Long-gap peripheral nerve repair through sustained release of a neurotrophic factor in nonhuman primates

See allHide authors and affiliations

Science Translational Medicine  22 Jan 2020:
Vol. 12, Issue 527, eaav7753
DOI: 10.1126/scitranslmed.aav7753

Growth factor guidance

Injuries to peripheral nerves that result in small gaps can heal after reapproximation; however, large gaps that occur after severe injuries require autograft implantation. As an alternative to autografts, Fadia et al. developed biodegradable polymer scaffolds that release a neurotrophic growth factor. In nonhuman primates, growth factor–eluting scaffolds led to increased nerve conduction velocity, greater Schwann cell recruitment, and similar functional recovery as compared to autograft treatment 1 year after median nerve injury. Results suggest that the acellular conduits could improve peripheral nerve regeneration.


Severe injuries to peripheral nerves are challenging to repair. Standard-of-care treatment for nerve gaps >2 to 3 centimeters is autografting; however, autografting can result in neuroma formation, loss of sensory function at the donor site, and increased operative time. To address the need for a synthetic nerve conduit to treat large nerve gaps, we investigated a biodegradable poly(caprolactone) (PCL) conduit with embedded double-walled polymeric microspheres encapsulating glial cell line–derived neurotrophic factor (GDNF) capable of providing a sustained release of GDNF for >50 days in a 5-centimeter nerve defect in a rhesus macaque model. The GDNF-eluting conduit (PCL/GDNF) was compared to a median nerve autograft and a PCL conduit containing empty microspheres (PCL/Empty). Functional testing demonstrated similar functional recovery between the PCL/GDNF-treated group (75.64 ± 10.28%) and the autograft-treated group (77.49 ± 19.28%); both groups were statistically improved compared to PCL/Empty-treated group (44.95 ± 26.94%). Nerve conduction velocity 1 year after surgery was increased in the PCL/GDNF-treated macaques (31.41 ± 15.34 meters/second) compared to autograft (25.45 ± 3.96 meters/second) and PCL/Empty (12.60 ± 3.89 meters/second) treatment. Histological analyses included assessment of Schwann cell presence, myelination of axons, nerve fiber density, and g-ratio. PCL/GDNF group exhibited a statistically greater average area occupied by individual Schwann cells at the distal nerve (11.60 ± 33.01 μm2) compared to autograft (4.62 ± 3.99 μm2) and PCL/Empty (4.52 ± 5.16 μm2) treatment groups. This study demonstrates the efficacious bridging of a long peripheral nerve gap in a nonhuman primate model using an acellular, biodegradable nerve conduit.


Peripheral nerve injuries can result from trauma, tumor extirpation, and iatrogenic injury. Trauma-related peripheral nerve injuries account for 2 to 5% of people entering level I trauma facilities in the United States and $150 billion in health care costs annually (1). Peripheral nerves can regenerate at a rate of 1 mm/day, and small nerve gaps (<8 mm) can be reapproximated and repaired using epineural coaptation to result in satisfactory nerve regeneration (2). For larger nerve injuries (>2 to 3 cm), direct anastomosis cannot be performed because this would cause excessive tension on the nerve (3). Therefore, the current standard of care for long-gap peripheral nerve injuries is autografting. This procedure involves harvesting a section of donor sensory nerve, often the sural nerve, and transplanting multiple bundled sensory nerves into the defect (4). However, autografts require an additional surgery to excise the peripheral nerve, which increases the chance of surgical complications. There could be donor-site complications such as neuroma formation and permanent loss in sensation (4, 5). The donor nerve may also be insufficient, depending on the size of the injury and other injuries to the patient. Autografts are not ideal for motor or mixed nerve injuries because the autografted sensory nerve differs in morphology from the lesioned motor nerve, resulting in diameter mismatch (6). Hence, there is a strong interest in developing alternatives for long-gap nerve repair.

Synthetic conduits have been examined as possible alternatives to autografts for nerve repair but are seldom used for long-gap nerve repairs because their counterpart, the autograft, produces superior results in terms of regaining function (4, 5, 79). If the nerve gap exceeds 3 cm, growth cones from the proximal stump are unable to respond to the signals released by the distal stump—neurotrophic and neurotropic factors that direct the sprouting axonal growth cones—resulting in poor nerve regrowth (2, 10, 11). Acellular human cadaver nerve grafts are available but have shown limited regenerative potential in nerve gaps of >3 cm (12, 13). To overcome the limitations of the current commercially available conduits, we developed a conduit that provides sustained release of a neurotrophic factor, glial cell line–derived neurotrophic factor (GDNF). Our previous studies examined nerve conduits containing GDNF in a rat sciatic nerve defect model (1419). These studies included optimization of conduit wall thickness, percent porosity, and pore size, as well as methods to create a sustained release of GDNF. In addition to our studies, others have also demonstrated the efficacy of a sustained or synchronous release of GDNF for axonal regeneration (9, 2024). The mechanism of GDNF in nerve repair has been well studied (25, 26). GDNF is up-regulated by Schwann cells of the distal sciatic motor nerve stump immediately after peripheral nerve injury (27); however, there is a decrease in GDNF expression after long-term denervation (28). GDNF is a mitogen and attractant for Schwann cells (29) and also reduces neuronal apoptosis (30). Schwann cells, recruited by GDNF, aid neurite outgrowth by guiding the growth cone physically and chemically toward its target motor unit.

Here, we used a double-walled microsphere strategy to deliver GDNF from the walls of a porous polymeric conduit. The inner portion of the poly(caprolactone) (PCL) nerve conduit is embedded with double-walled poly(lactic-co-glycolic acid)/poly(l-lactic acid) (PLGA)/(PLLA) microspheres containing GDNF. PLGA and PLLA are widely used polymers due to their biocompatibility and biodegradability (14, 31, 32). The coating of the PLGA core with a PLLA shell results in extended drug release as compared to single-walled PLGA microspheres (33, 34). This is relevant for an indication such as peripheral nerve repair, which can require weeks or months for the axons to bridge a large gap. We implanted the porous PCL nerve conduit containing GDNF-loaded double-walled microspheres in a critical-sized 5-cm median nerve defect in the rhesus macaque animal model and compared to reverse-polarity autograft and PCL nerve conduit with empty double-walled microspheres (negative control). Nerve regeneration through the PCL/GDNF conduit at 1 year was significantly improved compared to the PCL nerve conduit containing microspheres without GDNF (PCL/Empty) based on functional, electrophysiological, and histological analyses. Furthermore, PCL/GDNF showed increased Schwann cell presence and enhanced nerve conduction velocity (NCV) compared to autograft, and there was no significant difference in functional recovery between the PCL/GDNF conduit and the autograft. These promising results support the potential of the “off-the-shelf” PCL/GDNF conduit to regenerate large peripheral nerve gaps.


Nonhuman primate 5-cm median nerve defect model

To address the challenge of bridging large peripheral nerve gaps (>2 to 3 cm in humans), we implanted a polymeric nerve conduit combined with a microsphere delivery system in a 5-cm median nerve defect in rhesus macaques (Fig. 1A). The biodegradable nerve conduit consisted of PCL embedded with double-walled PLGA/PLLA microspheres encapsulating GDNF (Fig. 1, B to G). A scanning electron microscope (SEM) image of the PCL nerve conduit after one coating of PLGA/PLLA microspheres indicates the distribution of microspheres within the walls of the PCL nerve conduit (Fig. 1F). Photographs from the surgical implantation (Fig. 2, A, C, and E) and surgical explantation after 1 year (Fig. 2, B, D, and F) indicate that the PCL/GDNF nerve conduits appeared to be well integrated, and the PCL had not yet degraded, as expected.

Fig. 1 Study design and characterization of the PCL/GDNF nerve guide.

(A) Schematic depicting experimental design. (B) Photograph of the 5.2-cm PCL/GDNF nerve guide. (C) SEM of the nerve guide cross section embedded with double-walled microspheres. Mag, magnification. (D) Diagram of the PCL/GDNF nerve guide cross section. (E) SEM of a bisected double-walled PLGA/PLA microsphere. (F) SEM of microsphere adhesion to the initial PCL layer during the manufacturing process. (G) Higher magnification of a cross section of a double-walled PLGA/PLA microsphere embedded in the PCL wall [rectangle in (C)]. EHT, electron high tension; WD, working distance.

Fig. 2 The gross morphological differences between treatment groups at the time of graft implantation and 1-year postoperative (explantation).

Photographs of (A) exposed native nerve, (B) PCL/GDNF conduit explanted after 1 year, (C) implanted PCL/Empty conduit, (D) PCL/Empty conduit explanted after 1 year, (E) implanted PCL/GDNF conduit, and (F) autograft explanted after 1 year.

Functional recovery

To test functional recovery, we trained the nonhuman primates (NHPs) to use only their thumb and index finger to retrieve a sugar pellet from a modified Klüver board, which has four wells of varying different diameters (Fig. 3A) (35, 36). If the thumb and the index finger were used exclusively, then the pinch was recorded as a correct retrieval attempt (Fig. 3B); an incorrect pinch is shown in Fig. 3C. The modified Klüver board is characterized by wells of varying circumferences; therefore, the pinching behavior of the thumb and the index finger is reinforced throughout training sessions using clicker training (37). Surgery was performed after the NHPs used a correct pinch during >90% of retrieval attempts. Beginning at postoperative week 3 through week 50, the same functional pinching motion was analyzed, and the correct number of pinches was recorded. Comparative analyses were also conducted using functional data from week 50 against baseline functional data (Fig. 3D). The percent of successful pinches over the duration of the study among groups is shown in a linear regression plot (Fig. 3E). Trends in all groups correlated to increasing correct pinches. In the full factorial fit model regression, PCL/GDNF and reverse-polarity autograft showed greater recovery throughout as determined by least square means (P > 0.0001). On the basis of crossing of time and condition, regressions were also shown as significantly different (P > 0.001). Removing the PCL/Empty condition and comparing the PCL/GDNF and autograft conditions, no significant difference in recovery throughout was shown with either the least square means analysis (P = 0.583) or comparison of regressions (P = 0.584). This indicates that the PCL/GDNF group recovered their pinch function at a rate similar to the autograft group and at a significantly higher rate than the empty group. Postoperative functional accuracy is as follows: autograft (77.49 ± 19.28%), PCL/Empty (44.95 ± 26.94%), and PCL/GDNF (75.64 ± 10.28%). The PCL/GDNF and the autograft conditions demonstrated similar correct pinch percentages at 1 year (P = 0.985; Fig. 3D). The PCL/Empty condition had the lowest functional return, significantly less than both the PCL/GDNF and autograft conditions (P > 0.001 and P > 0.01, respectively). Videos of all treatment groups documented functional activity at baseline, immediate postoperative, and 1 year postoperative (movies S1 to S5).

Fig. 3 Results and assessment technique for the analysis of NHP functional recovery.

(A) Modified Klüver board with varying well diameters used for functional training and assessment. Well 1 has a diameter of 2.5 cm, and well 2 has a diameter of 0.5 cm. (B) Photograph of the correct pinching motion. (C) Photograph of the incorrect pinching motion. (D) Normalized functional bar graph comparing the NHPs’ 50-week functional recovery to their preoperative baselines. n.s., not significant. (E) Linear regression plot assessing functional recovery over 50 weeks for all treatment groups. n = 30 measurements per time point per NHP. Means represented with +SE/−SD. Adjusted P values presented as: **P < 0.01; ***P < 0.001 (select comparisons shown).

Electrophysiological analyses

Electrophysiological studies were conducted immediately before implantation (baseline native nerve) and immediately before explantation at 1 year (Fig. 4). Compound nerve action potentials (CNAPs) were obtained for the median nerve, and compound muscle action potentials (CMAPs) were obtained via intramuscular (IM) recordings for the abductor pollicis brevis (APB), abductor digiti minimi (ADM), and first dorsal interosseous (1DI) muscles. Representative waveforms for typical biphasic CNAP (top) and CMAP (bottom) responses at 1 year postoperative for native median nerve and each treatment group are shown in Fig. 4 (A to D). Using the measured amplitudes (millivolts) of the recorded peak-to-peak amplitude at various stimulation intensities, a sigmoidal stimulus-response curve was created to determine the suprathreshold stimulus intensity for generating maximal CNAP or CMAP responses. The peak latency values of the maximal, suprathreshold CNAP responses, and the distance between the stimulating and recording troughs were used to determine NCV. The NCV, although significantly reduced relative to the native nerve (70.75 ± 27.19 m/s), was improved in the PCL/GDNF group (31.41 ± 15.34 m/s) compared to the autograft (25.45 ± 3.96 m/s) (P < 0.01) and the PCL/Empty (12.60 ± 3.89 m/s) (P < 0.05) groups (Fig. 4E). The comparative analyses between the preoperative and postoperative NCV for autograft (implant, 91.34 ± 25.13 m/s; explant, 25.45 ± 3.96 m/s), PCL/Empty (implant, 61.93 ± 21.87 m/s; explant, 5.12 ± 6.74 m/s), and PCL/GDNF (implant, 60.43 ± 22.96 m/s; explant, 31.41 ± 15.34 m/s) groups are shown in Fig. 4F. All postoperative NCVs are significantly lower than the preoperative NCVs (P < 0.001) while also being significantly different among the three treatment groups (P < 0.001). The PCL/Empty group, on average, generated little to no CNAP activity at all stimulus intensities, likely due to very few axons being stimulated; in addition, only one of the four PCL/Empty-treated NHPs demonstrated a measurable CNAP (fig. S1).

Fig. 4 Summary of the electrophysiological findings for the regenerated nerves in all the groups.

Voltage waveforms over time in (A) native, (B) autograft, (C) PCL/Empty, and (D) PCL/GDNF groups. For each group, the top waveform, as indicated by “CNAP,” represents compound nerve action potential; and the bottom waveform, as indicated by “CMAP,” represents compound muscle action potential for the APB muscle. (E) The absolute median NCVs between native, autograft, PCL/Empty, and PCL/GDNF groups. (F) Absolute preoperative and postoperative median nerve conduction velocities between autograft, PCL/Empty, and PCL/GDNF groups. (G) The absolute median nerve–stimulated APB muscle amplitude changes between native, autograft, PCL/Empty, and PCL/GDNF groups. (H) Absolute preoperative and postoperative median nerve–stimulated APB amplitude changes between PCL/Empty and PCL/GDNF groups. n = 10 measurements per treatment group. Means represented with +SE/−SD. Adjusted P values presented as: *P < 0.05; **P < 0.01; ***P < 0.001.

The CMAP was obtained for the APB muscle (Fig. 4, G and H). The absolute median nerve-stimulated APB CMAP amplitude changes between native (21.54 ± 13.19 mV), autograft (6.00 ± 3.36 mV), PCL/Empty (0.70 ± 1.08 mV), and PCL/GDNF (1.11 ± 1.05 mV) groups were determined (Fig. 4G). The autograft group produced a greater CMAP than the other treatment groups (P < 0.001), although significantly less than the native group (P < 0.001). Figure 4H compares the pre- and postoperative differences in APB CMAP among autograft (implant, 19.21 ± 9.36 mV; explant, 6.00 ± 3.36 mV), PCL/Empty (implant, 21.45 ± 16.02 mV; explant, 0.70 ± 1.08 mV), and PCL/GDNF (implant, 24.41 ± 12.55 mV; explant, 1.11 ± 1.05 mV) groups. All postoperative APB CMAPs were significantly lower than their preoperative counterparts (P < 0.001) and were also significantly different among the three treatment groups (P < 0.001). Comparative analysis from Fig. 4H shows a statistical difference between the postoperative CMAPs of the PCL/Empty and the PCL/GDNF (P < 0.001) treatment groups, as well as a statistical significance between the autograft group with both the PCL/Empty and PCL/GDNF groups.

Histology and histomorphometry

Native nerves and explanted nerve conduits were subjected to various histological assessments to determine the density and health of the regenerated nerve fibers. In all instances, the distal nerve segment of the NHP’s nerve was analyzed (fig. S2).

Masson’s trichrome. Figure 5 shows prominent collagenous deposits around the epineurium in all groups. A prominent epineurium is observable in all groups except the PCL/Empty group, which displays disorganized collagenous deposition. In contrast, the native nerve, PCL/GDNF, and autograft conditions show a prominently red epineurium that is saturated with organized nerve fibers.

Fig. 5 Brightfield microscopy images of distal cross sections of the explanted native and regenerated nerves stained with Masson’s trichrome.

All images were obtained from the distal nerve segment. (A) Native nerve, (B) autograft, (C) PCL/Empty, and (D) PCL/GDNF at ×100 magnification (scale bars, 500 μm). (E) Native nerve, (F) autograft, (G) PCL/Empty, and (H) PCL/GDNF at ×400 magnification. Scale bars, 100 μm.

S-100. Immunofluorescent stain for the S-100 protein, a Schwann cell surface marker, was performed on the distal nerve segment for all groups (Fig. 6, A to D) and quantified using ImageJ (Fig. 6, E and F) (38). Schwann cells occupied a larger area within the distal nerve segment of the PCL/GDNF group (Fig. 6D), indicating a significantly higher Schwann cell presence compared to the other treatment groups (adjusted P > 0.001). This was similar to native nerve, although significantly less (adjusted P > 0.001) in both area (Fig. 6E) and perimeter (Fig. 6F) quantification. Average Schwann cell area fraction between native nerve (34.31 ± 106.30 μm2), autograft (4.62 ± 3.99 μm2), PCL/Empty (4.52 ± 5.16 μm2), and PCL/GDNF (11.60 ± 33.01 μm2) treatment groups was determined. Average Schwann cell perimeter between native nerve (19.92 ± 19.31 μm), autograft (7.77 ± 4.27 μm), PCL/Empty (7.11 ± 3.59 μm), and PCL/GDNF (12.50 ± 11.87 μm) treatment groups was also determined. As expected, the native nerves expressed the largest average Schwann cell area fraction and perimeter (P < 0.001) compared to the other groups; however, among treatment groups, the PCL/GDNF group expressed the largest average area of Schwann cell presence (P < 0.001).

Fig. 6 Fluorescent micrographs and quantification of Schwann cells in distal cross sections of the explanted native and regenerated nerves.

Sections were stained with S-100 (red, indicating Schwann cells) and 4′,6-diamidino-2-phenylindole (DAPI) (blue, indicating nuclei). (A) Native nerve, (B) autograft, (C) PCL/Empty, and (D) PCL/GDNF (×400 magnification). (E) Average Schwann cell area fraction between native, autograft, PCL/Empty, and PCL/GDNF treatment groups. (F) Average Schwann cell perimeter between native, autograft, PCL/Empty, and PCL/GDNF treatment groups. n = 400 measurements per sample. Means represented with +SE/−SD. Adjusted P values presented as: *P < 0.05; **P < 0.01; ***P < 0.001. Scale bars, 100 μm.

Neurofilament. Immunofluorescent staining on the distal nerve cross sections of all groups for neurofilament is shown in Fig. 7 (A to D). The PCL/Empty group had the lowest average density of neurofilament expression (Fig. 7C). The axons in the PCL/Empty group did not form fascicles; rather, the axons appeared isolated and scattered and were not wrapped in Schwann cells. This observation suggests that the PCL/Empty group was unable to preserve the gross morphological organization of the median nerve fibers. The PCL/GDNF group, however, exhibited considerable nerve fiber density and organization at the level of the distal nerve segment, which was comparable to the native nerve (Fig. 7D). Significantly greater axonal density and morphological organization was observed within the autograft group compared to other treatment groups, which was expected because the reverse polarity autograft group contained appropriate axonal and Schwann cell presence immediately after surgery. Average areas of individual axons from the distal nerve segments were calculated for all groups (Fig. 7, E and F). Measurements of individual axonal areas and radii of all specimens indicated significant differences among all groups. Average axonal area was determined for native nerve (3.79 ± 1.86 μm2), autograft (3.19 ± 1.67 μm2), PCL/Empty (2.88 ± 1.69 μm2), and PCL/GDNF (3.02 ± 1.61 μm2) treatment groups. Average axonal perimeter for native nerve (6.66 ± 1.80 μm), autograft (6.11 ± 1.66 μm), PCL/Empty (5.75 ± 1.74 μm), and PCL/GDNF (5.94 ± 1.62 μm) treatment groups was also determined. Average area and radius of the native nerve were largest (adjusted P > 0.001 compared to all conditions); autografted nerves showed the second largest area and radius measurements. Regarding regeneration in the absence of a nerve scaffold or Schwann cell laminae, the PCL/GDNF treated nerves did not differ significantly in area or radius from the PCL/Empty nerves (P = 0.201). The autograft group had the highest average neurofilament area and radius relative to the other treatment groups (adjusted P > 0.001, both area and radii).

Fig. 7 Fluorescent micrographs and quantification of neurons in distal cross sections of the explanted native and regenerated nerves.

Sections were stained with neurofilament (green, indicating neurons) and DAPI (blue, indicating cell nuclei). (A) Native nerve, (B) autograft, (C) PCL/Empty, and (D) PCL/GDNF (×400 magnification). (E) Average axonal area between native, autograft, PCL/Empty, and PCL/GDNF treatment groups. (F) Average axonal radius between native, autograft, PCL/Empty, PCL/GDNF treatment groups. n = 400 measurements per sample. Means represented with +SE/−SD. Adjusted P values presented as: ***P < 0.001. Scale bars, 100 μm.

g-Ratio. The g-ratio is indicative of the degree of fiber myelination (39). Figure 8 (A to D) shows representative images from each of the treatment groups; these images were used to quantify the g-ratio. Figure 8E reports the area-based analysis, and perimeter-based analysis is shown in Fig. 8F. Average g-ratio area between native nerve (0.424 ± 0.084 μm2), autograft (0.466 ± 0.081 μm2), PCL/Empty (0.473 ± 0.115 μm2), and PCL/GDNF (0.447 ± 0.094 μm2) treatment groups was determined. Average g-ratio perimeter between native nerve (0.464 ± 0.095 μm), autograft (0.498 ± 0.089 μm), PCL/Empty (0.482 ± 0.124 μm), and PCL/GDNF (0.471 ± 0.098 μm) treatment groups was also determined. In both area- and perimeter-based analyses, the myelination of the PCL/GDNF conduits was similar to the native nerves. The greatest myelination, as determined by area-based analysis, was found in the native nerves, as indicated by the lower areas and perimeters. However, according to perimeter-based analyses, there was no significant difference between the PCL/GDNF and native nerve, which showed a similar degree of fiber myelination (P > 0.999). There also was no significant difference between the autograft and the empty graft groups, indicating a similar but lower degree of myelination as assessed by perimeter-based analyses, when compared to the PCL/GDNF group or the native nerves.

Fig. 8 Degree of myelination as determined by g-ratio.

(A) Native nerve, (B) autograft, (C) PCL/Empty, and (D) PCL/GDNF cross sections were stained with osmium tetroxide for contrast and viewed under brightfield microscopy at ×400 magnification. Red rings indicate the myelin, whereas green indicates the axon fiber. (E) Average g-ratio area between native, autograft, PCL/Empty, and PCL/GDNF treatment groups. (F) Average g-ratio perimeter between native, autograft, PCL/Empty, PCL/GDNF treatment groups. n = 60 measurements per sample. Means represented with +SE/−SD. Adjusted P values presented as: *P < 0.05; **P < 0.01; ***P < 0.001. Scale bars, 100 μm.


This work demonstrates that the delivery of GDNF from polymeric nerve conduits can guide the growth of axons across a 5-cm median nerve defect in an NHP model. It is challenging to repair long peripheral nerve gaps due to the lack of cues to guide the axons to the distal stump. Several neurotrophic factors, including nerve growth factor, have been examined in preclinical studies to assess optimal axonal regeneration over the past several decades (14, 20, 4043). Although there is an abundance of studies using small animal nerve defect models to examine nerve conduits, there are few studies that examined large peripheral nerve injuries in the NHP model (4449). Hu et al. (45) investigated a chitosan-based nerve conduit containing bone marrow–derived stem cells in a 5-cm median nerve defect in rhesus monkeys, finding that the inclusion of stem cells in the conduit enhanced nerve regeneration. In another study, Jiang et al. (48) examined decellularized nerve allograft in a 4-cm ulnar nerve defect in rhesus monkeys and found that the nerve was repaired only when Schwann cells were added to the graft. These previous studies indicated that conduits without cells or growth factors cannot bridge a large (>3 cm) peripheral nerve gap.

Here, GDNF was selected for microsphere encapsulation and localized delivery due, in part, to known proliferative and migratory effects on Schwann cells (25, 26, 29). As the GDNF is slowly released from the nerve conduit, Schwann cells are recruited into the conduit and secrete additional factors, thus guiding and enhancing axonal regeneration. Previous studies have found GDNF to be a potent neurotrophic factor and have observed areas of regenerated nerve with a high GDNF concentration to exhibit the “candy store” effect, whereby growing axons become trapped and do not extend beyond the concentrated GDNF site (5055). It is possible that we do not observe the candy store effect due to the controlled release of GDNF, resulting in substantially lower overall GDNF concentration within the conduit (14, 15, 53). Furthermore, axons affected by the candy store effect would likely have produced neuromas due to axonal entrapment; however, no neuromas were found in any nerves upon excision.

The GDNF-encapsulated microsphere delivery system was selected on the basis of our results from in vitro and small animal in vivo studies (14, 15). This system was initially investigated due to the reduced burst release of encapsulated GDNF and linear controlled release of bioactive GDNF over more than 50 days (14). When the doubled-walled microspheres were incorporated into the PCL conduit, the GDNF release was further reduced due to the slow degradation kinetics of the PCL polymer (14, 15). In vivo studies with the PCL/GDNF conduit indicated an increased Schwann cell migration into the conduit (14, 15). The PCL/GDNF conduits also increased axonal regeneration (15) and end organ innervation as assessed by isometric twitch force measurements, which were comparable to the reverse-polarity autograft (14). On the basis of our previous success in the small animal studies, we pursued our evaluation of the PCL/GDNF guide in the large animal model described here. In addition, we examined the bioactivity of GDNF released from the guide after ethylene oxide (EtO) sterilization, determining that the GDNF remained active and about 14 ng was released from the PCL/GDNF nerve guide over 50 days (56).

In addition to GDNF-encapsulating microspheres, a porous PCL nerve conduit was also chosen to bridge the nerve gap, due in part to its slow degradation time. Through previous studies, we determined that the PCL conduit at 80% porosity allows adequate permeability to solutes and fluids while preventing a rapid release of GDNF (57). The slow degradation is an essential property of the PCL conduit as it can adequately guide neurite outgrowth over long distances for an extended period of time. In addition, the porosity provides flexibility to the conduit, allowing nerve repair over anatomical creases.

In general, PCL has a slower degradation rate and lower permeability than other commercially used polymers such as collagen and poly(glycolic acid). These features have been suspected to contribute toward adverse effects such as extrusion, fistula formations, and foreign body reactions (58, 59). However, we did not observe any negative consequences in either of our PCL groups in this NHP model. We have also performed numerous small animal studies with PCL conduits over the last decade and have not observed any adverse reactions to the PCL polymer or the porosity of the conduit (14, 15, 18, 6062).

The return of motor function, histological analysis, and electrophysiological analysis were examined after 1 year. The median nerve defect was studied because the median nerve innervates the APB muscle that controls the thumb, allowing the measurement of functional recovery throughout the duration of the study. The functional recovery results indicate that there was no significant difference in functional recovery between the PCL/GDNF and autograft groups. While assessing functional recovery, it is important to note the visual differences in the overall manual dexterity of the NHPs across each group. The NHPs assigned to the PCL/Empty group generally had resting stiffness in their thumbs and index fingers throughout the regeneration period, indicating poor motor function of the target muscles. In contrast, fine motor control of the thumb and the index finger was observed within the PCL/GDNF grafted NHPs and the autografted NHPs in terms of digit coordination and pinch force required to hold the pellet as early as 16 weeks. Comparative analysis of the functional data indicates that there is no significant functional difference between the PCL/GDNF group and the autograft group. Given the acceptable performance of the reverse-polarity autograft, these findings are important, demonstrating the efficacy of PCL/GDNF nerve conduit to bridge a large nerve gap with marked functional recovery.

The functional recovery tests were corroborated with both electrophysiology and histology. The NCV for the PCL/GDNF nerve conduit was significantly increased compared to autograft and PCL/Empty groups. Moreover, we observed a multipeaked CNAP waveform for the regenerated nerves in the autograft group, unlike the typical biphasic CNAP waveform recorded from either native or PCL/GDNF graft group. Because the CNAP is a population response aggregate from the action potentials generated by every individual axon in the mixed median nerve, including all sizes and all functional axon types, one explanation for this finding could be the unequal rate of remyelination for different subgroups of regenerating axons, likely determined by axon diameter and functional classification. Therefore, as different subpopulations of axons regenerate at different rates, each subgroup will conduct along the gap at differing mean NCVs such that the resulting CNAP has multiple waveform components across a broad range of peak latencies, and various proportions of axons in each group remain unmyelinated after peripheral nerve injury (63). This robust finding of multipeaked CNAP waveforms only for autograft CNAPs suggests that the remyelination process for autografts was delayed in contrast to a more complete remyelination for PCL/GDNF conduits. The reinnervation of the APB muscle distal to the injured nerve was assessed with CMAP recordings. On the basis of the similar electrophysiological findings between the PCL/GDNF conduit and autograft, the PCL/GDNF conduit was as effective at regenerating the median nerve. Similar to the CNAP, CMAPs are a population response of the aggregate total motor unit potentials generated by all muscle fibers within individual motor units. The amplitude of the CMAP serves as an estimate of motor unit numbers that regenerated, and similar CMAP amplitudes for autografts and PCL/GDNF grafts suggest that similar numbers of motor units regenerated for both grafts. Note that, although neither of the groups exactly replicated the native nerve CNAP or CMAP values, the PCL/GDNF group achieved the most robust biphasic morphology of the CNAP waveform. Also note that the PCL/Empty group, on average, generated little to no CNAP response even at supramaximal stimulation intensities, likely due to the small number of axons that had regenerated. Last, although the autograft group had a robust CNAP response, it did not generate a typical biphasic waveform as did the native and the PCL/GDNF groups. Thus, on the basis of the similar electrophysiological findings between the PCL/GDNF conduit and autograft, the PCL/GDNF conduit is effective at regenerating the median nerve.

Histological analysis of NHP nerve cross sections at the distal site of anastomosis was performed to measure axonal maturity, Schwann cell presence, gross structure evaluation, collagenous deposition, and myelination. Collagenous deposition with small islets of nerves was observed in all groups. The degree of collagen deposition appeared highest in the PCL/Empty group. Native nerve, PCL/GDNF conduit, and autograft groups showed visible fascicles consisting of axons, encapsulated by the epineurium at the distal site of anastomosis. However, the lack of fascicles in the PCL/Empty conduits depicts the lack of nerve maturity at the level of the distal nerve. Sustained GDNF release, such as from the PCL/GDNF conduit, likely promoted Schwann cell migration and proliferation and a subsequent increase in the rate of axonal regeneration. A robust Schwann cell presence was evident at the level of the distal nerve segment for the autografts and the PCL/GDNF conduits. Quantification of the area occupied by Schwann cells indicated a significantly higher concentration of Schwann cells in the PCL/GDNF group compared to the autograft. The Schwann cells in the reverse-polarity autografts were expected to be similar to the native median nerve because the Schwann cell basal lamina, which was retained throughout the study, may have accelerated Schwann cell migration (64). The histological and histomorphometric analyses of Schwann cells for PCL/GDNF conduits demonstrated robust Schwann cell presence within the PCL/GDNF conduit, likely due to sustained release of GDNF.

Axonal area, a direct indicator of axonal conduction velocity and an indirect measure of axonal maturity in the rhesus macaque, was also studied (65). Regenerating axons will have a smaller diameter and a higher resulting axoplasmic resistance, whereas fully matured axons will have a wider diameter and lower subsequent resistance. This phenomenon is quantified as the Hursh factor, which relates axonal diameter to the NCV (66). The axon diameter within a rhesus macaque NHP median nerve follows a bimodal distribution between 1 and 19 μm, with the major peak occurring at 3 μm (67). Thus, using diameter ranges of 1 to 3 μm, axonal areas of distal nerves were calculated. The quantified axonal area of the autografts was significantly higher than the PCL/GDNF conduits, suggesting a greater axonal maturity. This was expected because the autograft provides a cellular scaffold, which includes stimulatory factors such as Schwann cell laminae, nerve growth factors, and adhesion-promoting molecules. The results for the PCL/Empty group have a minimal amount of intact regenerated nerve fibers in the distal nerve segment, in comparison to the other treatment groups, and these nerves were likely not myelinated. The PCL/GDNF group regenerated a significant amount of nerve fibers, in comparison to the PCL/Empty group. The autograft group was able to regenerate a significant amount of nerve fibers while preserving the median nerve morphology.

Area- and perimeter-based g-ratios were analyzed to assess the degree of fiber myelination between two similar scalar modalities. Because perimeter and area are similar scalar modifications, the results between the two analyses were expected to be similar as well. The results suggest physiological differences within the nerve morphology and higher axonal myelination within the PCL/GDNF group compared to other treatment groups. This was expected as it has been shown that GDNF increases the myelination of nerve fibers, as well as the ratio of myelinated nerve fibers (26).

The combined results from functional assessment, electrophysiology, and histology provide supporting evidence for the efficacy of the PCL/GDNF conduits to bridge large peripheral nerve gaps. However, there are some limitations to the study design. In a traditional clinical setting, the autografting procedure would typically involve harvesting sural nerve from a patient to bridge the primary peripheral nerve gap. However, sural autografting was not possible with NHPs because their arm span exceeded their leg height. Because of difference in lengths of the appendages, there was insufficient sural nerve to bridge the 5-cm median nerve gap. In addition, using insufficient length of the sural nerve graft could result in excessive tension and lead to poor nerve regeneration (68, 69). To avoid false misrepresentation of the regenerative capacity of our autograft group, we used the excised 5-cm median nerve as a reverse-polarity autograft for our positive control. The reverse-polarity autograft is not the clinical gold standard; however, we were unable to use sural autograft due to the aforementioned limitations and chose to use the reverse-polarity autograft, which is the predominant positive control used in both small animal and large animal nerve defect studies (46, 7072). Furthermore, Tang et al. (73) recently demonstrated that there was no significant difference in performance of the sural nerve autograft versus reverse-polarity nerve autograft in a rat sciatic nerve defect model, further justifying the use of the reverse-polarity autograft as a clinical control.

A limitation of this study involves the test used to assess motor recovery, which consisted of a simplified pinch retrieval test using a modified Klüver board. We chose to use the Klüver board due to its simple effective design requiring minimal training, and because the pinch retrieval task critically depends on a minimal number of primary muscles acting synergistically (74, 75). Most other functional tests require involvement of many more muscles that cross proximal arm joints, which could confound the interpretation of the simpler (but still purposeful) pinch task used (75, 76). Our overall goal for this study was to use an array of evaluations (functional, electrophysiological, and histological) to interpret the regeneration of the median nerve and subsequent muscular reinnervation. The electrophysiological studies evaluate the reinnervation profile of individual hand muscles, and by comparing native to grafted CMAP amplitudes, it is possible to estimate the percentage of motor units present in the grafted subjects compared to the native nerves. In addition to the extensive electrophysiology performed on these animals, we also assessed the ability of stimulation proximal to the median nerve to elicit an APB thumb twitch. This suggests that the median nerve was innervating the APB muscle sufficiently to cause a muscular contraction. Through extensive electrophysiological studies on the NHPs, we demonstrated contractions of the APB muscle based on discreet stimulating pulses supplied to the proximal median nerve.

Another limitation of this study is the variability that was shown to exist between the individual NHPs for pinching correctness, as it was defined for assessing functional criteria. To account for this variability, we normalized the “correct pinch” values for each NHP such that the baseline value was 100%, and all subsequent values were adjusted accordingly. Although steps were taken to reduce this subjectivity by increasing the analysts for the functional assessments, the variability in data was present at time of statistical analyses.

Last, the type of nerve selected for transection also introduced variability in the electrophysiology and histomorphometry. The median nerve innervates the APB muscle and, together with the ulnar nerve, cross-innervates the 1DI muscle. In surgically transecting the median nerve, the innervation of the APB and 1DI motor units was removed. This lack of median nerve innervation of APB could allow compensatory mechanisms, such as new ulnar nerve cross-innervation of APB, to occur, measurable by testing ulnar nerve stimulation CMAPs after 1 year. New ulnar nerve innervation of APB has been previously reported (77). Although small in amplitude (~1 mV), indicating the relatively small number of ulnar nerve axons innervating the APB muscle at implant with the median nerve intact, APB CMAP amplitudes tripled (>3 mV) 1 year after median nerve denervation of APB, supporting that new ulnar nerve cross-innervation of APB had occurred in this model, involving about three times the number of ulnar nerve axons innervating APB. To some extent, this cross-innervation of APB from the intact ulnar nerve would have contributed to the postoperative performance observed.

Despite the study limitations, the data collected during the study support the efficacy of the PCL/GDNF nerve conduit for long-gap peripheral nerve repair. Comparative efficacy against autograft, the clinical standard of care, suggests the PCL/GDNF nerve conduit can be examined as a clinical alternative to the autograft.


Study design

We implanted the porous PCL nerve conduit containing GDNF-loaded double-walled microspheres in a critical-sized 5-cm median nerve defect in the rhesus macaque animal model and compared to reverse-polarity autograft and PCL nerve conduit with empty double-walled microspheres (negative control). Our laboratory has previously demonstrated the efficacy of the PCL/GDNF nerve conduit in a rat sciatic nerve defect model (14, 15). Slowly degrading double-walled microspheres consisting of a PLGA core and a PLLA shell (Fig. 1E) containing GDNF were embedded into the inner luminal wall of the conduit for sustained GDNF release (Fig. 1, C, F, and G). To create a critical-sized peripheral nerve injury, 5-cm median nerve defects were created in 12 NHP arms, which were randomly distributed into three groups of n = 4: PCL/Empty (negative control), PCL conduit with GDNF-encapsulated microspheres, and reverse-polarity autograft. A 5-cm segment of the median nerve was sharply excised and autograft, PCL, or PCL/GDNF nerve conduit was sutured to the proximal and distal ends of median nerve under no tension to bridge the defect. In the autograft group, the graft oriented such that the distal portion of the excised nerve faced the proximal stump and vice versa and microsurgically coapted to the median nerve, creating a reverse-polarity autograft. Conduit implantation surgeries and reverse-polarity autograft surgeries were performed in NHPs. The nerves were allowed to regenerate in all groups for 1 year before being explanted. During the regeneration period, functional analyses were carried out at regular intervals to assess recovery of hand function. In addition, electrophysiological studies were performed before the implantation of the treatment groups and the explantation of the regenerated nerve to determine nerve conduction characteristics of the regenerated nerve. Last, histology and histomorphometry were conducted on the excised native and regenerated nerves for comparative microscopic analyses.

PCL nerve conduit fabrication

The polycaprolactone nerve conduit fabrication was based on our previously conducted nerve repair study using a small animal model (see Supplementary Materials and Methods) (14, 15). Conduits were autoclaved using room temperature ethylene oxide sterilization, as previously described (56).

Surgical procedures

Following the guidelines of the University of Pittsburgh Institutional Animal Care and Use Committee, eight male Macaca Mulatta (rhesus macaque) weighing 7 to 13 kg (average weight, 11.2 ± 1.6 kg) and 6 to 10 years old (average age, 7.4 ± 1.2 years) NHPs without any known prior hand injuries were obtained for these studies. Animals were first anesthetized with ketamine (1.4 mg/kg) intramuscularly and then transferred to the sterile operating room. The animals were then intubated and exposed continuously to isoflurane gas anesthesia (1 to 2%) throughout the duration of the surgery. Surgeons who performed the surgeries were faculty and residents in plastic surgery with previous experience in peripheral nerve repair. Under sterile conditions, the median nerve (~1.5 mm in diameter) was transected about 2 cm proximal to the wrist. The median nerve was mobilized after transection within the muscle bed for about 10 cm to decrease potential tension at the repair site. A 5-cm median nerve section in the forelimb was resected using sterilized instruments. For all treatment groups, the grafts were sutured at the proximal and distal ends with 8-0 nylon perineural sutures. The skin was closed with 5-0 silk sutures and disinfected with povidone-iodine solution. Immediately after surgery, NHP arms were bandaged and casted (Vetcast Plus, 1″ Veterinary Casting Tape) to prevent disruption of the surgical site and to allow for tissue healing. NHPs were ambulatory within 1 to 2 hours after surgery. The cast was retained on the NHPs for 10 to 14 days after surgery. After 1 year, the implant was removed and replaced with a decellularized human allograft. After 30 days of recovery, followed by modified Klüver board training, surgery was then conducted on the other arm, with 12 of 16 arms used for this study.

Functional training and analysis

To measure functional recovery, NHPs were trained to use their thumb and index finger in a pinching motion to retrieve a banana pellet from a modified Klüver board before surgery (35, 36). The board is rectangular with four wells of varying diameter from 1.0 to 2.5 cm, with a depth of 1 cm. As the wells reduce in diameter, the NHPs were compelled to use their index finger and thumb to retrieve the pellet. During one functional training session, 30 total trials were conducted and video-recorded each week, and the behavioral data were analyzed by three blinded reviewers through the recorded video clips. The number of correct pinches and incorrect pinches by the NHP was recorded. As shown in Fig. 3B, for a pinch to be recorded as “correct,” the NHP must have used its thumb and index finger exclusively, in a pinching motion. In addition, the NHP must have enough force within its digits to hold the pellet between its thumb and index finger pads for the pinch to be recorded as correct. If the NHP used more than two digits, grabbed at the pellet using all digits, or scooped out the pellet using the index finger before grabbing it, the pinch was counted as an “incorrect” pinch. The NHPs were trained to eat with a pinch using the modified Klüver board to establish a baseline. Once sufficiently trained (which took ~3 weeks), each NHP’s baseline pinch accuracy for pellet retrieval was recorded before surgical implantation. Immediately after surgery, the NHPs were casted for 2 weeks. Functional training was conducted every week until week 50 to record the frequency of accurate pinches and extrapolate functional recovery from the data collected.


Electrophysiological studies were conducted before implantation and at explantation. During surgery, a stimulating trough was positioned several millimeters from the proximal site of median nerve anastomosis, and the recording trough was positioned several millimeters from the distal site of median nerve anastomosis (fig. S1). Distance between the two troughs was recorded. Each stimulation was recorded as an independent waveform, and the waveforms were overlaid to generate repeated measures of one recording channel of CNAPs and three channels of CMAPs. Using the recorded peak-to-peak amplitude at various stimuli, a sigmoidal stimulus-response curve was created to determine the suprathreshold stimulus intensity for generating each potential. The peak latency values (milliseconds) of the biphasic CNAP/CMAP were used along with the distance between the stimulating and recording troughs to generate NCV. In addition, the peak-to-peak CNAP amplitude was compared between all groups as a surrogate for the number of axons that regenerated across the gap. Stimulation currents of varying intensity (0 to 20 mA) with 0.3-ms pulse duration were delivered to the nerve. Distance between the two troughs was recorded, and care was taken to ensure that the distance between the stimulating and the recording troughs was greater than 5 cm. For CMAP recordings, pairs of sterile hook-wire electrodes (019-475300, Natus Medical Inc.) were inserted intramuscularly using a 27-gauge hypodermic needle into the bellies of the APB, 1DI, and ADM muscles to record the evoked CMAP responses. Compound nerve and muscle action potentials were recorded in response to escalating stimulus intensities of 0.05 to 20 mA. After completing electrophysiological studies on the median nerve, CMAP studies of the ipsilateral ulnar nerve were performed using the same methods and same three pairs of IM recording electrodes in APB, 1DI, and ADM. In this study, the ulnar nerve was used as a control group and to study the effect of denervation of the median nerve to account for the possibility that the ulnar nerve could sprout expanded cross-innervation into the APB muscle as a compensatory mechanism to achieve recovery of grip function. After completion of the electrophysiology studies, the corresponding nerve graft surgery was performed on the median nerve.

The median and ulnar nerve studies were recorded and analyzed using EPWorks software from Natus/Xltek. To determine the stimulation current that produced suprathreshold CNAP and CMAP responses, stimulus-response curves were produced for each preimplant and preexplant study. In addition, peak-to-peak CNAP and CMAP amplitudes were measured to analyze the largest voltage values to infer the degree of nerve regeneration and reinnervation that occurred. Because the CMAP/CNAP is a population response, we assumed that native nerve maximum amplitudes represented 100% of all median nerve motor axons contributing to the population response. Therefore, at explantation, the recorded maximum amplitude should estimate the percentage of conducting axons in the population of the regenerated nerve 1 year later. For NCVs, the latency of the initial, depolarizing peak of the CNAP was used to calculate conduction time that was divided into the measured distance between the two troughs to calculate NCV. The NCV and peak-to-peak amplitudes of the CNAPs and CMAPs, respectively, were used for comparative analysis between the preimplant and preexplant studies. Ten measurements at each time point per animal were analyzed.


After postfixation with osmium tetroxide, all nerves were sectioned and embedded within paraffin. For this study, the distal nerve segment was used exclusively for histological analysis. Detailed experimental methods for preparation and staining of the slides are detailed in Supplementary Materials and Methods.

Statistical analyses

All statistical analyses were performed in JMP Pro 14 except post hoc analyses for analyses of variance (ANOVAs) with unequal variances, which were performed in Minitab 14. Functional recovery as assessed by pinch was normalized to 100% in each condition, after which the values across all time points where adjusted to the condition’s respective normalization factor. A full factorial fit model regression was performed crossing time against condition with a = 0.05. To compare the PCL/GDNF and autograft conditions, the model was repeated, suppressing the PCL/Empty condition and adjusting the error to a = 0.033. For the linear regressions, correlations above R2 > 0.50 were considered strong, correlations within 0.50 ± R2 ± 0.25 were considered moderate, and correlations with R2 > 0.25 were considered weak. Repeated measures ANOVA were performed on data collected from electrophysiology, and one-way ANOVA were performed on S-100, neurofilament, and g-ratio analyses. Residuals from each dataset were assessed for normality consistent with assumption of ANOVA by Shapiro-Wilk W test. If residual distributions were found to be non-normal, the dataset was transformed using either normal quantile or cumulative probability transformation, and then, residuals were reassessed to ensure the normality condition was met. Conditions were then tested for homoscedasticity using Welch’s F test. Tukey comparisons were used in two-way ANOVA for assessing differences between least squares means. One-way ANOVA with Tukey-Kramer post hoc was performed with datasets of equal variances or with Games-Howell post hoc (Minitab) with datasets of unequal variances. Statistical significance was defined as P < 0.05. All charts are expressed as means with SE above (+) and SD (−) below the mean. Primary data are reported in data file S1.


Materials and Methods

Fig. S1. CMAP amplitude for APB after ulnar stimulation.

Fig. S2. Schematic of the histology study design.

Movie S1. Preoperative functional assessment.

Movie S2. Postoperative functional assessment at day 15.

Movie S3. Postoperative functional assessment at week 50 after autograft.

Movie S4. Postoperative functional assessment at week 50 after PCL/Empty scaffold.

Movie S5. Postoperative functional assessment at week 50 after PCL/GDNF treatment.

Data file S1. Primary data (Excel file).


Acknowledgments: We acknowledge L. Kokai, J. Hokanson, R. Schroth, M. McLaughlin, J. McAtee, D. Linger, A. Taylor, A. Dees, A. D. Narayanan, D. Gryboski, R. Liberatore, and G. Williamson for technical assistance; the University of Pittsburgh Center for Biologic Imaging for assistance with microscopy; and M. Luz at MedGenesis Therapeutix Inc. for supplying GDNF. We also acknowledge S. Cashman and M. Petts in the University of Pittsburgh Animal Facility for assistance with the surgeries. Funding: This work was supported by the Army, Navy, NIH, Air Force, VA, and Health Affairs to support the AFIRM II effort under award no. W81XWH-14-2-0003. The U.S. Army Medical Research Acquisition Activity, 820 Chandler Street, Fort Detrick MD 21702-5014 is the awarding and administering acquisition office. Opinions, interpretations, conclusions, and recommendations are those of the author and are not necessarily endorsed by the Department of Defense. R.S. was a recipient of Swiss National Foundation funding (Early Postdoc Mobility Grant). Author contributions: Experimental design: K.G.M., J.M.B., and D.J.C. SEM imaging: C.T.-R., J.M.B., and A.J.R. Nerve guide fabrication: C.T.-R., J.M.B., D.M.M., and W.N.S. Polymer microsphere fabrication: C.T.-R. and J.M.B. NHP surgeries: J.M.B., D.A.B., D.-Y.K., A.M.S., G.A.D., H.-T.L., M.W., R.S., F.M.E., I.B.J., K.M.W., M.A.S., and W.N.S. Histological staining: J.M.B., J.E.B., I.K.C., T.G., N.B.F., T.J., A.R.C., G.E.P., and A.A.G. Histological analysis: C.T.-R., J.M.B., J.E.B., N.B.F., A.R.C., G.E.P., and A.A.G. Functional analysis: J.M.B., I.K.C., N.B.F., G.A.D., J.M.B., and A.R.C. Functional training: J.M.B., I.K.C., G.A.D., T.G., C.M.M., D.M.M., and W.N.S. Electrophysiological testing: J.M.B., G.A.D., T.S., D.J.W., and D.J.C. Electrophysiological analysis: J.M.B., G.A.D., N.B.F., D.J.C., G.E.P., and A.R.C. Data analysis: K.G.M., J.M.B., G.A.D., N.B.F., and D.J.C. Interpretation of results: K.G.M., J.M.B., G.A.D., N.B.F., B.K.S., and D.J.C. Statistical analysis: B.K.S. Manuscript writing: K.G.M., N.B.F., B.K.S., J.M.B., G.E.P. and D.J.C. Project funding: K.G.M. Competing interests: There are two patents related to the PCL/GDNF conduit: 9,498,221 and 9,750,851. K.G.M. is the Founder of a new start-up company, AxoMax Technologies Inc. The authors declare no other competing interests. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.

Correction: Some sections of the Materials and Methods were inadvertently left out of the Supplementary Materials. The Supplementary Materials PDF was updated 4 February 2020.

Stay Connected to Science Translational Medicine

Navigate This Article