Injection of nerve growth factor (NGF) produces mechanical and thermal hypersensitivity in rodents and humans. Treatment with sequestering antibodies demonstrates the importance of NGF in various pain states, with efficacy seen in a number of animal pain models and in painful human conditions. However, these phenomena have not been evaluated in the context of using NGF-induced hypersensitivities as a model of pain.
NGF-induced behaviours were characterized using von Frey filament, pinprick and thermal endpoints and then pharmacologically evaluated with known reference agents.
Intraplantar NGF injection produced a dose-dependent increase in thermal sensitivity that lasted through 24 h post-injection and an immediate long-lasting (2 week) increase in mechanical sensitivity at the injection site, with no effects detected at secondary sites. NGF-induced mechanical sensitivity was pharmacologically characterized at 4 h and 1 week post-NGF injection. The nonsteroidal anti-inflammatory drugs (NSAIDs), celecoxib and diclofenac, were minimally effective against both thermal and mechanical endpoints. Gabapenitn and duloxetine were only moderately effective against thermal and mechanical hypersensitivity. Morphine was effective against thermal and mechanical endpoints at every time point examined. Treatment with the transient receptor potential vanilloid 1 (TRPV1) antagonist A-784168 partially attenuated NGF-induced thermal and mechanical sensitivity at all time points examined.
The results reported here suggest that effects of NGF on thermal and mechanical sensitivity in rats are similar to those reported in human and are partially driven by TRPV1. The rat NGF model may serve as a potential translational model for exploring the effects of novel analgesic agents.
Nerve growth factor (NGF) is a neurotrophic factor required by 80% of rat sensory neurons, mainly small-diameter peptide-containing neurons (Goedert et al., 1984; Lindsay, 1988), for survival through early post-natal development (Johnson et al., 1980; Yip et al., 1984; Johnson, 1986; Barde, 1989). Shortly after birth, the role of NGF changes to support, helping maintain the phenotype of myelinated nociceptors (Ritter et al., 1991). In adult, sensory neurons continue to express the NGF receptor TrkA, and NGF produced in peripheral tissues is retrogradely transported back to the cell body (Goedert et al., 1984; Verge et al., 1989), suggesting an ongoing physiological need. Treatment with anti-NGF antibodies leads to neuronal cell death during the neonatal period, a change in phenotype of myelinated nociceptors to low-threshold Aδ fibres in perinatal rats but has no noticeable effect on sensory neurons in mature rats (see Lewin and Mendell, 1993 for review).
NGF plays a significant role in a range of pain states (reviewed in Pezet and McMahon, 2006). Anti-NGF or -TrkA treatment ameliorates or reduces pain-like behaviours in animal models of neuropathic (Ramer and Bisby, 1999; Ro et al., 1999; Ugolini et al., 2007; Jing et al., 2009), inflammatory (Woolf et al., 1994; McMahon et al., 1995; Ma and Woolf, 1997; Banik et al., 2005; Shelton et al., 2005) and other pain states (Delafoy et al., 2003; Halvorson et al., 2005; Sevcik et al., 2005; Sabsovich et al., 2008). Importantly, anti-NGF treatment has shown efficacy in clinical trials for chronic low back pain (Katz et al., 2011), osteoarthritis (Lane et al., 2010; Schnitzer et al., 2011) and interstitial cystitis (Evans et al., 2011; Te, 2011).
Exogenous NGF produces pain in rodents (Lewin and Mendell, 1993; Lewin et al., 1994; Woolf et al., 1994; Andreev et al., 1995; Thompson et al., 1995; Amann et al., 1996; Svensson et al., 2010) and humans (Petty et al., 1994; Dyck et al., 1997; Svensson et al., 2003, 2008a,b; Andersen et al., 2008; Rukwied et al., 2010). NGF injection into human muscle induces long-lasting mechanical sensitivity (Svensson et al., 2003, 2008a; Andersen et al., 2008; Gerber et al., 2011) and mechanical and thermal sensitivity when injected into the skin (Rukwied et al., 2010; Weinkauf et al., 2011). Similar patterns are reported in rodents (Andreev et al., 1995; Amann et al., 1996; Bennett et al., 1998; Malik-Hall et al., 2005; Hathway and Fitzgerald, 2006) where thermal hypersensitivity appears to require sympathetic post-ganglionic terminals (Andreev et al., 1995). Currently, there is a large focus on translational models to help bridge the gap between preclinical and clinical pain studies. In an effort to develop NGF-induced pain as a translational model, we characterized the development of thermal and mechanical (primary and secondary) hypersensitivity after intraplantar NGF injection in the rat. NGF-induced behaviours were pharmacologically evaluated using reference agents with a range of mechanism of actions: duloxetine (serotonin-norepinephrine reuptake inhibitor; SNRI), morphine (opioid), diclofenac (NSAID), celecoxib (cyclooxygenase-2; COX-2 selective NSAID), gabapentin (anticonvulsant) and A-784168 (TRPV1 antagonist).
What's already known about this topic?
Nerve growth factor (NGF) plays an important role in various pain states, in both rodents and humans.
Injection of NGF induces pain in rodents and humans.
What does this study add?
This study spatially and temporally characterizes NGF-induced pain behaviours in the rat using mechanical and thermal endpoints.
NGF-induced pain behaviours were pharmacologically and mechanistically evaluated. The results support the contention that the NGF model may be a suitable translational model of pain.
Subjects were adult male Sprague-Dawley rats (Charles River, Wilmington, MA, USA) weighing 220 to 320 g at the time of testing. All animals were housed in groups of five on solid bottom Plexiglas cages (52 × 28 × 20 cm) with bedding material and had free access to food and water in a temperature-and humidity-controlled room on a 12-/12-h light/dark cycle in compliance with the Association for Assessment and Accreditation of Laboratory Animal Care. Experimental protocols were in compliance with the Ethical Guidelines of the International Association for the Study of Pain (Zimmermann, 1983) and were approved by the Institutional Animal Care and Use Committee at Abbott Laboratories.
2.2 NGF injection
A stock solution of recombinant human β-NGF (R&D Systems, Minneapolis, MN, USA, catalogue number 256-GF) was prepared by reconstituting in sterile saline to a concentration of 5 μg/30 μL. The stock solution was further diluted in sterile saline to give final working concentrations. NGF was injected intraplantar in the right hind paw at a volume of 30 μL for all doses. Following initial characterization, acute studies were performed 4 h after NGF injection and chronic studies were performed at 1 week post-NGF injection.
2.3 Reference agents
Morphine sulfate (Spectrum Chemical MFG Corp, New Brunswick, NJ, USA) was administered in a dose range of 0.3–3 mg/kg dissolved in sterile saline and dosed subcutaneous (s.c.), 2 mL/kg, 0.5 h before testing. Gabapentin (Spectrum Chemical MFG Corp, New Brunswick, NJ, USA) was dissolved in water, dosed intraperitoneal (i.p.) at 2 mL/kg, and was tested at 0.5 h post-administration due to the i.p. route of administration. Diclofenac (Sigma-Aldrich, St. Louis, MO, USA) was dissolved in water and dosed i.p., 2 mL/kg, 0.5 h before testing. Celecoxib (ChemPacific, Corp, Baltimore, MD, USA) was dissolved in PEG400 and dosed per os (p.o.) (overnight-fasted animals), 2 mL/kg, 1 h before testing. Duloxetine (ChemPacific) was dissolved in water and dosed p.o. (overnight-fasted animals), 2 mL/kg, 3 h before testing. Plasma and brain levels at the time of testing are given in Table 1. A-784168 (TRPV1 antagonist) was synthesized at Abbott Laboratories (Abbott Park, IL, USA), dissolved in 10%EtOH/20%Tween80/70%PEG400 and dosed p.o. (fed animals) at 2 mL/kg. All studies were conducted in a blinded fashion. None of the reference agents affected mechanical or thermal responses in naïve animals (data not shown), morphine notwithstanding (noted below).
Table 1. Reference agent exposure levels
Plasma and brain exposure levels of reference compounds taken at the time of testing. Values are listed as means ± SEM.
0.66 ± 0.11
0.17 ± 0.02
2.08 ± 0.35
0.46 ± 0.09
7.20 ± 1.40
1.20 ± 0.18
0.36 ± 0.10
2.45 ± 0.61
1.40 ± 0.26
20.71 ± 4.22
3.15 ± 0.89
44.92 ± 10.39
4.77 ± 0.31
4.50 ± 0.69
9.42 ± 2.15
7.85 ± 1.94
9.66 ± 1.81
9.26 ± 2.92
0.03 ± 0.01
0.002 ± 0.0002
0.13 ± 0.02
0.01 ± 0.001
0.59 ± 0.03
0.04 ± 0.01
0.70 ± 0.33
0.18 ± 0.09
2.74 ± 0.42
0.65 ± 0.06
4.35 ± 0.20
0.99 ± 0.03
2.4 Mechanical sensitivity
To assay mechanical sensitivity, tactile allodynia was measured using calibrated von Frey filaments (Stoelting, Wood Dale, IL, USA) as previously described (Chaplan et al., 1994). The rats were placed in inverted individual plastic containers (20 × 12.5 × 20 cm) on a suspended wire mesh grid and acclimated to the testing chamber for 20 min. Filaments were applied perpendicular to the selected hind paw with enough force to induce slight bending and held in this position for 6–8 s. A positive response included hind paw withdrawal from the stimulus or flinching and attention to the paw immediately following removal of the filament. The strength of the maximum filament used was 15 g. The 50% paw withdrawal threshold (PWT) was determined by increasing and decreasing the stimulus intensity of von Frey filaments using the Dixon up-down procedure (Dixon, 1965) and reported as percent of maximum effect (%MPE) as described previously (Mills et al., 2012).
Paw withdrawal response frequency evoked by mechanical stimulation was assayed using calibrated von Frey filaments (2, 6, 10 and 26 g). Filaments were applied to the plantar surface of each hind paw 10 times for 1–2 s with an interstimulus duration of 5–6 s in ascending order. Paw withdrawal response frequency is reported as the percent of withdrawals to each filament out of 10 applications.
The pinprick test was conducted using a blunt safety pin. A single stimulus was presented to the plantar surface of the injected paw with enough force to indent the skin but not enough force to break the integrity of the skin. The duration of paw withdrawal was recorded with a minimal time defined as 0.5 s and a maximal cut-off time set at 60 s. Data are reported as the mean of two stimuli presented with a minimum interval of 5 min.
2.5 Thermal sensitivity
Thermal sensitivity was determined using a commercially available thermal paw stimulator (University Anesthesia Research and Development Group, University of California, San Diego, CA, USA) described previously (Hargreaves et al., 1988). The rats were placed into individual plastic cubicles mounted on a glass surface maintained at 30 °C and allowed a 20-min acclimation period. A thermal stimulus in the form of radiant heat emitted from a focused projection bulb was then applied to the plantar surface of the hind paw. The stimulus current was held constant and the maximum time of exposure set at 20 s to limit possible tissue damage. The elapsed time until a brisk withdrawal of the hind paw from the thermal stimulus (i.e., paw withdrawal latency, PWL) was recorded automatically using photodiode motion sensors. The right and left hind paw of each rat was tested in three sequential trials at approximately 5-min intervals and the mean of the two shortest latencies was used as the PWL.
2.6 Statistical analysis
Data were analysed using Graph Pad Prism 5.0 (GraphPad Software, Inc., San Diego, CA, USA) by one- or two-way analysis of variance (ANOVA) as indicated with Bonferroni and Dunnett's post hoc tests for between group comparisons. Statistical analyses were conducted on raw data (PWTs and PWLs). Data are presented as mean ± standard error of the mean (SEM).
3.1 NGF injection
Injection of NGF at doses of 0.3–5 μg did not cause obvious erythema, calor or oedema. Nor did it induce nocifensive behaviours such as flinching, guarding, or biting of the injected paw. Consistent with previous reports (Andreev et al., 1995), nocifensive behaviours were only observed when the paw had been previously sensitized. For example, flinching behaviour was observed when NGF (3–10 μg/20 μL saline) was administered 110 min following a low dose of carrageenan (66.7 μg/20 mL saline; data not shown).
3.2 NGF-induced mechanical sensitivity
Intraplantar injection of NGF induced an immediate, long-lasting increase in mechanical sensitivity that resolved in a dose-dependent manner (Fig. 1A). As early as 1 h post-injection, all doses of NGF examined (0.3–5 μg) reduced PWTs with a maximum effect seen with the highest dose of NGF. The increase in sensitivity was seen through 1 week for all doses and was still evident at 2 weeks for the highest two doses (3 and 5 μg). In addition to shifts in PWTs, NGF induced a linear, leftward shift in frequency of responses to von Frey filaments (Fig. 1B). A maximum shift was seen with the highest dose of NGF at 1 h and 1 day post-injection, which followed a similar pattern of resolution as PWTs. In contrast to the effect seen on primary sensitivity, there was no evidence of secondary hypersensitivity with any dose or at any time point examined (Fig. 1C). NGF injection produced only a transient increase in sensitivity to noxious pinprick stimulation at 1 h at the site of injection (Fig. 2). There was no evidence of increased sensitivity to noxious stimuli at any other time point examined or at secondary sites. For each endpoint (PWTs, frequency and pinprick), there was no effect on the contralateral paw after NGF injection (data not shown).
3.3 NGF-induced thermal sensitivity
Intraplantar NGF injection induced an immediate, dose-dependent increase in thermal sensitivity (Fig. 3). The effect of NGF on thermal sensitivity was seen as early as 1 h post-injection for all doses and resolved in a dose-dependent manner. The temporal effect on thermal sensitivity was shorter in duration compared with the increase in mechanical sensitivity. Effects on thermal sensitivity lasted through 24 h for the highest dose (5 μg), which resolved by 48 h.
Since the temporal effects of exogenous NGF on mechanical sensitivity may be produced by different mechanisms, the effects of pharmacological agents were examined at 4 h and 1 week post-NGF injection. All compounds exhibited dose-dependent effects at the acute time point (Fig. 4A). Morphine produced robust efficacy, up to 82% at 3 mg/kg. Gabapentin also produced robust efficacy at 100 and 150 mg/kg, 67% and 71%, respectively. However, the NSAIDs, celecoxib and diclofenac, produced only modest effects of approximately 40% at the highest doses examined. Modest efficacy of 38% was seen with duloxetine at 60 mg/kg.
A similar pharmacological profile was seen at the 1 week time point (Fig. 4B), with morphine producing robust efficacy, up to 79% at 3 mg/kg. Celecoxib and diclofenac produced only minimal effects of approximately 25%. Gabapentin had somewhat less of an effect compared with the acute time point, producing a maximum effect of 51% at 150 mg/kg. Duloxetine produced a maximum effect of 56% at 60 mg/kg. Individual dose response curves for reference compounds were not statistically significant between the two time points (two-way ANOVA, p > 0.05).
To investigate the pharmacological response of NGF-induced thermal sensitivity, we examined thermal responses 4 h post-NGF injection (Fig. 4C). Morphine demonstrated the greatest efficacy, producing a greater than 100% effect (contralateral responses were also slightly elevated at the highest dose, data not shown). Gabapentin and duloxetine produced effects of 65% and 60% at the highest doses, respectively. Celecoxib and diclofenac produced little or no effect, 22% and 7%, respectively.
3.6 Evaluation with the TRPV1 antagonist A-784168
Since NGF is known to effect TRPV1 expression and the thermal hypersensitivity induced by exogenous NGF is thought to be partially attributed to NGF-mediated effects on TRPV1, we evaluated the effects of a TRPV1 antagonist on thermal and mechanical sensitivity (Fig. 5). Acutely, A-784168 at doses shown to be effective in vivo (Cui et al., 2006) partially inhibited the development of mechanical sensitivity, producing a 50% effect that lasted through 3 h post-NGF injection (Fig. 5A). A-784168 also attenuated the development of thermal hypersensitivity, producing a 53% inhibition that lasted through 3 h post-NGF injection (Fig. 5B). One week post-NGF injection, A-784168 partially reversed NGF-induced mechanical hypersensitivity with a similar dose–response curve as seen acutely (Fig 5C). There was not a statistically significant difference between time points or between acute and 1-week dose response curves (two-way ANOVA, p > 0.05). The pharmacokinetic profile for A-784168 is similar for the time points examined in the acute (1 h and 3.5 h) and the 7-day study (1.5 h and 4 h; data not shown).
4. Discussion and conclusions
Exogenous NGF whether injected intramuscular, intraplantar, or systemically is known to produce both mechanical and thermal hypersensitivity in rodents and humans. However, studies of intraplantar NGF-induced behaviours that assess primary and secondary mechanical hypersensitivity, pinprick hyperalgesia, thermal sensitivity and pharmacological evaluation with known analgesics have not been described. Here, we characterized and pharmacologically evaluated these NGF-induced behaviours in the rat with the idea of using NGF injection as a translational model of pain.
Consistent with previous results (Amann et al., 1996), intraplantar injection of NGF did not cause obvious oedema. In contrast, intraplantar NGF injection has been reported to produce plasma extravasation (Andreev et al., 1995) and to cause transient, calcitonin gene-related peptide (CGRP) and mast cell dependent oedema at doses comparable with doses reported here (Amann et al., 1996). NGF injection only produced nocifensive behaviours when the paw had been previously sensitized (e.g., with a low dose of carrageenan; data not shown). However, NGF injection did induce an immediate increase in mechanical and thermal sensitivity, contrary to what is seen after systemic (Lewin et al., 1993) or intradermal injection of cleaved NGF (octapeptide NGF) (Taiwo et al., 1991) or intradermal injection in humans (Rukwied et al., 2010; Weinkauf et al., 2011).
Spatial and temporal development of NGF-induced mechanical and thermal hypersensitivity appears to be dependent on the site and route of injection and on species examined. In adult rats, intraperitoneal injection of NGF induces mechanical sensitivity after 6 h and an increase in thermal sensitivity as early as 15 min post-administration. These effects appear to be independent of Aδ fibre sensitization observed in neonatal rats treated with NGF (Lewin et al., 1993). The observed increase in thermal sensitivity following intraplantar NGF injection is attenuated after sympathectomy, suggesting a peripheral, indirect site of action requiring normal sympathetic post-ganglionic terminals (Andreev et al., 1995). Here, and in previous rat studies (Andreev et al., 1995; Amann et al., 1996), mechanical and thermal sensitivity follow similar time courses regardless of the route of administration. Mechanical sensitivity persists for several days, taking much longer to resolve than heat hypersensitivity. This is similar to the mouse and human in that heat sensitivity resolves much faster than the long-lasting increase in mechanical sensitivity. Intramuscular injection in humans, as in rodents, produces an immediate increase in muscle tenderness that lasts for days that is not confined to the injection site. In humans, mechanical sensitivity after intradermal injection is delayed and long-lasting. Again, increased sensitivity to heat develops more rapidly than mechanical and is quicker to resolve.
In the current study, mechanical sensitivity after NGF injection was restricted to the injection site, and no secondary areas of sensitivity were observed. This is in contrast to other models, such as intraplantar or intradermal capsaicin injection that causes sensitivity at areas away from the injection site. This increase in sensitivity at secondary sites is presumably mediated and maintained by central sensitization. Since NGF does not produce secondary hypersensitivity, it could be considered as a peripheral sensitization model. However, others have found secondary hyperalgesia on the plantar aspect of the ipsilateral paw after injection of NGF into the biceps femoris muscle of mice, which they attributed to central sensitization (Hathway and Fitzgerald, 2006). Additionally, intramuscular NGF treatment leads to altered afferent connectivity within the dorsal horn (Lewin et al., 1992) and systemic administration in neonates produces changes in spinal reflex activity (Thompson et al., 1995). How the differences in route of NGF administration impacts the development of secondary hyperalgesia needs further investigation. The findings of the current study show that the lack of development of areas of secondary sensitivity away from the NGF injection site are consistent with findings reported following intradermal NGF injections in humans (Rukwied et al., 2010).
Since mechanical sensitivity may be driven by different mechanisms at early and late time points, we pharmacologically evaluated the acute and late phases separately. The mechanisms surrounding this long-lasting mechanical sensitivity are unclear. The acute response may be mediated through direct and/or indirect actions of NGF on nociceptors (Malik-Hall et al., 2005; Pezet and McMahon, 2006). As high levels of NGF are expected to be transient, the long-lasting effects on mechanosensation may be driven by NGF-induced changes in gene expression, such as Nav1.7 and Nav1.8 (Fjell et al., 1999a,b,c; Gould et al., 2000), P2X3, acid-sensing ion channel 3 (ASIC3), bradykinin B2 and mu opioid receptors (MORs) [reviewed in (Pezet and McMahon, 2006)]. NGF induces up-regulation of neuropeptides, such as substance P and CGRP (Donnerer et al., 1992; Pezet et al., 2001) and induces brain-derived neurotropic factor (BDNF) expression (Michael et al., 1997; Priestley et al., 2002), which could drive long-lasting sensitization. Mechanical hypersensitivity may be mediated through TrkA on primary afferents via downstream pathways such as MEK/ERK, PI3K and PLCγ (Malik-Hall et al., 2005). NGF-induced mechanical hypersensitivity appears to be independent of NMDA receptor function, whereas the late phase (>3 h) of thermal hypersensitivity is blocked by the NMDA receptor antagonist MK801 (Lewin et al., 1994). However, there does not seem to be ongoing primary afferent input into the spinal cord to drive central sensitization (Lewin and Mendell, 1993; Lewin et al., 1994).
Intraplantar NGF also induced an immediate increase in thermal sensitivity of the ipsilateral paw. COX products appear to play only a minimal role in NGF-induced thermal hypersensitivity as revealed by a lack of effect with indomethancin treatment (Amann et al., 1996), which is further supported here by the poor effects of diclofenac and celecoxib. Cultured dorsal root ganglion (DRG) neurons treated with NGF display increased TRPV1 currents (Shu and Mendell, 1999; Zhu et al., 2004). NGF increases TRPV1 expression (Ji et al., 2002; Donnerer et al., 2005; Xue et al., 2007) and treatment with anti-NGF antibodies attenuates inflammation-induced hyperalgesia and decreases TRPV1 expression (Ji et al., 2002; Cheng and Ji, 2008). Given the relationship between NGF and TRPV1, we examine the ability of A-784168, a characterized TRPV1 antagonist (Cui et al., 2006), to block NGF-induced mechanical and thermal sensitivity. The development of both mechanical and thermal hypersensitivity was partially attenuated with A-784168 pretreatment, suggesting that TRPV1 plays a role in the development of NGF-induced hypersensitivities. In addition to effects on TRPV1, NGF is known to induce mast cell degranulation, which may also drive hypersensitivities. Thus, the inability to completely attenuate the hypersensitivities by blocking TRPV1 may be due to the release of histamine and serotonin from mast cell degranulation. TRPV1 antagonism not only affected induction of thermal and mechanical sensitivity but also affected established mechanical hypersensitivity, suggesting an ongoing role for TRPV1 in NGF-induced mechanical hypersensitivity.
It is clear that endogenous NGF plays a significant role in inflammation and that exogenous NGF produces pain in rodents and humans. However, the utility of using NGF-induced behaviours as a translational model is not without caveats. First, inflammation is a complex cascade of events with multiple players. While increases in NGF may be a trigger for some of these events, there are other contributing factors. Second, the temporal differences in resolution between mechanical and thermal sensitivity suggests different mechanisms. Similarly, the temporal aspect of mechanical sensitivity may be driven by different mechanisms at different times (e.g., induction vs. maintenance). Finally, the doses of NGF used here (up to 5 μg), while comparable with other studies, likely produce tissue levels much higher than has been reported for endogenous NGF levels after an injury or in a disease state (Woolf et al., 1994; Matsuda et al., 1998; Saade et al., 2002; Banik et al., 2005).
In conclusion, we characterized the effects of intraplantar NGF injection in the adult rat and demonstrate that both primary mechanical and thermal sensitivity are increased and resolve in a dose-dependent manner. There was no evidence for increased sensitivity or flare in areas away from the injection site, suggesting that NGF-induced behaviours are driven by a peripheral mechanism. COX products appear to play a minimal role in the development and maintenance of mechanical or thermal hypersensitivity, whereas thermal and mechanical sensitivity are driven to a greater extent through activation of TRPV1. Future studies need to be directed at understanding the mechanisms driving NGF-induced changes to understand the utility of NGF-induced responses as a translational model of pain. Studies that use Na channel blockers, lignocaine (for comparison with human data), anti-NGF antibodies, immunohistochemical and electrophysiological evaluation will be important to provide further insight into mechanisms that drive NGF-induced pain responses.
All authors are employees of Abbott Laboratories.
All authors discussed the results and commented on the manuscript.
Charles Mills conceived, designed and supervised all experiments, wrote the manuscript and assimilated all comments from contributing authors and prepared all figures for publication. Trang Nguyen performed experiments, prepared data and statistical analysis and contributed to the writing and editing of the manuscript. Flobert Tanga performed experiments, prepared data and statistical analysis and contributed to the writing and editing of the manuscript. Chengmin Zhong performed experiments, prepared data and statistical analysis and contributed to the writing and editing of the manuscript. Donna Gauvin performed experiments, prepared data and statistical analysis and contributed to the writing and editing of the manuscript. Joseph Mikusa performed experiments, prepared data and statistical analysis and contributed to the writing and editing of the manuscript. E.G. rica Gomez performed experiments, prepared data and statistical analysis and contributed to the writing and editing of the manuscript. Anita Salyers performed experiments, prepared data and statistical analysis and contributed to the writing and editing of the manuscript. Anthony Bannon conceived, designed and supervised all experiments and contributed to the writing and editing of the manuscript.