Many forms of arthritis are accompanied by significant chronic joint pain. This study was undertaken to investigate whether there is significant sprouting of sensory and sympathetic nerve fibers in the painful arthritic knee joint and whether nerve growth factor (NGF) drives this pathologic reorganization.
A painful arthritic knee joint was produced by injection of Freund's complete adjuvant (CFA) into the knee joint of young adult mice. CFA-injected mice were then treated systemically with vehicle or anti-NGF antibody. Pain behaviors were assessed, and at 28 days following the initial CFA injection, the knee joints were processed for immunohistochemistry analysis using antibodies against calcitonin gene–related peptide (CGRP; sensory nerve fibers), neurofilament 200 kd (NF200; sensory nerve fibers), growth-associated protein 43 (GAP-43; sprouted nerve fibers), tyrosine hydroxylase (TH; sympathetic nerve fibers), CD31 (endothelial cells), or CD68 (monocyte/macrophages).
In CFA-injected mice, there was a significant increase in the density of CD68+ macrophages, CD31+ blood vessels, and CGRP+, NF200+, GAP-43+, and TH+ nerve fibers in the synovium, as well as a significant increase in joint pain–related behaviors. None of these findings were observed in sham-injected mice. Administration of anti-NGF reduced these pain-related behaviors and the ectopic sprouting of nerve fibers, but had no significant effect on the increase in density of CD31+ blood vessels or CD68+ macrophages.
These findings demonstrate that ectopic sprouting of sensory and sympathetic nerve fibers occurs in the painful arthritic joint and may be involved in the generation and maintenance of arthritic pain.
Although arthritis is the most common musculoskeletal disorder in the world, affecting ∼50 million people in the US alone (1), our understanding of and ability to treat arthritic joint pain and disease remains strikingly poor. A number of factors have frustrated efforts to understand what drives arthritic joint pain, including conflicting observations in epidemiologic studies, protracted disease duration, and the poor correlation between joint damage and arthritic pain (2). Currently, there is a notable lack of well-tolerated and effective analgesic therapies (3–5). Compounding these difficulties, human tissue used for experimental analyses is typically obtained at joint replacement surgery from patients with advanced disease, thereby limiting insight into the mechanisms that contribute to the development and maintenance of chronic arthritic joint pain in humans.
Arthritic joint disease presents a variety of symptoms that affect the use and function of the joint, including stiffness, inflammation, swelling, and pain (6). From the perspective of most patients, the symptom that has the greatest impact on their functional status is joint pain (1). Due to the dearth in understanding of the specific mechanisms that drive arthritic joint pain, there is currently no unifying theory as to what drives this pain, and what key processes need to be targeted to effectively attenuate it.
In the present study we used a mouse model of a painful arthritic joint, induced by injecting Freund's complete adjuvant (CFA) into the articular space of the knee joint of young adult mice. This procedure generates highly robust and reproducible pain-related behaviors, including increased flinching as well as reduced weight bearing and use of the arthritic joint. Using this model we tested the hypothesis that joint injury/inflammation is followed by release of neurotrophic factors that induce a robust and abnormal sprouting of sensory and sympathetic nerve fibers in the arthritic joint. These newly sprouted nerve fibers are not only present in a higher density per unit area than is found in the normal joint, but they are also present in inappropriate areas (i.e., synovium, meniscus, in the space normally occupied by articular cartilage) so that normally non-noxious loading of the joint will now be perceived as painful. This hypothesis would suggest that the extent of joint destruction alone will not predict the frequency and severity of pain, as a significant component of chronic arthritic joint pain would be maintained by an active and pathologic neurochemical and/or morphologic remodeling of sensory and sympathetic nerve fibers in the painful joint.
MATERIALS AND METHODS
Experiments were performed on a total of 40 adult male C57BL/6J mice (The Jackson Laboratory), initially at 8 weeks of age, weighing 20–25 gm. 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 were given autoclaved food and water ad libitum. All procedures were approved by the Institutional Animal Care and Use Committee at the Minneapolis Veterans Administration Medical Center.
A modified version of a previously validated model of arthritic inflammation of the knee joint (7) was produced by administering a single intraarticular injection of CFA (5 μg in 10 μl) every 7 days, over a period of 28 days (4 injections total). Briefly, mice were anesthetized using 2–3% isoflurane mixed with air. Saline solution (sham-injected animals) or CFA was injected using a 30-gauge ½-inch needle that was fitted with cannulation tubing such that only 2.5 mm of the needle could puncture the joint. Ten microliters of CFA or saline was injected through the patellar ligament into the articular space, using the femoral condyles as a guide.
Mice were divided into 4 groups (10 mice per group): naive, sham-injected, CFA plus vehicle–injected (vehicle [saline solution] administered intraperitoneally on days 0, 5, 10, 15, and 25 after the initial injection of CFA), and CFA plus anti–nerve growth factor (anti-NGF) (anti-NGF administered intraperitoneally on days 0, 5, 10, 15, and 25 after the initial injection of CFA).
Treatment with anti-NGF.
The anti-NGF sequestering antibody (monoclonal antibody 911; kindly provided by Dr. David Shelton, Rinat/Pfizer, San Francisco, CA), is effective in blocking the binding of NGF to both TrkA and p75 NGF receptors and inhibiting TrkA autophosphorylation (8). The anti-NGF antibody has a plasma half-life of 5–6 days in the mouse, and it does not cross the blood–brain barrier to an appreciable extent (9). The dose used (10 mg/kg, administered intraperitoneally) was selected due to its reported efficacy in attenuating skeletal pain in rodents (9). Anti-NGF was administered within 4 hours after initial intraarticular CFA injection and additionally on days 5, 10, 15, and 25 after initial CFA injection.
Behavioral measures of arthritic joint pain.
Behavioral measures of arthritic joint pain, including spontaneous pain (flinching) and stimulus-evoked pain (limb use and rotorod analysis), were performed on day 0 and 3, 10, 17, and 24 days after the initial CFA injection, and the ability of the animal to place weight on the arthritic limb versus the nonarthritic limb (dynamic weight bearing) was assessed on day 0, and 4, 9, 16, and 23 days after initial CFA injection. At least 6 animals were used in each behavioral experiment group.
The number of spontaneous flinches, representative of spontaneous nocifensive behavior, was recorded over a 2-minute observation period. Number of flinches was defined as the number of times the animal raised its hind paw.
Normal limb use during spontaneous ambulation over a 2-minute period in an open field was used as an indicator of stimulus-evoked pain. Limb use was scored on a scale of 5 to 0 (5 = normal use; 4 = partial limp, but not pronounced; 3 = pronounced limp; 2 = limp and guarding behavior; 1 = partial nonuse of limb in locomotor activity; 0 = complete lack of limb use).
Forced ambulation was determined during a 2-minute period using rotorod analysis (Columbus Instruments). The animals were placed on the rod using X2 speed, 8.0 acceleration, and 2.5 sensitivity control settings, and the score was recorded using the 5–0 limb use scale described above.
Dynamic weight bearing analysis of the arthritic limb was performed using a Dynamic Weight Bearing device (Bioseb), as similar pedobarographic analysis is used clinically to determine weight bearing in patients with arthritic knee joints (10). Using a synchronized video recording of the 5-minute test and a scaled map of the detected zones, each detection was validated by an observer and identified as a left or right and fore or hind paw.
Mice were deeply anesthetized by carbon dioxide asphyxiation, delivered using a compressed gas cylinder, on day 28 after the initial CFA injection, and 20 ml of 0.1M phosphate buffered saline (PBS; pH 7.4 at 4°C) was perfused intracardially, followed by 30 ml of 4% formaldehyde/12.5% picric acid solution in 0.1M PBS (pH 6.9 at 4°C). Ipsilateral and contralateral knee joints were harvested following perfusion and postfixed for at least 12 hours in the perfusion fixative. Postfixing, decalcification, and sectioning of the bones/joints were performed as previously described (11).
To qualitatively and quantitatively assess changes in the density and morphology of nerve fibers that innervate the knee joint, macrophage infiltration, and aberrant neovascularization of the synovium, 20-μm–thick frozen sections of the bone/joint were processed according to procedures we have described previously (11). Frozen bone/knee joint sections were incubated with an antibody against calcitonin gene–related peptide (CGRP) (polyclonal rabbit anti-rat CGRP; 1:10,000) (catalog no. C8198; Sigma) to label unmyelinated and thinly myelinated primary afferent sensory nerve fibers, and an antibody against neurofilament 200 kd (NF200) (chicken anti-NF200; 1:5,000) (catalog no. CH22104; Neuromics) to label myelinated primary afferent sensory nerve fibers. Sympathetic nerve fibers were labeled with an antibody against tyrosine hydroxylase (TH) (polyclonal rabbit anti-rat TH; 1:1,000) (catalog no. AB152; Chemicon). Sprouted nerve fibers were labeled with an antibody against growth-associated protein 43 (GAP-43) (rabbit anti–GAP-43; 1:1,000) (catalog no. AB5220; Millipore). Blood vessels were labeled with an antibody against platelet endothelial cell adhesion molecule (rat anti-mouse CD31; 1:500) (catalog no. 550274; BD PharMingen). Monocyte/macrophages were identified with an antibody against a myeloid glycoprotein (rat anti-mouse CD68; 1:2,000) (catalog no. MCA1957; AbD Serotec). Additionally, 4–5 sequential frozen bone sections from animals in each experimental group were cut (10 μm–thick) and stained with hematoxylin and eosin to visualize gross pathologic changes induced by CFA.
Quantification of nerve fiber density, sprouting, macrophage infiltration, and neovascularization.
Approximately 30 separate 20-μm–thick frozen sections were obtained from each knee joint. Three confocal images (Olympus, software version 5.0) from different sections separated by at least 100 μm were obtained for each marker. Sections were initially scanned at low power (×100) to identify areas with the highest capillary or nerve fiber density in the synovium (hot spots), and 1 image per section was acquired within the medial synovial hot spot. While nerve fibers and blood vessels were observed throughout the inflamed synovium, neovascularization and nerve sprouting were consistently present in the synovium adjacent to the meniscus, and therefore most of the hot spots were found in this area.
The average volume of the CFA-inflamed synovial specimens analyzed was 315 μm (length) × 315 μm (width) × 20 μm (depth). The z-stacked images were analyzed with Image-Pro Plus, version 6.0 (Media Cybernetics), and nerve fibers and blood vessels were manually traced to determine their length within each tissue section. Nerve sprouting or neovascularization was reported as density of nerve fibers or blood vessels per volume of synovium (mm/mm3) (11). CD68+ macrophages were quantified in each layer of the inflamed synovium from z-stacked images (×400 magnification) of each field of view using Imaris Pro, version 6.0 (Bitplane). Only CD68+ cells that displayed visible nuclei as determined by counterstaining with DAPI were counted. Data from at least 3 slices per knee joint were averaged and expressed as total number of CD68+ macrophages per tissue volume (mm3).
Statistical testing was performed with SPSS, version 12. Differences in behavioral results and immunohistochemical measures between experimental groups were compared by one-way analysis of variance. For multiple comparisons, Fisher's post hoc analysis of protected least significant difference was used. P values less than 0.05 were considered significant. In all cases, the investigator responsible for behavioral testing, plotting, measuring, and counting was blinded with regard to the experimental group identifier of each animal.
Infiltration of macrophages, aberrant neovascularization, sensory and sympathetic nerve fiber sprouting, and joint-related pain behaviors in a model of CFA-induced arthritic joint pain.
Changes in structures of the knee joint, including bone, synovium, and meniscus (Figure 1), were examined histologically and immunohistochemically in CFA-injected mice 28 days after the initial intraarticular injection, and the results were compared with those in naive and sham-injected mice. As results in naive and sham-injected animals were essentially identical, results from naive animals are not shown. In sham-injected mice, low levels of CD68+ macrophages and CD31+ blood vessels were observed in the synovial–meniscal interface (Figures 2–4). In addition, CGRP+, GAP-43+, and NF200+ sensory and TH+ sympathetic nerve fibers were evident at low levels in the bone proximal to the joint and the synovial–meniscal interface in the sham-injected mice (Figures 4 and 5).
In contrast, 28 days after the initial CFA injection, a substantial influx of CD68+ macrophages (Figures 2 and 4) and substantial neovascularization (Figures 3 and 4) were observed in the synovium of the arthritic joint. Interestingly, these newly formed blood vessels exhibited a nonstandard bifurcating pattern and uneven thickness of the vessel wall as compared to blood vessels in the synovium of sham-injected mice (Figure 3). In addition, robust sprouting of CGRP+, NF200+, and GAP-43+ sensory and TH+ sympathetic nerve fibers occurred in the synovium of the inflamed joint in CFA-treated mice (Figures 4 and 5). These newly sprouted nerve fibers in the synovium of the arthritic joint were found in higher density, appearing highly disorganized as compared to the primarily linear morphology of these nerve fibers in the synovium of sham-treated mice.
In addition to the cellular changes that occurred in the CFA-induced arthritic joint, significant pain-related behaviors were observed in mice treated with CFA plus vehicle as compared to sham-injected mice (Figure 6). Results from all pain-related behavioral analyses revealed a rapid escalation of these pain behaviors by day 3 after the initial CFA injection, with significance maintained at each behavioral test time point throughout the experimental period (day 24 post–initial CFA injection) as compared to sham-treated mice.
Attenuation of CFA-induced nerve fiber sprouting and arthritic joint pain by anti-NGF.
Quantification of nerve fiber density following early/sustained treatment with anti-NGF revealed that this therapy prevented the sprouting of CGRP+, NF200+, and GAP-43+ sensory and TH+ sympathetic nerve fibers in the synovium of the inflamed joint, as assayed 28 days after the initial CFA injection (Figures 4A–D and 5C, F, and I). Additional results suggested that early/sustained anti-NGF therapy did not affect the organization or density of CGRP+, NF200+, and GAP43+ sensory and TH+ sympathetic nerve fibers in the synovium of the contralateral knee joint as compared to the ipsilateral knee joint in sham-treated mice (data not shown).
Sustained treatment with anti-NGF significantly reduced spontaneous flinching pain behavior by day 3 after the initial CFA injection, and this pain reduction was observed at each behavioral test time point throughout the 24-day experimental period (Figure 6A). Hind limb use and rotorod analysis results in CFA-injected mice showed that these pain-related behaviors were attenuated by ∼50% with addition of anti-NGF to the treatment regimen (Figures 6B and C). The pain-related level of dynamic weight bearing of the hind limb (both percent of weight on the ipsilateral hind limb and percent of time spent on the ipsilateral hind limb) was reduced by ∼75% by day 23 in animals that received early/sustained treatment with anti-NGF (Figures 6D and E), while the contralateral limb remained unaffected (Figure 6F).
Lack of effect of anti-NGF on macrophage infiltration or aberrant neovascularization in joints with CFA-induced arthritis.
Quantification of the total length of CD31+ blood vessels per tissue volume within the synovium of the inflamed joint on day 28 post–initial CFA injection suggested that early/sustained anti-NGF treatment does not significantly reduce the number of vessels as compared to the number in CFA-injected mice treated with vehicle (Figure 4E). Similarly, quantification of the total number of CD68+ macrophages per tissue volume within the synovium of the inflamed joint on day 28 indicated that early/sustained anti-NGF treatment does not significantly reduce the number of macrophages as compared to that in mice receiving CFA plus vehicle (Figure 4F).
Although chronic joint pain can be caused by a very diverse group of injuries and disorders as well as by aging, significant pain and impairment of physical function are common features of most joint disorders (12, 13). Management of skeletal pain is often not completely effective due to a high incidence of dose-limiting side effects with the major therapies currently used to treat this pain, i.e., nonsteroidal antiinflammatory drugs and opioids (4, 5, 14).
Our present understanding of what drives joint and skeletal pain is that as the joint and adjacent bone are injured due to trauma and/or aging, nerves that innervate the bone are first activated and sensitized by factors released by stromal/inflammatory/immune cells. As the joint continues to deteriorate, these “sensitized” nerve fibers then become activated when noxious or non-noxious mechanical stimuli are applied to the joint. When the cartilage deteriorates to the point where it is no longer intact, bone-on-bone interactions can occur, which may induce direct mechanical stimulation of these sensitized nerve fibers (6, 15). These changes in the discharge pattern of peripheral nerve fibers have also been shown to lead to neurochemical and cellular changes in the spinal cord and higher centers of the brain (i.e., central sensitization) that contribute to the generation and maintenance of chronic joint pain (16, 17).
While the above-described mechanisms certainly contribute to arthritic joint pain, the present data demonstrate that the nerve fibers that innervate the joint are not simply static structures, but can undergo remarkable reorganization in terms of altered morphology, an increase in the density of nerve fibers per unit area, and sprouting into areas of the joint that are either poorly innervated or not innervated. The populations of nerve fibers that undergo sprouting include CGRP+ and NF200+ sensory nerve fibers, which correspond to unmyelinated/thinly myelinated and myelinated nerve fibers, respectively. In the present study, robust sprouting of TH+ postganglionic sympathetic nerve fibers was observed in the inflamed synovium of the painful arthritic joint, and anti-NGF therapy blocked this sprouting. Several reports have suggested that sympathectomy attenuates disease progression and/or pain in arthritis (18–20), although sympathectomy-induced enhancement of disease progression and/or pain in arthritis has also been reported (21–23). Whether these differences are due to the different species, models of arthritis, or methods of performing sympathectomy that were used in these studies remains unclear. In light of these findings, future studies to more fully elucidate the role of sympathetic nerve fibers in driving disease progression and pain in arthritis are clearly warranted.
It has previously been shown that the majority of both sensory and sympathetic nerve fibers that innervate the skeleton express TrkA (24). Activation of TrkA, which is the cognate receptor for NGF, has been shown to induce sprouting in both developing and adult sensory and sympathetic nerve fibers (25, 26). In the present study, administration of an anti-NGF antibody, which sequesters NGF and prevents its binding to TrkA, largely blocked the sprouting of sensory and sympathetic nerve fibers in the arthritic joint and significantly attenuated arthritic joint pain. In addition, we did not observe any changes in nerve fiber sprouting or pain behaviors in the contralateral joint, although this phenomenon has been previously described to occur in other pain states (27).
Additionally, administration of anti-NGF did not affect disease progression as evaluated by neovascularization and macrophage infiltration in the inflamed knee joint. These findings are consistent with the results of previous studies in rodents, which demonstrated that CFA-induced knee joint inflammation and hind paw edema are not significantly modified by anti-NGF treatment (9, 28).
Based on the data presented here we hypothesize that, in the adult arthritic joint, NGF released from inflammatory/immune/stromal cells induces a marked sprouting of TrkA+, but not TrkA−, sensory and sympathetic nerve fibers. These newly sprouted sensory and sympathetic nerve fibers have a distinctive morphology and are present at a higher density of nerve fibers/unit area than is found in the normal joint. As these newly sprouted fibers are present in areas of the joint that are subject to significant stress and load bearing, and NGF has been shown to induce a marked sensitization and alteration in the phenotype of sensory and sympathetic nerve fibers, these changes may contribute to arthritic joint pain.
Pathologic nerve sprouting similar to that observed in the present mouse model of the painful arthritic joint is found in other nonskeletal and skeletal pain states, both in rodents and in humans. Aberrant nerve sprouting has been shown to be present in humans with nonskeletal chronic pain states including interstitial cystitis, vulvodynia, and irritable bowel disease (29). In addition, in humans with chronic discogenic pain, growth of CGRP+ nerve fibers into normally aneural and avascular areas of the intervertebral disc has been demonstrated (30). Studies have also shown that significant sprouting of CGRP+ nerve fibers occurs following bone fracture in rats as well as in the arthritic joints of other animals, including humans (31–35). Furthermore, it has been suggested that significant sprouting of TrkA+ nerve fibers can occur after tumor infiltration of the skeleton (11). Whether similar nerve sprouting can occur in bone following total knee replacement is not known. However, if it does occur, this may partially explain why some patients experience unsuccessful relief of pain following total knee replacement (36).
While sprouting of TrkA+ nerve fibers clearly can occur in the painful arthritic joint, the specific endogenous stromal, inflammatory, and immune cells that are the major source of NGF have not yet been defined. In addition, whether the availability of NGF in the injured and/or aged joint is a major determinant of the sprouting of TrkA+ nerve fibers, and whether there is a correlation between peripheral nerve fiber sprouting and the generation and/or maintenance of chronic arthritic joint pain, remain to be elucidated.
Previous data suggest that release of NGF can contribute to the sensitization of TrkA+ sensory nerve fibers, as NGF–TrkA activation of intracellular signaling cascades in the adult specifically modulate the sensitivity of primary afferent nociceptors to mechanical, thermal, and chemical stimuli in vitro and in vivo (37, 38). NGF also binds TrkA receptors expressed on the peptidergic fiber terminal itself, resulting in sensitization or increased expression of a number of receptors and channels at the membrane surface, including transient receptor potential vanilloid channel 1, acid-sensing ion channel 3, bradykinin receptors, voltage-gated sodium and calcium channels, and putative mechanotransducers, that may contribute to hypersensitivity after inflammation (25).
Following the period of immediate hypersensitivity with NGF release after tissue injury, early transcriptional changes also occur in the sensory signaling pathway. As NGF also signals via retrograde transport of the internalized NGF–TrkA complex, there is a delay (from hours to days) before some of the contribution of NGF to hypersensitivity is observable. After retrograde transport to the dorsal root ganglia, the signal from the NGF–TrkA complex can produce changes in sensory phenotype through the switching on (and off) of gene promoters, which leads to increased synthesis of neuropeptides (e.g., substance P, CGRP, and brain-derived neurotrophic factor), and of receptors and ion channels that are expressed by nociceptors (39–42). Taken together, these data demonstrate that while NGF-induced sprouting of TrkA+ nerve fibers may be involved in driving skeletal pain, NGF-induced sensitization and alteration of the phenotype of TrkA+ nerve fibers may play a significant role in the generation and maintenance of chronic arthritic joint pain.
Findings of several preclinical studies have suggested that blocking of NGF–TrkA can be efficacious in attenuating skeletal pain. Therapies blocking the NGF–TrkA pathway reversed the established hyperalgesia in rodent models of autoimmune arthritis (9) and osteoarthritis (OA) (43), suggesting that NGF is involved in prolonged hyperalgesia. Additionally, a role of NGF in maintenance of hypersensitivity in chronic injury has been demonstrated using several models of bone cancer (44, 45) and a model of bone fracture (46).
One rather unique aspect of the sensory innervation of bone and joint, which may explain in part why anti-NGF therapy is effective in relieving both malignant and nonmalignant skeletal pain, is that the majority of unmyelinated (CGRP+) and myelinated (NF200+) sensory nerve fibers that innervate bone and joint appear to express TrkA (24). Accordingly, few unmyelinated nonpeptidergic (IB4+/RET+) nerve fibers are present in bone (47, 48). Therefore, therapies that target NGF or TrkA may be particularly efficacious in relieving bone pain, as the majority of nociceptors express TrkA and respond to NGF.
Recent clinical trials of an anti-NGF (humanized anti-NGF monoclonal antibody) therapy have demonstrated that this treatment is efficacious in relieving pain due to OA (49, 50). In those trials, and in preclinical studies, anti-NGF was anti-hyperalgesic (i.e., normalizing a decreased nociceptive threshold) as opposed to analgesic (i.e., increasing a normal and sensitized nociceptive threshold). However, recent clinical trials in elderly patients with OA have been halted due to the need for earlier-than-expected joint replacement in a small subset of participants (49). It remains unclear whether this earlier-than-expected joint replacement in patients being treated with anti-NGF was simply due to greater use of the diseased joint or was due to unforeseen adverse effects on the bone itself, such as a decrease in either the formation or the maintenance of the vascular supply of the bone.
The present findings suggest that anti-NGF has a profound effect on blocking sprouting of TrkA+ nerve fibers and pain in the arthritic joint, but does not alter the formation or maintenance of CD31+ blood vessels or the influx of CD68+ macrophages in arthritic or normal joint and bone. These results demonstrate that NGF activation has a significant role in driving arthritic joint and skeletal pain. Several important questions, pertaining to, e.g., the efficacy of NGF–TrkA blockade in removing ectopic sprouting once it occurs, the sources of NGF in the arthritic joint, and whether there is a significant sympathetic component to this pain, remain to be answered.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Mantyh had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Ghilardi, Jimenez-Andrade, Mantyh.
Acquisition of data. Freeman, Jimenez-Andrade, Coughlin, Kaczmarska, Castaneda-Corral, Bloom.
Analysis and interpretation of data. Ghilardi, Freeman, Jimenez-Andrade, Kuskowski, Mantyh.