© 2003 Nature Publishing Group http://www.nature.com/naturemedicine ARTICLES Multiple actions of systemic artemin in experimental neuropathy Luis R Gardell1, Ruizhong Wang1, Chris Ehrenfels2, Michael H Ossipov1, Anthony J Rossomando2, Stephan Miller2, Carolyn Buckley2, Amber K Cai2,3, Albert Tse2, Susan F Foley2, BangJian Gong2, Lee Walus2, Paul Carmillo2, Dane Worley2, Carol Huang2,3, Thomas Engber2, Blake Pepinsky2, Richard L Cate2, Todd W Vanderah1, Josephine Lai1, Dinah W Y Sah2 & Frank Porreca1 The clinical management of neuropathic pain is particularly challenging. Current therapies for neuropathic pain modulate nerve impulse propagation or synaptic transmission; these therapies are of limited benefit and have undesirable side effects. Injuries to peripheral nerves result in a host of pathophysiological changes associated with the sustained expression of abnormal pain. Here we show that systemic, intermittent administration of artemin produces dose- and time-related reversal of nerve injury–induced pain behavior, together with partial to complete normalization of multiple morphological and neurochemical features of the injury state. These effects of artemin were sustained for at least 28 days. Higher doses of artemin than those completely reversing experimental neuropathic pain did not elicit sensory or motor abnormalities. Our results indicate that the behavioral symptoms of neuropathic pain states can be treated successfully, and that partial to complete reversal of associated morphological and neurochemical changes is achievable with artemin. The glial cell line–derived neurotrophic factor (GDNF)-related family includes GDNF1, neurturin2, persephin3 and artemin (also known as neublastin or enovin)4,5. These factors form ternary complexes with the GDNF family receptor accessory proteins (GFR-α1, GFR-α2, GFR-α3 and GFR-α4), which each bind preferentially to a different ligand3,6–15, and the tyrosine kinase receptor RET (encoded by the RET proto-oncogene), which mediates signaling. Artemin supports survival of cultured sensory neurons12,13 apparently by interacting with GFR-α3, which is largely restricted in adults to the peripheral nervous system6,9,16. GFR-α3 is expressed by a subpopulation of unmyelinated sensory neurons, some or all of which are also immunoreactive for RET, trkA, calcitonin gene-related peptide (CGRP), peripherin and the vanilloid receptor VR-1, and bind isolectin B4 (IB-4)17. Neuropathic pain is characterized by persistent pain, hyperalgesia and allodynia18,19 and is poorly controlled by current therapies, severely impacting the patient’s quality of life. The trophic effect of artemin on sensory neurons and the restricted expression of GFR-α3 to nociceptive sensory neurons suggests that artemin might selectively impact the pathophysiologic mechanisms that promote pain, potentially normalizing behavioral pain and other changes associated with the neuropathic pain state. Here we show that intermittent, systemic artemin reverses the behavioral manifestation of neuropathic pain as well as the associated morphological and biochemical changes produced by nerve injury. Our findings point to the potential application of artemin as a treatment for multiple facets of human neuropathic pain. RESULTS Systemic artemin prevents and reverses neuropathic pain We assessed the effects of artemin on spinal nerve ligation (SNL)induced tactile and thermal hypersensitivity, with subcutaneous administration initiated at the time of surgery. No significant treatment effects were seen in sham-operated animals across the testing period (Fig. 1). SNL with concurrent artemin or vehicle produced significant (P ≤ 0.05) tactile and thermal hypersensitivity within 24 h (Fig. 1a,b). Only rats with SNL and artemin treatment exhibited a progressive reversal of tactile and thermal hypersensitivity; by days 10–14, their response thresholds remained significantly (P ≤ 0.05) and maximally elevated above post-SNL baseline (Fig. 1). We next examined whether systemic artemin could reverse established neuropathic pain. No significant treatment effects were seen in sham-operated rats across the 4-week testing period (Fig. 2a,b). In rats with SNL, hypersensitivity persisted with vehicle treatment but progressively reversed with artemin treatment such that by days 10–14, their response thresholds remained significantly (P ≤ 0.05) and maximally elevated above post-SNL baseline (Fig. 2a,b). After artemin cessation on day 14, hypersensitivity was reestablished; by days 18 and 21, tactile and thermal responses, respectively, were not significantly different from those of vehicle-treated controls (Fig. 2a,b). In contrast, 1Department of Pharmacology, University of Arizona Health Sciences Center, Tucson, Arizona 85724, USA. 2Biogen Inc., 14 Cambridge Center, Cambridge, Massachusetts 02142, USA. 3Present addresses: A.K.C. (11 Smith Hill Road, Lincoln, Massachusetts 01773, USA) and C.H. (874 Yorkchester Drive, Houston, Texas 77079, USA). Correspondence should be addressed to F.P. (frankp@u.arizona.edu). Published online 5 October 2003; doi:10.1038/nm944 NATURE MEDICINE VOLUME 9 | NUMBER 11 | NOVEMBER 2003 1383 © 2003 Nature Publishing Group http://www.nature.com/naturemedicine ARTICLES when artemin administration was continued through day 28, the artemin-induced reversal of tactile and thermal hypersensitivity was maintained (Fig. 2c,d). Some of the morphological and neurochemical changes induced by SNL do not manifest fully until day 10 after surgery20. In light of that fact, we studied the possibility that artemin could reverse the SNL-induced neuropathic pain state after full manifestation, by examining on day 26 the effects of starting artemin administration on day 14. This artemin treatment resulted in substantial and statistically significant (P ≤ 0.05) reversal of tactile threshold, from 3.22 ± 0.27 g on day 3 to 9.50 ± 2.75 g on day 26, in contrast to 3.92 ± 0.21 g on day 26 in vehicle-treated SNL rats. a Figure 1 Normalization of SNL-induced tactile and thermal hypersensitivity by systemic artemin, with administration beginning at the time of surgery. (a,b) Response thresholds to innocuous mechanical (von Frey filaments; a) and thermal (radiant heat; b) stimuli. Key in b also applies to a. Arrows indicate when each injection was administered. *, P ≤ 0.05 for SNL/artemin group compared with SNL/vehicle group. Data represent mean ± s.e.m.; n = 6–7 per group. BL, baseline. Systemic artemin reversal of neuropathic pain is dose-related We next explored the dose dependence of the artemin effect by subcutaneously administering graded doses of the protein (0.03, 0.1, 0.6, 1.0 and 2.0 mg/kg). SNL rats receiving artemin showed progressive reversal of injury-induced tactile and thermal hypersensitivity (Fig. 3a,b) with a clear dose-response relationship (Fig. 3c,d). The doses producing a 50% effect (A50 values) were 0.27 mg/kg (95% confidence interval, 0.16–0.45) and 0.12 mg/kg (95% confidence interval, 0.08–0.19) against tactile and thermal hypersensitivity, respectively. No alterations in sensory or motor thresholds in naive rats resulted from subcutaneous artemin at doses up to fivefold higher (2.0 and 5.0 mg/kg) than the maximally efficacious dose. No significant effects were observed in the tail flick, hot plate and rotarod tests during the first hour (measurements at 10-min intervals) and 48 h after artemin challenge at either dose (data not shown). Artemin normalizes evoked transmitter release post-SNL We evaluated the possible effects of artemin on SNL-induced functional neurochemical alterations by measuring capsaicin-induced CGRP release in the ipsilateral spinal dorsal quadrant. Recent work showed a significant enhancement of capsaicin-induced CGRP a release in tissues from SNL rats 10–14 d after surgery21. We assayed spinal cord tissue from rats that received sham or SNL surgery and vehicle or artemin treatment for capsaicininduced CGRP release. No differences in basal release of CGRP were found among the four Figure 2 Lack of tolerance to repeated artemin treatment in neuropathic pain. (a,b) Systemic artemin beginning 3 d after surgery reversed SNL-induced tactile (von Frey; a) and thermal (noxious heat; b) hypersensitivity, which was reestablished after cessation of artemin administration on day 14. Key in b also applies to a. Arrows indicate administration of artemin. (c,d) Systemic artemin given over 28 d maintained normalization of SNL-induced tactile (c) and thermal (d) hypersensitivity. Key in d also applies to c. *, P ≤ 0.05 compared with day 1. Data represent mean ± s.e.m.; n = 6–8 rats per group. BL, baseline. 1384 b c treatment groups. CGRP release induced by 1 µM capsaicin from tissues from vehicle-treated, sham-operated rats was normalized as 100 ± 32.6%, serving as the baseline control level (Fig. 4a). Capsaicininduced CGRP release from tissues of vehicle-treated SNL rats was significantly (P ≤ 0.05) elevated compared with tissues from sham-operated rats21 (Fig. 4a). Capsaicin-evoked CGRP release from tissues obtained from artemin-treated, sham-operated rats (normalized as 100 ± 20% to serve as the baseline control level) did not differ significantly from CGRP release in tissues from vehicle-treated, shamoperated rats. Capsaicin-induced CGRP release from tissues taken from artemin-treated SNL rats was not significantly elevated above sham-operated levels (Fig. 4a). Artemin normalizes spinal dynorphin post-SNL SNL is known to induce changes in spinal neurochemistry that are thought to be important in the maintenance of neuropathic pain. To determine whether systemic artemin could normalize these changes, we assayed spinal cords from vehicle- or artemin-treated, shamoperated or SNL rats for dynorphin content in the dorsal quadrant b d VOLUME 9 | NUMBER 11 | NOVEMBER 2003 NATURE MEDICINE ARTICLES © 2003 Nature Publishing Group http://www.nature.com/naturemedicine a b c d Figure 3 Dose-related reversal of neuropathic pain by artemin. (a,b) Response thresholds to innocuous mechanical (von Frey; a) and noxious thermal (radiant heat; b) stimuli in SNL rats that received various doses of artemin on the days indicated by the arrows. (c,d) Data were converted to percent reversal of tactile (c) and thermal (d) hypersensitivity to generate dose-response curves. *, P ≤ 0.05 compared with day 2. Data represent mean ± s.e.m.; n = 4–6 per group. BL, baseline. ipsilateral to the side of surgery at 14 d (ref. 20). SNL rats receiving vehicle showed a significant (P ≤ 0.05) elevation in spinal dynorphin content over sham-operated, vehicle-treated controls (Fig. 4b). Artemin treatment of SNL rats completely normalized spinal dynorphin content, which was not significantly different from that in artemin-treated, sham-operated rats (Fig. 4b). Prodynorphin immunoreactivity at the spinal lumbar level 5 (L5) was consistent with the dynorphin biochemical measurements (Fig. 5). Artemin normalizes SNL-induced morphological changes We used immunofluorescence to examine possible artemin normalization of SNL-induced changes in isolectin B-4 (IB-4) binding and purinergic receptor (P2X3), CGRP, substance P, galanin and neuropeptide Y (NPY) expression in ipsilateral L5 dorsal root ganglia (DRG), 14 d after surgery, from SNL or sham-operated rats receiving vehicle or artemin. These markers were unaltered by vehicle or artemin treatment in DRG neurons and ipsilateral spinal cords of sham-operated rats. In contrast, artemin partially normalized SNLinduced reductions in IB-4 binding and P2X3, CGRP and substance P expression, as well as SNL-induced increases in galanin and NPY expression (Fig. 5, Table 1 and Supplementary Figs. 1 and 2 online). Changes in the expression and distribution of sodium channels such as Nav1.8 in the injured peripheral nerve may contribute to neuropathic pain22. The possibility that artemin might normalize Nav1.8 migration to the injury site was explored by examining Nav1.8 immunoreactivity in the ipsilateral L5 DRG and sciatic nerve on day 14 after SNL. Artemin reduced both the decrease in Nav1.8-immunoreactivity in the DRG (Table 1) and the increase in Nav1.8 immunoreactivity in the sciatic nerve (Fig. 5). Vehicle or artemin was given to sham-operated or SNL rats through day 14 after surgery, after which treatment was stopped for 1 week. These studies assessed whether the immunohistochemical changes in the DRG would return after artemin was discontinued, concomitant with the return of pain. After artemin was discontinued, SNL-induced changes in IB-4 binding and P2X3, CGRP, substance P, galanin, NPY and Nav1.8 expression in DRG neurons and spinal cord all returned and were indistinguishable from those on post-SNL days 14 and 21 in vehicle-treated rats (Fig. 5, Table 1 and Supplementary Figs. 1, 2 and 3 online). We also assessed the expression of GFR-α3 and RET in ipsilateral L5 DRG from sham-operated or SNL rats receiving vehicle or artemin treatment. Previous work17 has shown that in the normal adult rat, virtually all GFR-α3-immunoreactive DRG neurons are also RET-immunoreactive, but only a small fraction of normal DRG neurons are GFR-α3-immunoreactive and would therefore respond NATURE MEDICINE VOLUME 9 | NUMBER 11 | NOVEMBER 2003 directly to artemin. SNL induced a substantial increase in the percentage of DRG neurons that were GFR-α3-immunoreactive in both vehicle- and artemin-treated rats (Fig. 5, Table 1 and Supplementary Fig. 1 online), suggesting that more sensory neurons can respond directly to artemin treatment after injury. RET immunoreactivity and distribution among small and large DRG neurons was not significantly affected23,24 by surgery or artemin treatment (Table 1 and Supplementary Fig. 1 online). To further define the broader subpopulation of DRG neurons that were GFR-α3-immunoreactive after SNL, DRG sections were doublelabeled for GFR-α3 and either IB-4 binding or substance P, CGRP or P2X3 immunoreactivity. GFR-α3 sensory neurons in normal adult rat DRG17 are small, correspond to the unmyelinated subpopulation, express RET and comprise the TrkA-immunoreactive, IB-4-negative sensory neurons, as well as the small, previously described24,25 population of TrkA-immunoreactive, IB-4-binding sensory neurons. Although cells that express TrkA and cells that express RET are mostly distinct subsets of sensory neurons, a significant percentage of cells express both RET and TrkA24, and it is this subset of sensory neurons that expresses GFR-α3 (ref. 17) in the normal adult rat. After SNL and vehicle treatment, in contrast to the normal adult rat, only a small fraction of these GFR-α3-immunoreactive neurons were substance P-immunoreactive or bound IB-4, and less than half of GFR-α3immunoreactive neurons were CGRP- or P2X3-immunoreactive (see Supplementary Table 1 and Supplementary Figs. 4 and 5 online). Artemin treatment of SNL rats restored IB-4 binding to most GFR-α3immunoreactive neurons and P2X3 expression to approximately half of GFR-α3-immunoreactive neurons, and partially restored substance P and CGRP expression to GFR-α3-immunoreactive neurons. a b Figure 4 Artemin normalizes SNL-induced neurochemical changes. (a,b) Systemic artemin normalized capsaicin-induced release of CGRP (a) and dynorphin (b) upregulation after SNL. Data represent mean ± s.e.m. of the percentage of induced release (a) or dynorphin content (b) obtained from sham-operated rats. *, P ≤ 0.05 compared with sham-operated rats; n = 7 per group. 1385 © 2003 Nature Publishing Group http://www.nature.com/naturemedicine ARTICLES substance P, galanin, NPY and Nav1.8 expression) in DRG neurons, on distribution of Nav1.8 in peripheral nerves and on spinal dynorphin content and capsaicin-induced CGRP release show that artemin treatment results in both peripheral and central normalization of the neuropathic pain state. The primary site of artemin action is the GFR-α3 subset of sensory neurons. SNL upregulated GFR-α3 expression, so that the majority of DRG neurons became GFR-α3-immunoreactive. In contrast, DRG RET expression did not change, similar to previous observations with sciatic axotomy24. Although technical limitations prevented double-labeling for GFR-α3 and RET, many of the GFR-α3-immunoreactive and RET-immunoreactive DRG neurons were of smaller size after SNL, suggesting a substantial overlap of the receptor components for artemin in many injured small DRG neurons. This suggests that multiple functional classes of injured sensory neurons may respond directly to and mediate the effects of artemin on neuropathic pain. Although we cannot exclude the possibility that large neurons expressed low levels of GFR-α3 after SNL, SNL most likely upregulated GFR-α3Figure 5 Systemic artemin normalized immunohistochemical markers in the ipsilateral L5 DRG, dorsal immunoreactivity in small neurons that conspinal cords and sciatic nerves of SNL rats. Immunofluorescent labeling was done 14 d after sham or comitantly lost IB-4 binding and substance P, SNL surgery and treatment with vehicle or artemin starting on day 3 after surgery. Scale bar = 100 µm CGRP and P2X3 immunoreactivity, which for DRG (first two rows), 50 µm for sciatic nerve (third row) and 200 µm for spinal cord (bottom row). were in part restored by artemin treatment. After SNL and artemin treatment, GFR-α3 Similarly, SNL led to a substantial decline in substance P, CGRP and expression remained upregulated. Intrathecal GDNF partly reversed P2X3 immunoreactivity and IB-4 binding in RET-immunoreactive axotomy-induced upregulation of DRG GFR-α1 and GFR-α3 (ref. neurons; artemin treatment fully restored CGRP and P2X3 24). Spinal nerve ligation permanently separates injured neurons from immunoreactivity and partially restored IB-4 binding and substance P their distal target tissues, which are a source of multiple neurotrophic immunoreactivity in those neurons (see Supplementary Table 1 and factors. Exogenous supply of one factor may not fully compensate for Supplementary Figs. 6 and 7 online). this, resulting in continued overexpression of neurotrophic factor DISCUSSION This study is the first demonstration of a systemically administered growth factor used to reverse and normalize experimental neuropathic pain. We found that artemin (i) produced a time- and doserelated reversal of nerve injury–induced tactile and thermal hypersensitivity, with a similar time course whether administered concurrently with nerve injury or after the establishment of injuryinduced pain; (ii) elicited a full reversal of nerve injury–induced sensory thresholds, an effect that persisted as long as artemin was administered (at least 28 d); (iii) did not produce sensory or motor disturbances at doses significantly higher than those relevant to the modification of neuropathic pain; (iv) partially or completely normalized multiple neurochemical and morphological consequences of nerve injury in both the peripheral and central nervous systems; and (v) reversed nerve injury–induced tactile hypersensitivity when treatment was initiated after full manifestation of the neuropathic pain state. Thus, systemic artemin not only produces symptomatic reversal of experimental neuropathic pain, but also exerts multiple actions on biochemical and morphological features associated with the neuropathic pain condition. The broad effects of systemic artemin administration on multiple immunohistochemical markers (IB-4 binding and P2X3, CGRP, 1386 Table 1 Administration and discontinuation of artemin changes the percentage of L5 DRG neurons immunoreactive for markers Marker Sham/vehicle Sham/artemin SNL/vehicle SNL/artemin GFR-α3 Discontinued 29.1 ± 3.4b 33.6 ± 1.3b 27.9 ± 2.3b 32.0 ± 4.1b 59.1 ± 5.7a 54.0 ± 4.3a 50.7 ± 4.3a 49.3 ± 3.1a RET Discontinued 58.9 ± 3.9 61.3 ± 4.3 51.5 ± 3.7 62.5 ± 15.2 56.9 ± 7.4 60.5 ± 5.3 54.7 ± 3.7 58.8 ± 2.8 IB-4 Discontinued 62.3 ± 2.4b 62.1 ± 1.9b 63.3 ± 1.4b 67.2 ± 3.0b 5.6 ± 0.5a 5.4 ± 1.4a 46.4 ± 5.9a,b 4.4 ± 0.8a P2X3 Discontinued 45.2 ± 0.8b 40.8 ± 3.6b 44.4 ± 0.9b 41.6 ± 4.4b 25.5 ± 2.1a 27.4 ± 1.8a 41.3 ± 0.7b 29.2 ± 3.0a CGRP Discontinued 49.0 ± 2.0b 49.9 ± 2.3b 48.0 ± 4.1b 50.4 ± 2.3b 27.0 ± 3.2a 30.6 ± 2.0a 37.9 ± 3.6a,b 31.1 ± 0.9a Substance P Discontinued 21.0 ± 1.2b 22.1 ± 1.5b 20.1 ± 2.0b 24.0 ± 1.4b 6.4 ± 0.4a 7.4 ± 0.5a 11.6 ± 0.9a,b 8.3 ± 0.7a Galanin Discontinued 6.8 ± 0.9b 6.6 ± 0.6b 5.5 ± 0.2b 6.2 ± 0.6b 42.8 ± 1.9a 40.1 ± 3.8a 17.5 ± 1.4a,b 38.7 ± 4.1a NPY Discontinued 0.1 ± 0.1b 0.0 ± 0.0b 0.0 ± 0.0b 0.1 ± 0.1b 32.1 ± 3.1a 35.4 ± 2.9a 17.2 ± 3.2a,b 33.5 ± 3.1a Nav1.8 Discontinued 61.5 ± 2.5b 61.8 ± 4.3b 60.1 ± 2.5b 61.9 ± 4.1b 38.3 ± 1.6a 37.1 ± 1.2a 53.6 ± 1.5a,b 40.3 ± 2.6a aP ≤ 0.05 compared with sham/vehicle. bP ≤ 0.05 compared with SNL/vehicle. VOLUME 9 | NUMBER 11 | NOVEMBER 2003 NATURE MEDICINE © 2003 Nature Publishing Group http://www.nature.com/naturemedicine ARTICLES receptors. Alternatively, higher artemin doses may be required for normalization of GFR-α3 expression. We found that artemin treatment partially reversed SNL-induced NPY upregulation in DRG neurons, and that NPY is upregulated exclusively in medium- to large-diameter sensory neurons after SNL26. This observation is consistent with an indirect effect of artemin on large sensory neurons, possibly through paracrine effects within the DRG27 or through central polysynaptic circuits. Our observations are consistent with a crucial role for small GFRα3-expressing sensory neurons in maintaining tactile hyperalgesia. Intrathecal substance P-SAP (saporin), which destroys neurokinin (NK)-1-positive cells in lamina I, could block SNL-induced tactile hypersensitivity28, showing that small fiber input is crucial. In addition, selective ‘knockdown’ of Nav1.8 with antisense oligodeoxynucleotides reversed established SNL-induced tactile allodynia, again suggesting a crucial role for small sensory fibers22,29. We found that SNL-induced potentiation of capsaicin-induced CGRP release was normalized by systemic artemin treatment in SNL rats. Again, technical limitations prevented double-labeling studies for CGRP and GFR-α3, but it seems likely that there is significant overlap between the GFR-α3-, RET- and CGRP-immunoreactive subpopulations of DRG neurons after SNL, as the majority of these immunoreactive neurons were of small size. Thus, artemin treatment may directly affect injured CGRP-containing neurons to modulate induced peptide release. In addition, adjacent segments (such as L4 sensory afferents) are likely to contribute to induced release. We also found that artemin normalizes spinal dynorphin content, providing an additional basis for the normalization of induced CGRP release. Increased CGRP release after SNL can also be prevented by experimental manipulations that prevent SNL-induced dynorphin upregulation, including lesions of the dorsolateral funiculus or deletion of rostral ventromedial medulla cells expressing µ-opioid receptors using a dermorphinsaporin construct, or by manipulations that prevent spinal dynorphin activity with antiserum to dynorphin21. The normalization of spinal dynorphin by systemic artemin administration is also notable because upregulation of spinal dynorphin has been consistently linked to the maintenance of neuropathic pain20,30,31. Substance P-SAP destruction of NK-1-expressing cells in lamina I (a predominantly spinoparabrachial projection) disrupts descending excitatory pathway(s) to the spinal dorsal horn, normalizing neuropathic pain32. As SNLinduced upregulation of spinal dynorphin depends on descending facilitation from the rostral ventromedial medulla30, it seems likely that artemin-induced normalization of spinal dynorphin may be an indirect effect arising from actions on small primary afferent fibers that innervate NK-1-expressing cells in the spinal dorsal horn. The fact that central changes associated with the neuropathic pain state are normalized by artemin suggests that these pathological changes in the central nervous system33,34 are secondary to pathophysiological changes in first-order sensory neurons, and require altered afferent input35,36 for their maintenance. The association between neuropathic pain behavior and neurochemical status was observed in SNL rats receiving artemin and exhibiting behavioral and partial immunohistochemical normalization, and after artemin discontinuation in SNL rats in which neuropathic pain behavior and immunohistochemical phenotype characteristic of nerve injury had returned. Although artemin treatment essentially normalized neuropathic pain behavior, there was only partial normalization of morphological markers; thus, these morphological changes do not correspond precisely to changes in pain behavior. One possible interpretation is that morphological changes do not represent the mechanism underlying neuropathic pain behavior. NATURE MEDICINE VOLUME 9 | NUMBER 11 | NOVEMBER 2003 Another possibility is that some morphological changes can be present without behavioral manifestation Our findings show that artemin normalizes multiple pathophysiological changes associated with the neuropathic pain state. Such a mechanism contrasts greatly with other systemically administered agents such as gabapentin, which produces only symptomatic relief, probably through calcium channel modulation and with a duration measured in hours37. Artemin’s effects are consistent with the finding that continuous intrathecal GDNF reverses the behavioral consequences of experimental neuropathic pain with a time course similar to that seen here38, decreases nerve injury–induced ectopic discharge in large myelinated afferents and normalizes sodium channel expression in DRG after injury. However, it is not clear whether dysregulation of sodium channels underlies neuropathic pain behavior, or whether other changes in gene transcription, translation or posttranslational modification may contribute. Ectopic firing in small, unmyelinated C fibers might also contribute substantially to the neuropathic pain state39, and it remains to be determined how artemin or GDNF might regulate the injury-induced spontaneous discharges in C fibers38. Nevertheless, we observed that artemin partially corrected the altered expression and distribution of Nav1.8 in sensory neurons, representing a potential mechanism of action. A similar effect was observed with antisense knockdown of Nav1.8, which prevented neuropathic behaviors and SNL-induced redistribution of this channel in the sciatic nerve22,29. Currently, few pharmacological agents are systemically effective against neuropathic pain. Opioids provide some relief but are limited by tolerance and unacceptable side effects40–42. Gabapentin is only effective in approximately half the patient population, provides modest pain relief and is limited by side effects including somnolence, dizziness and fatigue43. The robust effect of artemin on multiple facets of experimental neuropathic pain suggests that artemin can broadly normalize cellular pathology without impacting normal sensory and motor functions. Prolonged artemin treatment also resulted in persistent normalization of neuropathic pain behavior, which may translate to clinical benefit with chronic dosing. In addition, a delayed treatment paradigm, in which artemin administration began after neurochemical changes had fully manifested, showed that artemin can indeed reverse neuropathic pain. In clinical settings, drug treatment typically is not initiated until well after the onset of neuropathic pain, so the ability of a therapeutic to reverse established pain may translate to increased clinical utility. Finally, the restricted expression of the artemin receptor GFR-α3 to sensory neurons in the adult11,13,44 makes it an attractive target, as it reduces the likelihood of potential side effects. Taken together, our results indicate that systemic artemin may represent a new approach effective treatment of clinical neuropathic pain. METHODS Animal treatments and surgery. Male Sprague-Dawley rats (Harlan) were studied after approval from the Institutional Animal Care and Use Committee. Rat artemin (113 amino acids) was refolded from Escherichia coli inclusion bodies and purified to >98% homogeneity. Purified artemin migrated as a reducible dimer by SDS-PAGE and eluted as a single peak (24 kDa) by size exclusion chromatography and reverse-phase high-performance liquid chromatography. The purified product was confirmed to contain the characteristic cysteine-knot disulfide pattern seen in GDNF, and to be fully active in vitro by assaying receptor binding, cell-based c-RET kinase activation8 and sensory neuronal survival. Artemin (1 mg/kg) was injected subcutaneously on a Monday, Wednesday and Friday schedule beginning on day 3 after surgery, unless otherwise indicated. Nerve injury to the L5 and L6 spinal nerves was done according to the procedure of Kim and Chung45. Sham surgery was identical to SNL but without actual ligation. 1387 © 2003 Nature Publishing Group http://www.nature.com/naturemedicine ARTICLES Behavioral assays. Paw withdrawal latency to a noxious thermal stimulus was assessed by determining the response latency to noxious radiant heat projected onto the plantar aspect of the hindpaw. A maximal cutoff of 40 s prevented tissue damage. Tactile withdrawal thresholds were measured by probing the hindpaw with calibrated von Frey filaments (applied to the plantar surface of the paw using the up-down method as described46). Tactile behavior was confirmed in rats used for immunohistochemical and biochemical assessments. Nociceptive testing was done by determining tail withdrawal latency from a 52 °C water bath or response latency from a 55 °C hot plate. Cutoff latencies of 10 and 30 s for the tail flick and hot-plate tests, respectively, prevented tissue injury. Motor coordination was determined by the rotarod test (Columbus Instruments), which measured the duration of balance maintenance on a rod turning at 10 r.p.m. The cutoff latency for that test was 120 s. CGRP release assay. The evoked release of CGRP was done according to Vasko and colleagues47 and as described earlier48. Minced (0.2-mm) ipsilateral dorsal quadrant of spinal cord (relative to the side of surgery) was superfused at 0.5 ml/min with oxygenated modified Kreb’s buffer (37 °C, pH 7.4). After 45 min, superfusate was collected in 3-min intervals 15 min before the addition of capsaicin, for 6 min during infusion of 1 µM capsaicin and then for 27 min afterward. The tubes were preincubated with 100 µl of antibody to CGRP (Bachem/Peninsula,) for 24 h at 4 °C, mixed with 100 µl of [125ITyr0]CGRP(28–37) and 50 µl of goat antiserum to rabbit antibodies, coupled to ferric beads and incubated for 24 h. Bound [125I]CGRP was separated from free tracer through immunomagnetic separation (PerSeptive Diagnostics). Induced release (CGRP level above baseline) was expressed as the percentage of an appropriate within-group control. Data for each fraction were expressed as fmol CGRP per mg tissue and evaluated for deviation from normal distribution. We used ANOVA with Fisher Least Significant Difference test to determine when the level of CGRP release was significantly above basal (unstimulated) levels, and thus corresponded to induced release. Data were normalized to the appropriate sham-operated control group. Dynorphin assay. Spinal dynorphin content was measured in the ipsilateral dorsal quadrant of the spinal cord (relative to the side of surgery) as described30,31,48. Spinal tissue was homogenized in 1 N acetic acid, incubated for 30 min at 95 °C and centrifuged at 14,000 g for 20 min at 4 °C. Protein concentrations were determined with the bicinchoninic acid method. Immunoassays were done using a commercial enzyme immunoassay system for dynorphin A(1–17) (Bachem/Peninsula). Immunohistochemistry. Spinal tissue was fixed and stained as described30,48. Frozen lumbar sections (10 mm for DRG, 20 mm for spinal cord) were incubated with rabbit antibodies to CGRP (1:10,000; Bachem/Peninsula), galanin (1:10,000; Bachem/Peninsula), GFR-α3 R11 (2 µg/ml; ref. 17), Nav1.8 (1:1,000; ref. 22; gift from S. Novakovic, Roche Biosciences), NPY (1:10,000; Bachem/Peninsula), P2X3 (1:5,000; Neuromics), RET 1429 (2 µg/ml; ref. 49) or substance P (1:10,000; Bachem/Peninsula), or guinea-pig antibody to prodynorphin (1:10,000; Neuromics). Secondary antibodies were Cy3-conjugated goat antibody to rabbit IgG (1:1,000; Jackson ImmunoResearch Laboratories), rhodamine red X–conjugated donkey antibody to rabbit IgG (1:200; Jackson ImmunoResearch Laboratories) or Alexa Fluor 594–conjugated goat antibody to guinea-pig IgG (1:1,000; Molecular Probes). For IB-4 histochemistry, the sections were directly incubated in FITC-conjugated lectin IB-4 (1:1,000; Vector Labs). Stained cells and total cells (visualized with DAPI or ethidium bromide50) were counted on randomly selected DRG sections. Statistical analysis. Statistical comparisons between treatment groups were done using ANOVA followed by Fisher Least Significant Difference test. Pairwise comparisons were made with Student t-test. Significance was set at P ≤ 0.05. Dose-response curves were generated by converting responses to percentage of maximum possible effect (% MPE) using the equation % MPE = (test score – baseline score) ÷ (cutoff score – baseline score) × 100. Regression analyses of the log dose-response curves produced A50 values and 95% confidence intervals. Note: Supplementary information is available on the Nature Medicine website. 1388 ACKNOWLEDGMENTS The authors thank S. Burgess, L. Majuta, K. Vault and C. Zhong for technical assistance; M. McAuliffe, B. Coleman and C. Tonkin for DNA sequencing; R. Boynton, A. Kaffashan, D. Mo, D. Wen and C. Young for protein sequencing and characterization; and the Biogen Media Preparation Group. COMPETING INTERESTS STATEMENT The authors declare competing financial interests (see the Nature Medicine website for details). Received 24 July; accepted 13 September 2003 Published online at http://www.nature.com/naturemedicine/ 1. Lin, L.F., Doherty, D.H., Lile, J.D., Bektesh, S. & Collins, F. GDNF: a glial cell linederived neurotrophic factor for midbrain dopaminergic neurons. Science 260, 1130–1132 (1993). 2. Kotzbauer, P.T. et al. Neurturin, a relative of glial-cell-line-derived neurotrophic factor. Nature 384, 467–470 (1996). 3. Milbrandt, J. et al. Persephin, a novel neurotrophic factor related to GDNF and neurturin. Neuron 20, 245–253 (1998). 4. Masure, S. et al. Enovin, a member of the glial cell-line-derived neurotrophic factor (GDNF) family with growth promoting activity on neuronal cells. Existence and tissuespecific expression of different splice variants. Eur. J. Biochem. 266, 892–902 (1999). 5. Rosenblad, C. et al. In vivo protection of nigral dopamine neurons by lentiviral gene transfer of the novel GDNF-family member neublastin/artemin. Mol. Cell. Neurosci. 15, 199–214 (2000). 6. Worby, C.A. et al. Identification and characterization of GFR-alpha3, a novel coreceptor belonging to the glial cell line-derived neurotrophic receptor family. J. Biol. Chem. 273, 3502–3508 (1998). 7. Treanor, J.J. et al. Characterization of a multicomponent receptor for GDNF. Nature 382, 80–83 (1996). 8. Sanicola, M. et al. Glial cell line-derived neurotrophic factor-dependent RET activation can be mediated by two different cell-surface accessory proteins. Proc. Natl. Acad. Sci. USA 94, 6238–6243 (1997). 9. Naveilhan, P. et al. Expression and regulation of GFR-alpha3, a glial cell line-derived neurotrophic factor family receptor. Proc. Natl. Acad. Sci. USA 95, 1295–1300 (1998). 10. Jing, S. et al. GDNF-induced activation of the ret protein tyrosine kinase is mediated by GDNFR-alpha, a novel receptor for GDNF. Cell 85, 1113–1124 (1996). 11. Baloh, R.H. et al. GFR-alpha3 is an orphan member of the GDNF/neurturin/persephin receptor family. Proc. Natl. Acad. Sci. USA 95, 5801–5806 (1998). 12. Baloh, R.H. et al. TrnR2, a novel receptor that mediates neurturin and GDNF signaling through Ret. Neuron 18, 793–802 (1997). 13. Baloh, R.H. et al. Artemin, a novel member of the GDNF ligand family, supports peripheral and central neurons and signals through the GFR-alpha3-RET receptor complex. Neuron 21, 1291–1302 (1998). 14. Enokido, Y. et al. GFR- alpha-4 and the tyrosine kinase Ret form a functional receptor complex for persephin. Curr. Biol. 8, 1019–1022 (1998). 15. Klein, R.D. et al. A GPI-linked protein that interacts with Ret to form a candidate neurturin receptor. Nature 387, 717–721 (1997). 16. Widenfalk, J., Tomac, A., Lindqvist, E., Hoffer, B. & Olson, L. GFR-alpha3, a protein related to GFR-alpha1, is expressed in developing peripheral neurons and ensheathing cells. Eur. J. Neurosci. 10, 1508–1517 (1998). 17. Orozco, O.E., Walus, L., Sah, D.W., Pepinsky, R.B. & Sanicola, M. GFR-alpha3 is expressed predominantly in nociceptive sensory neurons. Eur. J. Neurosci. 13, 2177–2182 (2001). 18. Payne, R. Neuropathic pain syndromes, with special reference to causalgia and reflex sympathetic dystrophy. Clin. J. Pain 2, 59–73 (1986). 19. Merskey, H. & Bogduk, N. 2nd edn. (eds. Merskey, H. & Bogduk, N.) Classifications of Chronic Pain: Descriptions of Chronic Pain Syndromes and Definitions of Pain Terms 40–43 (IASP Press, Seattle, 1994). 20. Malan, T.P. et al. Extraterritorial neuropathic pain correlates with multisegmental elevation of spinal dynorphin in nerve-injured rats. Pain 86, 185–194 (2000). 21. Gardell, L.R. et al. Increased evoked excitatory transmitter release in experimental neuropathy requires descending facilitation. J. Neurosci. 23, 8370–8379 (2003). 22. Lai, J. et al. Inhibition of neuropathic pain by decreased expression of the tetrodotoxin-resistant sodium channel, NaV1.8. Pain 95, 143–152 (2002). 23. Bennett, D.L. et al. A distinct subgroup of small DRG cells express GDNF receptor components and GDNF is protective for these neurons after nerve injury. J. Neurosci. 18, 3059–3072 (1998). 24. Bennett, D.L. et al. The glial cell line-derived neurotrophic factor family receptor components are differentially regulated within sensory neurons after nerve injury. J. Neurosci. 20, 427–437 (2000). 25. Molliver, D.C., Radeke, M.J., Feinstein, S.C. & Snider, W.D. Presence or absence of TrkA protein distinguishes subsets of small sensory neurons with unique cytochemical characteristics and dorsal horn projections. J. Comp. Neurol. 361, 404–416 (1995). 26. Ossipov, M.H. et al. Selective mediation of nerve injury-induced tactile hypersensitivity by neuropeptide Y. J. Neurosci. 22, 9858–9867 (2002). 27. Acheson, A. & Lindsay, R.M. Non target-derived roles of the neurotrophins. Philos. Trans. R. Soc. Lond. B Biol. Sci. 351, 417–422 (1996). VOLUME 9 | NUMBER 11 | NOVEMBER 2003 NATURE MEDICINE © 2003 Nature Publishing Group http://www.nature.com/naturemedicine ARTICLES 28. Nichols, M.L. et al. Transmission of chronic nociception by spinal neurons expressing the substance P receptor. Science 286, 1558–1561 (1999). 29. Porreca, F. et al. A comparison of the potential role of the tetrodotoxin-insensitive sodium channels, PN3/SNS and NaN/SNS2, in rat models of chronic pain. Proc. Natl. Acad. Sci. USA 96, 7640–7644 (1999). 30. Burgess, S.E. et al. Time-dependent descending facilitation from the rostral ventromedial medulla maintains, but does not initiate, neuropathic pain. J. Neurosci. 22, 5129–5136 (2002). 31. Wang, Z. et al. Pronociceptive actions of dynorphin maintain chronic neuropathic pain. J. Neurosci. 21, 1779–1786 (2001). 32. Suzuki, R., Morcuende, S., Webber, M., Hunt, S.P. & Dickenson, A.H. Superficial NK1-expressing neurons control spinal excitability through activation of descending pathways. Nat. Neurosci. 5, 1319–1326 (2002). 33. Porreca, F., Ossipov, M.H. & Gebhart, G.F. Chronic pain and medullary descending facilitation. Trends Neurosci. 25, 319–325 (2002). 34. Ossipov, M.H., Lai, J., Malan, T.P., Jr., Vanderah, T.W. & Porreca, F. Tonic descending facilitation as a mechanism of neuropathic pain. in Neuropatic Pain: Pathophysiology and Treatment (eds. Hansson, P.T., Fields, H.L., Hill, R.G. & Marchettini, P.) 107–124 (IASP Press, Seattle, 2001). 35. Devor, M. & Seltzer, Z. Pathophysiology of damaged nerves in relation to chronic pain. in Textbook of Pain (eds. Wall, P.D. & Melzack, R.) 129–164 (Churchill Livingstone, London, 1999). 36. Liu, X., Eschenfelder, S., Blenk, K.H., Janig, W. & Habler, H. Spontaneous activity of axotomized afferent neurons after L5 spinal nerve injury in rats. Pain 84, 309–318 (2000). 37. Vollmer, K.O., von Hodenberg, A. & Kolle, E.U. Pharmacokinetics and metabolism of gabapentin in rat, dog and man. Arzneimittelforschung 36, 830–839 (1986). 38. Boucher, T.J. et al. Potent analgesic effects of GDNF in neuropathic pain states. Science 290, 124–127 (2000). 39. Wu, G. et al. Early onset of spontaneous activity in uninjured C-fiber nociceptors after NATURE MEDICINE VOLUME 9 | NUMBER 11 | NOVEMBER 2003 injury to neighboring nerve fibers. J. Neurosci. 21, RC140 (2001). 40. McQuay, H.J. et al. Opioid sensitivity of chronic pain: a patient-controlled analgesia method. Anaesthesia 47, 757–767 (1992). 41. Portenoy, R.K., Foley, K.M. & Inturrisi, C.E. The nature of opioid responsiveness and its implications for neuropathic pain: new hypotheses derived from studies of opioid infusions. Pain 43, 273–286 (1990). 42. Rowbotham, M.C., Reisner-Keller, L.A. & Fields, H.L. Both intravenous lidocaine and morphine reduce the pain of postherpetic neuralgia. Neurology 41, 1024–1028 (1991). 43. Rose, M.A. & Kam, P.C. Gabapentin: pharmacology and its use in pain management. Anaesthesia 57, 451–462 (2002). 44. Trupp, M., Raynoschek, C., Belluardo, N. & Ibanez, C.F. Multiple GPI-anchored receptors control GDNF-dependent and independent activation of the c-Ret receptor tyrosine kinase. Mol. Cell. Neurosci. 11, 47–63 (1998). 45. Kim, S.H. & Chung, J.M. An experimental model for peripheral neuropathy produced by segmental spinal nerve ligation in the rat. Pain 50, 355–363 (1992). 46. Chaplan, S.R., Bach, F.W., Pogrel, J.W., Chung, J.M. & Yaksh, T.L. Quantitative assessment of tactile allodynia in the rat paw. J. Neurosci. Methods 53, 55–63 (1994). 47. Chen, J.J., Barber, L.A., Dymshitz, J. & Vasko, M.R. Peptidase inhibitors improve recovery of substance P and calcitonin gene-related peptide release from rat spinal cord slices. Peptides 17, 31–37 (1996). 48. Gardell, L.R. et al. Sustained morphine exposure induces a spinal dynorphin-dependent enhancement of excitatory transmitter release from primary afferent fibers. J. Neurosci. 22, 6747–6755 (2002). 49. Ehrenfels, C.W., Carmillo, P.J., Orozco, O., Cate, R.L. & Sanicola, M. Perturbation of RET signaling in the embryonic kidney. Dev. Genet. 24, 263–272 (1999). 50. Guo, A., Vulchanova, L., Wang, J., Li, X. & Elde, R. Immunocytochemical localization of the vanilloid receptor 1 (VR1): relationship to neuropeptides, the P2X3 purinoceptor and IB-4 binding sites. Eur. J. Neurosci. 11, 946–958 (1999). 1389