GEO sequence is driven by the
Rous sarcoma virus (RSV) long terminal repeat (LTR) and is transcribed as a multicistronic mRNA. The IRES directs cap-independent
translation of the mRNA by providing internal binding sites for ribosomes and ensures efficient coexpression of the BDN F and GEO genes
(Jang and Wimmer, 1990; Ghattas et al., 1991; K im et al., 1992; Morgan
et al., 1992). The vector was used to transfect the packaging cell line c2
(Mann et al., 1983; Miller, 1990), and neomycin resistant clones were
selected in the presence of 600 mg /ml G418 (Life Technologies, Grand
Island, N Y). A clone (c2-BDN F) that produced the highest viral titer
was propagated and stored. Frozen aliquots of virus were gradually
thawed on ice, mixed with 8 mg /ml polybrene, and applied onto rapidly
dividing rat primary fibroblasts isolated from abdominal skin. Fresh growth
medium was replenished after 4 hr, and the transduced cells were selected
with G418. Transgene expression by the engineered fibroblasts was monitored by X-gal histological staining for b-galactosidase (Liu et al., 1997a).
Clones containing high percentages (.90%) of X-gal-positive cells were
propagated, stored, and used in transplantation experiments. Several methods, including Western blotting (Liu et al., 1997b), slot blot, immunocytochemistry, and bioassay with embryonic day 8 (E8) chicken DRG explant
(Horie et al., 1991), were used to verify the production of biologically active
BDNF (see below).
Cell culture. Primary fibroblasts (Fb) and BDN F-expressing fibroblasts
(Fb/ BDN F) were cultured as described previously (Liu et al., 1998). For
surgery or stock, the cells were grown on 100 mm uncoated tissue culture
dishes (Becton Dickinson Labware, Franklin Lakes, NJ) and split weekly
at 1:10 ratio into fresh medium. T wenty-four hours before surgery, cells
were labeled with the nuclear dye bisBenzimide (Sigma Aldrich Co.,
Irvine, England) as described (Menei et al., 1998). On the day of surgery,
confluent cultures of cells were washed with HBSS (Life Technologies),
trypsinized, gently triturated, counted, washed, pelleted (900 rpm for 5
min), and resuspended in growth medium at a concentration of 10 5
cells/ml. The cells were maintained on ice during surgery. After each
surgery, some of the remaining cells were stained with Trypan Blue
(Sigma Aldrich), and the rest were replated and stained by X-gal histochemistry to verif y viability and transgene expression. For in vitro histochemical and immunocytochemical staining, Fb and Fb/ BDN F were
seeded into adjacent chambers of eight-chamber LabTek glass slides
(Nalge Nunc, Naperville, IL), cultured for 3– 4 d, and fixed with 4%
paraformaldehyde or 0.5% glutaraldehyde (Electron Microscopy Sciences, Ft. Washington, PA). Unless specified, culture supplies were
purchased from Fisher Scientific (Pittsburgh, PA).
Western blot and slot blot anal ysis. To verif y BDN F expression, immunoblotting with polyclonal anti-human-BDN F antibody (see Table 2) was
performed according to the procedure described previously (Liu et al.,
1997b). Briefly, the day before harvest, cells grown on 24-well plates
(Becton Dickinson) were washed with HBSS, fed with 500 ml serum-free
DM EM, and cultured for another 24 hr. The conditioned medium was
collected and mixed with an equal volume of 23 sample buffer containing
125 mM Tris, 4% SDS, 20% glycerol, and 10% 2-mercaptoethanol, pH
6.8. The cell layer was gently scraped off and homogenized with a
Teflon –glass homogenizer in 5 vol of homogenization buffer (50 mM
Tris, pH 7.5, 2 mM EDTA, 1 mM PMSF, 25 mM leupeptin, 1.0% aprotinin). The homogenate was centrif uged at 15,000 3 g for 10 min at 4°C.
The supernatant was then mixed with an equal amount of 23 sample
buffer. Samples containing media or cell homogenates were loaded onto
adjacent lanes and separated by 15% SDS-PAGE, then transferred onto
nitrocellulose (NC) membranes, and processed for Western blotting. The
NC membranes were blocked with 5% nonfat dry milk in TTBS buffer
(0.1% T ween 20, 150 mM NaC l, 50 mM Tris-HC l, pH 7.6) and incubated
overnight with primary antibodies for BDN F (1:50 dilution). After three
rinses with TTBS buffer, the membranes were incubated with HRPconjugated secondary antibody (Jackson ImmunoResearch Laboratories,
West Grove, PA; diluted 1:4000) for 1 hr. The immunoreactivity was
visualized by chemiluminescence with ECL reagents (Amersham, Arlington Heights, IL). Recombinant human-BDN F (Regeneron Pharmaceutical, Tarrytown, N Y) and cells that had not been genetically modified were used in the analysis as controls for the specificity of the
antibody. For slot blot analysis, recombinant human BDN F and supernatant samples (100 ml) from Fb or Fb/ BDN F were applied onto NC
membrane with a slot blot apparatus, dried overnight, and processed for
immunostaining with the BDN F antibody.
BDNF bioassay. Conditioned media from Fb or Fb/ BDN F were tested
for the production of bioactive BDN F using an E8 chicken DRG explant
bioassay that has been described before (Horie et al., 1991). Briefly,
confluent cultures of cells were split onto 50 mm culture dishes at 1:3
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
4372 J. Neurosci., June 1, 1999, 19(11):4370–4387
Table 1. Experimental groups
1 month (n)
Animal groups
FG
2 months (n)
BDA
E xperimental group I–tracing analysis (total n 5 72)
Hx1FB/BDNF
6
6
Hx1FB
6
6
Hx1GF
6
6
E xperimental group II–behavioral analysis (total n 5
Hx1FB/BDNF
Hx1FB
Hx1GF
Tracing control groups (total n 5 9)
Normal1FG
Normal1BDA
FG
BDA
6
6
6
18)
6
6
5
6
6
6
3
6
FG, Fluorogold retrograde tracing; BDA, biotinalyted dextran amine anterograde
tracing; Hx1FB/ BDNF, hemisection and BDNF-producing fibroblast transplant;
Hx1FB, hemisection and unengineered fibroblast transplant; and Hx1GF, hemisection and Gelfoam transplant.
ratio (Becton Dickinson). The cells were then cultured for 24 hr. On the
second day, after three rinses with HBSS, low serum (0.1% FC S) medium was used to feed the cells. The cells were then cultured for another
24 hr, and conditioned media were collected for bioassay. Fertile eggs
were purchased from SPAFAS (Preston, C T). The eggs were incubated
in a humidified incubator at 37°C for 8 d before use. Eggs containing E8
embryos were opened, and the embryos were removed to a sterile Petri
dish containing prewarmed (37°C) DM EM with 0.1% heat-inactivated
goat serum. Meninges and connective tissue were removed, and lumbar
DRG were gently dissected out and embedded in a 12-well culture plate
containing 600 ml of collagen gel that was prepared by mixing solutions
A (CH3C OOH containing 0.3% rat tail type I collagen; Upstate Biotechnology, Lake Placid, N Y), B (103 DM EM), and C (2.2 gm NaHC O3 ,
4.77 gm H EPES in 100 ml 0.05N NaOH) at a ratio of 4:1:1. Three DRGs
were placed into each well of the 12-well plate, and the gel solution was
allowed to solidif y by incubation at 37°C for 20 min. Conditioned medium (400 ml) from Fb, Fb/ BDN F, or culture medium alone was then
applied onto wells containing the DRG explants. The conditioned media
were used without dilution or diluted at 1:2 and 1:10. Neurite outgrowth
from the DRG was examined at 24 – 48 hr and compared with 15, 45, and
450 ng /ml recombinant human-BDN F.
Immunosuppression with c yclosporin A. C yclosporin A (C sA) injection
solution (Sandoz Pharmaceuticals, East Hanover, NJ) was administered
subcutaneously at a dose of 1 mg /100 gm body weight. The daily C sA
injection started 3–5 d before the transplantation procedures and continued for 2 weeks after operation. After this, oral C sA solution (Sandoz,
East Hanover, NJ) was administered via the drinking water (50 mg /ml)
and continued throughout the survival period.
Animal groups. A total of 105 female Sprague Dawley rats (250 –300
gm; Taconic, Germantown, N Y) were studied. All procedures were
approved by the institutional animal welfare committee and were in
accord with the National Institutes of Health guidelines for the care and
use of laboratory animals. Animals were divided into three groups (Table
1). Rats in experimental group I (n 5 72) (Table 1, Fig. 2) received a
partial hemisection of the right side of the spinal cord and a transplant of
Fb/ BDN F, Fb, or Gelfoam alone (n 5 24, for each transplant group).
The RST of these rats was then traced retrogradely by fluorogold (FG,
n 5 36) or anterogradely by biotinylated dextran amine (BDA, n 5 36).
These animals were examined at 1 or 2 months (n 5 36 for each time
point). Rats in experimental group II (Table 1) received the identical
spinal cord lesion as experimental group I and a transplant of Fb/ BDN F,
Fb, or Gelfoam alone (n 5 6 for each group). They were tested weekly
for 2 months for recovery of control of forelimb movement. They then
received a second lesion at C2 to section the right dorsolateral quadrant
and were tested weekly for another 2 months before they were killed and
anatomical analysis was performed. These animals did not receive FG or
BDA injection, because the tracing procedures introduce additional C NS
lesions that may interfere with the interpretation of behavioral results.
Animals in the tracing control group (Table 1) received FG (n 5 3) or
BDA injections (n 5 6). An additional group of six animals received
Figure 2. Schematic diagram of the experimental paradigm. Animals
received a right C3– 4 partial hemisection that disrupted the axons from
the left RN. A, Drawing of a spinal cord cross section; the lesion and
transplant are represented by the shaded area. Immediately after the
spinal cord lesion, Gelfoam, Fb, or Fb/BDNF cells were grafted into the
lesion cavity. RST regeneration was studied using either BDA anterograde tracing or FG retrograde tracing ( B).
spinal cord lesions and Fb/ BDN F transplants and were killed at 1 week
to evaluate transgene expression.
Surg ical procedures. Rats were anesthetized with an intraperitoneal
injection of acepromazine maleate (0.7 mg / kg; Fermenta Animal Health
Co., Kansas C ity, MO), ketamine (95 mg / kg, Fort Dodge Animal
Health, Fort Dodge, IA), and xylazine (10 mg / kg, Bayer Co., Shawnee
Mission, K S), and underwent laminectomy at the C3– 4 level to expose
one spinal cord segment. After hemostasis was achieved, the spinal cord
midline and the dorsal root entry zone were identified. A microscalpel
was used to open the dura and pia mater and to make a shallow incision
in the right dorsal spinal cord. A fine-tipped glass-pulled microaspiration
device was then used to extend the lesion laterally and ventrally (Fig. 2).
Such a lesion completely disrupted the lateral f uniculus (containing the
RST) and partially lesioned the ipsilateral ventral f uniculus and gray
matter but left the dorsal columns intact (Fig. 2 A). The rostrocaudal
extent of the lesion cavity was ;2–3 mm. A piece of Gelfoam soaked with
Fb/ BDN F, Fb cells, or growth medium alone was implanted into the
cavity, and then another 10 ml of cells suspended in growth medium were
slowly injected onto the Gelfoam with a 10 ml Hamilton syringe attached
to a glass pipette (tip diameter 50 mm). The dura was closed with
interrupted 10-A silk sutures, and the muscle and skin were closed in
layers. Immediately after the completion of the procedure (within 10 min
of the spinal cord lesion), all rats received a bolus intravenous injection
of methylprednisolone (30 mg / kg; Pharmacia and Upjohn Company,
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
J. Neurosci., June 1, 1999, 19(11):4370–4387 4373
Table 2. Primary antibodies
T ype
Dilution
Anti-human BDNF
Anti-b-gal
RT-97
chicken pAb IgY
pAb IgG
mAb IgG1
1:100
1:1000
1:100
Anti-5-HT
Anti-CGRP
pAb IgG
pAb IgG
1:10,000
1:10,000
Anti-DbH
Anti-ChaT
Anti-GFAP
pAb IgG
pAb IgG
pAb IgG
1:5000
1:500
1:500
OX-42
mAb IgG
1:500
ED-1
mAb IgG
1:500
Kalamazoo, M I) through the tail vein. After the surgery, animals were
kept on heating pads, closely observed until awake, and then returned to
their home cages. For re-lesion experiments, animals in experimental
group II (Table 1) were anesthetized and subjected to a second laminectomy at C2, followed by removal of the right dorsolateral quadrant using
the same procedure as described above. For retrograde tracing with FG,
3 d before they were killed, anesthetized animals received another
laminectomy 3– 4 segments caudal to the initial lesion-transplant site; 1 ml
2% FG (Fluorochrome, Englewood, C O) was pressure-injected into each
side of the spinal cord, and animals were killed 3 d later. For BDA
anterograde tracing, 15 d before kill, the animals were anesthetized and
positioned on a stereotaxic apparatus, and a dental drill was used to make
a burr hole according to the coordinates described previously (Mori et
al., 1997). One microliter of 10% BDA (Molecular Probes, Eugene, OR)
was slowly injected over 2–3 min as five 200 nl pulses using a 10 ml
Hamilton syringe. The needle was left in place for another 20 min and
gradually withdrawn over 2–3 min.
T issue preparation. Before they were killed, animals were anesthetized
with an intraperitoneal injection of sodium pentobarbital (100 mg / kg;
Abbott Laboratories, North Chicago, IL) and perf used through the heart
with 200 ml of physiological saline (FG animals) or normal saline (BDA
animals) followed by 500 ml of ice-cold 4% paraformaldehyde in 0.1 M
phosphate buffer, pH 7.4. The entire brain and spinal cord were dissected
out and immersed in 0.1 M phosphate buffer (PB) at 4°C overnight
followed by cryoprotection in 30% sucrose (in 0.1 M PB containing 0.5
mM Thimerosal) for 3–5 d. Spinal cord and brain tissue were serially
blocked, embedded in OC T compound (Fisher Scientific, Pittsburgh,
PA), and kept at 280°C before being cut into 20 mm (spinal cord) or 40
mm (brain) sections on a cryostat and mounted onto gelatin-coated slides.
Histolog y and immunoc ytochemistr y. X-gal histochemical and immunocytochemical staining procedures were described before (Liu et al.,
1997a). Briefly, for immunocytochemistry (ICC) staining of cultured
cells, Fb/ BDN F or Fb cells seeded in adjacent chambers on Lab-Tek
slides were stained with an anti-BDN F antibody and /or an anti-b-gal
antibody to check for transgene expression. The experiments were repeated at least three times. For ICC staining of spinal cord tissue, 20 mm
sections were stained with primary antibodies listed in Table 2. The
reactions were performed either with an ABC kit (Vector Labs, Burlingame, CA) or with fluorescent secondary antibodies. The fluorescent
secondary antibodies, including FI TC -conjugate donkey anti-rabbit
IgG(H1L), Texas Red-conjugate donkey anti-rabbit IgG(H1L), FI TC conjugate goat anti-mouse IgG1IgM, and Texas Red-conjugate goat
anti-mouse IgG (H1L) (diluted 1:100) were purchased from Jackson
ImmunoResearch. The specificity of the immunostaining was verified in
sister cultures or adjacent sections by omitting primary or secondary
antibodies.
Detection of BDA-labeled fibers. BDA-labeled fibers were detected
either by staining with an ABC elite kit (Vector) and visualized with
DAB as a chromagen or with FI TC – avidin (Vector). These procedures
were modified from the protocols described by Brosamle and Schwab
(1997). For the ABC elite reaction, slides were rinsed three times, 30 min
Sources
Promega (Madison, W I)
59 3 39 Inc. (Boulder, C O)
Boehringer Mannheim
(GmbH, Germany)
Eugene Tech. (Ridgefield, NJ)
Peninsula Laboratories (Belmont, CA)
Eugene Tech.
Incstar (Stillwater, M N)
Biomedical Technologies
(Stoughton, M A)
Harlan Bioproducts for Science
(Indianapolis, I N)
Harlan Bioproducts for Science
each, in TBST (50 mM Tris-buffered saline containing 0.5% Triton
X-100, pH 10.0), then incubated overnight with an avidin –biotin –peroxidase complex at room temperature. On the second day, after three 30
min rinses in TBST and a short rinse in 50 mM Tris buffer, the sections
were reacted with Sigma Fast-DAB compounds according to the manufacturer’s instructions (Sigma, St. L ouis, MO), dehydrated, and coverslipped with DPX (Fluka Chemie AG, Buchs, Switzerland). For FI TC –
avidin reaction, sections were rinsed three times for 30 min with TBST,
incubated overnight with FI TC – avidin (1:200 dilution), rinsed three
times for 30 min with TBST, and coverslipped with Vectashield (Vector).
Behavioral testing. All rats were examined for forelimb use during
spontaneous vertical exploration, a test that is highly sensitive to chronic
limb use asymmetries (Schallert and Lindner, 1990; Schallert and Jones,
1993; Jones and Schallert, 1994; McDermott et al., 1995; Kozlowski et al.,
1996; Choi-L undberg et al., 1998). Repeated testing does not influence
the asymmetry score, because weight-shifting movements initiated by the
forelimbs are typically used by the animal in its home cage.
The rats were placed in a clear Plexiglas cylinder (20 cm in diameter and
30 cm high) for 5 min. The cylinder encourages use of the forelimbs for
vertical exploration. A mirror was placed at an angle behind the cylinder so
that the forelimbs could be viewed at all times. The testing session was
videotaped, and forelimb usage was scored blindly at a later date.
The following behaviors were scored: (1) independent use of the left or
right forelimb for contacting the wall of the cylinder during a f ull rear, to
initiate a weight-shifting movement, or to regain center of gravity while
moving laterally in a vertical posture along the wall; and (2) simultaneous
or near-simultaneous use of both the left and right forelimb to contact
the wall of the cylinder during a f ull rear and for lateral movements along
the wall.
Each behavior was expressed in terms of (1) percentage use of the
contralateral (nonimpaired) forelimb relative to the total number of
ipsilateral, contralateral, and simultaneous (both) limb use observations;
(2) percentage use of the ipsilateral (impaired) forelimb relative to the
total number of ipsilateral, contralateral, and simultaneous (both) limb
use observations; and (3) percentage simultaneous (both) limb use relative to the total number of ipsilateral, contralateral, or simultaneous
(both) limb use observations.
During a rear, the first limb to contact the wall with clear weight
support (without the other limb contacting the wall within 0.5 sec) was
scored as an independent wall placement for that limb. After the first
limb contacted the wall, a delayed placement of the other limb on the wall
while the first limb remained anchored on the wall was counted as an
additional movement and scored as simultaneous (both). For example, if
an animal placed its contralateral limb on the wall, followed by delayed
contact with both forelimbs, the animal would receive a score of one
“contralateral” and one “both” for that sequence. If only one forelimb
contacted the wall, all lateral movements thereafter were each scored as
independent movements of that limb until the other forelimb contacted
the wall with weight support, at which point one “both” was scored. If the
rat continued to explore the wall laterally in a rearing posture while
alternating both limbs on the wall, a “both” was recorded, and every
4374 J. Neurosci., June 1, 1999, 19(11):4370–4387
additional combination of two-limb movements (wall stepping) received
a “both” score. Thus, both paws must be removed from the vertical
surface before another movement can be scored. If the animal removed
both forelimbs from the wall during a rear and then immediately resumed
wall exploration, the movements were again scored as independent (left
or right) or simultaneous (both) as described above.
Baseline behavior was measured before surgery, and all the animals
were tested weekly for 13 weeks after surgery. T wo-way ANOVA (treatment 3 preferred limb) was performed to test for differences between
animal groups, and one-way ANOVA was used to test for differences
within a treatment group.
Image anal ysis and statistics. Images were captured using a Photometric
Sensys K AF-1400 CCD camera (Photometric, T ucson, AZ) and a DC 330 CCD color video camera (DAGE-MTI, Michigan C ity, I N) attached
to a Leica DMRBE microscope (Wetzlar, Germany) and processed on a
Macintosh Power PC 8500 with N IH image, I P Lab (Scanalytics, Fairfax,
VA), and Photoshop (Adobe System Inc., San Jose, CA) image analysis
software packages. In experimental group I animals (Table 1), because
the FG-labeled cells were sparse in the injured red nucleus (RN), to
avoid underestimating the number of regenerated RN cells, the FGlabeled neurons were counted in every section throughout the rostrocaudal extent of the RN, spanning 1000 mm from the caudal pole (Kobayashi
et al., 1997; Diener and Bregman, 1998). However, in tracing control
animals and on the intact side of experimental animals, because RN was
packed with brightly labeled neurons, to avoid overestimating, neurons
were counted in every other section, and the number was multiplied by 2
to calculate the total number of neurons in each RN. Only those cells
with identifiable nuclei, nucleoli, and characteristic neuronal morphology
were counted. Adjacent sections were always compared with each other to
avoid repeated counting. Images containing FG-labeled cells were captured
at 1003 magnification, and the cross-sectional area of the neurons was
measured using NIH image software. All statistical analyses were performed using Microsoft Excel software (Microsoft, Redmond, WA).
RESULTS
In vitro transgene expression by the genetically
engineered fibroblasts
The retroviral vector LIG/BDNF (Fig. 1) encodes three genes
that were essential for this experiment. BDNF was the gene of
interest; b-gal/neo was a fusion gene composed of two reporter
genes (b-galactosidase and neomycin resistance). The neo gene
enabled selection of the transfected producer cells and establishment of the packaging cell line c2-BDN F. The b-gal gene allowed
monitoring of in vitro and in vivo transgene expression and identification of the donor cells using a highly specific histological
staining procedure (X-gal histochemistry). This vector design
included the IRES sequence that linked the BDNF gene with the
two reporter genes (Fig. 1). It has been shown that IRES directs
cap-independent translation of the mRNA by providing internal
binding sites for ribosomes (Jang and Wimmer, 1990; Ghattas et
al., 1991; K im et al., 1992; Morgan et al., 1992). The BDNF gene
and the two reporter genes were therefore driven by a single LTR
promoter. This design avoided the problem of independent gene
expression associated with vectors that use multiple promoters to
drive multiple gene expression (Ghattas et al., 1991). Because the
IRES sequence ensures efficient (85–90%) coexpression (Ghattas
et al., 1991; our unpublished data), the reliable coexpression of
the gene of interest (BDN F) and a reporter gene (b-gal) offered
us a convenient way to monitor transgene expression both in vitro
and in vivo. As shown in Figure 3A–E, BDNF antibody specifically stains nearly all of the Fb/ BDNF cells (Fig. 3A) but none of
the unmodified primary fibroblasts (Fig. 3B). The same BDNF
antibody has been used by other investigators to demonstrate in
vitro BDNF expression by genetically engineered Schwann cells
(Menei et al., 1998). When double-stained with BDNF and b-gal
antibodies, virtually all cells in the culture were BDNF positive,
as shown in Figure 3C (see also Fig. 3A); ;80 –90% of cells were
b-gal positive (Fig. 3D), and virtually all b-gal positive cells were
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
Figure 3. Analysis of in vitro transgene expression. Cultured Fb/ BDNF
(A, C–E) or Fb ( B) cells were stained with an anti-BDNF antibody (A, B)
or double-labeled with anti-BDNF ( C) and anti-b-gal antibody ( D). E,
From the same visual field as C and D; cells are labeled with a nuclear dye
(49,6-diamidino-2-phenylindole) to reveal the entire cell population in the
culture. Conditioned media from Fb ( F) or Fb/ BDNF ( G) cells were
analyzed for production of biologically active BDNF using an E8 chicken
DRG explant assay. Recombinant human BDNF was used as a positive
control ( H ). The rate that Fb/ BDNF cells secrete BDNF was calculated
relative to control recombinant BDNF using a slot blot assay ( I ). Recombinant BDNF (rBDNF) was loaded onto a slot blot apparatus at 2000,
400, 80, 16, 3.2, and 0.064 ng (lanes 1– 6, respectively) and compared with
20 ml conditioned media from Fb and Fb/ BDNF (5000 cells in triplicates).
Fb/ BDNF cells secrete BDNF at a rate of 12.8 ng/10 6 cells per 24 hr,
whereas Fb cells do not secrete detectable levels of BDNF. Scale bars,
100 mm.
BDNF positive, but some BDNF positive cells were not b-gal
positive (Fig. 3C,D). On the basis of the in vitro immunocytochemical staining, we concluded that the Fb/ BDNF cells expressed both BDNF and b-gal transgenes and that the IRES
sequence therefore ensured high levels of coexpression of BDNF
and b-gal, with a preference for BDNF. This result justified the
use of X-gal histochemistry to monitor in vivo transgene
expression.
To verify that Fb/ BDNF cells secreted BDNF, we tested
homogenates of Fb/ BDNF or Fb cells and media conditioned by
them using Western blot analysis. Anti-BDNF antibody detected
a protein band in conditioned medium from Fb/ BDNF cells but
not from unmodified cells. This protein had the same apparent
molecular weight on a SDS-PAGE gel as commercially available
recombinant human-BDN F. As expected, anti-BDNF antibody
also detected the same band in homogenates of Fb/ BDNF but
not Fb cells (data not shown). Therefore we conclude that Fb/
BDNF cells express and secrete BDNF.
To test whether BDNF secreted by Fb/ BDNF cells was biologically active, we analyzed the conditioned media from Fb/
BDNF or Fb cells using the standard E8 chicken DRG bioassay.
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
J. Neurosci., June 1, 1999, 19(11):4370–4387 4375
As shown in Figure 3F–H, both conditioned medium from Fb/
BDNF cells (Fig. 3G) and commercially available recombinant
human-BDN F (Fig. 3H ) induced neurite outgrowth, indicating
bioactivity of the BDNF secreted by Fb/ BDNF cells. Conditioned medium from unmodified fibroblasts failed to induce neurite outgrowth (Fig. 3F ).
To measure the levels of BDNF produced by Fb/BDNF cells, we
tested their conditioned media using slot blot analysis relative to
the levels of control recombinant BDNF. We calculated from the
data shown in Figure 3I that Fb/BDNF cells secrete BDNF into
medium at a rate of 12.8 ng/10 6 cells per 24 hr, whereas unmodified
fibroblasts do not secrete any detectable levels of BDNF.
Spinal cord lesion, transplant survival, and in vivo
transgene expression
Spinal cord tissue from the lesion only and lesion plus transplant
groups was examined for lesion extent, the transplant survival,
host–graft apposition, and scar tissue formation. In animals that
received a Gelfoam implant, 1 month after surgery Gelfoam was
reabsorbed, leaving a CSF-filled cyst with collapsed dura (Fig.
4 A). Both types of cell transplants completely filled the lesion
cavity in the host spinal cord (Figs. 4 B,C, 5). There was no
obvious morphological difference between the two types of grafts.
Both consisted of densely packed cells with the morphological
characteristics of fibroblasts (Fig. 4 E). The donor cells were
supported by host blood vessels that are present throughout the
transplants (Fig. 4 D). The grafts were almost always completely
apposed to the host tissue, showing excellent tissue apposition
without interruption by cysts (Figs. 4 B –D, 5) or scar tissue at the
graft–host interface (Figs. 4 D, 5). The lack of scar tissue was also
evident in animals receiving only Gelfoam transplants (Fig. 4 A).
We attributed the good graft survival and absence of scar formation to the efficient CsA immunosuppression protocols and to the
inhibition of inflammation and perhaps prevention of secondary
cell death by methylprednisolone (Taoka and Okajima, 1998).
We used X-gal histochemical staining to monitor survival of
the donor cells and in vivo transgene expression. Figure 5 presents
serial sections from a typical Fb/ BDNF transplant, showing that
the donor cells formed a homogeneous cell column that stained
robustly for the presence of b-galactosidase. Two months after
grafting, many cells in the transplant remain X-gal positive, but
the staining was much less intense (data not shown). The presence of donor cells in the graft area, at longer survival times, was
also verified by their bright bisBenzimide nuclear staining (data
not shown).
Host response to the cell transplants
We used various immunocytochemical markers (Table 2) to analyze the host response to the Fb or Fb/ BDNF transplants. In
animals receiving Fb/ BDNF transplants, numerous host axons,
stained with a monoclonal neurofilament antibody RT-97, were
present throughout the grafts and at the graft–host interface (Fig.
6 B). In contrast, these fibers penetrated Fb transplants only
sparsely and superficially (Fig. 6 A). To analyze the source of
these penetrating axons, we stained the graft with antibodies for
CGRP (Fig. 6C,D), serotonin (Fig. 6 E–G), dopamine-bhydroxylase (DbH), and choline acetyltransferase (ChAT). We
also stained for the presence of BDA anterogradely labeled RST
axons in the graft (described in the next section). CGRP- (Fig.
6C,D), serotonin- (Fig. 6 E–G), and BDA-labeled (see Figs. 9, 10)
fibers were present in the Fb/ BDNF transplants, indicating axon
ingrowth from dorsal root, raphe nuclei, and RN. Most serotonin
Figure 4. Photomicrographs of cervical spinal cord sections showing
lesion and transplant in different animal groups. Animals received a right
C3– 4 partial hemisection and a transplant of Gelfoam ( A), Fb ( B), or
Fb/ BDNF ( C). The animals were killed 1 month after surgery. Spinal
cord tissue was cut into cross sections and stained with cresyl violet. D, E,
High-power images of C showing the host–graft interface ( D) and the
characteristic morphology of fibroblasts in the transplants ( E). Scale bars:
A–C, 500 mm; D, E, 100 mm.
input is known to originate from raphe nuclei, and the projection
is bilateral; therefore, the extent to which serotoninimmunoreactive profiles represented regenerating axons or
sprouting from unlesioned local axons cannot be distinguished.
ChAT-or DbH-positive axons were present in the host tissue
near, but not in, either type of transplant (data not shown).
To characterize the host immune response, we performed
additional immunocytochemical studies, using anti-GFAP antibody to identif y astrocytes (Fig. 7A,B), OX-42 antibody to identify microglia and macrophages (Fig. 7C,D), and ED-1 antibody
4376 J. Neurosci., June 1, 1999, 19(11):4370–4387
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
Figure 5. Photomicrographs of cervical spinal cord sections showing transgene expression in Fb/ BDNF transplants. Animals received a right C3– 4
partial hemisection and a transplant of Fb/ BDNF cells. Serial sections of spinal cord tissue were stained with X-gal histochemistry and lightly
counterstained with cresyl violet. In B, the sections (spaced by 500 mm) were serially reconstructed to show the extent of the lesion and the transplant.
A, C, High-power images from B (arrowheads). A, Host–graft interface and numerous blood vessels in the transplant. C, Host–graft integration and the
intense X-gal staining, suggesting high levels of transgene expression. One week survival. Scale bars: A, C, 200 mm; B, 500 mm.
to identif y activated microglia and macrophages (Fig. 7E,F ).
These studies demonstrated astrocytes, macrophages, and microglia on the border of the transplants but not within (Fig. 7). Glial
scar formation around the grafts was modest (Fig. 7A,B). The
immune response at the graft–host interface and within the grafts
was similar in the three groups of animals with Gelfoam, Fb, or
Fb/ BDNF transplants, probably because they received the same
immunosuppression treatment.
RST anterograde tracing with BDA
To study whether cell transplants induced regeneration from
axotomized RN neurons, we determined the distribution of RST
axons after injection of the anterograde tracer BDA into the
magnocellular portion of the lesioned RN. This tracer was chosen
because of the high resolution, which enables identification of
axons (Brosamle and Schwab, 1997; Z’Graggen et al., 1998).
Figure 8 shows the location of RST axons in normal animals. In
the cervical region, the RST axons occupy a wedge-shaped area in
the superficial dorsolateral white matter. The medial border of
the tract is separated from the dorsal gray matter by a narrow
band of ;100 mm, which includes axons of the spinocervical tract
(Fig. 8) (Brown, 1974). Ventrally the RST reaches approximately
to the level of the base of the dorsal horn, although a few scattered
axons are positioned lateral to the ventral horn (Fig. 8). Axonal
branches arise perpendicular to parent RST axons and enter lam-
inae V–VII of the gray matter. These results are consistent with
previous studies (Waldron and Gwyn, 1969; Brown, 1974; Tracey,
1995). Both DAB and FITC labeled the BDA-traced axons efficiently, but DAB offered greater resolution and revealed smallcaliber axons that were not visible by FITC stain (data not shown).
However, FITC was superior to DAB when observing spinal cord
cross sections because it more readily distinguished axons from red
blood cells and allowed the host and graft to be distinguished on
the basis of a fluorescent image without a counterstain.
To visualize the regenerated RST axons in the lesiontransplant region, cross, sagittal, and horizontal sections were
stained for the presence of BDA-labeled fibers. As shown in
Figures 9 and 10, the spinal cord lesion completely disrupted the
right RST. Numerous BDA-labeled RST axons had regenerated
into the transplant (Fig. 9A–C). Many axons cut in cross section
were present along the host–graft interface (Fig. 9D), and axon
branches (sectioned longitudinally) entered the gray matter (Fig.
9E). Double-labeling with FITC –avidin and GFAP immunofluoresence demonstrated that axons at the host–graft interface
intermingled with processes of activated astrocytes (Fig. 9F–H ).
FITC staining underestimated the number of RST axons that
regenerated into the transplants because most of them were
smaller-caliber axons and not well stained by FITC (Fig. 10 I,J ).
Sagittal and horizontal spinal cord sections (Fig. 10) confirmed
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
J. Neurosci., June 1, 1999, 19(11):4370–4387 4377
Figure 6. Photomicrographs of cervical spinal cord sections showing host axon growth into Fb/ BDNF or Fb transplants. Spinal cord cross sections from
animals receiving Fb ( A) or Fb/ BDNF ( B) transplants were stained with the RT-97 antibody and show cell grafts ( g), the host gray matter ( h), and the
host–graft interface (dashed lines). Numerous axons are present within the Fb/ BDNF transplant and at the host–graft interface ( B), whereas host axons
penetrate Fb transplant sparsely and superficially ( A). One month survival. Scale bars, 100 mm. In C and D, a spinal cord cross section from a Fb/BDNF
recipient was stained with an anti-CGRP antibody. A dorsal root had regenerated into the transplant and elongated toward the dorsal horn, which was
partially disrupted by the transplantation procedures, as intended. Arrowheads outline the graft. D, Higher-power view of axons that had reached the
dorsal horn. T wo month survival. Scale bars: C, 200 mm; D, 100 mm. In E–G, a spinal cord cross section from an Fb/ BDNF recipient was stained with
an anti-serotonin antibody. Numerous serotonin-immunoreactive fibers are present throughout the transplant ( E). At higher power (F, G), the axons
show the characteristic “beads on a string” morphology. In E arrowheads outline the graft–host interface. One month survival. Scale bars, 100 mm.
4378 J. Neurosci., June 1, 1999, 19(11):4370–4387
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
Figure 7. Photomicrographs of cervical spinal cord sections showing host immune response. Spinal cord cross sections from animals receiving
Fb/ BDNF transplants were immunostained with GFAP (A, B), OX-42 (C, D), and ED-1 (E, F ) antibodies. A mild astrocytic activation is visible along
the graft–host interface, but few if any astrocytes are in the graft (A, B). Macrophages and activated microglia accumulate at the graft–host interface,
but few are in the transplants ( C–F). One month survival. Scale bars, 100 mm.
that the spinal cord lesion disrupted the entire RST and that all of
the axotomized axons stopped at the host–graft border in recipients of unmodified fibroblasts (Fig. 10 A). Some of the axons,
however, had grown across the interface of Fb/BDNF transplants
(Fig. 10 B,D,H ), elongated caudally (Fig. 10 B,E,H,I,J ), and exited
the graft at the caudal graft–host interface (Fig. 10 F–H ). Although
many larger-caliber axons stopped at the host–graft border, numerous small-caliber axons continued directly into the transplant (Fig.
10 I,J ).
Serial sections of spinal cord caudal to the grafts were studied
to determine the caudal extent of regenerated BDA-labeled axons and their path of regeneration. Numerous transversely sectioned, BDA-labeled axons had regenerated to the upper thoracic
spinal cord, which was four to five segments caudal to the transplant, but their number was small compared with normal. Regenerating axons were diffusely located and not as well organized as
the normal RST axon. Most of them were localized to the white
matter, but a few were present in the gray matter. In general the
location of the regenerated RST axons deviated only slightly from
normal (Fig. 11, compare A, B), although in some cases BDAlabeled axons were present diffusely throughout the ipsilateral
lateral funiculus (data not shown). Axonal branches arising perpendicular to the main stem frequently projected toward laminae
V–VII of the gray matter (Fig. 11 B–D) and bore small terminal
bouton-like varicosities (Fig. 11C,D), similar to those observed
previously on corticofugal axons labeled with BDA (Z’Graggen
et al., 1998). Many of these branches grow toward interneurons in
the gray matter, and the terminal bouton-like structures were
close to these neurons (Fig. 11C). The maximum length of RST
regeneration and the number and distribution of regenerating
axons differed among animals. Similar findings were reported by
other investigators studying CST regeneration (Schnell and
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
J. Neurosci., June 1, 1999, 19(11):4370–4387 4379
difference between the 1 and 2 month groups. Three of the 2
month-survival animals were eliminated from the study because
in one the lesion was incomplete, one lacked a transplant, and in
the third the tracing technique failed. In animals receiving Fb or
Gelfoam transplants, few if any BDA-labeled axons were present
either in the transplants or at the host–graft interface. No BDApositive axons were detected caudal to the transplant (data not
shown). In animals receiving Fb/ BDNF transplants, RST axons
rostral to the transplants also showed a considerable amount of
sprouting (Fig. 10 B,C), whereas sprouting was minimal in control
animals (Fig. 10 A).
RST retrograde tracing with FG
Figure 8. Photomicrographs of cervical spinal cord sections showing the
distribution of the RST in normal animals after BDA anterograde tracing.
BDA was injected into the maganocellular portion of the left RN. C ervical spinal cord sections were stained with DAB and demonstrate the
discrete location of RST in the superficial dorsolateral quadrant ( A). B,
Higher-power view of the morphology and organization of RST axons. In
B arrows point to axon branches in the gray matter. Scale bars, 100 mm.
Schwab, 1990; Bregman et al., 1995). We do not know the reasons
for these differences, but we speculate that they may be attributable to differences in interactions between the grafts and hosts.
Among the 12 animals that received Fb/ BDNF grafts and BDA
tracing (Table 1, E xperimental group I), four of the six 1 monthsurvival animals showed BDA-labeled axons caudal to the transplant: one animal to low cervical, two animals to upper thoracic,
and one animal to mid thoracic levels. The rate of RST axon
regeneration was estimated to be 1–1.5 mm /d. E xamination of the
lesion-transplant site confirmed the completeness of the lesion
and the presence of healthy grafts. Because two animals that
showed no BDA labeling in the caudal spinal cord also had no
labeling in the CNS tissue rostral to the transplant, the apparent
failure of RST regeneration in these animals may therefore have
been attributable instead to the failure of the tracing technique.
Three of the six 2 month-survival animals (Table 1, E xperimental
group I) showed BDA labeling similar to the four 1 monthsurvival animals, with the most caudal levels that contained
BDA-positive fibers being upper to mid thoracic. Therefore the
extent of RST regeneration (the most caudal spinal cord level in
which BDA-labeled axons is detected) showed no consistent
We also studied RST regeneration using the FG retrograde
tracing technique. The FG was injected bilaterally three to four
segments (1–1.5 cm) caudal to the transplant to avoid diffusion of
FG into the transplant. We nevertheless found FG in several
grafts, probably because of diffusion via the CSF, and these
animals were consequently eliminated from the study. Figure
12 A–C showed the typical pattern of RSN labeling after bilateral
injections of FG into the low cervical region of normal rats. The
injections labeled both red nuclei equally. In animals with a
right-sided lesion and Gelfoam transplant, no labeled cells were
present in the left RN, whereas labeling in the right RN resembled that of normal animals (Fig. 12 D–F ). In animals receiving a
lesion and Fb graft, FG labeling was infrequently observed in the
contralateral RN (Fig. 12G–I ). In contrast, in Fb/ BDNF recipients numerous neurons were brightly labeled by FG throughout
the rostrocaudal extent of the axotomized RN (Fig. 12 J–L). The
number of FG-labeled RN neurons was counted in all four groups
of animals (Fig. 13). In normal animals, ;3000 neurons were
labeled in each RN. After cervical axotomy and Gelfoam implant, ;10 cells were labeled in the contralateral RN; these cells
may represent neurons that project ipsilaterally (,1% of total
RSN). With Fb transplant, 30 – 40 cells were retrogradely labeled
in the contralateral RN. Although this number was not significantly different from that of animals receiving only Gelfoam
implants, Fb grafts may have provided a permissive environment
that allowed a very small percentage of axotomized neurons to
regenerate to the caudal spinal cord. Approximately 175–200
neurons were labeled in the contralateral RN in the presence of
Fb/ BDNF grafts. This number represents 7–10% of the total RN
neuron population (Fig. 13). The number of labeled neurons in
animals with 2 month-survival was not significantly different from
that of the 1 month-survival animals for any of the groups (Fig. 13).
Fetal spinal cord transplants secure a partial rescue of RN
neurons injured by a C3– 4 axotomy but fail to prevent their
atrophy (Mori et al., 1997). To analyze whether cell atrophy also
occurred in the regenerated RN neurons, the cross-sectional area
of FG-labeled RN neurons was measured and compared across
experimental and control groups. We found no difference among
them, indicating that the regenerated RN neurons had not atrophied in animals receiving Fb/ BDNF or Fb transplants (data not
shown). The normal soma size of the few FG-labeled neurons in
Gelfoam recipients was not surprising because these neurons
probably project ipsilaterally and were not axotomized.
Behavior analysis
When placed in a cylinder, normal rats spontaneously reared and
explored the wall of the cylinder using a single forepaw alone
(50%) or both forepaws together (50%). We calculated the percentage of wall exploratory behavior that was initiated by right or
4380 J. Neurosci., June 1, 1999, 19(11):4370–4387
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
Figure 9. Photomicrographs of cervical spinal cord sections showing BDA anterograde tracing of RST axons in the lesion–transplant site. The animal
received a right cervical hemisection and an Fb/ BDNF transplant. The left RN was anterogradely traced with BDA 15 d before killing. T wo month
survival after transplantation. A, Section through the transplant 1000 mm from its rostral pole. The section was stained with DAB as chromagen for
BDA-labeled fibers and counterstained with cresyl violet. Numerous BDA-labeled axons are present in the transplant (arrowheads) and at the graft–host
interface (large arrow) and send off branches into the gray matter (small arrows). However, most of the BDA-labeled axons are obscured by the
counterstain. B, Adjacent section stained with FITC to identif y BDA-labeled fibers, no counterstain. C–E, Higher magnification of the corresponding
regions in B. C, Numerous BDA-labeled axons in the transplant. Most are cut transversely. Arrowheads point to smaller-caliber axons, and arrows point
to larger-caliber axons. In D, many transversely sectioned larger-caliber axons are labeled at the graft–host interface (arrows) and send off branches
perpendicular to the main stem toward the gray matter (arrowheads). E, Higher-power view of axon branches that enter the gray matter (arrows). F–H,
Section double-labeled with FITC – avidin ( F) and an anti-GFAP antibody ( G). H, Merged image of F and G. Numerous RST axons are intermingled
with processes of activated astrocytes. Scale bars: A, 200 mm; C, F, 100 mm. Scale bar in A applies to B; scale bar in C applies to D and E; scale bar in
F applies to G and H.
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
J. Neurosci., June 1, 1999, 19(11):4370–4387 4381
Figure 10. Photomicrographs of cervical spinal
cord sections showing regeneration of RST axons
through Fb/ BDNF transplants. Animals received
a right cervical hemisection and Fb ( A) or Fb/
BDNF transplants ( B –J). The left RN was anterogradely traced with BDA 15 d before killing. One
month survival after transplantation. Spinal cord
tissue was cut into sagittal ( A–H) or horizontal (I,
J ) sections. For all sections lef t is rostral, and right
is caudal. A, All BDA-labeled RST axons are
interrupted by the lesion–transplant and failed to
enter an Fb transplant. B, Some RST axons regenerated into an Fb/ BDNF transplant; the dashed
line indicates the rostral graft–host interface. C,
Region rostral to the transplant. Numerous axon
branches are evident, suggesting sprouting induced by the transplant. D, E, Higher magnifications of regions from B. D, BDA-labeled axons
that have penetrated the rostral graft–host interface. E, RST axons deeply within the transplant. F,
Region in the host white matter immediately caudal to the transplant. The dashed line indicates the
caudal graft–host interface. BDA-labeled axons
exit the transplant and elongate caudally. Some
axons bear varicosities resembling terminal boutons. G, Terminal bouton-like structure at higher
power. H, BDA-labeled axons stained by FITC.
Regenerated axons (arrows) pass through the rostral graft–host interface and continue for several
millimeters through the transplant and the caudal
graft–host interface. I, J, Higher-power views of
many smaller-caliber and some larger-caliber RST
axons in a rostrocaudal direction in an Fb/BDNF
transplant. Scale bars, 100 mm. The scale bar in E
applies to C, D, I, and J.
left forepaws alone or both forepaws together (see Fig. 15).
Hemisection at the upper cervical level produced asymmetry in
forelimb use; hemisected rats or rats with nonmodified transplants rarely used the forelimb ipsilateral to the injury and did not
show recovery during the 8-week observation period (Figs. 14C,
15A). Animals with Fb/ BDNF transplants used the injured forelimb more frequently than the hemisection or Fb alone groups,
resulting in more symmetrical limb use (Figs. 14 A, 15A). At 1
4382 J. Neurosci., June 1, 1999, 19(11):4370–4387
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
used the ipsilateral forelimb alone ;10% of the time. In addition,
control (hemisection and Fb transplant) groups held the forepaw
ipsilateral to the lesion in a strongly flexed position (Fig. 14C),
whereas the posture was closer to normal in animals with Fb/
BDNF transplants (Fig. 14 A). These results clearly demonstrate
functional recovery of injured limb usage in the Fb/ BDNF group;
improvement was present by 1 week after transplant and reached
a plateau between week 3 and 4. Control groups did not recover
(Fig. 15A).
To study whether functional recovery was mediated by regenerated RST axons, animals in E xperimental group II (Table 1)
were subjected to a second lesion that removed the right dorsolateral quadrant at C2 just rostral to the transplant. Forelimb
function was then tested for another 5 weeks. The second lesion
almost completely abolished the recovered function in Fb/ BDNF
animals (Figs. 14 B, 15B) but had little effect on the Fb or
Gelfoam recipients (Figs. 14 D, 15B). The second lesion also
caused animals receiving Fb/ BDNF transplants to lose their
nearly normal forepaw posture and to hold the forepaw ipsilateral
to the lesion in a strongly flexed position similar to that of Fb and
Gelfoam animals (Fig. 14 A,B), The second lesion had little effect
on forepaw posture of Fb and Gelfoam recipients (Fig. 14C,D).
DISCUSSION
Figure 11. Photomicrographs of upper-thoracic (A, B) and mid-thoracic
(C, D) spinal cord showing BDA-labeled RST axons. A, Cross section
from a normal animal demonstrating the normal RST location and organization. Arrows point to axon branches in the gray matter. B, Cross
section from an animal with an Fb/ BDNF transplant. One month survival. Numerous BDA-labeled axons are present in the lateral funiculus,
but their location is aberrant and more diffuse than normal. A few
transversely sectioned BDA-labeled axons are also present in the gray
matter (arrows). One axon branch arises perpendicular to the main stem
and enters lamina V II (arrowheads). C, Axon branches in the gray matter
with varicosities resembling terminal boutons. D, Terminal bouton-like
structures at higher power. Scale bars: A–C, 100 mm; D, 25 mm. Scale bar
in A also applies to B.
week after surgery, both Fb and hemisection control animals used
only the unaffected forelimb (contralateral to lesion) in exploring
the wall ( p , 0.05). The Fb/BDNF-treated rats performed much
of the exploration with the good limb, but they also used both
forelimbs together ;10% of the times. This suggests that although the animals did not use the affected limb independently,
they could use it to support the unaffected limb as early as the first
week after surgery. By 2 weeks, the Fb/ BDNF-treated animals
used both forelimbs together as often as they used the unaffected
forelimb alone ( p . 0.05), indicating further recovery. We observed a similar pattern of limb use at 3 weeks after surgery,
except that rats with Fb/ BDNF transplants sometimes used the
limb ipsilateral to the lesion alone. Approximately 5% of the total
movements were made by the ipsilateral forelimb alone in the
Fb/ BDNF group, whereas neither the Fb nor the hemisection
group animals used that forelimb independently. By 4 and 8
weeks after surgery, some of the animals in the Fb/ BDNF group
In the present study we report that intraspinal transplants of
primary fibroblasts genetically modified to express BDNF enhanced regeneration of the RST. When grafted into a cervical
spinal cord lesion, the donor cells survived well and expressed the
transgenes for at least 2 months. The axons of at least 7% of the
severed RN neurons regenerated through and around the transplants, extended for long distances in the white matter caudal to
the transplant, and sent off axon branches to normal RST target
regions. Behavioral testing revealed significant functional recovery in limb usage, which may be partially mediated by the RST
regeneration. In contrast, we found no anatomical or behavioral
evidence for RST regeneration in animals that received Gelfoam
or unmodified fibroblast transplants.
Critical elements that allow RST regeneration
Several features of the grafts and their interaction with the host
are likely to have contributed to RST regeneration. First, the
excellent graft survival and tissue apposition provided the regenerating RST axons with a continuous terrain for growth. A well
integrated substrate has been considered of critical importance
for regeneration (Bregman et al., 1997). Almost all of the transplants entirely filled the lesion cavity, were intimately apposed to
the host tissue, and thus formed an interface with the host that
was not interrupted by cysts or scars. The absence of scar formation may reflect the mild host immune reaction attributable to
immunosuppression by CsA and the neuronal protective effects of
methylprednisolone and BDNF (Novikova et al., 1996). The mild
astrocytic activation and accumulation of macrophages and activated microglia along the host–graft interface may also have
provided a favorable substrate for regenerating RST axons. We
found numerous BDA-labeled RST axons at the graft–host interface and intermingled with processes of activated astrocytes.
Activated CNS immunocompetent cells have been proposed to
act as permissive substrates for regenerating axons (Kawaja and
Gage, 1991; Davies et al., 1997; Prewitt et al., 1997; Rabchevsky
and Streit, 1997).
The second crucial feature was the relatively permissive environment provided by the fibroblast grafts (T uszynski et al., 1994;
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
J. Neurosci., June 1, 1999, 19(11):4370–4387 4383
Figure 12. Photomicrographs of midbrain showing FG retrograde tracing of RN neurons. Neurons in both RNs were retrogradely traced by injection
of FG into both sides of the spinal cord in normal animals ( A–C) or in recipients of Gelfoam ( D–F), Fb ( G–I), or Fb/ BDNF ( J–L) transplants. Survival
after transplantation was 1 month. All sections were taken ;480 mm from the caudal pole of RN. B, E, H, K, Higher-power views of the left RN regions
corresponding to A, D, G, and J. C, F, I, L, Higher-power views of the right RN regions corresponding to A, D, G, and J. In normal animals both RNs
are equally labeled ( A–C). In Gelfoam recipients virtually no RN neurons are labeled by FG on the left (D, E), whereas labeling on the right is normal
(D, F ). In Fb transplant recipients, very few RN neurons are labeled in the left RN (G, H ), whereas the right RN is normally labeled (G, I ). In recipients
of Fb/ BDNF transplants, numerous RN neurons are labeled in the left RN (J, K ), and labeling is normal in the right RN (J, L). Scale bars, 100 mm. The
scale bar in J applies to A, D, G, and J; the scale bar in K applies to B, C, E, F, H, I, K, and L.
Nakahara et al., 1996). We found that 30 – 40 RN neurons were
FG labeled in animals receiving unmodified fibroblast transplants, which was considerably larger although not different statistically than the ;10 labeled cells in Gelfoam recipients. Similar
numbers of RN neurons regenerated into peripheral nerve grafts
placed into a comparable lesion (Richardson et al., 1984; Kobayashi et al., 1997). Unmodified fibroblast grafts may therefore
offer the RST a growth-permissive environment similar to that
provided by a peripheral nerve graft.
The most important feature, however, was the local delivery of
BDNF. Fb/ BDNF transplants were homogeneously and robustly
stained by X-gal histochemistry at 1 week, and many cells remained X-gal positive for at least 2 months, suggesting that the
engineered cell transplants acted as an abundant and sustained
source of BDNF that promoted the survival of RN neurons and
regeneration of RST axons. This suggestion is supported by our
findings that at least 7% of RN neurons regenerated caudal to the
transplant in animals receiving Fb/ BDNF transplants, in contrast
to 1% in Fb recipients, and that these transplants induced profuse
raphe-spinal and dorsal root axon ingrowth, whereas growth into
Fb grafts was sparse. Previous reports have demonstrated the
effectiveness of BDNF in promoting RST regeneration. For ex-
ample, as many as 200 neurons regenerated into a peripheral
nerve graft when BDNF was administered adjacent to RN
perikarya (Kobayashi et al., 1997), and RN neurons failed to
regenerate into Schwann cell grafts unless BDNF was administered or the Schwann cells were engineered to express BDNF (Xu
et al., 1995a; Menei et al., 1998). E xogenous BDNF also dramatically increased the number of chronically injured RN neurons
that regenerated into a peripheral nerve graft (Ye and Houle,
1997). The enhanced regenerative response of RN neurons was
likely mediated through trkB, the high-affinity receptor for
BDNF, which is expressed by these neurons (Kobayashi et al.,
1997).
Mechanisms of RST regeneration in spinal cord
Stimulated by BDNF, the axotomized RN neurons grew directly
through the Fb/ BDNF transplants and caudally for up to 40 –50
mm through host white matter. This pattern of growth differs
from that of axotomized CST neurons, which responded to grafts
of fibroblasts genetically modified to express N T-3 by growing
through host gray matter both around the transplants and caudally for up to 9 mm (Grill et al., 1997). The differing patterns of
growth shown by these two supraspinal systems suggest intrinsic
4384 J. Neurosci., June 1, 1999, 19(11):4370–4387
Figure 13. Bar graph comparing numbers of FG-labeled RN neurons
among animal groups. The FG-labeled RN neurons in normal animals
and animals that had received Fb/ BDNF, Fb, or Gelfoam transplants
were counted and compared by one-way ANOVA, followed by Fisher’s
post hoc test 1 or 2 months after transplantation. Significantly more RN
neurons (;7%) were labeled contralateral to surgery in animals receiving
Fb/ BDNF transplants than in those receiving Fb or Gelfoam (,1%). p ,
0.00001; n 5 3 for normal; n 5 5 for Fb/ BDNF 1m, Hx 1m, and Hx 2m;
n 5 6 for Fb/ BDNF 2m, Fb 1m, and Fb 2m.
differences in the regenerative capacities of these neurons and is
consistent with other evidence showing that RN neurons regenerate more robustly than CST neurons in several different environments (Richardson et al., 1984; Tetzlaff et al., 1994; Ye and
Houle, 1997; Dyer et al., 1998). Observations made in other
systems also demonstrate intrinsic differences in neuronal regenerative capacity. After injury of the larval lamprey spinal cord, for
example, only some populations of reticulospinal neurons regenerate (Davis and McC lellan, 1994; Jacobs et al., 1997). Similarly,
the neurons in the reticular nucleus of the adult rat thalamus, but
not the neurons in other thalamic nuclei, regenerate into peripheral nerve grafts (Morrow et al., 1993), and only the CGRPcontaining neurons of the adult rat DRG appear to regenerate
into fetal spinal cord transplants (Itoh et al., 1996). It seems likely
that the more vigorous regenerative response of RN neurons
allowed them to grow directly through the grafts and host white
matter, whereas CST neurons were able to grow only through the
more permissive substrate provided by gray matter.
The limited CST growth observed caudal to genetically modified fibroblast grafts is within the possible range of diffusion of
N T-3 and may result from a tropic effect (Grill et al., 1997).
Although BDNF may also have exerted a tropic influence, the
greater length of regeneration through white matter that we
observed for RST axons requires additional explanations. We
found that the regenerated axons occupied the general location of
the RST in normal white matter and terminated in regions of
cervical gray matter that are the normal targets of these axons.
This observation is consistent with the notion that the regenerating axons responded to cues similar to those that operate during
development. The regenerating axons appear to have been able to
grow in response to BDNF and these or other cues despite the
well known inhibitory influence of CNS myelin. However, the
numerous in vivo studies of myelin-associated neurite growth
inhibitors have primarily focused on CST axons (Savio and
Schwab, 1990; Schnell and Schwab, 1990, 1993; Schnell et al.,
1994; Bregman et al., 1995; Z’Graggen et al., 1998), and whether
they exert similar effects on RST axons has not been examined. It
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
Figure 14. Photograph comparing forelimb use. Animals were analyzed
in a cylinder test to study preferred forelimb use. A, C, Seven weeks after
transplantation, Fb/ BDNF recipients used their forelimb ipsilateral to the
lesion to explore the environment ( A), whereas Fb and Gelfoam recipients rarely did so ( C). The forepaw posture in animals with Fb/ BDNF
transplants was nearly normal ( A), but Fb or Gelfoam recipients kept the
forepaw ipsilateral to the lesion strongly flexed ( C). B, D, Seven weeks
after the relesion at C2, animals with Fb/ BDNF transplants lost the
normal forelimb posture and the ability to use the forelimb ipsilateral to
the lesion ( B). In contrast, the relesion had little effect on forelimb
posture and forelimb use in animals with Fb transplants ( D). A, B,
Photographs of the same animal with an Fb/ BDNF graft taken before and
after the relesion. C, D, From the same animal with an Fb graft before
and after the relesion. Arrows in A–D point to the forelimb ipsilateral to
the lesion.
is possible, therefore, that RST axons are less susceptible to
inhibition by myelin or, as suggested for other developing and
adult axons whose axons grow through adult white matter (Wictorin et al., 1990a,b, 1992; Wictorin and Bjoklund, 1992; Davies et
al., 1993, 1994, 1997; Li and Raisman, 1993; Oudega and Hagg,
1996; Li et a., 1997, 1998), that regenerating RST axons lack or
downregulate the relevant receptors. The RST regeneration that
we observed shares several characteristics with the growth
through white matter reported for these other axons, including
rapid growth rate (up to 1–2 mm /d) (Davies et al., 1994), growth
of at least several centimeters (Wictorin et al., 1992), and the
ability to find targets (Wictorin et al., 1990a, 1992; Davies et al.,
1994, 1997). We speculate that in response to BDNF and the
favorable substrate provided by the transplants, the intrinsically
robust regenerative capacity of RN neurons allowed their axons
to grow past the lesion site and through host white matter in
response to cues that remained after injury and were relatively
unaffected by the inhibitory influence of CNS myelin. Future
experiments will determine whether additional systems of supraspinal axons are similarly responsive to neurotrophic factors
supplied by transplants.
Recovery of function correlates with RST regeneration
Our behavioral results are consistent with the idea that regenerated RST axons contributed to recovery of forelimb motor func-
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
J. Neurosci., June 1, 1999, 19(11):4370–4387 4385
axons. In addition, the animals showed no further functional
recovery during the second month after surgery, and we observed
no consistent difference in RST regeneration between animals
that survived for 1 or 2 months. Additional mechanisms may
contribute to the functional recovery in animals receiving Fb/
BDNF transplants: (1) regeneration and sprouting by several
descending pathways, including raphe-spinal and vestibulospinal
axons that are known to respond to BDNF; (2) recovery of
sensory function mediated by sprouting of dorsal roots; (3) preservation of segmental neuronal circuitry; and (4) facilitation of
local segmental function by BDNF (Jakeman et al., 1998). However, the recovery does not appear to be attributable to the neuroprotective effect of BDNF on motor neurons because we found
no qualitative difference in the cervical motor neuron pools in
animals receiving transplants of Fb/BDNF, Fb, or Gelfoam alone.
Our finding that a second lesion that cut RST axons rostral to
the transplants apparently permanently abolished the recovered
behavior in BDNF-transplant-treated animals suggests that RST
regeneration is a major contributor to the recovery. However, we
cannot rule out contributions from other descending pathways
whose axons may also have regenerated or sprouted.
In summary, we have demonstrated that cellular delivery of
BDNF by genetically engineered fibroblasts promoted longdistance regeneration of rubrospinal axons that at least partially
mediate recovery of forelimb usage after cervical spinal cord
injury.
REFERENCES
Figure 15. Behavioral analysis of forelimb use. Three weeks after transplantation, Fb/ BDNF recipients showed significant recovery of use of the
injured limb alone or together with the uninjured limb, whereas Fb or
Gelfoam recipients showed negligible use of the injured limb alone ( A).
The relesion abolished most of the forelimb use in Fb/ BDNF recipients
but had little effect on the Fb or Gelfoam recipients ( B). n 5 6 for
Fb/ BDNF and Fb, n 5 5 for Hx.
tion. Animals receiving Fb/ BDNF transplants showed significant
improvements in usage of the forelimb ipsilateral to the lesion. In
contrast, those receiving unmodified fibroblasts or Gelfoam alone
relied almost entirely (;90%) on the contralateral limb and
failed to recover independent use of the impaired forelimb (Fig.
15A). Similar deficits have been reported after severe unilateral
damage to the forelimb region of the sensorimotor cortex or basal
ganglia. In these studies, independent use of the impaired forelimb for wall support and weight shifting along vertical surfaces
did not recover substantially over a 6 month period (Kozlowski et
al., 1996; Schallert and Kozlowski, 1998). Moreover, the functional recovery in Fb/BDNF recipients followed a distinct progression during the first few weeks after surgery. At 1 week the
rats did not use the forelimb ipsilateral to the lesion independently in this task, but by 3– 4 weeks they used this limb alone
5–10% of the time and together with the contralateral limb ;50%
of time, similar to control rats. This progression of functional
recovery correlates well with the anatomical recovery that we
observed in the rubrospinal system. The RST axons regenerated
at a rate of 1–1.5 mm /d during the first month and sent branches
with terminal bouton-like structures into the normal RST target
region in cervical gray matter. This correlation suggests that the
functional recovery in forelimb motor control can be attributed at
least partially to the regeneration and reinnervation by RST
Bernstein-Goral H, Bregman BS (1993) Spinal cord transplants support
the regeneration of axotomized neurons after spinal cord lesions at
birth: a quantitative double-labeling study. E xp Neurol 123:118 –132.
Bregman BS, Kunkel-Bagden E, Schnell L, Dai HN, Gao D, Schwab M E
(1995) Recovery from spinal cord injury mediated by antibodies to
neurite growth inhibitors. Nature 378:498 –501.
Bregman, BS, McAtee M, Dai HN, Kuhn PL (1997) Neurotrophic factors increase axonal growth after spinal cord injury and transplantation
in the adult rat. E xp Neurol 148:475– 494.
Brosamle C, Schwab M E (1997) C ells of origin, course, and termination
patterns of the ventral, uncrossed component of the mature rat corticospinal tract. J Comp Neurol 386:293–303.
Brown LT (1974) Rubrospinal projections in the rat. J Comp Neurol
154:169 –187.
C aroni P, Schwab M E (1988) Antibody against myelin-associated inhibitor of neurite growth neutralizes nonpermissive substrate properties of
C NS white matter. Neuron 1:85–96.
Chen A, Xu XM, Kleitman N, Bunge MB (1996) Methylprednisolone
administration improves axonal regeneration into Schwann cell grafts
in transected adult rat thoracic spinal cord. E xp Neurol 138:261–276.
Cheng H, C ao Y, Olson L (1996) Spinal cord repair in adult paraplegic
rats: partial restoration of hind limb function. Science 273:510 –513.
Choi-L undberg DL, Lin Q, Schallert T, Crippens D, Davidson BL, Chang
YN, Chiang YL, Qian J, Bardwaj L, Bohn MC (1998) Behavioral and
cellular protection of rat dopaminergic neurons by an adenoviral vector
encoding glial cell line-derived neurotrophic factor. E xp Neurol
154:261–275.
David S, Aguayo AJ (1981) Axonal elongation into peripheral nervous
system “bridges” after central nervous system injury in adult rats.
Science 214:931–933.
Davies SJ, Field PM, Raisman G (1993) Long fibre growth by axons of
embryonic mouse hippocampal neurons microtransplanted into the
adult rat fimbria. Eur J Neurosci 5:95–106.
Davies SJ, Field PM, Raisman G (1994) Long interfascicular axon
growth from embryonic neurons transplanted into adult myelinated
tracts. J Neurosci 14:1596 –1612.
Davies SJ, Fitch MT, Memberg SP, Hall AK, Raisman G, Silver J (1997)
Regeneration of adult axons in white matter tracts of the central
nervous system. Nature 390:680 – 683.
Davis Jr GR, McClellan AD (1994) Long distance axonal regeneration
of identified lamprey reticulospinal neurons. E xp Neurol 127:94 –105.
4386 J. Neurosci., June 1, 1999, 19(11):4370–4387
Diener PS, Bregman BS (1994) Neurotrophic factors prevent the death
of C NS neurons after spinal cord lesions in newborn rats. NeuroReport
5:1913–1917.
Diener PS, Bregman BS (1998) Fetal spinal cord transplants support
growth of supraspinal and segmental projections after cervical spinal
cord hemisection in the neonatal rat. J Neurosci 18:779 –793.
Dyer JK, Bourque JA, Steeves JD (1998) Regeneration of brainstemspinal axons after lesion and immunological disruption of myelin in
adult rat. E xp Neurol 154:12–22.
Gage F H, Wolff JA, Rosenberg MB, Xu L, Yee JK, Shults C, Friedmann
T (1987) Grafting genetically modified cells to the brain: possibilities
for the future. Neuroscience 23:795– 807.
Ghattas IR, Sanes JR, Majors JE (1991) The encephalomyocarditis virus internal ribosome entry site allows efficient coexpression of two
genes from a recombinant provirus in cultured cells and in embryos.
Mol C ell Biol 11:5848 –5859.
Grill R, Murai K, Blesch A, Gage F H, T uszynski M H (1997) C ellular
delivery of neurotrophin-3 promotes corticospinal axonal growth and
partial functional recovery after spinal cord injury. J Neurosci
17:5560 –5572.
Himes BT, Goldberger M E, Tessler A (1994) Grafts of fetal central
nervous system tissue rescue axotomized C larke’s nucleus neurons in
adult and neonatal operates. J Comp Neurol 339:117–131.
Honmou O, Felts PA, Waxman SG, Kocsis JD (1996) Restoration of
normal conduction properties in demyelinated spinal cord axons in the
adult rat by transplantation of exogenous Schwann cells. J Neurosci
16:3199 –3208.
Horie H, Bando Y, Chi H, Takenaka T (1991) NGF enhances neurite
regeneration from nerve-transected terminals of young adult and aged
mouse dorsal root ganglia in vitro. Neurosci Lett 121:125–128.
Itoh Y, Waldeck RF, Tessler A, Pinter MJ (1996) Regenerated dorsal
root fibers form functional synapses in embryonic spinal cord transplants. J Neurophysiol 76:1236 –1245.
Iwashita Y, Kawaguchi S, Murata M (1994) Restoration of function by
replacement of spinal cord segments in the rat. Nature 367:167–170.
Jacobs AJ, Swain GP, Snedeker JA, Pijak DS, Gladstone LJ, Selzer M E
(1997) Recovery of neurofilament expression selectively in regenerating reticulospinal neurons. J Neurosci 17:5206 –5220.
Jakeman LB, Wei P, Guan Z, Stokes BT (1998) Brain-derived neurotrophic factor stimulates hindlimb stepping and sprouting of cholinergic
fibers after spinal cord injury. E xp Neurol 154:170 –184.
Jang SK , Wimmer E (1990) C ap-independent translation of encephalomyocarditis virus RNA: structural elements of the internal ribosomal
entry site and involvement of a cellular 57-kD RNA-binding protein.
Genes Dev 4:1560 –1572.
Jones, TA, Schallert T (1994) Use-dependent growth of pyramidal neurons after neocortical damage. J Neurosci 14:2140 –2152.
Joosten EA (1997) Corticospinal tract regrowth. Prog Neurobiol
53:1–25.
Kawaja MD, Gage F H (1991) Reactive astrocytes are substrates for the
growth of adult C NS axons in the presence of elevated levels of nerve
growth factor. Neuron 7:1019 –1030.
Kawaja MD, Rosenberg MB, Yoshida K, Gage F H (1992) Somatic gene
transfer of nerve growth factor promotes the survival of axotomized
septal neurons and the regeneration of their axons in adult rats. J Neurosci 12:2849 –2864.
Kim DG, Kang HM, Jang SK , Shin HS (1992) Construction of a bif unctional mRNA in the mouse by using the internal ribosomal entry site of
the encephalomyocarditis virus. Mol C ell Biol 12:3636 –3643.
Kobayashi NR, Fan DP, Giehl KM, Bedard AM, Wiegand SJ, Tetzlaff W
(1997) BDNF and NT-4/5 prevent atrophy of rat rubrospinal neurons
after cervical axotomy, stimulate GAP-43 and Ta1-tubulin mRNA
expression, and promote axonal regeneration. J Neurosci 17:9583–9595.
Kozlowski DA, James DC, Schallert T (1996) Use-dependent exaggeration of neuronal injury following unilateral sensorimotor cortex lesions.
J Neurosci 16:4776 – 4786.
Li Y, Raisman G (1993) Long axon growth from embryonic neurons
transplanted into myelinated tracts of the adult rat spinal cord. Brain
Res 629:115–127.
Li Y, Field PM, Raisman G (1997) Repair of adult rat corticospinal tract
by transplants of olfactory ensheathing cells. Science 277:2000 –2002.
Li Y, Field PM, Raisman G (1998) Regeneration of adult rat corticospinal axons induced by transplanted olfactory ensheathing cells. J Neurosci 18:10514 –10524.
Liu Y, Himes BT, Moul J, Huang W L, Chow SY, Tessler A, Fischer I
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
(1997a) Application of recombinant adenovirus for in vivo gene delivery to spinal cord. Brain Res 768:19 –29.
Liu Y, Xia J, Ma DL, Faber DS, Fischer I (1997b) Tau-like proteins in
the nervous system of goldfish. Neurochem Res 22:1511–1516.
Liu Y, Himes BT, Tryon B, Moul J, Chow SY, Jin H, Murray M, Tessler
A, Fischer I (1998) Intraspinal grafting of fibroblasts genetically modified by recombinant adenoviruses. NeuroReport 9:1075–1079.
Mann R, Mulligan RC, Baltimore D (1983) Construction of a retrovirus
packaging mutant and its use to produce helper-free defective retrovirus. C ell 33:153–159.
McDermott PE, Frydel B, Ullman MD, Schallert T, Emerich DF (1995)
Implantation of encapsulated catecholamine and GDNF-producing
cells in rats with unilateral dopamine depletions and Parkinsonian
symptoms. E xp Neurol 132:62–76.
McTigue DM, Horner PJ, Stokes BT, Gage F H (1998) Neurotrophin-3
and brain-derived neurotrophic factor induce oligodendrocyte proliferation and myelination of regenerating axons in the contused adult rat
spinal cord. J Neurosci 18:5354 –5365.
Menei P, Monteromenei C, Whittemore SR, Bunge RP, Bunge MB
(1998) Schwann cells genetically modified to secrete human BDNF
promote enhanced axonal regrowth across transected adult rat spinal
cord. Eur J Neurosci 10:607– 621.
Miller AD (1990) Retrovirus packaging cells. Hum Gene Ther 1:5–14.
Miya D, Giszter S, Mori F, Adipudi V, Tessler A, Murray M (1997)
Fetal transplants alter the development of function after spinal cord
transection in newborn rats. J Neurosci 17:4856 – 4872.
Morgan RA, Couture L, Elroy-Stein O, Ragheb J, Moss B, Anderson W F
(1992) Retroviral vectors containing putative internal ribosome entry
sites: development of a polycistronic gene transfer system and applications to human gene therapy. Nucleic Acids Res 20:1293–1299.
Mori F, Himes BT, Kowada M, Murray M, Tessler A (1997) Fetal spinal
cord transplants rescue some axotomized rubrospinal neurons from
retrograde cell death in adult rats. E xp Neurol 143:45– 60.
Morrow DR, C ampbell G, Lieberman AR, Anderson PN (1993) Differential regenerative growth of C NS axons into tibial and peroneal nerve
grafts in the thalamus of adult rats. E xp Neurol 120:60 – 69.
Nakahara Y, Gage FH, Tuszynski MH (1996) Grafts of fibroblasts genetically modified to secrete NGF, BDNF, NT-3, or basic FGF elicit
differential responses in the adult spinal cord. Cell Transplant 5:191–204.
Novikova L, Novikov L, Kellerth JO (1996) Brain-derived neurotrophic
factor reduces necrotic zone and supports neuronal survival after spinal
cord hemisection in adult rats. Neurosci Lett 220:203–206.
Oudega M, Hagg T (1996) Nerve growth factor promotes regeneration
of sensory axons into adult rat spinal cord. E xp Neurol 140:218 –229.
Prewitt C M, Niesman IR, Kane C J, Houle JD (1997) Activated macrophage/microglial cells can promote the regeneration of sensory axons
into the injured spinal cord. E xp Neurol 148:433– 443.
Rabchevsky AG, Streit WJ (1997) Grafting of cultured microglial cells
into the lesioned spinal cord of adult rats enhances neurite outgrowth.
J Neurosci Res 47:34 – 48.
Richardson PM, McGuinness UM, Aguayo AJ (1982) Peripheral nerve
autografts to the rat spinal cord: studies with axonal tracing methods.
Brain Res 237:147–162.
Richardson PM, Issa V M, Aguayo AJ (1984) Regeneration of long spinal axons in the rat. J Neurocytol 13:165–182.
Rosenberg MB, Friedmann T, Robertson RC, T uszynski M, Wolff JA,
Breakefield XO, Gage F H (1988) Grafting genetically modified cells
to the damaged brain: restorative effects of NGF expression. Science
242:1575–1578.
Savio T, Schwab M E (1990) Lesioned corticospinal tract axons regenerate in myelin-free rat spinal cord. Proc Natl Acad Sci USA
87:4130 – 4133.
Schallert T, Jones TA (1993) “Exuberant” neuronal growth after brain
damage in adult rats: the essential role of behavioral experience.
J Neural Transplant Plast 4:193–198.
Schallert T, Kozlowski DA (1998) Brain damage and plasticity: userelated enhanced neural growth and overuse-related exaggeration of
injury. In: C erebrovascular disease (Ginsberg MD, Bogousslavsky J,
eds), pp 611– 619. New York: Blackwell Science.
Schallert T, Lindner MD (1990) Rescuing neurons from trans-synaptic
degeneration after brain damage: helpf ul, harmf ul or neutral in recovery of function? C an J Psychol 44:276 –292.
Schnell L, Schwab M E (1990) Axonal regeneration in the rat spinal cord
produced by an antibody against myelin-associated neurite growth
inhibitors. Nature 343:269 –272.
Liu et al. • Intraspinal Grafts of BDNF Fibroblasts Promote Regeneration
Schnell L, Schwab M E (1993) Sprouting and regeneration of lesioned
corticospinal tract fibres in the adult rat spinal cord. Eur J Neurosci
5:1156 –1171.
Schnell L, Schneider R, Kolbeck R, Barde YA, Schwab M E (1994)
Neurotrophin-3 enhances sprouting of corticospinal tract during development and after adult spinal cord lesion. Nature 367:170 –173.
Schwab M E, Bartholdi D (1996) Degeneration and regeneration of axons in the lesioned spinal cord. Physiol Rev 76:319 –370.
Shibayama M, Hattori S, Himes BT, Murray M, Tessler A (1998)
Neurotrophin-3 prevents death of axotomized C larkes nucleus neurons
in adult rat. J Comp Neurol 390:102–111.
Snyder EY, Senut MC (1997) The use of nonneuronal cells for gene
delivery. Neurobiol Dis 4:69 –102.
Stichel CC, Muller HW (1998) E xperimental strategies to promote axonal regeneration after traumatic central nervous system injury. Prog
Neurobiol 56:119 –148.
Taoka Y, Okajima K (1998) Spinal cord injury in the rat. Prog Neurobiol
56:341–358.
Tessler A, Fischer I, Giszter S, Himes BT, Miya D, Mori F, Murray M
(1997) Embryonic spinal cord transplants enhance locomotor performance in spinalized newborn rats. In: Advances in neurology, Vol 72,
neuronal regeneration, reorganization and repair (Seil FJ, ed), pp
291–303. Philadelphia: Lippincott-Raven.
Tetzlaff W, Kobayashi NR, Giehl KM, Tsui BJ, C assar SL, Bedard AM
(1994) Response of rubrospinal and corticospinal neurons to injury
and neurotrophins. Prog Brain Res 103:271–286.
Tracey DJ (1995) Ascending and descending pathways in the spinal
cord. In: The rat nervous system, Ed 2 (Paxinos G, ed.), pp 67– 80. San
Diego: Academic.
T uszynski M H, Peterson DA, Ray J, Baird A, Nakahara Y, Gage F H
(1994) Fibroblasts genetically modified to produce nerve growth factor
induce robust neuritic ingrowth after grafting to the spinal cord. E xp
Neurol 126:1–14.
Waldron HA, G wyn DG (1969) Descending nerve tracts in the spinal
J. Neurosci., June 1, 1999, 19(11):4370–4387 4387
cord of the rat. I. Fibers from the midbrain. J Comp Neurol
137:143–153.
Whittemore SR, Snyder EY (1996) Physiological relevance and functional potential of central nervous system-derived cell lines. Mol Neurobiol 12:13–38.
Wictorin K, Bjorklund A (1992) Axon outgrowth from grafts of human
embryonic spinal cord in the lesioned adult rat spinal cord. NeuroReport 3:1045–1048.
Wictorin K, Brundin P, Gustavii B, Lindvall O, Bjorklund A (1990a)
Reformation of long axon pathways in adult rat central nervous system
by human forebrain neuroblasts. Nature 347:556 –558.
Wictorin K, C larke DJ, Bolam JP, Brundin P, Gustavii B, Lindvall O,
Bjorklund A (1990b) E xtensive efferent projections of intra-striatally
transplanted striatal neurons as revealed by a species-specific neurofilament marker and anterograde axonal tracing. Prog Brain Res
82:391–399.
Wictorin K, Brundin P, Sauer H, Lindvall O, Bjorklund A (1992) Long
distance directed axonal growth from human dopaminergic mesencephalic neuroblasts implanted along the nigrostriatal pathway in
6-hydroxydopamine lesioned adult rats. J Comp Neurol 323:475– 494.
Xu XM, Guenard V, Kleitman N, Aebischer P, Bunge MB (1995a) A
combination of BDNF and NT-3 promotes supraspinal axonal regeneration into Schwann cell grafts in adult rat thoracic spinal cord. E xp
Neurol 134:261–272.
Xu XM, Guenard V, Kleitman N, Bunge MB (1995b) Axonal regeneration into Schwann cell-seeded guidance channels grafted into
transected adult rat spinal cord. J Comp Neurol 351:145–160.
Ye JH, Houle JD (1997) Treatment of the chronically injured spinal cord
with neurotrophic factors can promote axonal regeneration from supraspinal neurons. E xp Neurol 143:70 – 81.
Z’Graggen WJ, Metz GAS, Kartje GL, Thallmair M, Schwab M E (1998)
Functional recovery and enhanced corticof ugal plasticity after unilateral pyramidal tract lesion and blockade of myelin-associated neurite
growth inhibitors in adult rats. J Neurosci 18:4744 – 4757.