Roles of ES Cell-Derived Gliogenic Neural Stem/Progenitor Cells in Functional Recovery after Spinal Cord Injury

Transplantation of neural stem/progenitor cells (NS/PCs) following the sub-acute phase of spinal cord injury (SCI) has been shown to promote functional recovery in rodent models. However, the types of cells most effective for treating SCI have not been clarified. Taking advantage of our recently established neurosphere-based culture system of ES cell-derived NS/PCs, in which primary neurospheres (PNS) and passaged secondary neurospheres (SNS) exhibit neurogenic and gliogenic potentials, respectively, here we examined the distinct effects of transplanting neurogenic and gliogenic NS/PCs on the functional recovery of a mouse model of SCI. ES cell-derived PNS and SNS transplanted 9 days after contusive injury at the Th10 level exhibited neurogenic and gliogenic differentiation tendencies, respectively, similar to those seen in vitro. Interestingly, transplantation of the gliogenic SNS, but not the neurogenic PNS, promoted axonal growth, remyelination, and angiogenesis, and resulted in significant locomotor functional recovery after SCI. These findings suggest that gliogenic NS/PCs are effective for promoting the recovery from SCI, and provide essential insight into the mechanisms through which cellular transplantation leads to functional improvement after SCI.


Introduction
Because the adult central nervous system (CNS) has limited potential for regeneration, spinal cord injury (SCI) results in severe dysfunction, such as paraplegia and tetraplegia. With the aim of regenerating the injured spinal cord, various intraspinal cellular transplants have been investigated, especially in the sub-acute phase after injury. This period, between the acute and chronic phases, is marked by the minimal expression of cytokines, and is likely to be amenable to transplantation therapy [1,2,3,4,5]. Embryonic stem (ES) cells, with their indefinite replication potential, pluripotency, and genetic flexibility, have attracted great interest, and methods for inducing their neural differentiation have been extensively studied [6]. ES cell-derived neural progenitors are currently one of the most promising cell sources for cell transplantation therapy for treating SCI. Although previous studies demonstrated that the transplantation of mouse ES cellderived embryoid bodies [7] or human ES cell-derived oligodendrocyte progenitor cells [8] promotes overall functional recovery after SCI, the types of neural progenitor cells most effective for treating sub-acute phase SCI has been uncertain.
We recently reported that a low concentration of retinoic acid (10 28 M: low-RA) can efficiently induce caudalized neural progenitors in embryoid bodies (EBs) [9], and we established a neurosphere-based culture system of ES cell-derived neural stem/ progenitor cells (NS/PCs) from low-RA-treated EBs, with midbrain to hindbrain identities [10]. These ES cell-derived primary neurospheres (PNS) mainly exhibit neurogenic differentiation potentials, whereas passaged secondary neurospheres (SNS) are more gliogenic, corresponding to changes in CNS development, in which neurogenic NS/PCs predominate early in gestation and gliogenic NS/PCs predominate in mid-to-late gestation. Here, taking advantage of this difference between neurogenic PNS and gliogenic SNS, we transplanted PNS and SNS into the injured spinal cord, examined the differentiation and growth properties of the grafted cells, and compared their effects on angiogenesis, axonal regeneration, and functional recovery after SCI. We also examined the survival and growth of the transplanted ES cell-derived NS/PCs using in vivo, live, bioluminescent imaging (BLI) to evaluate the tumorigenicity and safety of the grafted cells.

Establishment of a Stable ES Cell Line Expressing CBRluc Luminescence and Venus Fluorescence
We first established an ES cell line that constitutively expresses the click beetle red-emitting luciferase (CBRluc) [11] and Venus [12] by introducing a CAG-CBRluc-IRES-Venus plasmid (Fig. 1A) into EB3 ES cells (CCV-ES cells) [13]. CCV-ES cells and their progenies were detected by both BLI [3,14,15] and fluorescence microscopy. To induce NS/PCs from ES cells and obtain PNS and SNS, we used a neurosphere-based culture system that we recently reported [10] (Fig. 1B), as described in Materials and Methods. More than 99% of the undifferentiated CCV-ES cells expressed Venus fluorescence by flow cytometry (Fig. 1D and E), and CCV-ES cell-derived PNS (CCV-PNS) and SNS (CCV-SNS) showed steady fluorescence that was detectable by fluorescence microscopy (Fig. 1C). Approximately 80% of the cells in the CCV-PNS and -SNS were positive for Venus by flow cytometry (Fig. 1D and E). bioluminescence imaging (BLI) revealed CBRluc expression in both CCV-PNS and -SNS, and we confirmed that the photon counts were in direct proportion to the cell numbers in vitro (Fig. 1F). We also confirmed that the CCV-ES cells could generate PNS and SNS similar to EB3-ES cells (Fig. 1C).

Distinct Differentiation Potentials of PNS and SNS In Vitro
We next examined the in vitro differentiation potentials of the PNS and SNS derived from EB3-and CCV-ES cells. PNS and SNS derived from EB3-and CCV-ES cells were allowed to differentiate in medium without FGF2 on poly-L-ornithine/ fibronectin coated coverslips for 5 days, and then processed for immunocytochemistry. We examined the frequency of colonies consisting of bIII tubulin-positive neurons, GFAP-positive astrocytes, and/or O4-positive oligodendrocytes, and found that the EB3-and CCV-PNS colonies predominantly differentiated into neurons, although a small number of colonies contained both neurons and glia (Fig. 1G). In contrast, most of the EB3-and CCV-SNS colonies differentiated into both neurons and glia, including astrocytes and oligodendrocytes, or into only glial cells (Fig. 1G), demonstrating that the ES cell-derived PNS and SNS had distinct differentiation potentials in vitro (Fig. 1H). Moreover, EB3-and CCV-ES cell-derived neurospheres exhibited similar differentiation properties, confirming that the transgene in the ES cells had negligible effects on differentiation (Fig. 1H).
We also examined the SNS formation rates to determine the self-renewing ability of the ES cell-derived PNS. We cultured CCV-PNS at a low cell density (2.5610 4 cells/ml), transferred them into 96-well plates at one neurosphere/well, dissociated the neurospheres, and cultured them again with FGF2 to form secondary neurospheres. Most of the CCV-PNS generated secondary neurospheres (79/90; 87.7%; from more than three independent experiments), confirming their ability to self-renew.

Transplanted SNS Prevented Atrophic Change and Demyelination after SCI
A contusive SCI was induced at the Th10 level of C57BL6 mice, and 5610 5 cells of CCV-PNS or CCV-SNS, or PBS as a control, were injected into the lesion epicenter 9 days after injury. We refer to these, respectively, as the PNS, SNS, and control groups. After 6 weeks, histological analyses were performed. We first examined atrophic changes of the injured spinal cord by Hematoxylin-eosin (H-E) staining ( Fig. 2A and B). The transverse area of the spinal cord at the lesion epicenter was significantly larger in the SNS group than in the control group, suggesting that SNS transplantation prevented atrophy of the injured spinal cord (Fig. 2E). Luxol Fast Blue (LFB) staining revealed significantly greater preservation of the myelinated areas in the SNS group compared with the control (both 2 and 6 weeks after injury) and PNS groups ( Fig. 2C and D), from 1 mm rostral to 1 mm caudal to the epicenter (Fig. 2F). Notably, there was a significantly spared rim of white matter in the SNS group, even at the lesion epicenter, whereas the control group exhibited severely demyelinated white matter throughout the lesioned area (2 mm rostral and caudal to the lesion epicenter) (Fig. 2C and D).
Transplanted PNS and SNS survived in the injured spinal cord and did not form tumors The photon count measured by bioluminescence imaging (BLI) quantifies only living cells, since the luciferin-CBR-luciferase reaction depends on oxygen and ATP. The successful transplantation of CCV-PNS and -SNS was confirmed immediately after transplantation using BLI, and the average signal intensity was 2.261.6610 5 photons/mouse/sec in 22 transplanted mice. Images obtained weekly thereafter for 6 weeks showed that the signal intensity dropped sharply within the first week after transplantation, but remained at 20% of the initial photon count in both the PNS and SNS transplantation groups throughout the remaining period. Although the signal intensity at 1 week was significantly higher in the PNS group (62.4%) than in the SNS group (29.5%), there was no significant difference in the signal intensity between the PNS (12.6%) and SNS (18.9%) groups at 6 weeks, suggesting there was a similar number of live grafted PNSand SNS-derived cells within the injured spinal cord 6 weeks after transplantation. Thus, similar numbers of grafted PNS and SNS cells may have survived in the injured spinal cord, although the possibility that the grafted cells proliferated differently in the two groups 1 to 6 weeks after transplantation cannot be excluded. Notably, a rapid increase in signal intensity, which would have suggested tumor formation, was not observed during this time period ( Fig. 3A and B). Consistently, histological analysis confirmed that both the CCV-PNS-and CCV-SNS-derived Venus-positive cells survived without forming tumors ( Fig. 3C-F). Quantitative analysis of the Venus-positive area revealed that there was no significant difference of the number of survived grafted cells between PNS and SNS groups 6 weeks after transplantation (Fig. 3G). Moreover, the data of BLI correlated well with Venus-positive area (Pearson's correlation coefficient: 0.759, p = 0.04, Fig. 3H).

PNS and SNS Grafted onto Injured Spinal Cord Exhibited Differentiation Potentials Similar to Those Observed In Vitro
To examine the differentiation characteristics of CCV-PNS and -SNS grafted onto the injured spinal cord, we performed immunohistochemical analyses, and determined the proportion of cells immunopositive for each cell type-specific marker among the Venus-positive grafted cells [3].   Transplanted SNS, but Not PNS, Enhanced Angiogenesis after SCI To examine the effects of CCV-PNS and -SNS transplantation on angiogenesis after SCI, sagittal and axial sections of the injured spinal cord were examined immunohistochemically for aSMA (a marker for smooth muscle cells) or PECAM-1 (a marker for endothelial cells). While a few aSMA-positive cells were observed at and near the lesion site in sagittal sections of both the control and PNS groups, significantly more aSMA-positive cells were found in the SNS group ( Fig. 5A and B). These aSMA-positive cells accumulated near Venus-positive grafted cells ( Fig. 5C and D). Similarly, significantly more PECAM-1-positive blood vessels were observed at the lesion site in the SNS group than in the PNS and control groups ( Fig. 5E-J, Y). To clarify the underlying angiogenic signals, we examined the expression of an angiogenic growth factor, vascular endothelial growth factor (VEGF), in the grafted spinal cord by immunohistochemistry. Although a VEGFpositive area was observed at the lesion epicenter in all three groups 6 weeks after injury ( Fig. 5K-M), it was significantly broader in the SNS group than in the other groups (Fig. 5Z). Furthermore, we found many GFAP/VEGF double-positive host astrocytes, which were negative for Venus (GFP) (Fig. 5N-Q), and a few Venus (GFP)/GFAP-positive graft-derived astrocytes expressing VEGF (Fig. 5R-X).

Transplanted SNS, but Not PNS, Promoted Axonal Regrowth and Enhanced Functional Recovery
To examine the effects of CCV-PNS and -SNS transplantation on axonal regrowth after SCI, we performed immunohistochemical analyses of the injured spinal cord for NF-H (RT97), 5hydroxytryptamine (5-HT), and growth-associated protein-43 (GAP43). While few NF-H-positive axons were observed at the  rim of the lesion epicenter in the control and PNS groups, there were significantly more NF-H-positive neuronal fibers in the SNS group at the lesion epicenter and perilesional area (Fig. 6A, B, and E). 5-HT-positive serotonergic fibers, which are descending raphespinal tract axons [16,17], were observed at the sites caudal to the lesion epicenter in all three groups 6 weeks after injury (Fig. 6C). Quantitative analysis revealed that there were significantly more 5-HT-positive fibers at the site 4 mm caudal to the lesion epicenter (Th10 level), which was approximately at the L1 level, in the SNS group compared with the other groups (Fig. 6F).
While few GAP43-positive axons [18,19,20] were detected caudal to the lesion epicenter in the control and PNS groups, there were significantly more GAP43-positive fibers in the SNS group in the ventral region 1 mm caudal to the lesion epicenter (Fig. 6D, G), suggesting that transplantation of the gliogenic SNS, but not of the neurogenic PNS, promoted axonal regeneration in the injured spinal cord.
We also observed NF-H-positive neuronal fibers extending along with the GFAP-positive immature astrocytes, which may have been partially derived from the grafted Venus-positive cells, and crossing the perilesional area in the SNS group ( Fig. 7A and B). Furthermore, the SNS-derived Venus-positive cells differentiated into MBP-positive oligodendrocytes, which myelinated NF-Hpositive fibers (Fig. 7C). Electron microscopy also revealed active remyelination in the SNS group ( Fig. 7D-F). The grafted cells were in small groups containing 50-100 cells (Fig. 7D). The axons at these sites were enwrapped by myelin sheathes of various thicknesses and numbers of lamellae, which were contributed by the grafted cells, as confirmed by immunolabeling with an anti-GFP antibody to distinguish the transplanted cells from the locally surviving recipient cells (Fig. 7E). A much higher magnification revealed nanogoldlabeled Venus-positive spots in the outer and inner mesaxons of the myelin cytoplasm. Some axons close to the lesion epicenter had undergone considerable re-myelination, and were enwrapped in spirals of more than ten layers of compacted lamellae (Fig. 7F).
Finally, we monitored the locomotor functional recovery in all three groups using the BMS scoring scale [21]. The contusive SCI resulted in complete paralysis (BMS score 0) on day 1, followed by gradual recovery with a plateau around 3 weeks. Although there was no significant difference in the BMS scores among the control, PNS, and SNS groups on day 14, the SNS group exhibited significantly better functional recovery than the PNS and control groups on day 21 and thereafter. On the other hand, there was no significant difference in the BMS scores between the PNS and control groups (Fig. 7G). From a clinical perspective, the recovery of the SNS group to levels exhibiting frequent to consistent weightsupported plantar steps and occasional forelimb-hindlimb coordination was especially noteworthy.

Discussion
Methods for effectively inducing neural differentiation from pluripotent ES cells have been extensively studied [6] and are   expected to be applied in cell replacement therapies for SCI [22]. However, detailed investigations of the optimal cell sources for promoting recovery from SCI are lacking. We recently developed an ES cell culture system that recapitulates the temporal progression of NS/PCs from the FGF-responsive early neurogenic NS/PCs to the EGF-responsive late gliogenic NS/PCs, consistent with CNS development in vivo [10,23] (Fig. 1G, H). Taking advantage of this difference in differentiation tendency, here we examined the distinct effects of the neurogenic PNS and gliogenic SNS on recovery following SCI.
One of the mechanisms underlying this developmental stagedependent gliogenic transition of NS/PCs is the epigenetic regulation of glial cell-specific genes. The gradual demethylation of CpGs around the Stat3 recognition sequence in the GFAP promoter is thought to be involved in the developmental stagedependent increase in transcription of the GFAP gene and the acquisition of astrocytic differentiation potentials [24,25,26]. Interestingly, this process is also observed in our ES cell-derived neurosphere system, in which the proportion of unmethylated CpGs in this region gradually increases during the development of ES cells into secondary neurospheres [10]. This may explain why the in vitro differentiation potentials of both the PNS and SNS were preserved even after their transplantation into injured spinal cord, despite its gliogenic environment (Fig. 4A-F) [27]. Since there was no significant difference in the numbers of grafted PNS and SNS in the injured spinal cord 6 weeks after transplantation, the difference in the in vivo differentiation potentials of the grafted neurospheres was the critical factor influencing the functional recovery after SCI. More than 70% of the grafted SNS cells differentiated into GFAP-positive astrocytes or APC-positive oligodendrocytes, and the engraftment of these cells led to improved functional recovery (Fig. 4F). In contrast, engrafted PNS cells, which mainly differentiated into neurons, did not promote functional recovery.
Determining the exact mechanisms through which the transplanted SNS, or glial cells, improved the recovery of the traumatically injured CNS has been challenging. The engrafted SNS cells could promote a wide range of effects, and here we showed positive effects of their transplantation on tissue sparing, myelination, angiogenesis, and axonal regeneration compared with the control group, and on myelination, angiogenesis, and axonal regeneration compared with the PNS group. One possible explanation for the functional recovery observed in the SNS group is that the SNS-derived astrocytes provided axonal guidance cues. This idea is supported by previous studies in which glial progenitors or glial progenitor-derived astrocytes were engrafted [28,29,30,31]. Immature astrocytes purified from the postnatal CNS have been shown to promote extensive neurite growth from a variety of neurons [32,33].
Although the reactive astrocytes in glial scar tissue express proteoglycans that can inhibit axonal growth, and have been shown to play a major role in the formation of misaligned scar tissue at sites of injury [34,35,36,37,38], we and others have previously shown that reactive astrocytes also have pivotal roles in the repair of injured tissue and recovery of motor function in the subacute phase after SCI, by sealing off injured areas and preventing the further spread of damage. They also produce an array of neurotrophic and growth factors [39]. Moreover, some astrocytes in the host spinal cord acquire stem-cell properties after injury and hence represent a promising cell type for initiating repair [40]. In combination with host astrocytes, immature astrocytes generated by the grafted SNS may express axonal growth-supporting molecules such as laminin, fibronectin, nerve growth factor (NGF), neurotrophin-3 (NT-3), vasoactive intestinal polypeptide (VIP), and activity-dependent neurotrophic factor (ADNF) [41] with minimal expression of chondroitin sulfate proteoglycans (CSPGs) [42]. In addition, SNS transplantation 9 days after SCI, between the acute and chronic phases, is likely to prevent grafted cells from differentiating into glial scar-forming reactive astrocytes due to their minimal expression of cytokines [3,43] and instead generate immature astrocytes, which provide cues for axonal regeneration. In fact, our immunohistochemical analysis revealed NF-H-positive neuronal fibers aligned with GFAP-positive fibers within the lesion site of the SNS group, suggesting that the SNS transplants promoted the alignment of regenerating axons with the fibers of astrocytes, which in turn promoted axonal growth into and out of the SNS grafts ( Fig. 7A  and B). In addition, the 5-HT-raphespinal system of the spinal cord has been shown to represent axonal regeneration after spinal cord injury [16,17], and the apparent regeneration and/or sparing of host 5-HT-positive fibers elicited by the grafting of SNS may have contributed to the observed functional recovery, since these fibers were not observed in the control or PNS groups (Fig. 6C  and F).
Another possible explanation for the functional improvement in the SNS group is the enhancement of angiogenesis, since angiogenesis is reported to promote endogenous repair and support axonal outgrowth after SCI [44]. Under hypoxic conditions, astrocytes express angiogenic growth factors, including VEGF [45,46]. We revealed that transplanted SNS, but not PNS, enhanced angiogenesis after SCI (Fig. 5A-Z). We observed many host astrocytes (Fig. 5N-Q), and a few SNS-graft-derived astrocytes that expressed VEGF (Fig. 5R-X), suggesting that the SNS transplants promoted VEGF expression in both the host-and graft-derived GFAP-positive astrocytes. The increase in blood vessels elicited by the transplantation of ES cell-derived gliogenic NS/PCs may have improved axonal growth and prevented atrophy of the injured spinal cord.
The functional improvement might also be due to remyelination by SNS-derived oligodendrocytes, as supported by previous transplantation studies of ES cell-derived NS/PCs or oligodendrocyte progenitor cells (OPCs) [8,47]. While the neurogenic PNS dominantly differentiated into Hu-positive neurons (Fig. 4F), the gliogenic SNS differentiated into APC-positive oligodendrocytes that provided MBP-positive sheathes and promoted myelination after SCI (Fig. 2C, D, F, and Fig.7C-F).
In summary, here we took advantage of our recently established neurosphere-based culture system of ES cell-derived NS/PCs, in which PNS and SNS exhibit neurogenic and gliogenic potentials, respectively, and found that SNS cells were the most effective for promoting recovery after SCI. We showed that grafted SNS generated approximately equal numbers of GFAP-positive astrocytes and APC-positive oligodendrocytes in vivo. Both of these glial cell types may have contributed to the functional recovery, through trophic effects and the promotion of angiogenesis and axonal regeneration by immature astrocytes, and possibly through remyelination by grafted oligodendrocyte progenitor cells. Notably, the transplantation of PNS did not improve the functional recovery after SCI. These findings provide critical information for clinical trials using human ES-and induced pluripotent stem cell (iPS)-derived NS/PC transplantation for SCI.
Moreover, our results suggest that ES cell-derived NS/PCs cultured for relatively long periods may provide sufficient amounts of efficient glial donor cells for cell transplantation therapies. This strategy may also prevent the contamination of tumorigenic undifferentiated ES cells that occurs during long-term culture under serum-free conditions, and support the development of safe embryonic stem cell-based treatment strategies for spinal cord injury. Although both the CCV-PNS-and CCV-SNS-derived Venus-positive cells survived without forming tumors for 6 weeks after transplantation in this study, careful observation for a longer period will be necessary to assess the possibility of tumor formation.
In the near future, other types of pluripotent stem cells, such as nuclear transfer ES (ntES) and iPS cells, which avoid the risk of immunological rejection and ethical concerns, will need to be evaluated to examine the applicability of human ES cells and human iPS cells in clinical applications.

ES Cell Culture and Differentiation
Mouse ES cells (EB3) [13] grown on gelatin-coated (0.1%) tissue-culture dishes were maintained in standard ES cell medium and used for EB formation as previously described, with slight modifications [9,13,48]. For neural induction, ES cells were dissociated into single cells with 0.25% trypsin-EDTA and cultured in bacteriological dishes for 6 days, to allow the formation of EBs. A low concentration of RA (low-RA; 10 28 M, Sigma) was added on day 2 of EB formation. The EBs were dissociated into single cells with 0.25% trypsin-EDTA and cultured in suspension at 5610 4 cells/ml for 7 days in Media hormone mix (MHM) medium with 20 ng/ml FGF2 (Peprotech) and 2% B27 (Invitrogen), to obtain primary neurospheres (PNS). These PNS were dissociated into single cells with TripleLE Select (Invitrogen) and cultured again in suspension at 5610 4 cells/ml for 7 days under the same conditions, to form secondary neurospheres (SNS) (Fig. 1B) [10]. For differentiation analysis, PNS and SNS were allowed to differentiate on poly-L-ornithine/fibronectin-coated coverslips for 5 days, followed by immunocytochemistry. The frequency of colonies consisting of bIII tubulin-positive neurons, GFAP-positive astrocytes, and O4-positive oligodendrocytes (N, A, O: colonies containing more than 20 positive cells are in capital letters; n, a, o: colonies containing fewer than 19 cells are in lowercase letters) is presented as the percentage of total colonies (50 colonies each) from three independent experiments.

Transfection of CAG-CBRluc-IRES-Venus
To visualize transplanted cells by both fluorescence and luminescence, we established an ES cell line that constitutively expresses a click beetle red-emitting luciferase variant (CBRluc) [11] and Venus [a yellow fluorescence protein (YFP) mutant] [12] by transfecting a linearized CAG-CBRluc-IRES-Venus plasmid (CCV; Fig. 1A) into EB3 ES cells using lipofectamine2000 (Invitrogen). Stably transfected ES cells were selected by G418 (200 mg/ml), subcloned, and screened by the expression of both CBRluc and Venus. The Venus could be detected by antibodies against EGFP.

Flow Cytometry
Undifferentiated ES cells, PNS, and SNS were dissociated and processed for flow cytometric analysis by FACS Calibur (Becton-Dickinson). The Venus-positive cells were counted and are presented as the percentage of the total number of cells, excluding dead cells stained by propidium iodide.

Spinal Cord Injury Model and Transplantation
Adult female C57BL/6J mice (20-22 g) were anesthetized via intraperitoneal (i.p.) injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). After laminectomy at the 10th thoracic spinal vertebra (T10), a contusive SCI was induced at the same level using a commercially available SCI device (IH impactor, Precision Systems and Instrumentation, Lexington, KY), as described previously [49]. This device creates a reliable contusion injury by rapidly applying a force-defined impact (60 kdyn) with a stainless steel-tipped impounder. The initial touch point of the impactor with the dura was determined (using the vibrator mode of the impactor tip), and from there a 1.5-mm displacement was applied to the spinal cord. Force curve readings revealed an average value of 6360.5 kdyn.
Nine days after the injury, CCV-PNS (n = 11) or -SNS (n = 11) that had been cultured for 7 days were partially dissociated and transplanted into the lesion epicenter using a glass micropipette (5610 5 cells/mouse) and stereotaxic injector (KDS 310, Muromachi-kikai, Tokyo, Japan). An equal volume of PBS was injected into the control group (n = 11). Hind limb motor function was evaluated for 6 weeks after SCI using the locomotor rating test of the Basso-Mouse-Scale (BMS), as described previously [21]. Well-trained investigators, blinded to the treatments, performed the behavioral analysis, determining the BMS scores at the same time each day. All animal experiments were approved by the ethics committee of Keio University, and were in accordance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health, Bethesda, MD).

Bioluminescence Imaging (BLI)
BLI was performed using a Xenogen-IVIS 100 cooled CCD optical macroscopic imaging system (SC BioScience, Tokyo, Japan) [3,50]. To examine the effective expression of CBRluc in vitro, we used a CCD-based macroscope detector to determine the luminescence intensity of cultures with various numbers of cells (0 to 5610 5 cells per well) in the presence of D-luciferin (150 mg/ml). The integration time was fixed at a 5-min duration for each image, and the signals were reported as photons/cells/sec. For in vivo BLI, D-luciferin was injected i.p. into mice (150 mg/kg body weight), and serial images were acquired 15-40 min. later, until a maximum signal intensity was obtained with the field-of-view, which was set at 15 cm. We found this time window to be optimal, since the signal intensity peaked 15-40 min after D-luciferin administration, and was followed by a 15-min plateau (data not shown). All images were analyzed with Igor (WaveMetrics, Lake Oswego, OR) and Living Image software (Xenogen, Alameda, CA), and the optical signal intensity was expressed as photon flux, in units of photons/sec/cm 2 /steradian. The results were displayed as a pseudocolor photon count image superimposed on a grayscale anatomic image. To quantify the measured light, we defined a region of interest (ROI) over the cell-implanted area and examined all the values within the same ROI. The signal intensity of the engrafted cells was measured weekly for 6 weeks after transplantation.

Histological Analyses
Animals were anesthetized and transcardially perfused with 4% paraformaldehyde in 0.1 M PBS 6 weeks after transplantation. The spinal cords were removed, embedded in OCT compound (Sakura Finetechnical Co., Ltd.), and sectioned in the sagittal/ axial plane at 20 mm on a cryostat (Leica CM3050 S). The injured spinal cords from the three groups were histologically evaluated by Hematoxylin-eosin (H-E) staining, Luxol Fast Blue (LFB) staining, and immunohistochemistry. The injured spinal cord from the vehicle control group 2 weeks after SCI was also evaluated by LFB staining and immunohistochemistry for 5-HT. Both cultured cells and tissue sections were stained with the following primary antibodies: anti-GFP (rabbit IgG, 1:500, MBL), anti-bIII tubulin (mouse IgG, 1:1000, Sigma), Alexa488-conjugated anti-bIII tubulin (mouse IgG, 1:4000, Covance), anti-Hu (human IgG, For immunohistochemistry with anti-Venus, VEGF, -NF-H, -5-HT, and -GAP43 antibodies, we used a biotinylated secondary antibody (Jackson Immunoresearch Laboratory, Inc.), after exposure to 0.3% H 2 O 2 for 30 minutes at room temperature to inactivate endogenous peroxidase. The signals were enhanced with the Vectastain ABC kit (Vector Laboratories, Inc.). Nuclei were stained with Hoechst33258 (10 mg/ml, Sigma). The samples were examined with a universal fluorescence microscope (Axiocam, Carl Zeiss) or a confocal laser scanning microscope (LSM510, Carl Zeiss).
For immunoelectron microscopy, frozen sections were incubated with nanogold-conjugated anti-rabbit secondary antibody (1:100 Invitrogen) followed by incubation with the primary anti-GFP antibody. After enhancement with HQ-Silver kit (Nanoprobes Inc.), sections were postfixed with 0.5% osmium tetroxide, dehydrated through ethanol, and embedded in Epon. Ultrathin sections were stained with uranyl acetate and lead citrate, observed under a transmission EM (JEOL model 1230), and photographed with a Digital Micrograph 3.3 (Gatan Inc.).

Quantitative Analyses of Stained Tissue Sections through Transplanted Spinal Cord
To quantify HE-, LFB-, or immunostained sections, images were obtained by a universal fluorescence microscope (Axiocam, Carl Zeiss), manually outlined, and quantified by Micro Computer Imaging Device (MCID; Imaging Research Inc., St. Catharines, Ontario, Canada). Constant threshold values were maintained for all the analyses with MCID. HE-stained images were taken at the lesion epicenter and 2, 1, and 0.5 mm rostral and caudal to the epicenter in axial sections at 625 magnification (n = 5, each). To analyze the LFB-positive area after transplantation, we automatically captured four regions from each animal in axial sections at the lesion epicenter and 2 mm and 1 mm rostral and caudal to the epicenter at 6200 magnification. Analyses were performed 2 weeks or 6 weeks after SCI for the vehicle-control group and 6 weeks after for the PNS and SNS groups. The total myelinated area was quantified by MCID using light intensity gain counting (n = 3, each). For the Venus (GFP)-positive area after transplantation, we captured in midsagittal sections the epicenter at 625 magnification from each animal (6 weeks after SCI for the vehicle-control, PNS, and SNS groups), and quantified the total Venus-positive area (n = 6, each). NF-H-stained images were taken at the epicenter and 4, 3, 2, 1, and 0.5 mm rostral and caudal to the epicenter in axial sections at 650 magnification, and the NF-H-positive areas were quantified using light intensity gain counting (n = 5, each). VEGFstained images were taken at the lesion epicenter in axial sections at 650, and the VEGF-positive areas were quantified using light intensity gain counting (n = 3, each). To analyze the 5-HT-positive area after transplantation, we automatically captured five regions from each animal in axial sections 4-mm caudal to the lesion epicenter (Th10 level), which was approximately at the L1 level, a non-lesion site, at 6200 magnification. The analysis was performed 2 weeks or 6 weeks after SCI for the vehicle-control group and 6 weeks after for the PNS and SNS groups. The total 5-HT-positive area was quantified (n = 3, for each condition). For the GAP43positive area after transplantation, we captured the ventral regions in midsagittal sections 1 mm caudal to the epicenter at 6200 magnification from each animal (6 weeks after SCI for the vehiclecontrol, PNS, and SNS groups), and quantified the total GAP43positive area (n = 4). To quantify the proportion of cells positive for each cell type-specific marker in vivo, we selected representative midsagittal sections and automatically captured five regions within 500 mm rostral and caudal to the lesion epicenter at 6200. The engrafted cells in each section that were positive for both Venus and each cell type-specific marker were counted (n = 4). The PECAM-1positive blood vessels were counted manually in axial sections of the lesion epicenter at 6200 magnification (n = 3, each).

Statistical Analysis
All data are presented as the mean 6 s.e.m. An unpaired twotailed Student's t-test was used for the BLI analyses and Venusstained analysis, and in vivo differentiation assays. ANOVA followed by the Turkey-Kramer test for multiple comparisons among the three transplantation groups was used for the in vivo differentiation analysis and PECAM-1-, VEGF-, 5HT-, and GAP43-stained analysis. Repeated measures two-way ANOVA followed by the Turkey-Kramer test was used for HE-, LFB-, NF-H-stained and BMS analysis. Pearson's correlation coefficient was used for correlation of the results of BLI analysis and the quantification of Venus-positive area. In all statistical analyses, the significance was set at P,0.05.