Effects of removal of dietary polyunsaturated fatty acids on plasma extravasation and mechanical allodynia in a trigeminal neuropathic pain model

No obvious change could be detected in spontaneous behavior in animals fed different diets, and after CCI or sham surgery. Food intake was also apparently unaffected by diet or surgery, and this was attested by the absence of significant differences in weight gain from six weeks of age to the day of perfusion between the rats fed regular rat chow (RD) (from 218.3 ± 52.0 g to 362.4 ± 27.1 g) and those fed modified diet (MD) (from 198.5 ± 21.9 g to 363.4 ± 19.0 g). Likewise, there were no weight differences at perfusion between the unoperated, naive group (363.4 ± 19.0 g), and the sham (356.4 ± 30.4 g) and CCI groups (340.2 ± 13.5 g).

The baseline response threshold to mechanical stimuli was similar in both diet groups (2.25 ± 0.32 g in RD vs. 1.94 ± 0.35 g in MD, expressed as bending force needed to elicit 50% of positive responses), and exhibited a highly significant decrease after CCI on the ipsilateral (Kruskal-Wallis, p < 0.001) but not the contralateral side in both. However, this decline in threshold appeared earlier and was more pronounced in the MD group, which differed very significantly from the RD group at 8 days postsurgery (0.43 ± 0.16 g vs. 1.96 ± 0.45 g; p = 0.005). At 15 days the difference showed statistical tendency (0.22 ± 0.11 g vs. 0.61 ± 0.26 g; p = 0.059), and by 26 days after surgery a small difference was still observed (0.10 ± 0.04 g vs. 0.32 ± 0.15 g), which did not reach significance. Side comparisons for each force tested also revealed a different time course of mechanical allodynia for RD and MD groups, showing an earlier onset of higher responsiveness in rats fed MD (at day 8 postsurgery), but the same pattern of side differences in MD and RD 26 days after surgery (Fig. 1). In the right side, contralateral to the CCI, the response threshold after surgery was somewhat lower in rats fed MD compared to those fed RD, although this difference did not reach statistical significance.

CAP-induced plasma extravasation in unoperated animals kept on MD (n = 14) was almost twice as large as that in the RD group (n = 14; Fig. 2). Pooling both sides for each group, extravasation values were 12.38 ± 5.13 for RD, and 21.57 ± 9.46 for MD. This 72% difference was very significant even when each side was compared separately (p = 0.012 for the left sides, p < 0.001 for the right). Similar values were obtained when comparisons between diet groups were made for the sham-operated animals (9.8 ± 4.4 for RD, n = 9; 24.9 ± 11.7 for MD, n = 9; p = 0.002).

The effects of topical application of vehicle (VEH) or saline (SAL) instead of CAP on naive rats was examined only in the group fed MD, on the assumption that, in the improbable case that those treatments altered basal extravasation, such effect would be more marked under no "pain-protecting" feeding conditions. Our results (plotted in Fig. 3 for the left side) showed no differences in extravasation (p > 0.60 in all cases) following application of VEH (6.27 ± 1.72) or SAL (6.89 ± 3.09) on either side, while both values differed significantly (p < 0.002) from the extravasation elicited by CAP.

Nerve injury underlying different neuropathic conditions is known to reduce the flare response to topical application of substance P, histamine or capsaicin [28, 29]. A related effect, consisting of reduction of vasodilatation and decreased plasma extravasation, has been reported in widely used models of neuropathy in the sciatic territory in rats, such as the CCI of the sciatic nerve [9, 11] or the spinal nerve ligation [30]. The axon reflex-dependent vasodilatation is normally mediated by antidromic activation of a subpopulation of Aδ and C fibers [8, 10], but Aβ fibers activated from injured or irritated target tissues may also play a role, by centrally sensitizing C nociceptors [31]. Plasma extravasation, however, depends not on Aδ fibers [8], but on C fibers, because it is elicited by specific antidromic stimulation of polymodal C fibers [32, 33], and is prevented by specifically blocking axoplasmic transport in C fibers with colchicine [34].

We have shown that topical application of CAP to the vibrissal pad also elicits neurogenic extravasation, which is reduced ipsilaterally to a CCI-IoN. However, under standard dietetic conditions, the actual levels of plasma extravasation are increased bilaterally after performing a unilateral constriction. This results in a net extravasation increase in the constricted side, when compared with the same side in naive or sham groups. To our knowledge, similar intergroup comparisons are mostly lacking concerning extravasation after chronic constriction or partial nerve injury, not only in trigeminal fields, but also in the more commonly studied sciatic or saphenous territories. However, Yonehara and Yoshimura [11] showed a significant net reduction in EB released by CAP application or nerve stimulation into the perfusate collected through a subcutaneous cannula, 7 days after sciatic CCI. This difference with our results may be due not only to a local damage created by the inserted cannula in that study, but also to a different degree of nerve damage between the sciatic and the trigeminal CCI models. Yonehara and Yoshimura [11] used Bennett and Xie's [7] four-ligature model of sciatic CCI, which is reported to determine a nearly complete loss of myelinated fibers and a substantial loss of unmyelinated axons, two weeks after ligation [35]. A similar quantitative analysis is not yet available for the CCI-IoN model, which used two loose ligatures [5], nor for our variant, which used a single ligature. Qualitative observations by Anderson et al. [36] after a single ligature, however, suggested a relatively moderate fiber loss three weeks after CCI-IoN. In contrast, we have found that four relatively tight ligatures applied to the IoN resulted in a substantial fiber loss and a marked net reduction of EB extravasation, but also -as previously reported by others [37] – a less pronounced and less consistent mechanical allodynia (unpublished findings).

The appearance of contralateral effects after a unilateral nerve injury has been known for a number of years. They consist of a diversity of phenomena, from altered gene expression to a range of functional and anatomical changes in homotopical contralateral peripheral nerves, sensory ganglia or motoneurons [38]. Acute contralateral neurogenic responses, such as tissue edema or plasma extravasation into the synovial space, have been described following the induction of unilateral inflammation [39‐41]. More recently, Kelly et al. [42] showed that five days after unilateral induction of inflammation in the rat knee joint, EB extravasation in the contralateral knee joint was increased, and this increase was accompanied by an elevated rate of spontaneous activity in CAP-sensitive fibers of the saphenous nerve (just above the incorporation of the medial articular nerve). These observations are consistent with our finding of an increased EB extravasation in the vibrissal pad contralateral to the CCI-IoN. Although a conclusive explanation for the contralateral effects is yet to be obtained, it has been shown that a pharmacological blockade of C and Aδ nociceptive fibers and/or sympathetic fibers in the sciatic nerve blocks the sustained peripheral vasodilatation elicited by a contralateral sciatic CCI [43], supporting the early claim that both neural components participate in what was called reflex neurogenic inflammation [39]. The role of the sympathetic system for explaining the increase in peripheral extravasation after partial limb nerve lesion is debated [30, 44], but it has been shown that SNL induces sympathetic fibers sprouting not only in the ipsilateral but also the contralateral DRG [45‐47]. Less likely, however, would be a similar role for the sympathetic innervation in the trigeminal territory after unilateral CCI-IoN, given the absence of sympathetic fiber sprouting in the trigeminal ganglion after nerve injury (IoN or inferior alveolar nerve constriction: [24]). Moreover, the fraction of sympathetic fibers composing the trigeminal nerve is 2.5 times lower than in limb nerves [48].

The use of corn or soy as sources of dietary fat has been associated to decreased expression of mechanical allodynia following partial sciatic nerve ligation (PSL, [16‐18]). Both types of fat contain similarly high levels of linoleic acid (58%), a short chain (18 carbons) PUFA of the ω-6 family with 2 double bonds, but different levels of α-linolenic acid (0.7% in corn, 6.8% in soy), a short chain PUFA of the ω-3 family with 3 double bonds [19, 49]. All the above mentioned studies tested the rats for the presence of allodynia for up to 10–14 days after PSL. After CCI-IoN we confirmed that the long-term absence of PUFAs from the diet elicited a decrease in mechanoresponsive thresholds, and an overt mechanical allodynia 8 and 15 days postsurgery, as judged from comparisons with responses from the contralateral, undamaged side. At 15 days, rats fed RD also started to display allodynia, but with response thresholds still higher than rats fed MD. By 26 days postconstriction, however, the allodynia reached the same level irrespective of the diet used, suggesting that the "protective" effect of the PUFAs-rich diet would be temporary. In the absence of a more prolonged follow-up, it remains to be investigated whether the total duration of CCI-connected neuropathic symptoms is affected by dietary lipid content.

The removal of PUFAs from the diet also had a hitherto unreported effect on neurogenic inflammation, consisting of, 1, an increase of CAP-induced plasma extravasation in the vibrissal pads of uninjured animals, and 2, an ipsilateral reduction following CCI-IoN, while maintaining high levels of extravasation in the contralateral side. Although the mechanisms of this effect remain elusive, the relationships existing between dietary lipids and inflammation may shed some light on them. Dietary fatty acids, which have been used to treat different inflammatory conditions [20, 22], have a complex influence on inflammatory responses, which is mainly exerted by modifying the production of inflammation-related eicosanoids and cytokines. The simplest types of unsaturated fatty acids of the ω-6 and ω-3 families cannot be synthesized in mammals, but once ingested, are desaturated further and elongated, giving rise to a series of long chain PUFAs, with different, and in some aspects opposite effects on inflammatory processes [19, 50]. While the ω-6 PUFAs boost the production of potent proinflammatory eicosanoids (such as prostaglandins and leukotrienes) and cytokines (such as interleukins 1 and 6, and TNFα), ω-3 PUFAs antagonize inflammatory responses by altering the eicosanoid production through metabolic competition with the ω-6 PUFAs, by indirectly blocking the expression of proinflammatory genes such as NF-κB, or by generating antiinflammatory mediators, such as the recently described resolvins, docosatrienes and neuroprotectins [50, 51].

Dietary ω-3 PUFAs also alleviate chronic pain by mechanisms other than their antiinflammatory effects, including inhibition of neuronal protein kinases and blocking of voltage-gated calcium channels involved in both inflammatory and NP processing, general reduction of the sympathetic tone, and, perhaps, improvement of pain-associated affective disorders and other psychiatric conditions through largely unexplained mechanisms [20, 52‐54]. Moreover, the absence of dietary α-linolenic acid may enhance the formation of lysophosphatidic acid [55], which is considered a critical mediator in the development of NP, through direct actions on the primary sensory neurons [56, 57], and by inducing demyelination and ephaptic cross-talk between fibers in dorsal roots and peripheral nerves [58].

In contrast, because of their putative proinflammatory effects, the possible effects on pain control of dietary ω-6 PUFAs have been essentially neglected. However, a clear-cut opposite effect of dietary short chain PUFAs on pain is questionable. Firstly, in naive rats it was an appropriate ratio of ω-3:ω-6 fatty acids in diet (around 1:5), rather than the absolute amount of ω-3 intake, which was associated to an elevation of the withdrawal threshold to acute thermal noxious stimuli [13, 21]. This is consistent with the fact that, because of the complex metabolic interaction between the ω-families of dietary fatty acids, their final outcome on the tissue fatty acid composition depends more on the ratios between their different classes than on their absolute amounts in the diet [19]. Secondly, a significant positive correlation was recently reported between the total intake of α-linolenic acid (an ω-3 PUFA) and thermal hyperalgesia (but not mechanical allodynia) after a PSL [49]. Moreover, the presence of both ω-3 and ω-6 PUFAs' families in the intake could help reduce the excitability of sodium channels involved in pain [59], particularly in injured nerves [60]. In our study, the lipid content of the olive oil used in MD to replace the fat present in RD had just one-eighth of linoleic acid (6.3% vs. 50%), and practically lacked α-linolenic acid (traces vs. 4.5%; [61]), making olive oil a well-suited placebo in studies of the antiinflammatory effect of PUFAs [62].

Adult male Sprague-Dawley rats (n = 108) from our own colony, originating from Harlan (Harlan Iberica, Barcelona), were used. They were divided into 14 groups, combining the various treatment protocols: Unilateral (left) CCI-IoN, sham operation, or no surgery (naive); bilateral application of CAP cream (1.6%), VEH (base cream) or SAL to the vibrissal pad; and maintenance on RD or MD. Animals' weight was recorded at six weeks of age, at the surgery day and at the perfusion day.

Rats were housed under standard colony conditions (4 rats per cage). Food and water were supplied ad libitum, and the animals were kept under a reversed 12:12 h dark/light cycle. All experiments were carried out in accordance with the Guidelines on Ethical Standards for Investigation of Experimental Pain in Animals [65] and the European Community's Council Directive 86/609/EEC, and the study was approved by the Ethical Committee of our institution.

Animals were i.m. anesthetized with a mixture of ketamine 55 mg/Kg, xylazine 15 mg/Kg, and atropine 0.2 mg/Kg. The left IoN was exposed as described before [66, 67]. One single ligature was tied loosely around the distal part of the nerve using a polypropylene monofilament suture (Surgipro 6.0). The mild degree of constriction apparently did not block the circulation through the superficial epineural vasculature of the IoN, as proposed by Bennett and Xie [7] for the sciatic CCI, although the fan-like geometry of the multi-fascicled IoN undoubtedly resulted in different degrees of constriction on different fascicles. Special care was taken to avoid an excessive compression of the nerve, which hampered the development of allodynia and resulted in more severe fiber loss (results not shown). The wound was closed using silk sutures. In sham-operated rats the IoN was exposed on the left side using the same procedure but the nerve was left untouched.

Before every stimulation session, rats were habituated for one hour daily to the testing room environment during 2–3 days. The animals were placed in clear acrylic cages (30 × 30 × 20 cm) in order to reduce their field of exploration. When the rat was still, with the four paws placed on the ground, it was gently handled to ensure that it became familiar with the experimenter and the approximation of the test filaments, in order to minimize stress.

Tactile sensitivity in the vibrissal area was measured with a set of eight calibrated von Frey monofilaments (North Coast Medical, Inc. Morgan Hill, CA) ranging from 0.02 g to 4.0 g. Each filament was applied to the bending point five times at 15–30 s intervals, on five different points of each vibrissal pad. The stimulus intensity was presented in a sequential ascending order. The lowest force that evoked 50% of withdrawal responses was considered the threshold. Responses to mechanical stimuli were explored bilaterally before the CCI surgery, and 8, 15 and 26 days after surgery.

Eight and twenty-six days after CCI-IoN, and 1–2 days following the last behavioral test, rats were anesthetized with pentobarbital (Dolethal, 50 mg/Kg, i.p.), had their snouts shaved, and then they received an i.v. injection of EB (Sigma; 50 mg/Kg of a 50 mg/ml solution in 0.9% saline) in the tail vein. Immediately after injection, the appropriate topical treatment (CAP, VEH, SAL) was applied bilaterally on the whisker pad. After 10 minutes the animals were perfused through the ascending aorta with saline (0.9% NaCl), and the trigeminal pads were quickly removed and transferred to an oven at 56°C for 2 days. Formamide (Sigma) was used to extract the EB out of the tissue [34], and local extravasation was estimated from the EB concentration measured with a spectrophotometer (Molecular Devices) at a wavelength of 620 nm. Because of the inherent intertrial variability, it was necessary to create a standard curve for each experimental session, against which the experimental results could be normalized. The absorbance values over a range of EB-formamide concentrations were adjusted to a sigmoidal curve through non-linear regression, and the concentration of EB corresponding to saturation was used as 100% reference for all experimental values obtained within the same session. The group means and dispersion values were then represented as percentages of saturation. Moreover, various animals from different groups were coded and processed in every measuring session.

After CCI surgery animals were coded by subcutaneous electronic tags, so that all behavioral tests and EB procedures were carried out with the experimenter blinded to the type of diet consumed and treatment received by the rats. For all statistical analysis Statgraphics Plus 4.0 were used. Side comparisons within the same group were made with a two-tailed t-test. For intergroup comparisons with factors treatment (CAP, VEH or SAL) and manipulation (CCI, sham or naive) a two-way analysis of variance (ANOVA) was applied when parametric conditions were met, or the non-parametric Kruskal-Wallis test otherwise. If multiple comparisons yielded significance, two-by-two comparisons were made with a t-test or Wilcoxon, as appropriate. The significance level was set at p < 0.05.