Exuberant sprouting of sensory and sympathetic nerve fibers in nonhealed bone fractures and the generation and maintenance of chronic skeletal pain

Abstract

Skeletal injury is a leading cause of chronic pain and long-term disability worldwide. While most acute skeletal pain can be effectively managed with nonsteroidal anti-inflammatory drugs and opiates, chronic skeletal pain is more difficult to control using these same therapy regimens. One possibility as to why chronic skeletal pain is more difficult to manage over time is that there may be nerve sprouting in non-healed areas of the skeleton that normally receive little (mineralized bone) to no (articular cartilage) innervation. If such ectopic sprouting did occur, it could result in normally nonnoxious loading of the skeleton being perceived as noxious and/or the generation of a neuropathic pain state. To explore this possibility, a mouse model of skeletal pain was generated by inducing a closed fracture of the femur. Examined animals had comminuted fractures and did not fully heal even at 90+ days post fracture. In all mice with nonhealed fractures, exuberant sensory and sympathetic nerve sprouting, an increase in the density of nerve fibers, and the formation of neuroma-like structures near the fracture site were observed. Additionally, all of these animals exhibited significant pain behaviors upon palpation of the nonhealed fracture site. In contrast, sprouting of sensory and sympathetic nerve fibers or significant palpation-induced pain behaviors was never observed in naïve animals. Understanding what drives this ectopic nerve sprouting and the role it plays in skeletal pain may allow a better understanding and treatment of this currently difficult-to-control pain state.

1. Introduction

Chronic skeletal diseases are prevalent and their impact on global health and disability is pervasive in both the developing and developed world. Importantly, if skeletal pain is not adequately controlled, it frequently has secondary effects, including the loss of mobility, bone and muscle mass, cardiovascular function, and cognitive health, all of which can significantly diminish the patient’s functional status and quality of life [12,46,73,99].

In general, the incidence of skeletal pain tends to increase with age. In large part, this is because the mass, quality, and strength of the human skeleton peaks at 25–30 years of age, in both males and females, and then declines thereafter [30,70]. Currently, the average lifespan of humans in both the developing and developed world is increasing to >80 years in several countries. Thus, the burden that skeletal pain will exact on individuals and society is expected to increase markedly in the coming decades [78,91]. Additionally, lifestyle influences such as rising obesity rates and reduction in daily physical activity also reduce skeletal health.

One major reason chronic skeletal pain is frequently debilitating is that we currently have very few therapies that can attenuate this pain without significant, unwanted side effects. While both nonsteroidal anti-inflammatory drugs (NSAIDs) and opiates have been shown to be effective in attenuating acute skeletal pain, long-term use of these agents for managing chronic nonmalignant skeletal pain is problematic [17,31]. Long-term NSAID use can induce significant renal, liver, and gastrointestinal toxicity and has also been suggested to inhibit bone healing [7,9]. In the last two decades there has been a marked increase in the use of opiates to treat a variety of chronic musculoskeletal conditions, but several recent systematic reviews suggest that opiates are more effective at controlling acute rather than chronic skeletal pain [67,92]. Opioids have been shown to be effective in reducing pain by approximately 30% in the initial 12 weeks following skeletal injury [5,25,44,63]. However, for patients experiencing skeletal pain lasting more than 12 weeks, opioid-induced pain relief frequently did not differentiate from placebo, and several studies suggest that long-term opioid therapy for nonmalignant skeletal pain has negative effects on the patient’s functional status [83,92].

In light of the data suggesting that NSAIDs and opiates are more effective at relieving acute vs chronic skeletal pain, an important question is whether there is a different anatomical substrate(s) that drives acute vs chronic skeletal pain. In the present report, we use a mouse model of nonmalignant fracture pain to explore if changes in the pattern of sensory and sympathetic innervation occur in the femur when rapid and effective bone healing does not occur.

2. Materials and methods

2.1. Animals

Experiments were performed with 40 adult male C57BL/6J mice (3–4 weeks old, Jackson Laboratories, Bar Harbor, ME) weighing 20 to 25 g. The mice were divided into 3 experimental groups: 1) naïve mice whose femur did not have anything done to it; 2) sham mice who had a metal pin placed in the intramedullary space but did not receive a fracture; 3) fractured mice who had both an intramedullary pin and a fracture. The mice were housed in accordance with the National Institutes of Health guidelines under specific pathogen-free conditions in autoclaved cages maintained at 22 °C with a 12-hour alternating light/dark cycle and access to food and water ad libitum. All procedures adhered to the guidelines of the Committee for Research and Ethical Issues of the International Association for the Study of Pain [102] and were approved by the Institutional Animal Care and Use Committee at the University of Arizona (Tucson, AZ, USA).

2.2. Pin placement surgery

Prior to surgical implantation of a pin into the mouse femur, an intraperitoneal injection of 100 mg/kg ketamine and 10 mg/kg xylazine was administered to provide a 20-minute period of deep anesthesia. The ventral portion of the left knee was shaved and the surgical field was prepared in a sterile manner with Betadine (Purdue Products, Stamford, CT, USA) and ethanol solution. An incision of approximately 10 mm was made in the skin with a number 11 blade. The vastus lateralis muscle was split with forceps to easily move the patellar ligament and give access to the distal head of the femur. A 30-gauge needle was used to core between the femoral condyles and into the medullary canal and immediately radiographed to ensure proper coring. Then a precut 0.011-inch-diameter and 11-mm-length stainless steel wire (Small Parts Inc., Miami Lakes, FL, USA) was inserted into the medullary space for fracture stabilization. Dental amalgam (Dentsply, Milford, DE, USA) was used to secure the implant and seal the hole. Following irrigation with sterile saline solution, the muscle was closed with quick absorbable suture (Ethicon, Somerville, NJ, USA), and the skin was closed with three simple interrupted 6–0 silk nonabsorbable sutures (Ethicon, Somerville, NJ, USA). Finally, topical bacitracin was applied around the incision and the mice received 1 mL of sterile saline solution. All surgeries were performed by the same individual. The mice were allowed a 21-day recovery period before the fracture was performed.

2.3. Fracture production

Closed mid-diaphyseal unilateral fractures of the left femur were performed 21 days after pin placement surgery. The animals were deeply anesthetized by an intraperitoneal injective mixture of xylazine and ketamine (10/1 mg/kg) before fractures were performed using a 3-point bending device (BBC Specialty Automotive Center, Linden, NJ, USA). The resulting fractures were primarily comminuted, an umbrella term used by the American Academy of Orthopedic Surgery for a nonsimple fracture composed of multiple bone fragments. Anterior-posterior radiographs were made of each mouse immediately post fracture to verify that a mid-diaphyseal fracture had been produced. Mice that met any of the exclusion criteria were immediately euthanized and were not used for this study. Exclusion criteria were adapted from Gerstenfeld and colleagues [26], and included fractures located too far from the mid-diaphyseal region of the femur, dislodged pins, and non-visible fracture (via radiograph) after impact. After recovery from anesthesia, mice were allowed unrestricted movement and limb weight-bearing.

2.4. Spontaneous and palpation-evoked pain behavior evaluation

Spontaneous pain-related behaviors are a measure of ongoing pain, whereas palpation-evoked pain-related behaviors are a measure of mechanical-induced or limb usage-induced pain. Normally nonnoxious palpation of a fractured limb is a common clinical criterion used to assess fracture healing by the presence or absence of pain. All mice were assessed for ongoing (spontaneous) and palpation-induced fracture pain behaviors on the day of euthanasia, which ranged from 85 days post fracture to 204 days post fracture. Mice were placed in small raised Plexiglass chambers (11.5 × 6.8 × 7.5 cm) with a wire grid floor and allowed to acclimate for 30 minutes (until cage exploration and major grooming activities ceased), after which their movements were videotaped from below using Sony Handycam DCR-SR68 cameras (Sony Electronics Inc., San Diego, CA, USA). Time spent in “nocifensive” behavior was assessed during a 5-minute observation period (between minutes 15 and 20 of the filmed behavior). All mice were observed by a researcher with extensive knowledge of mice pain behaviors and who was blinded to the experiment.

During the 5-minute behavioral recording period, the time spent in a variety of nocifensive behaviors were recorded for each mouse. These nocifensive behaviors included: guarding, bites, vocalizations, and fighting [33]. Additionally, flinches were recorded during the 5-minute observation period, though flinches were tabulated separately from nocifensive behaviors. Flinches were defined as brief raisings of the hind paw aloft while the mouse was not ambulatory [33].

Palpation-induced behavior assessments were developed to mirror the clinical criterion of using normally nonnoxious palpation to assess fracture healing based on the presence or absence of pain [8,16,49,74]. Palpation-induced nocifensive behaviors were provoked by manually palpating the distal femur of naïve or fractured mice over a 2-minute period [84]. The palpation-induced nocifensive behaviors and flinches were filmed and recorded in the same fashion as the spontaneous nocifensive behaviors and flinches.

2.5. Radiology

High resolution X-ray images of normal or fractured femurs were obtained several days before surgery, immediately before and immediately after fracture surgery, and immediately following behavioral assessments on the day of euthanasia, using a Faxitron MX-20 digital cabinet X-ray system (Faxitron/Bioptics, Wheeling, IL, USA). Mice were lightly anesthetized with ketamine/xylazine (0.005 mL/g, 50 mg/10 kg, subcutaneously) to enable consistent placement of the animal for radiological assessment. Faxitron settings were optimized for radiological assessment of cortical or trabecular bone destruction. The settings used were: 4.5-second exposure, 26 kV, sharp assist (4), contrast (1), center (5445), width (2584), and 4× magnification.

2.6. Preparation of tissue for immunohistochemistry and histology

At days ranging from 85 to 204 post fracture, mice were deeply anesthetized with CO₂ delivered from a compressed gas cylinder and perfused intracardially with 20 mL of 0.1 M phosphate-buffered saline (PBS, pH = 7.4 at 4 °C) followed by 30 mL of 4% formaldehyde/ 12.5% picric acid solution in 0.1 M PBS (pH 6.9 at 4 °C). After perfusion, the hind limbs were removed and postfixed for 24 hours in the same perfusion fixative solution. Following fixation, the femurs and tibias were separated by manually bending the knee to expose the patellar tendon. The patellar tendon and the lateral and medial collateral ligaments were scored with a size-11 scalpel, revealing the intramedullary space of the knee joint. The scalpel was then inserted in a vertical fashion into the intramedullary space, severing the anterior and posterior cruciate ligaments and pushing through the entire knee joint. The remaining muscle holding the tibia and femur together was then cut and excess muscle was removed from the femur, using care to preserve an even layer of approximately 3–5 mm of muscle on the entire bone. Following this dissection, the contralateral and ipsilateral (fracture) femurs were decalcified for approximately 2–4 weeks in 10% ethylenediaminetetraacetic acid (EDTA) (PBS, pH 7.4 at 4 °C). Every 5–7 days following initial submersion in EDTA, the EDTA solution was changed and decalcification was monitored radiographically with a Faxitron MX-20 digital cabinet X-ray system (Faxitron/Bioptics, Tucson, AZ, USA). Following total decalcification, each femur was cryoprotected in 30% sucrose at 4 °C for at least 48 hours before being sectioned.

2.7. Immunohistochemistry and histology

To characterize the innervation and surrounding microenvironment at late time points post fracture, we processed the murine femora immunohistochemically and histologically. Briefly, the femur was frozen, condyles of femur facing down, in a peel-away embedding mold (22 × 30 mm, Electron Microscopy Sciences, Hatfield, PA, USA) with a small amount (~0.5 cm) of optimal cutting temperature (OCT) embedding medium (Sakura Finetek USA Inc., Torrance, CA, USA). The femur was held in place with plastic tweezers, while the embedding mold was resting on a flat sheet of dry ice. Once the OCT surrounding the femur had turned white (indicating that it had frozen), the tweezers were removed and more OCT was added to the mold (approx. 1.5 cm total). The mold was then covered with a small piece of dry ice until the entire block solidified. Next, the block was placed in a −20 °C freezer for 20 minutes to bring the mold to a consistent temperature and ensure complete freezing. In order to cut the tissue, the mold was peeled away from the frozen OCT block containing the femur, and the block was frozen onto a dry-ice-chilled metal mounting plate using OCT. The top of the frozen block touched the surface of the mounting block. The block was then inserted into the cryostat (Bright OTF5000; Hacker Industries Inc., Winnsboro, SC, USA) so that the blade cut the distal area of the femur first (we have found that cutting from this area of the bone preserves the integrity of the tissue), and the orientation of the block was adjusted to ensure that the entire longitudinal bone section was captured. Sections were cut at 20 and 10 µm serially and thaw mounted, 2 sections of bone per gelatin-coated slide. Slides of 10-µm sections were stained with Safranin O and hematoxylin & eosin (H & E) for anatomical reference with X-rays and immunofluorescent staining.

In order to perform immunofluorescent staining, the slides were dried at room temperature (RT) for 30 minutes, washed in 0.1 MPBS 3 times for 10 minutes each (3 × 10), blocked with 3% normal donkey serum (Jackson ImmunoResearch, Cat# 017-11-121; West Grove, PA, USA) in PBS with 0.3% Triton-X 100 (Sigma Chemical Co., Cat # X100; St. Louis, MO, USA) for 60 minutes, and then incubated overnight with primary antibodies made in 1% normal donkey serum and 0.1% Triton-X 100 in 0.1 M PBS at RT. Peptide-rich sensory nerve fibers were labeled with an antibody against calcitonin gene-related peptide (CGRP; polyclonal rabbit anti-rat CGRP; 1:10,000; Cat #8198; Sigma Chemical Co.). Myelinated sensory nerve fibers were labeled with an antibody against neurofilament 200 kDa (NF200; chicken anti-NF200; 1:5000; Cat #CH22104; Neuromics, Edina, MN, USA). Sympathetic nerve fibers were identified using an antibody against tyrosine hydroxylase (TH; polyclonal rabbit anti-rat TH, 1:1000, Cat #AB152; Millipore, Temecula, CA, USA). Sprouted nerve fibers were labeled with an antibody against growth-associated protein 43 (GAP43; rabbit anti-GAP43, 1:1000, Cat #AB5220; Millipore, Temecula, CA, USA). Monocytes and macrophages were identified with an antibody against a myeloid glycoprotein (cluster of differentiation 68 [CD68]; rat anti-mouse CD68, 1:2000, Cat# MCA1957; AbD Serotec, Raleigh, NC, USA). (Note: see Table 1 for extended information on antibodies). After primary antibody incubation, preparations were washed 3×10 minutes each in PBS and incubated for 3 hours at RT with secondary antibodies conjugated to fluorescent markers (Cy3/Dylight 488/Dylight 649; 1:600/1:200/1:400; Jackson ImmunoResearch). Preparations were then washed 3×10 minutes each in PBS and dehydrated through an alcohol gradient (2 minutes each; 70%, 80%, 90%, and 100%), cleared in xylene (2×2 minutes), and coverslipped with di-n-buty-lphthalate-polystyrene-xylene (Sigma Chemical Co.). Preparations were allowed to dry at RT for 12 hours before imaging.

Table 1.

Table of primary antibodies used for immunological staining.

Antigen	Immunogen	Manufacturer, species raised in, mono/polyclonal, catalogue and lot number	Dilution used	Reference
CGRP	Synthetic rat Tyr-CGRP conjugated to keyhole limpet hemocyanin (37 amino acids)	Sigma Chemical Co., rabbit, polyclonal, Cat# C8198, Lot# 059K4841	1:10,000	[28,51,76]
NF200	NF-H isolated from bovine spinal cord (hyperphosphorylated, axonal form)	Neuromics, chicken, polyclonal, Cat#CH22104, Lot#400719	1:5000	[61,88]
TH	Tyrosine hydroxylase from rat phenochromocytoma denatured with sodium dodecyl sulfate (complete sequence)	Millipore, rabbit, polyclonal, Cat#AB152, Lot#2066692	1:1000	[77,97]
GAP43	Recombinant rat GAP43 (complete sequence; 266 amino acids)	Millipore, rabbit, polyclonal, Cat#AB5220, Lot#JBC1766780	1:1000	[47,90]
CD68	Purified Concanavalin A acceptor glycoprotein from P815 cell line	AbD Serotec, rat, monoclonal, Cat#MCA1957, Lot#0108	1:2000	[19,69]

CGRP, calcitonin gene-related peptide; NF, neurofilament; TH, tyrosine hydroxylase; GAP43, growth-associated protein 43; CD, cluster of differentiation.

2.8. Bright field and laser confocal microscopy

Bright field images of histologically stained sections were acquired using an Olympus BX51 microscope fitted with an Olympus DP70 digital CCD camera and an UPlanSApo 10×/0.40 objective. Confocal images were acquired with an Olympus Fluoview FV1000 (Olympus, Melville, NY, USA) system equipped with Multiline Argon (458, 488, 515 nm), Green HeNe (543 nm), Red HeNe (633 nm) lasers and with an Olympus Fluoview FV1200 system (Olympus) equipped with LD (405, 440, 473, 559, 635 nm), Multiline Argon (457, 488, 515 nm), and HeNe(G) (534 nm) lasers. Each Fluoview system was equipped with multiple excitation and emission filters. CGRP, NF200, TH, GAP43, and CD68 markers were visualized using excitation beams of 488, 599, and 635 nm, and emissions were detected using BA505-540, B575-620, and BA655-755 emission filters. Nuclear staining (4′,6-diamidino-2-phenylindole [DAPI]) was visualized using an excitation beam of 405 nm and emissions were detected using a BA430-470 emission filter. Sequential acquisition mode was used to reduce bleed-through from fluorophores. Images were obtained using Olympus UPlanApo 40×/1.30 and 60×/1.42 (FV1000) and UPlanFL N 40×/1.30 and PlanApo N 60×/1.42 (FV1200) oil objectives. The average volume of data that were collected was 211.7 µm × 211.7 µm × 40 µm, with each Z-axis slice being 1 µm/slice.

2.9. Nerve fiber density and neuroma quantification

For quantification, frozen sections were used, as cross-sectional analysis allowed for the visualization of the bone’s anatomy (such as the condyles and growth plate), which enabled the observer to locate the same anatomical area when quantifying different animals. The number of animals used for frozen section quantification was: n = 5 for naïve and n = 10 for nonhealed fracture. For each given marker (ie, CGRP, NF200, TH, and GAP43), three images were obtained for each of the 3 bone compartments: periosteum, bone marrow, and mineralized bone. Thus, a total of nine images were taken for each primary marker per bone. Nerve fibers were manually traced using ImageJ and added together for a total nerve fiber length (mm). To measure the surface area (mm²) of each tissue in each region, we analyzed the same tissue sections from which nerve fiber counts were obtained. The area of bone was measured from the frontal section of the femur and digital images were acquired using an Olympus FV100 and FV1200 confocal microscope (further described in “Materials and methods”). These measurements were used to determine the fiber density (mm²) for each of the 3 compartments: periosteum, mineralized bone, and bone marrow.

To quantify the extent of formation of neuroma-like structures, frozen sections were examined with a fluorescent microscope and these structures were manually counted and totaled from the entire 20-µm-thick section. Three different sections, each at least 0.1 mm apart, were evaluated per animal. A neuroma-like structure was defined as satisfying all three of the following characteristics: 1) a disordered mass of blind ending axons (CGRP-IR, NF200-IR, or TH-IR) that has an interlacing and/or whirling morphology; 2) a structure with a size of more than 10 individual axons that is at least 20 µm thick and 70 µm long; and 3) a structure that is never observed in a normal bone [10,20,65,93].

2.10. Three-dimensional rendering of confocal imaging

In addition to the 2D analysis described previously, sensory and sympathetic neurons were visualized using a commercially available software program, Filament Tracer (Imaris, Bitplane AG, South Windsor, CT, USA). Images were acquired using an Olympus FV1200 microscope and a 60×/1.42 PlanApo N objective to capture the innervation at the highest resolution possible. Optical slices were 211.7 µm × 211.7 µm in the XY and ranged from 30 to 36 µm in the Z (0.1 µm interval step slices). The raw data were saved as 16-bit, Olympus Image Binary files and were then imported into Imaris and further analyzed for detailed qualitative 3D characterization. Previously published parameters for the accurate, automated, detection and tracing of nerve fibers could not be found using Filament Tracer with NF200⁺ sensory and TH⁺ sympathetic nerve fibers. To generate the 3D reconstructions for nerve fiber tracing, we selected a region of the naive and nonhealed, fractured femur that displayed NF200⁺ and TH⁺ nerve fibers. Once the Olympus Image Binary file was imported into the Imaris software, the macro “Filaments” (denoted by a green leaf icon) was used to generate a 3D solid mask of the original confocal signal (Fig. 8A, B, D, E). The NF200 and TH confocal signal (ie, voxels) was used to define the spatial location of the nerve fibers. The dendrite starting point diameter was set at 10.0 µm and the dendrite seed point diameter was set at 0.500 µm. The starting point low threshold was 246.071, and the starting point high threshold was automatically determined by Imaris. The dendrite seed point threshold was 469.285. The Filaments tool then traced the nerve fibers and created a 3D cylindrical outline of the nerve fiber network (Fig. 8B, E). After volume rendering, the Imaris tool Snapshot was used to create images for publication purposes (Fig. 8). Volume-rendered images were also animated using the Animation tool in the Imaris software (Supplementary Video 1). No further image processing was done on images or animations.

Fig. 8.

Three-dimensional (3D) rendering of sensory and sympathetic nerve fibers in the marrow space of naïve and nonhealed, fractured femurs. High-power 2-dimensional confocal images of decalcified, frozen sections of the marrow space from (A) naïve and (D) nonhealed, fractured femurs, immunostained for neurofilament 200kD (NF200⁺) sensory (green) and tyrosine hydroxylase ([TH⁺] sympathetic, red) nerve fibers. The NF200 antibody was raised in chicken and is usually expressed in myelinated sensory fibers. The TH antibody was raised in rabbit and is expressed in sympathetic nerve fibers. Two-dimensional confocal images of NF200⁺ and TH⁺ nerve fibers (A, D) were rendered as 3D, real-diameter cylinders using Imaris imaging software in the (B) naïve and (E) nonhealed, fractured femur (further described in “Materials and methods”). 3D rendered images were magnified to qualitatively characterize NF200⁺ sensory and TH⁺ sympathetic nerve association in the naïve (C) and nonhealed, fractured femur (F). Note the interlaced, web-like appearance of the nerve fibers in the nonhealed, fractured femur (F). This close association is never observed in the naïve femur where NF200⁺ and TH⁺ nerve fibers are relatively rare in the bone marrow and are almost never found adjacent to each other. The great majority (70%, data not shown) of NF200⁺ nerve fibers also expressed calcitonin gene-related peptide, but these nerve fibers never colocalized with TH. Confocal images of the femur were projected from 36 optical sections at 0.1-µm intervals with a 60× objective.

2.11. Statistical analysis

A Mann-Whitney rank-sum test (or Wilcoxon rank-sum test) was used to compare ongoing (spontaneous) and palpation-induced nocifensive behaviors and flinches between naïve and fractured animal groups. Significance level was set at P < 0.05. A Student’s t-test was used to compare nerve fiber density of sensory and sympathetic neurons between naïve femurs and nonhealed fracture femurs. Significance was set at P < 0.05. In all cases, the investigator responsible for behavioral testing and plotting of data was blind to the experimental situation of each animal.

3. Results

3.1. Radiographic comparison of femurs from naive and fractured C57 animals

The femurs of age-matched naïve mice were radiographically similar in appearance at each time point examined (Fig. 1A). Naïve mice are defined as mice whose femur has had nothing done to it. After pin placement and surgical recovery, experimental fractures were performed at the mid-diaphysis of the femur. The 3-point fracture device primarily resulted in comminuted fractures. Comminuted is an umbrella term for a nonsimple fracture that is composed of multiple bone fragments (Fig. 1B). Radiographs were taken immediately after fracture to characterize fracture healing and then once per week until euthanasia (radiographs not shown). After adequate time was provided for fracture healing to occur [64], radiographs of fractured femurs revealed that appropriate healing of the bone had not occurred. Indeed, radiographic evaluation of the nonhealed, fractured femurs revealed aberrant bone remodeling following fracture, as indicated by web-like, radiopaque lines on the radiograph, the lack of adequate bone resorption, and the persistence of the initial fracture callus (Fig. 1C).

Fig. 1.

Radiographic images of the same femur at 3 time points: naïve (before fracture), immediately post fracture, and 90 days post fracture. (A) Anterior/ posterior (AP) view radiograph of a naïve mouse femur acquired immediately before fracture. (B) AP view radiograph of acute, comminuted femoral fracture acquired immediately after fracture with an intramedullary pin. As defined by the American Academy of Orthopedic Surgeons, comminuted is an umbrella term for a nonsimple fracture composed of multiple bone fragments. White arrows indicate fracture lines. Red outlined boxes indicate where bright field and confocal images for Figs. 4–7 were acquired. (C) AP view radiograph of nonhealed, comminuted fracture acquired 90 days post fracture (C). Note that in (B) there is an extensive series of fracture lines and in (C), the same bone 90 days after fracture, complete union of the mineralized bone has yet to occur and the partly mineralized, cartilaginous callus (yellow outline) has yet to be fully resorbed. It should be mentioned that the contrast settings used for (A), (B), and (C) were the same and were used specifically to enhance the visualization of the fracture callus.

3.2. Animals with nonhealed femoral fractures generated increased spontaneous and palpation-evoked pain-related behaviors compared to naïve animals

Spontaneous and palpation-evoked pain behaviors were evaluated on the day of euthanasia, which ranged from 85 days to 204 days post fracture, to assess the level of nocifensive pain in naïve (n = 3) and nonhealed, fractured (n = 14) animals. Spontaneous and palpation-evoked pain-related behaviors (nocifensive behaviors and flinching) were analyzed over a 5-minute period in naïve and nonhealed, fractured mice. Animals with nonhealed, fractured femurs spent a significantly greater time in spontaneous and palpation-evoked nocifensive behaviors (Fig. 2A) and had a significantly greater number of spontaneous and palpation-evoked flinches (Fig. 2B) compared to naïve animals. Naïve animals exhibited minimal spontaneous nocifensive behaviors and flinches (Fig. 2A, B) and had a slight increase (not statistically significant) in palpation-evoked flinches (Fig. 2B).

Fig. 2.

Comparison of spontaneous and palpation-induced pain behaviors (nocifensive and flinches) in naïve animals and animals with late-stage, nonhealed femoral fractures. Histograms demonstrating average time spent in spontaneous and palpation-evoked nocifensive behaviors (A) and flinches (B). Naïve mice (n = 3) exhibited basal levels of spontaneous and palpation-induced behaviors. Mice with nonhealed fractures (n = 14) exhibited a significant increase in the time spent in spontaneous and palpation-induced nocifensive behaviors, and had an increased number of flinches as compared to naïve animals. Note that palpation of the nonhealed, fractured femur resulted in almost a fivefold increase in the time spent in nocifensive behaviors. Number above histogram bars indicates the number of animals. Error bars represent the mean ± SEM; *P < 0.05 nonhealed fracture vs naïve.

3.3. Profuse sprouting and ectopic reorganization of sensory and sympathetic nerve fibers occurs in the marrow space and periosteum of the nonhealed, fractured femur

To investigate the extent of sensory and sympathetic nerve fiber innervation in the nonhealed fracture, decalcified, frozen bone sections from naïve and nonhealed, fractured mouse femurs were labeled with fluorescent antibodies raised against primary antibodies for: CGRP, a marker for peptide-rich C fibers and some Ad sensory nerve fibers; NF200, a marker for myelinated sensory nerve fibers; TH, a marker for sympathetic nerve fibers; and neuron GAP43, a marker for axonal remodeling and regeneration. Additionally, bone sections were labeled with DAPI, a marker that binds A-T rich regions of DNA that assisted the observer with anatomical orientation. Given the extent of aberrant bone remodeling in animals with nonhealed fractures, analysis of the sensory and sympathetic innervation in the fractured femur was not confined to the fracture site; innervation was characterized throughout the bone.

The density of sensory and sympathetic nerve fibers in naïve and nonhealed, fractured femurs were quantified in the marrow space (bone marrow and mineralized bone) and periosteum in 5 naïve femurs and in 10 nonhealed, fractured femurs. In the marrow space of nonhealed, fractured femurs, there was a significant increase in the density of CGRP⁺, NF200⁺, TH⁺, and GAP43⁺ (Fig. 3A–D, respectively) nerve fibers, compared to the naïve marrow space as measured as length-per-unit area. Similarly, there was a significant increase in the density of NF200⁺ sensory (Fig. 3B) and TH⁺ sympathetic (Fig. 3C) nerve fibers in the periosteum of nonhealed, fractured femurs compared to the naive periosteum.

Fig. 3.

Density of sensory and sympathetic nerve fibers in the marrow space and periosteum of naïve and nonhealed fractured femurs. Histograms demonstrating the density of (A) CGRP⁺, (B) NF200⁺, (C) TH⁺, and (D) GAP43⁺ nerve fibers per mm² area in the marrow space (bone marrow and mineralized bone) and periosteum. Note that in the marrow space and periosteum of the nonhealed fracture, there is an increase in nerve fiber density compared to the naive femur. In the marrow space there was a statistically significant increase in nerve fiber density of all markers observed in bone. Nerve fiber density was determined by measuring the total length of nerve fibers per unit area in the marrow space and periosteum. Numbers above histogram bars indicate the number of animals. Bars represent the mean ± SEM. *P < 0.05 nonhealed fracture vs naive. CGRP, calcitonin gene-related peptide; NF, neurofilament; TH, tyrosine hydroxylase; GAP43, growth-associated protein 43.

The immunostained slides of naive and nonhealed fractured femurs were also observed qualitatively to characterize the sensory and sympathetic innervation in the marrow space (bone marrow and mineralized bone) and periosteum. CGRP⁺ sensory nerve fibers exhibited remarkable sprouting in the marrow space and periosteum of the nonhealed, fractured femur compared to the same compartments in the naïve femur. In the naïve femur, CGRP⁺ sensory nerve fibers primarily ran along the long axis of the bone (in both the bone marrow and periosteum) and had a linear morphology (Fig. 4A, B). Additionally, CGRP⁺ nerve fibers in the naïve femur were predominately found in bundles consisting of no more than 1–3 separate nerve fibers. In marked contrast, CGRP⁺ innervation in the nonhealed, fractured femur was highly heterogeneous and often appeared as neuroma-like structures (see “Materials and methods” for description) (Fig. 4C, D). CGRP⁺ neuroma-like structures were never observed in the naïve femur.

Fig. 4.

Sprouting of CGRP⁺ sensory nerve fibers and the formation of neuroma-like structures in the nonhealed, fractured femur. High-power confocal images of decalcified, frozen sections of the bone marrow and periosteum of the (A, B) naïve femur and (C, D) fractured femur immunostained with calcitonin gene-related peptide (CGRP, red) antibodies raised in rabbit, a marker for peptide-rich C fibers and 4’,6-diamidino-2-phenylindole (DAPI, blue), a marker for DNA. In the naïve marrow space and periosteum, CGRP⁺nerve fibers ran along the long axis of the bone and were never found in bundles consisting of more than 1–3 nerve fibers. In the (C) bone marrow and (D) periosteum of fractured femurs, there was a substantial increase in CGRP⁺ innervation. Note the increased heterogeneity and density of CGRP⁺ nerve fibers in bone marrow and periosteum of fractured femurs compared to naïve femurs. Additionally, we observed neuroma-like structures (see definition in “Materials and methods”); these structures were never observed in the normal femur. Confocal images of the femur were projected from 40 optical sections at 1-µm intervals with a 60x objective.

Confocal images of TH⁺ sympathetic nerve fibers in the bone marrow of the naïve femur demonstrated a distinctive cylindrical morphology (Fig. 5A). In the periosteum of naïve femurs, TH⁺ nerve fibers ran along the long axis of the bone and intermittently exhibited the characteristic cylindrical morphology (Fig. 5B). TH⁺ sympathetic nerve fibers mirrored the pathological changes observed in CGRP⁺ sensory nerve fibers. In both the bone marrow and periosteum of nonhealed, fractured femurs, there was a profuse and exuberant sprouting of TH⁺ sympathetic nerve fibers (Fig 5C, D). Sprouted TH⁺ nerve fibers had a highly heterogeneous, random morphology and no longer displayed the characteristic cylindrical morphology as observed in the naive tissue. Additionally, sprouted TH⁺ fibers formed neuroma-like structures, which were never observed in the naïve femur.

Fig. 5.

TH⁺ sympathetic innervation in the naïve vs the nonhealed, fractured femur. High power images of decalcified, frozen sections of the femur from (A, B) naïve and (C, D) fractured mice immunostained with tyrosine hydroxylase (TH, yellow) antibodies raised in rabbit, a marker for sympathetic nerve fibers and 4′,6-diamidino-2-phenylindole (DAPI, blue), a marker for DNA. In naïve femurs, TH⁺ nerve fibers innervated the bone marrow and periosteum with a characteristic cylindrical morphology (A, B). In the nonhealed, fractured femur, TH⁺ nerve fibers no longer exhibit this distinctive morphology and instead exhibit a highly disorganized and random morphology resembling neuromas. These neuroma-like structures were not observed in naïve femurs. Confocal images of the femur were acquired from frozen sections and projected from 40 optical sections at 1-µm intervals with a 60 × objective.

3.4. Pathologic sprouting of nerve fibers and the aberrant formation of mineralized bone in the marrow space of nonhealed, fractured mice

As previously shown with CGRP⁺ sensory and TH⁺ sympathetic nerve fibers, NF200⁺ sensory innervation in the marrow space of naïve (Fig. 6A) femurs typically ran along the long axis of the femur and consisted of nerve bundles containing 1–3 nerve fibers (Fig. 6A). In marked contrast, NF200⁺ innervation of the marrow space in nonhealed, fractured femurs was highly disorganized (Fig. 6B) and had a significant increase in nerve fiber density per unit area (Fig. 3B) compared to the naïve bone. In addition to observing ectopic sensory and sympathetic innervation in the nonhealed fractured femur, radiographs of these same bones revealed aberrant bone remodeling in the marrow space. The same immunologically stained sections were stained histologically with H & E and Safranin O to better characterize this bone remodeling and the accompanying ectopic nerve sprouting. Safranin O histological staining of the naïve marrow space exhibited a homogenous environment (Fig. 6C). However, in the marrow of the nonhealed fractured femur, there was profuse and aberrant growth of mineralized bone in the normally homogenous marrow space (Fig. 6D). Importantly, the sensory and sympathetic neuroma-like structures in the nonhealed state were always observed immediately adjacent to the mineralized bone “pockets.”

Fig. 6.

Formation of neuroma-like structures and aberrant mineralized bone growth in the marrow space of nonhealed, fractured femurs. High-power confocal images of decalcified, frozen sections of the marrow space (bone marrow and mineralized bone) from the (A) naïve and (B) nonhealed fractured femur immunostained with neurofilament 200kD (NF200, green) antibodies raised in chicken, a marker for myelinated sensory nerve fibers and 4’,6-diamidino-2-phenylindole (DAPI, blue), a marker for DNA. There is a marked increase in NF200⁺ nerve fiber density in the nonhealed fracture compared to the naïve femur and the appearance of neuroma-like structures (B). Low power-bright field images of decalcified, frozen sections of the marrow space from the (C) naïve and (D) nonhealed fractured femur histologically stained with Safranin O, a histological stain or cartilage. The marrow space of a naïve femur is typically homogenous and appears as a violet structure (C). Note that in the marrow space of the nonhealed fracture, however, there is a profuse infiltration of mineralized bone, denoted by its teal stain and black arrows. Confocal images of the femur were acquired from frozen sections and projected from 40 optical sections at 1-µm intervals with a 60× objective.

GAP43⁺ innervation of the naïve marrow space had a linear morphology and primarily innervated the marrow as individual nerve fibers (Fig. 7A). In the marrow space of the nonhealed fracture, there was profuse, heterogeneous sprouting of GAP43⁺ nerve fibers and the formation of neuroma-like structures (Fig. 7B). Similar to Safranin O stained sections, H & E stained sections of the naïve femur revealed the marrow space to be homogenous and composed of hematopoietic and lymphoid tissue and adipose cells (Fig. 7C). However, the marrow space of non-healed, fractured femurs was highly heterogeneous and consisted of aberrant mineralized bone “pockets” and small amounts of cartilage. Furthermore, the neuroma-like structures of all observed immunological markers were always detected immediately adjacent to these “pockets.” Mineralized bone “pockets,” cartilage, and neuroma-like structures were never observed in the marrow space of naïve femurs. Interestingly, while there was a significant increase in sensory nerve fiber sprouting in the nonhealed fracture and a pathologic change in bone remodeling, macrophage levels remained the same between nonhealed fractured and naïve bones (Supplemental Figure 1).

Fig. 7.

Ectopic mineralized bone formation and sprouting of nerve fibers in the marrow space of nonhealed, fractured femurs. High-power images of decalcified, frozen sections of the femur from (A) naïve and (B) nonhealed fractured mice immunostained with neuron growth-associated protein 43 (GAP43, orange) antibodies raised in rabbit, a marker for axonal remodeling and regeneration and 4’,6-diamidino-2-phenylindole (DAPI, blue), a stain for DNA. Compared to naïve femurs, there was a significant increase in the density of GAP43⁺ nerve fibers in the bone marrow of nonhealed, fractured femurs. Low-power bright-field images of decalcified, frozen sections of the marrow space from the (C) naïve and (D) nonhealed fractured femur histologically stained with hematoxylin and eosin (H & E) for better visualization of the anatomical structure of the femur. The bone marrow is typically comprised of hematopoietic tissue, adipose cells, and lymphoid tissue, and appears as a homogenous, basophilic structure (C). Note that in the marrow space of the nonhealed fracture, mineralized bone (eosinophilic, denoted by black arrows) has infiltrated the normally homogenous bone marrow and is closely localized with sprouted neurofilament 200 kD (NF200⁺) nerve fibers (B, D). Additionally, the sprouted GAP43⁺ nerve fibers in the nonhealed fracture were found immediately adjacent to this mineralized bone and had a neuroma-like appearance. Confocal images of the femur were acquired from frozen sections and projected from 40 optical sections at 1-µm intervals with a 60x objective.

3.5. Sensory and sympathetic nerve fiber association in the marrow space of nonhealed, fractured femurs

As previously described above, NF200⁺ sensory nerve fibers had a linear morphology and ran along the long axis of the bone as individual fibers and TH⁺ sympathetic nerve fibers exhibited their distinctive cylindrical morphology; these nerve fibers had little association in the naïve state (Fig. 8A). In marked contrast, in the nonhealed state, NF200⁺ and TH⁺ nerve fibers in the marrow space of 2 out of 2 sectioned femurs simultaneously stained for NF200 and TH were found to have a strong association with each other (Fig. 8D). To further characterize this association, the 2D confocal images were imported into Imaris 3D rendering software, where the 2D confocal data were reconstructed into a real-diameter, 3D outline of the original confocal image (see “Materials and methods” for detailed description). Utilization of the Imaris 3D reconstruction allowed for more accurate characterization of the morphology of the sprouted NF200⁺ and TH⁺ nerve fibers. In addition to having a highly heterogeneous morphology, these sprouted neurons were found immediately adjacent to each other, exhibiting an interlaced, web-like morphology (Fig. 8F). This observation was never detected in the naïve marrow space where NF200⁺ sensory and TH⁺ sympathetic nerve fibers rarely associate with each other. Additionally, qualitative observation of NF200⁺ nerve fibers in the marrow space revealed NF200⁺ colocalization with CGRP, a peptide expressed by peptidergic, C-sensory fibers; approximately 70% of NF200⁺ fibers also expressed CGRP. TH expression was never colocalized with NF200⁺ or CGRP⁺ nerve fibers (data not shown) in the naïve state.

4. Discussion

4.1. Normal innervation of the bone: functional and nociceptive implications

In marked contrast to skin, which is richly innervated by a variety of sensory nerve fibers [2,45,87,95,103], bone and joint tissues are innervated primarily by a unique population of nociceptive nerve fibers (Ad and peptide-rich C fibers) [41]. Functionally, the specificity of innervation in the skeleton suggests that a deep tissue like bone requires less nociceptor “redundancy” for the detection of injury. Although naïve bone and joint tissues appear to be innervated by the same subpopulations of nociceptive nerve fibers, the density, pattern, and morphology of nerve fibers in each tissue is remarkably different [4,14,53,62]. Thus, in the young, healthy skeleton, there are no sensory nerve fibers in the articular cartilage, a very low density in mineralized bone, low to moderate density in bone marrow, and a very high density in the periosteum [14]. This sensory innervation of the skeleton appears to be tightly regulated in order to minimize sensory innervation in skeletal compartments that undergo significant mechanical stress and loading (ie, articular cartilage and mineralized bone) and maximize sensory innervation in compartments that do not undergo mechanical loading, but instead detect injurious mechanical distortion (ie, periosteum) [14,62,68].

4.2. Sensory and sympathetic innervation of the bone after fracture

Following fracture, movement of the bone can be remarkably painful. The initial pain probably arises primarily from mechanical distortion and activation of the normally silent mechanosensitive sensory nerve fibers in the periosteum expressing ion channels Piezo 2, transient receptor potential A1 channel (TRPA1), TRPV1, and TRPV4 [22,56,68,82]. Following the initial sharp, arresting pain, a hematoma is formed around the fracture site and inflammatory cells release immune mediators such as prostaglandins, bradykinins, histamine, and nerve growth factor [6,29,32,34,48,52,55,57,60] that sensitize nerve fibers present in and around the fracture site [35,66,81]. This sensitization is perceived as a dull, aching pain, with sharp increases in the pain if the periosteum undergoes further mechanical distortion.

Following injury to the young healthy bone there is a small but noticeable sprouting of sensory and sympathetic nerve fibers around the site of bone fracture [35,58,59,101]. As simple fractures rapidly heal in the young, healthy skeleton, these newly sprouted nerve fibers are pruned back to normal levels as the cartilaginous callus undergoes mineralization and then resorption. However, in the present study we show that when rapid and desired mineralization and resorption of the callus does not occur, profuse ectopic sprouting of both sensory and sympathetic nerve fibers occurs in the bone marrow space and periosteum and this ectopic sprouting remains 90+ days post fracture. Furthermore, we report that in nonhealed fractures with significant nerve sprouting, normally nonnoxious palpation of the bone is now perceived as noxious. As previous data have suggested that the density of sensory inner-vation of the skeleton is correlated to the severity of pain [37,71,72], the present data suggest that the presence of ectopic nerve sprouting in the facture callus, which is the primary load-bearing structure in the nonhealed fracture, may be contributing to use-dependent skeletal pain.

Previous studies have also noted that significant sprouting of nerve fibers occurs in the injured and/or diseased skeleton. Studies including osteoarthritis, temporomandibular joint disorder, and degenerative intervertebral disc (IVD) have also reported the sprouting of sensory and sympathetic nerve fibers into normally avascular, cartilaginous tissue [15,23,24,39,40,42,50,89,96]. Studies have also reported sprouting of CGRP⁺, Substance P⁺, and GAP43⁺ nerve fibers into the callus that forms around the bone fracture in rats [35,58,59] and in the synovium of arthritic joints of humans and mammals [13,94,100]. Furthermore, it is well documented in sarcoma, prostate, and breast cancers in bone that there is ectopic sprouting of sensory and sympathetic nerve fibers in the bone marrow, mineralized bone, and periosteum [10,27,38]. These sprouted nerve fibers, along the margin of the viable tumor, have a unique morphology, organization, and high density that is never observed in the normal bone and are similar to the ectopic sprouting of nerve fibers in the present study.

4.3. A permissive environment for nerve sprouting in the nonhealed fracture

Taken together, the previous studies suggest that the microenvironment in which sensory and sympathetic nerve fibers innervate plays an important role in maintaining proper afferent function. In our current model, we observed the close association of neuroma-like structures with mineralized bone “pockets” in the normally homogenous marrow space of nonhealed, fractured femurs. Disruption of a normally homogenous microenvironment can create a permissive environment allowing for the sprouting of sensory and sympathetic nerve fibers, and ultimately, the generation of chronic skeletal pain. Currently there is a deficiency of literature that addresses the specific cellular factors that maintain a nonpermissive or permissive environment for the presence and sprouting of sensory and sympathetic nerve fibers in the skeleton. However, a few potential mediators have been identified using an in vitro approach. The secreted cartilage-specific proteoglycan core protein aggrecan and a secreted class of semaphorins, Semaphorin 3A, have been implicated in inhibiting nerve ingrowth in the healthy IVD [43]. In models of degenerative IVD in vitro, Semaphorin 3A and aggrecan were expressed in reduced quantities and were associated with increased nerve ingrowth into normally aneural tissue [42,43,96]. On the other hand, both interleukin 1 and tumor necrosis factor alpha have been associated with the release of nerve growth factor, neurogenesis, and nerve sprouting in vitro in the painful IVD [1,3,23,50]. It remains to be determined what other factors are involved in regulating the ingrowth and maintenance of sensory and sympathetic nerve fibers in the in vivo skeleton.

4.4. Nerve sprouting and chronic skeletal pain

Given the above observations of ectopic sprouting in the injured or diseased skeleton, the major question is to what extent does the ectopic nerve sprouting observed in present and previous studies contribute to chronic skeletal pain? Previous data indicating that nerve fibers are injured in osteoarthritis, low back pain, and bones with metastatic tumors suggest that a component of malignant and nonmalignant skeletal pain can be neuropathic in origin [75,86]. These findings and data showing that a variety of different skeletal pains can be attenuated by gabapentin [11,21,80,98], which is approved for the treatment of neuropathic pain, suggest that injury to sensory nerve fibers may play a role in several types of skeletal pain [21,54].

Increased or aberrant sensory and sympathetic interaction may also be driving neuropathic pain in models of complex regional pain syndrome [18,36,79,85], a disorder that most commonly occurs following traumatic skeletal injury. In the present study, we detected the close association of NF200⁺ sensory and TH⁺ sympathetic innervation in the marrow space of nonhealed fractures, which we never observed in the normal mouse femur. Developing a better understanding of what drives this ectopic sprouting of sensory and sympathetic nerve fibers following injury to the skeleton may provide insight into the mechanisms that drive chronic skeletal pain and what needs to be targeted to attenuate this frequently difficult-to-control pain state.

4.5. Conclusions and limitations

In the present report, we show that comminuted bone fractures do not fully heal, there is ectopic sprouting and increased density of both sensory and sympathetic nerve fibers in the nonhealed fracture, and neuroma-like structures form near the fracture site. The ectopic nerve sprouting was often detected adjacent to mineralized bone “pockets” in the normally homogenous marrow space. Importantly, all of the animals with nonhealed fractures displayed both ectopic nerve sprouting and significant pain behaviors upon palpation of the nonhealed fracture. In contrast, similar aberrant nerve growth or palpation-induced pain behaviors were never observed in normal femurs. Understanding the functional significance of sensory and sympathetic nerve sprouting and determining whether nerve sprouting is a key participant in the transition from acute to chronic skeletal pain may provide insight into the mechanisms that generate and maintain skeletal pain.

There are several limitations concerning the present study. First, only one species and one load-bearing bone were examined. Second, we examined only fracture-induced skeletal pain models in bones of young adult mice, whereas bone fractures and nonhealing of bone is most common in older individuals where bone strength and rapid bone healing is diminished. Third, our model of fracture was induced using a 3-point bending device and can only be used to characterize a purely mechanical fracture. It remains unclear whether the model can be used to study other forms of fracture and orthopedic injury. Fourth, females generally have a lower peak bone mass than males, a higher incidence of osteoporosis, and more fragility fractures than males. Thus, even with the confounding issue of the estrous cycle potentially influencing pain behaviors, performing similar experiments in older animals and female animals will help us understand the effects that age and gender have on the nerve sprouting and desired bone healing following bone fracture.