Neuronal Differentiation of Human Adipose Tissue–Derived Stem Cells for Peripheral Nerve Regeneration In Vivo

Nerve lesions in the peripheral nervous system are common injuries that affect 2.8% of trauma patients and often result in permanent disability.^1,2 If primary repair of nerve lesions cannot be achieved, the gold standard is the autologous nerve graft. However, this treatment has drawbacks of suboptimal functional recovery rates, limited graft tissue supply, loss of function at the donor site, and donor site morbidity.^3,4 A promising alternative to direct suturing or interposing of an autograft is a tissue-engineered nerve construct that bridges the nerve gap and facilitates the complex interplay between cells, growth factors, and extracellular matrix required for regeneration. Studies have demonstrated that embryonic stem cells and adult neural stem cells isolated from the brain and spinal cord could differentiate into neurons after implantation into an injured nerve, but ethical and logistical concerns have hindered their therapeutic use until now.⁴ Some recent studies have described mesenchymal bone marrow–derived stem cells (BMSCs) that exhibited the capacity to differentiate into many mesodermal cell types, including neural cell lineages.^5-10 Although these are more plentiful and more accessible than neural progenitor cells, bone marrow aspiration incurs donor site morbidity and often yields limited numbers of cells.¹¹

Similar to BMSCs, adult human adipose tissue–derived stromal cells (hASCs) have also been differentiated into various mesodermal cell types but are advantageous in that they are more accessible, have lower donor site morbidity, and are more plentiful.^11,12 Previously, we demonstrated that hASCs could differentiate into neuronlike cells in vitro using defined culture conditions on a time scale useful for in vivo nerve regeneration.¹³ The aim of this study is to evaluate the functionality of a tissue-engineered peripheral nerve construct composed of a nerve guidance channel and differentiated hASCs in a rat sciatic nerve model. We hypothesize that these hASCs will help bridge a critical size peripheral defect and promote functional nerve regeneration by differentiating into neural cells.

After approval from the human subject protection committee, human adipose tissue was acquired from patients undergoing liposuction procedures. The hASCs were isolated and expanded from the tissue as previously described and were maintained in culture medium (high-glucose Dulbecco's modified Eagle medium [DMEM]; Cambrex, Walkersville, Maryland) supplemented with 10% fetal bovine serum, 1-IU/mL penicillin, 100-mg/mL streptomycin, and 0.25-mg/mL amphotericin B (Mediatech, Herndon, Virginia) at 37°C and 5% carbon dioxide.¹³ Once 70% confluent, the medium was switched to control medium (high-glucose DMEM was changed to low-glucose DMEM) 3 days prior to carrying out the differentiation protocol.

The subconfluent cultures of experimental hASCs were washed with phosphate-buffered saline and standardized for cell number. These cells were placed in neuronal induction medium (DE-1 medium), a modification of previously published protocols, which consisted of high-glucose DMEM supplemented with 10% fetal bovine serum, 0.5mM isobutylmethylxanthine, 1μM dexamethasone (Sigma-Aldrich, St Louis, Missouri), 10μM insulin (Invitrogen, Carlsbad, California), 200μM indomethacin (Sigma-Aldrich), 1% antibiotic/antimycotic solution, 1% nonessential amino acid, 1% N-(2-hydroxyethyl)piperazine-N′-(2-ethanesulfonic acid), and 1% L-glutamine.⁵

Sixty-four athymic nude female rats (rnu/rnu rats; Harlan Laboratories, Indianapolis, Indiana) were used to avoid immunologic response to implanted hASCs. The animals were anesthetized (with ketamine hydrochloride, xylazine hydrochloride, and atropine sulfate intraperitoneally). A 3-cm skin incision was made, and the underlying gluteus muscle was dissected in a muscle-sparing technique. Once the sciatic nerve was exposed, a 13-mm nerve gap was created. Group 1 served as a positive control consisting of 10 rats that were treated with 13-mm isografts harvested from the hindlegs of 5 donor animals. Groups 2 through 5 received silastic conduits that were 15 mm in length, with a wall thickness of 0.38 mm and an inner diameter of 1.25 mm (Pharmaseal, Irwingdale, California). Nerve ends were sutured in place such that they were 1 mm inside the conduit, resulting in a 13-mm nerve gap. The lumens of these conduits were filled with 10⁵ hASCs suspended in 360 μL of culture medium. Group 2 was a negative control and consisted of 20 rats that received nondifferentiated hASCs in culture medium. Group 4 consisted of 20 rats that received differentiated hASCs with DE-1 medium. Groups 3 and 5 consisted of 12 rats each and received the same treatments as groups 2 and 4, respectively, except their medium was renewed 14 and 28 days after implantation of the nerve construct. To renew the medium, animals were anesthetized and the surgical site was reopened. The fluid within the nerve construct was aspirated with a 1-mL tuberculin syringe, and new culture medium was administered with a 27-gauge needle. All experiments were carried out in accordance with the Institutional Animal Care and Use Committee at the University of California, Irvine, and were consistent with federal guidelines.

Functional analyses prior to and 1, 2, 3, and 4 months after surgery were carried out. Animals were trained and tested in a walkway (length, 95 cm; width, 12 cm) as previously described.¹⁴ Using 3 to 6 footprints per animal/time point/group, the sciatic functional index (SFI) score was calculated. The extensor postural thrust (EPT) measures the amount of force that a rat can exert onto a digital scale as it extends its hind limb and is an indication of the performance of the soleus and gastrocnemius muscles. The thrusts from the uninjured limb and the injured limb were determined, and the percentage of motor deficit was calculated as previously described.¹⁴ Sensory evaluation was achieved by testing the withdrawal response mediated by myelinated Aδ fibers in response to mechanical stimulation on the plantar skin. A withdrawal reflex of less than 1 second in response to pinprick was recorded as a positive result. At the same time points, 4 animals in groups 2 through 5 and 2 animals in group 1 were harvested to evaluate the wet weights of the gastrocnemius and soleus muscles. Because these muscles atrophy without innervations by the sciatic nerve, the ratios of harvested wet muscle weights of the injured leg to those of the uninjured leg indicated the degree of muscle atrophy.¹⁴ All animals were functionally tested (SFI, EPT, and sensory evaluation) during baseline measurements (n = 20), but owing to harvest at 7, 14, and 21 days and 2 months, n = 12 for each group for functional testing at 1 month, n = 8 at 2 months, and n = 4 at 3 and 4 months. For end points regarding harvested tissue (muscle weights and histomorphometry), n = 4 for all times consistently.

At each time, rats were killed with carbon dioxide and nerve tissue was harvested for histological assessment as previously described.¹⁵ After isolation, the nerves were fixed in 4% glutaraldehyde (Fisher Chemicals, Fairlawn, New Jersey), embedded in epoxy resin, cut, and stained with toluidine blue. An oil microscope was used to view cross-sections of the nerve 2 mm distal to the injury site by bright field optics. A charged-coupled device camera was used to create digital images of the nerve cross-sections. At a magnification of ×500, 3 nonoverlapping fields for each nerve segment were analyzed using computer image analysis software to measure axon number, nerve fiber density, axon diameter, fiber diameter, connective proliferation index, and number of small fibers (diameter <6 μm). Using these parameters, the myelin thickness, g-ratio (axon diameter–total fiber diameter ratio), and myelin thickness–axon fiber diameter ratio were calculated.

Differences between groups were statistically analyzed using 1-way analysis of variance for gastrocnemius and soleus atrophy because a statistically normal distribution is assumed. In contrast, the SFI, EPT, and sensory evaluations required a nonparametric test, so they were tested using a 2-tailed rank sum test (Mann-Whitney). P < .05 was considered statistically significant.

At 1 month, all groups had a significant decline in SFI score, which is indicative of lack of function. At 3 months, groups 1 and 5 showed much higher SFI scores than the other groups. The disparity between hASCs in DMEM and DE-1 medium was even more apparent 4 months after surgery, with groups 1 and 5 having SFI scores significantly different from those of groups 2, 3, and 4 (P < .05). Group 1 had the best SFI scores at 4 months (mean [SD], −30.4 [10.9]), but they were not statistically different from the SFI scores of group 5 (mean [SD], −35.1 [12.3]). Also, groups 3 and 5, which had their medium renewed, showed better improvements in SFI score than groups 2 and 4, respectively (Figure 1).

Similar to the SFI scores, all groups demonstrated motor deficit during the first 2 months. However, 3 months after surgery, groups 1, 4, and 5 showed decreases in motor deficit compared with negative controls. Groups 2 and 3 showed minimal improvement even after 4 months. Groups 1, 4, and 5 were significantly different from groups 2 and 3 (P < .05). Groups 1 and 4 had similar improvements in motor deficit to about 38%, while the experimental group 5 showed the lowest motor deficit (mean [SD] EPT value, 29.5 [7.812]). The groups with renewed medium had significantly better EPT values than corresponding groups that did not have their medium renewed (Figure 2).

The ratio of wet weights of the gastrocnemius and soleus muscles of the injured leg to those of the uninjured leg is indicative of muscle atrophy, an indirect parameter for nerve regeneration. In all groups, muscle weights increased with reinnervation. In reference to the gastrocnemius muscle wet weights, group 5 had the highest ratio after 4 months (mean [SD], 77 [6]). This value was significantly greater than those of all other groups at every time (P < .05) (Figure 3). The soleus muscles were not significantly different between any of the groups (Figure 4). Renewing the medium had no significant effect in both muscles.

After 1 month, no animal showed a positive withdrawal from mechanical stimulation. At 2 months, where no appreciable improvements in SFI score and EPT were recorded, a positive sensation was found in groups 1 and 5. After 3 months, all groups demonstrated a positive result. Finally, at 4 months, group 1 demonstrated the greatest sensory improvement (Figure 5).

The histomorphometry of groups 1, 2, 4, and 5 are demonstrated. The number of axons increased within each group but without significant differences (Figure 6). However, the renewal of DE-1 medium after 14 and 28 days resulted in a statistically significant increase in axons from 1 to 2 months and from 2 to 4 months, unlike in group 4, where DE-1 medium was not renewed. The number of small fibers (diameter <6 μm) demonstrated a distribution very similar to the number of axons (Figure 7). Early, at days 7 and 14, we saw significantly fewer small nerve fibers than axons, which indicates that these axons are instead associated with larger fibers. Because small nerve fibers (diameter <6 μm) are considered to be regenerating fibers during nerve regeneration, the large fibers might be original nerve fibers that have not been removed by macrophages. The nerve fiber density also correlated well with the previous 2 parameters (Figure 8). Although not significantly different, group 5 had the most axons, most small nerve fibers, and greatest nerve fiber density after 2 and 4 months. In all 3 parameters, the number of axons and fiber density started low and steadily grew. Fiber diameter was larger than 6 μm at days 7 and 14. On day 28, fiber diameter was highly variable. Most fibers were thinner than 6 μm at 2 months but then regained size by 4 months (Figure 9). At 4 months, the fiber diameters of groups 4 and 5 were significantly wider than those of groups 1 and 2. The axon diameter (Figure 10) and myelin thickness (Figure 11) demonstrated patterns comparable to the fiber diameter. Both axons and myelin were thicker at the early times, became significantly thinner at 2 months, and regained thickness at 4 months. The myelin thickness of group 5 was greater than that of the other groups at 4 months. The myelin thickness–axon diameter ratio (Figure 12) and g-ratio (Figure 13) did not show any statistically significant differences between any of the groups.

This study demonstrates an improvement of the sciatic peripheral nerve regeneration over a critical size defect by means of hASCs in DE-1 medium within silicone tubing. Silicone tubing was used as a nerve guidance channel instead of a bioactive material to minimize variables other than the use of hASCs in differentiation medium for our in vivo results. Disadvantages of using synthetic materials such as silicone were never observed within the 4-month observation period.⁴

Contrary to our hypothesis, no hASCs were found to have differentiated histologically. Furthermore, no formation of synapticlike connections between hASCs and the growth cone were observed. It is important to note that improved regeneration is evident in the experimental groups receiving hASCs and differentiation medium (groups 4 and 5). The histomorphological analysis demonstrates that fibers have large diameters and thick myelin sheaths at earlier times. These large fibers are low in number and are likely to be nerve fibers of the sciatic nerve stump that have not been removed by macrophages. At months 2 and 4, the histological picture becomes dominated by small fibers. These small regenerating nerve fibers are responsible for the nerve fiber density at these times. During our observation period, nerves began to reach histomorphological parameter levels typical of normal sciatic nerves. Normal rat sciatic nerves approximately have a mean (SD) of 7115 (413) fibers, a mean myelinated fiber diameter of 6.5 μm, and a nerve fiber density of 11 882 fibers/mm².¹⁶ These histological improvements of regenerated axons correlate well with the overall functional improvement of the sciatic nerve.

The end-organ functional recovery of the regenerated nerve is the clinically relevant outcome. The functional assessment with the SFI has been extensively used and correlates very well with overall functional recovery.¹⁴ Although no major improvements of the SFI score were found after 2 months, the use of induced hASCs in differentiation medium that was renewed actually rivaled the SFI score of the isograft treatment after 4 months. Nie et al¹⁷ designed a tissue-engineered nerve filled with differentiated ectomesenchymal stem cells in collagen within a 10-mm nerve gap and achieved SFI results that were not statistically different from those of an autograft treatment after 4 months. However, after 4 months, their values stayed above an index of 40 and were inferior to autografts. In 2007, Chen et al¹⁸ used BMSCs in silicone tubes to treat a 15-mm gap and also obtained comparable SFI results 10 weeks after surgery, but they did not perform autograft controls. Their experimental values also stayed above an index of 40. These studies and others performed SFI measurements without a baseline value before surgery, which limits the information gained from SFI measurements.^14,15 Because SFI is a dynamic parameter, we believe it should be interpreted over time in regard to changes before and after surgery. Our study revealed marked improvements of SFI scores after 3 and 4 months. Importantly, group 5, the experimental group that had their medium renewed, had a better SFI score on average than the autografts after 4 months.

Although the SFI is the current standard for measuring functional recovery, the variability in SFI measurements requires an alternative test to examine the motor recovery. The results of the EPT coincide with the results of the SFI, as there were considerable improvements after 3 and 4 months. In contrast to the SFI, the EPT showed that group 4 actually achieved functional recovery similar to that of the isograft. Group 5 exceeded the functional recovery of both group 4 and the isograft. The SFI and EPT indicate that functional recovery was incomplete after 4 months. Nevertheless, the use of hASCs appears to be advantageous. In 2004, Kim et al¹⁹ used acellular nerve allograft and attained significant improvements, but these improvements were only 16% to 26% of that of normal nerves after 6 months. In contrast, group 5 shows a much better recovery in EPT after only 4 months. It appears that our cellularized tissue constructs could be a more promising approach than their acellular counterparts.

The gastrocnemius muscle regains mass directly proportional to the extent of reinnervation. Although our experimental group reached a significantly improved outcome, it was still lower than native, normal innervated muscle. This outcome suggests that regeneration is not yet complete. Additionally, these indicate that nerve regeneration and regain of function most likely take longer than 4 months, and a longer observation period is needed to determine whether full recovery is ever achieved. Nevertheless, the muscle weights of the experimental groups rivaled those of the isograft. Importantly, group 5 actually had muscle weights significantly larger than those of the isograft. This result contrasts with the SFI outcomes, as group 5 had a slightly lower average SFI score than the isograft group. The positive result in the gastrocnemius muscle weights indicates a lower degree of muscle atrophy due to greater reinnervation. In contrast to a study by Clavijo-Alvarez et al²⁰ in 2007, the use of hASCs in silicone tubing appears to be superior to acellular solutions. Overall, group 5 demonstrated the greatest motor recovery, as it had results either comparable with or superior to those of the isografts in every parameter. Because our hASCs do not differentiate into neurons, these histomorphological and functional outcomes suggest that hASCs may help improve motor nerve regeneration through the production of neurotrophins such as neural growth factor, which is primarily responsible for the regeneration of motor neurons.⁴

The sensory evaluation demonstrates a faster recovery of sensory neurons in the experimental group compared with controls. The results were comparable with those of Kim et al¹⁹ in 2004; however, in their study, animals with 10-mm sciatic nerve defects treated with acellular nerve grafts had positive reflexes already after 2 months. This discrepancy in sensory response could be attributed to the larger nerve gap of 13 mm in our study. The positive reflex appears to be evident earlier than the onset of detectable motor activity. This might suggest that sensory neurons are quicker to regenerate than motor neurons. Furthermore, because neural growth factor primarily improves motor neurons, the regeneration of sensory neurons implies that hASCs may have expressed other neurotrophins such as neurotrophin 3 and extracellular matrix molecules.²¹

The hASCs offer significant advantages over other stem cell types. They are readily accessible, avoid the risks of obtaining neural stem cells directly from the brain, and do not have the pain and complications of obtaining BMSCs through bone marrow aspirations. The hASCs can be expanded in culture for many passages in vitro without cell senescence, making immortalization unnecessary. This study demonstrates successful results with respect to direct and indirect functional assessment of the nerve. Despite having to withdraw our hypothesis that hASCs would differentiate into neural cells, this study proposes that regeneration may happen through neurotrophin production. Furthermore, because renewing the medium appears to improve parameters of nerve regeneration, it is believed that the differentiation medium plays an important role in nerve healing. Altogether, our peripheral nerve construct strongly promotes nerve regeneration and appears to be comparable with the current gold standard of autograft transplantation. More research is needed to find out whether these hASCs express neurotrophins, whether the DE-1 medium is able to induce expression of these factors, and whether these factors are in fact driving the regeneration process. Further studies need to observe the bioactivity changes to determine which factors promote regeneration, find optimal routes of medium renewal, and analyze this system within porous and biodegradable nerve conduits.

Correspondence: Thomas Scholz, MD, Aesthetic and Plastic Surgery Institute, University of California, Irvine, 200 S Manchester Ave, Ste 650, Orange, CA 92868-3298 (tscholz@uci.edu).

Accepted for Publication: April 15, 2010.

Author Contributions:Study concept and design: Scholz. Acquisition of data: Scholz and Krichevsky. Analysis and interpretation of data: Scholz, Sumarto, and Evans. Drafting of the manuscript: Scholz and Sumarto. Critical revision of the manuscript for important intellectual content: Scholz, Krichevsky, and Evans. Statistical analysis: Scholz. Administrative, technical, and material support: Scholz and Krichevsky. Study supervision: Scholz and Evans.

Financial Disclosure: None reported.

Previous Presentation: This paper was presented at the 81st Annual Meeting of the Pacific Coast Surgical Association; February 13, 2010; Maui, Hawaii.