OUP user menu

Delayed inflammation in rat meninges: implications for migraine pathophysiology

Uwe Reuter, Hayrunnisa Bolay, Inger Jansen-Olesen, Alberto Chiarugi, Margarita Sanchez del Rio, Richard Letourneau, Theoharis C. Theoharides, Christian Waeber, Michael A. Moskowitz
DOI: http://dx.doi.org/10.1093/brain/124.12.2490 2490-2502 First published online: 1 December 2001


Nitric oxide (NO) has been implicated in migraine pathogenesis based on the delayed development of typical migraine headache 4–6 h after infusing the NO donor nitroglycerin [glyceryl trinitrate (GTN)] to migraineurs. Furthermore, inhibiting the synthesis of NO by treatment with a NO synthase (NOS) inhibitor attenuates spontaneous migraine headaches in 67% of subjects. Because NO has been linked to inflammation and cytokine expression, we investigated the delayed consequences of brief GTN infusion (30 min) on the development of meningeal inflammation in a rat model using doses relevant to the human model. We found dose-dependent Type II NOS [inducible NOS (iNOS)] mRNA upregulation in dura mater beginning at 2 h and an increase in the corresponding protein expression at 4, 6 and 10 h after infusion. Type II NOS immunoreactivity was expressed chiefly within resident meningeal macrophages. Consistent with development of a delayed inflammatory response, we detected induction of interleukin 1β in dura mater at 2 and 6 h and increased interleukin 6 in dural macrophages and in rat cerebrospinal fluid at 6 h after GTN infusion. Myeloperoxidase-positive cells were rarely found. Leakage of plasma proteins from dural blood vessels was first detected 4 h after GTN infusion, and this was suppressed by administering a specific Type II NOS inhibitor [L-N(6)-(1-iminoethyl)-lysine (L-NIL)]. In addition to cytokine induction, macrophage iNOS upregulation and oedema formation after GTN infusion, dural mast cells exhibited granular changes consistent with secretion at 4 and 6 h. Because iNOS was expressed in dural macrophages following topical GTN, and in the spleen after intravenous injection, the data suggest that the inflammatory response is mediated by direct actions on the dura and does not develop secondary to events within the brain. Our findings point to the importance of new gene expression and cytokine expression as fundamental to the delayed response following GTN infusion, and support the hypothesis that a similar response develops in human meninges after GTN challenge.

  • migraine
  • nitroglycerin
  • inflammation
  • nitric oxide
  • macrophages
  • ED-2 = a membrane antigen expressed on resident macrophages
  • FITC = fluoroisothiocyanate
  • GTN = glyceryl trinitrate
  • IL = interleukin
  • LPS = lipopolysaccharide
  • L-NIL = l-N(6)-(1-iminoethyl)-lysine
  • L-NMMA = n(G)-monomethyl-l-arginine
  • NO = nitric oxide
  • NOS = NO synthase
  • NOS Type II = Type II NOS
  • iNOS = inducible Type II NOS
  • TNC = trigeminal nucleus caudalis


Despite the relatively high prevalence of migraine headache, the physiological mechanisms and biochemical pathways initiating headaches are largely unknown. Inflammation within dura mater has been proposed as fundamental to the pathogenesis of migraine headaches (Moskowitz, 1992). Anti-inflammatory drugs such as acetylsalicylic acid and non-steroidal anti-inflammatory drugs are reportedly useful prophylactic and acute anti-migraine treatments and even corticosteroids are used on occasion (Buring et al., 1990; Schulman and Silberstein, 1992; Diener, 1999). Components of the inflammatory response such as vasodilation, oedema and mast cell degranulation are evoked by electrical or chemical trigeminal nerve stimulation following neuropeptide release from perivascular axons innervating the dura mater (Markowitz et al., 1987; Buzzi et al., 1991; Dimitriadou et al., 1991). Moreover, meningeal oedema was recently observed in a preliminary study in humans during migraine attacks using single photon emission computed tomography (Pappagallo et al., 1999).

Nitric oxide (NO), a short-lived vasodilator and weak oxygen radical, has been implicated in the genesis of migraine headache beginning with the observations of Wolff (Wolff, 1929). More recently, nitroglycerin (GTN) caused 11 out of 17 migraineurs (66%) to develop delayed migraine attacks 4–6 h after intravenous infusion (Olesen et al., 1993), whereas control subjects only experienced a short-lived (minutes) immediate headache (as did the migraineurs). In addition, NO is linked to spontaneous migraine attacks based on the efficacy of a NO synthase (NOS) inhibitor, n(G)-monomethyl-l-arginine (L-NMMA), to decrease pain in delayed onset headache (and accompanying photo- and phonophobia) in 10 out of 15 treated subjects (Lassen et al., 1997).

These human studies in a limited population, plus reports of debilitating headaches in people working in factories making explosives (Trainor and Jones, 1966), implicate NO in the genesis of headache, albeit in an unclear way. More indirect evidence is provided by studies showing that platelet levels of cyclic GMP, the second messenger of NO, as well as NO metabolites such as nitrate/nitrite are increased in migraineurs and rise further during attacks (Stepien et al., 1998, Shimomura et al., 1999). In migraineurs, especially with aura, NO production in platelets is increased following stimulation with collagen as compared with controls, and this increase varies during the ovarian cycle (Gallai et al., 1996; Sarchielli et al., 1996).

Based on the studies above, it appears that a subpopulation of migraineurs is susceptible to GTN-induced delayed headache, the features of which resemble spontaneous migraine headache. GTN is an organic nitrate with a short plasma half-life (1–4 min), although its half-life can reach 2 h in a lipophilic tissue like brain (Torfgard et al., 1991). GTN is metabolized to NO within cells by a combination of glutathione-S-transferase, cytochrome P450 and thiol reactions. NO is synthesized endogenously by nitric oxide synthases (NOS) using l-arginine as a substrate (for review, see Yun et al., 1996). At least three NOS isoforms are known, encoded by distinct genes. Endothelial (Type III) and neuronal (Type I) NOS are Ca2+ dependent and constitutively expressed. Although only Type II NOS [inducible NOS (iNOS) or NOS Type II] was classically thought to be transcriptionally regulated, more recent studies show that the expression of the other isoforms can also be induced (Sasaki et al., 2000). Type II NOS can be expressed in macrophages activated by pro-inflammatory stimuli such as bacterial lipopolysaccharides or cytokines (Forstermann and Kleinert, 1995). iNOS can generate NO for several hours or days at 100- to 1000-fold greater quantities (μM) than its constitutive counterparts, amounts that can promote tissue injury and inflammation.

Several studies have examined the effects of GTN on meninges and brain. An increase in NOS-immunoreactivity nerve fibres in dura and beading within these fibres has been observed following GTN administration (Knyihar-Csillik et al., 1999). In other studies, GTN infusion caused a long-lasting increase of cortical NO within physiological levels, increase of cortical blood flow and a reduced level of superoxide in the rat and cat (Read et al., 1997, 1999).

Cytokines and NO-generating enzymes are commonly co-expressed following challenge with an inflammatory stimulus (Van der Meide and Schellekens, 1996; MacMicking et al., 1997). Different cytokines may stimulate the same activity, which provides functional redundancy within the inflammatory and immune systems. Pro-inflammatory cytokines such as interleukin (IL)-1β and IL-6 contribute to oedema and Aδ-fibre sensitization (Goldblum et al., 1988; Herbert and Holzer, 1994; Yamasaki et al., 1995; Opree and Kress, 2000), which appear to be important features in the pathophysiology of migraine.

Based on the data implicating GTN and NO in migraine and inflammation, we investigated the delayed effects of GTN infusion on gene expression and inflammatory markers in rodent dura mater. The dura mater and particularly its attendant large blood vessels are implicated in migraine pain and are among the few intracranial tissues that evoke pain when stimulated. We describe novel pathophysiological events in rat meninges, which may also develop during migraine attacks after GTN infusion in susceptible individuals. These include upregulation of pro-inflammatory mediators, macrophage activation, oedema formation and mast cell degranulation by a direct action on meningeal tissues. Hence, these studies use a previously described model by Read and colleagues (Read et al., 1997, 1999) to elucidate the pathophysiological mechanisms in meninges by which NO promotes the development of migraine headaches.

Material and methods

Animal preparation

Animal experiments were approved by the Massachusetts General Hospital, Committee on Research, Subcommittee on Research Animal Care. A total of 247 rats were used for this study. Adult male Sprague–Dawley rats (200–250 g, Charles River, Wilmington, Mass., USA) were anaesthetized with pentobarbital (50 mg/kg i.p.; Veterinary Laboratories, Lenexa, Kan., USA). The animals were placed on a heating blanket and the temperature was controlled and maintained at 37.0 ± 0.5°C using a rectal probe. After disinfecting the skin with an iodine preparation solution, the left femoral vein was exposed and a polyethylene catheter (PE 50, Becton Dickinson, Franklin Lakes, NJ, USA) was inserted for the administration of glyceryl trinitrate (GTN) (2 μg/kg/min for 30 min), vehicle (10% ethylene glycol in saline) or saline (0.9% NaCl). The infusion rate was set at 1 ml/h. In a subset of animals (n = 16) a PE 50 catheter was placed in the ipsilateral femoral artery to monitor arterial blood pressure and heart rate (MacLab software, ADInstruments, Mountain View, Calif., USA). Blood samples were occasionally taken to monitor arterial blood gases. Following the surgical procedure, GTN, vehicle or saline was infused for 30 min. Catheters were then removed and the blood vessels were tied. The wounds were sutured and 5% lidocaine paste was applied to the skin around the scar. The animals were placed in a warming chamber for 2 h at 31°C and then put back into their cage with food and water ad libitum. In a subset of GTN- and vehicle-infused animals (n = 3 per group), rectal temperature was also measured when the rats were taken out of the warming chamber. Rats were sacrificed with an overdose of pentobarbital and perfused transcardially as described below.

To establish direct effects of GTN on dura mater following topical application, rats (n = 4) were anaesthetized as described above, placed on a heating blanket and fixed in a stereotaxic frame (Stoelting, Wood Dale, Ill., USA). The skin of the scalp was opened by a midline incision and retracted laterally. One small burr hole (diameter ~2–3 mm) was carefully drilled 2 mm posterior and 1.5 mm laterally of the bregma under permanent cooling with saline. The bone in the burr hole was retracted so that the intact dura mater was exposed. The burr hole was then filled with 0.9% NaCl for the following 30 min. The NaCl was removed and the dura mater was exposed to either 10 μM GTN or vehicle for 30 min. After carefully removing the fluid, the dura was washed several times with 0.9% NaCl. The bone flap was placed back in the burr hole. A small piece of sterile sponge (Upjohn, Kalamazoo, Mich., USA) was placed above the craniotomy and the skin of the scalp was closed by suturing. Lidocaine paste was applied to the scar. The rats were treated as described above and sacrificed after 6 h. Dura mater was used for iNOS immunohistochemistry.

Reverse transcriptase–polymerase chain reaction

Rats (n = 30) were perfused transcardially with 250 ml cold 0.9% NaCl (4°C). The dura was carefully dissected out and placed in an autoclaved Eppendorf tube for immediate freezing in liquid nitrogen. Total RNA was extracted using the Trizol (Gibco, Rockville, Md., USA) extraction method. Concentration and purity of RNA were determined spectrophotometrically (Beckman Spectrophotometer BU 640) by measuring absorbency at 260/280 nm. RNA was reverse-transcribed to first strand DNA with a Superscript II™ kit (Life Technologies, Rockville, Md., USA) using random priming. PCR (polymerase chain reaction) amplification of cDNA was performed in a programmable thermal cycler (MJ Research, Cambridge, Mass., USA) with the following primers: iNOS forward primer 5′-CGCCAAGAACGTGTTCACCA-3′, iNOS reverse primer 5′-AGCAGGCACACGCAATGATG-3′, GADPH forward primer 5′-TAAAGGGCATCCTGGGCTACACT-3′, GADPH reverse primer 5′-TTACTCCTTCGAGGCCATGTAGG- 3′. The reaction mixture contained 1× reaction buffer, 200 μmol deoxynucleotides, 0.05 units/μl Red Taq DNA polymerase (Sigma, St Louis, Mo., USA), 1.5 nM MgCl2, 0.5 μM of each primer and 1 μl cDNA. PCR was performed at 95°C for 5 min followed by 38 cycles through 95°C for 45 s, 60°C for 30 s, and 72°C for 45 s (iNOS) and 30 cycles through 94°C for 30 s, 55°C for 15 s and 72°C for 1 min (GADPH). PCR products were separated by gel electrophoresis using a 1% agarose gel in TAE buffer (0.04 M Tris-acetate, 0.001 M EDTA) containing ethidium bromide (10 mg/ml; Fisher Scientific, Pittsburgh, Pa., USA) and visualized with an imaging system (Gel Doc, Biorad, Hercules, Calif., USA). RNA extracted from rat dura mater of rats treated with lipopolysaccharide (LPS) (5 mg/kg i.p.; Sigma) was used for positive control of iNOS expression.

Western blot analysis

Rats (n = 89) were perfused transcardially with 250 ml cold 0.9% NaCl. The dura was dissected out and frozen in 2-methylbutane at –40°C and stored at –80°C until further processed. Dura mater samples (10 mg) were homogenized with an electric grinder on ice in 250 μl buffer containing 10 mM Hepes (pH 7.4), 0.42 M KCl, 5 mM MgCl2, 1 mM EDTA (ethylenediaminetetraacetic acid), 1 mM EGTA (ethyleneglycoltetraacetic acid), 1 mM PMSF (phenylmethylsulphonyl fluoride), 1 mM DTT (dithiothreitol) and a protease inhibitor cocktail (dilution 1 : 50; Sigma, St Louis, Mo., USA). The specimen was centrifuged at 20.800 g for 30 min at 4°C. Tissue homogenization was continued as recently described (Matsushita et al., 2000). Whole cell lysates were separated on Tris–glycine gels (Invitrogen Corporation, Carlsbad, Calif., USA) and transferred onto a nitrocellulose membrane. Blots were incubated overnight with rabbit polyclonal anti-iNOS serum (1 : 1250; Santa Cruz Biotechnology, Santa Cruz, Calif., USA) and then probed for 2 h with an anti-rabbit horseradish peroxidase coupled secondary antibody (1 : 7500; Amersham, Little Chalfont, Bucks., UK). An enhanced chemoluminescence system (ECL+PLUS; Amersham) was used for visualization. Membranes were also probed (2 h) for β-actin, a housekeeping protein, with a mouse polyclonal anti β-actin antibody (1 : 5000; Sigma) followed by 2 h incubation with horseradish peroxidase coupled anti-mouse IgG.

A rabbit anti IL-1β antibody (1 : 1000, Santa Cruz) was used for the determination of IL-1β in dura mater. Type II NOS and IL-1β protein expression were examined 30 min, 1 h, 2 h, 4 h, 6 h and 10 h after GTN infusion (n = 3 per group; 2 μg/kg/min for 30 min).

Optical density measurement for IL-1β/iNOS was performed by dividing the intensity of the IL-1β/iNOS bands by the intensity of the house-keeping protein (β-actin, 1 : 5000; Sigma) at each time point.

The anti-IL-1β and anti-actin antibodies recognized only a single band at the expected size. To test for cross-reactivity, the NOS Type II antibody was pre-absorbed with the immunizing peptide. The iNOS band was eliminated. Protein extracted from dura mater of LPS-treated rats (5 mg/kg, i.p.; Sigma) was used as a positive control for iNOS and IL-1β induction.


Rats (n = 46) were perfused transcardially with 250 ml 0.9% NaCl followed by 1.5 l of fresh cold 4% paraformaldehyde in 0.1 M PBS (phosphate-buffered saline) (pH 7.4). The skull with attached dura was post-fixed overnight (4°C) in 4% paraformaldehyde after which the dura mater was dissected out. To block endogenous peroxidase activity, the specimen was placed in 0.3% H2O2 in 0.1 M PBS for 30 min and then washed 3 × 10 min in 0.1 M PBS (pH 7.4). Non-specific antibody binding was blocked by incubation in 10% normal donkey serum, 0.3% Triton-X in 0.1 M PBS for 2 h at 4°C. Whole-mount dura mater was incubated overnight with a polyclonal rabbit anti-iNOS antibody (1 : 1000; BD Transduction Laboratory, Franklin Lakes, NJ, USA) in 2% normal donkey serum and 0.3% Triton-X at 4°C. Whole dural mounts were then washed 3 × 10 min in PBS, and incubated for 1 h with FITC (fluoroisothiocyanate)-labelled donkey anti-rabbit IgG (1 : 200) at room temperature. For macrophage staining (McMenamin, 1999), the dura mater was incubated overnight in polyclonal mouse anti-ED-2 (a membrane antigen expressed on resident macrophages) serum (1 : 1000; at 4°C; RDI, Flanders, NJ, USA) followed by 1 h in Cy3-labelled donkey anti-mouse IgG (1 : 200, RT). IL-6 immunohistochemistry was performed as described above using a goat anti-rat IL-6 antibody (1 : 500; R&D Systems, Minneapolis, Minn., USA) and a donkey anti-goat FITC-labelled secondary antibody (1 : 200). All secondary antibodies were obtained from Jackson ImmunoResearch Laboratories (West Grove, Pa., USA). Sections were placed on slides, air-dried and mounted with a glycerol-containing mounting medium (SlowFade®; Molecular Probes, Eugene, Oreg., USA). Prior to use, all secondary antibodies were tested for non-specific staining by omitting the primary antibodies in both GTN- and vehicle-treated tissues. No non-specific staining was observed with any of the secondary antibodies. The cross-reactivity of both (iNOS and IL-6) primary antibodies was also tested. There was no positive staining in the untreated dura mater.

For light microscopy of myeloperoxidase-positive cells (rabbit anti-myeloperoxidase IgG, 1 : 1000; Dako, Carpinteria, Calif., USA), a biotinylated secondary antibody (goat anti-rabbit IgG, 1 : 800; Vector Technology, Houston, Tex., USA) was used. The reaction product was visualized using the DAB (3,3′diaminobenzidine chloride) method as previously described by Keller and Marfurt (Keller and Marfurt, 1991). Sections were placed on a slide, air-dried, dehydrated in ascending alcohol concentrations and mounted with Permount mounting medium (Fisher Scientific, Pittsburgh, Pa., USA). Dura mater from LPS-treated animals was used as positive control for iNOS induction.

For mast cell staining, the tissue was prepared as described above. After perfusion-fixation of the animal, the dura mater was dissected out and washed in PBS (pH 7.4; 3 × 10 min) followed by incubation in methylene blue (1 g/mldistilled H2O; J. T. Baker Chemical, Philipsburg, NJ, USA) for 1–2 min at room temperature. Dura mater whole mounts were washed several times in distilled water, dehydrated and mounted with a Permount mounting medium. Mast cells, stained blue/purple, were counted in 12 random fields along the middle meningeal artery on one side of the dura mater (the other side was used for iNOS immunohistochemistry). The number of mast cells with signs of degranulation (scattered blue staining/granules outside the cell membrane) were counted and expressed as a percentage of all mast cells within this visual field. Numbers were averaged for all 12 random fields and expressed as mean (± standard deviation).

Electron microscopy

Animals (n = 12) were sacrificed immediately, 4 and 6 h after GTN (2 μg/kg/min for 30 min) or vehicle infusion by intracardiac perfusion with isotonic phosphate buffer (50 ml) followed by fixative solution (300 ml) containing 2% paraformaldehyde and 2.5% glutaraldehyde in PBS (pH 7.4). After perfusion and decapitation, the skull with attached dura was placed in the same fixative overnight. The dura was then removed and further processed as described previously (Dimitriadou et al., 1991; Buzzi et al., 1992).

IL-6 determination

Rats (n = 12) were prepared as described above. Following vehicle or GTN infusion, animals were allowed to recover in the warming chamber (31°C). They were re-anaesthetized after 4 and 6 h with pentobarbital (50 mg/kg, i.p), and a 30.5-gauge needle was introduced into the cisterna magna after a midline skin incision was made from the occipital protuberance to the cervical area. About 100 μl of CSF were withdrawn, carefully avoiding blood contamination. Samples were kept frozen at –80°C until analysed. IL-6 measurements in CSF were performed by enzyme-linked immunosorbent assay (R&D Systems) according to the manufacturer's instructions.

Plasma protein extravasation

Rats (n = 38; 200–250 g) were infused with GTN (2 μg/kg/min for 30 min) or vehicle as described above. They were re-anaesthetized with pentobarbital (50 mg/kg, i.p) 4 h later and 50 μCi/kg [125I]BSA (bovine serum albumin) (100 μCi/ml saline 0.9%; stock solution; NEN® Life Science Products, Boston, Mass., USA) were injected via the femoral vein. Fifteen minutes later the thorax was opened, the descending aorta clamped, and the right atrium incised for drainage. Normal saline (0.9%) was perfused via the left ventricle at a constant pressure of 100 mmHg for 2 min. The dura mater above the brain convexity was removed and the area of the transverse sinus was dissected out and discarded. The remaining dura was weighed and counted for radioactivity in a gamma counter (Cobra II, Packard Instruments, Downers Grove, Ill., USA). Radioactivity (c.p.m./mg dura wet weight) was compared between GTN-infused and control animals. To investigate the importance of NOS isoforms to the protein leakage after GTN or vehicle infusion, animals received n(G)-monomethyl-l-arginine (L-NMMA) (20 mg/kg, i.v.; Sigma) 90 min and 210 min after the end of GTN infusion, or the selective iNOS inhibitor l-N(6)-(1-iminoethyl)-lysine (L-NIL, 4 mg/kg, i.p.; United Biomedical, Haappauge, NY, USA), after 120 min. L-NIL exhibits about a 28-fold greater selectivity for iNOS (IC50 = 3.3 μM) than for the rat brain constitutive enzyme (IC50 = 92 μM) (Moore et al., 1994). The L-NIL dose in our study was chosen because L-NIL administered orally before the injection of LPS reduced the plasma NOx response with an ED50 of 3 mg/kg and an ED90 of 30 mg/kg (Fletcher et al., 1998).

Statistical analysis

Data in this manuscript are expressed as mean values and standard deviations. Comparisons between groups were performed using one-way ANOVA followed by Tukey's protected t-test. Statistical significance was assumed when P was <0.05.


Physiological data

Because of the well-established acute vascular and haemodynamic effects of GTN, several physiological parameters were measured before, during and shortly after GTN infusion (2 μg/kg/min for 30 min; n =10; Table 1). These data did not differ from vehicle-infused animals (n =10); in particular, mean arterial blood pressure did not decrease during and after GTN infusion, indicating that delayed changes in gene expression (at 2 μg/kg/min) and inflammation were not caused by hypotension or other early haemodynamic events.

View this table:
Table 1 (A)

Physiological parameters in GTN- and vehicle-infused rats

pHpCO2 (mmHg)pO2 (mmHg)Temperature (°C) baselineTemperature (°C) after 2 h.
Vehicle (n = 10)7.38 ± 0.0143.5 ± 1.686.2 ± 4.236.9 ± 0.337.3 ± 0.2 (n = 3)
GTN (n = 10)7.38 ± 0.0242.8 ± 1.988.2 ± 2.936.9 ± 0.337.2 ± 0.2 (n = 3)
Vehicle (n = 10)GTN (n = 10)
Rats were infused with vehicle or GTN (2 μg/kg/min). (A) Rectal temperature, pH, pCO2 (arterial pCO2) and pO2 (arterial pO2) values were measured 15 min after the beginning of the infusion. Rectal temperature was also measured after 2 h. (B) Mean arterial blood pressure (MABP) and heart rate (HR) were measured at baseline (2 min before infusion) and at the indicated time points after the beginning of the infusion. Values are given as mean (± 1 SD).
Baseline112 ± 11381 ± 60114 ± 19389 ± 46
1–2 min112 ± 11387 ± 56115 ± 19381 ± 43
5–6 min111 ± 10384 ± 55109 ± 19386 ± 44
25–26 min113 ± 12398 ± 57115 ± 16406 ± 34

Expression of pro-inflammatory cytokines in dura mater and CSF after GTN infusion

NO and cytokines contribute to inflammation and sensitization of primary afferents (Goldblum et al., 1988; Herbert and Holzer, 1994; Yamasaki et al., 1995; Opree and Kress, 2000). We therefore investigated whether GTN infusion altered IL-1β and IL-6 expression in dura mater. GTN caused a biphasic increase in IL-1β protein with a rapid rise at 1 h, followed by a second peak at 6 h (Fig. 1). Vehicle infusion did not cause IL-1β expression at 1, 2, 4 and 6 h. Scattered IL-6 immunoreactivity was distributed along the middle meningeal artery 6 h after GTN, but not vehicle infusion (Fig. 2). Since IL-6 expression is often present in activated macrophages (Tang et al., 1999), we double-stained for IL-6 and ED-2, a membrane antigen expressed on resident macrophages. IL-6 immunoreactivity almost exclusively co-localized with ED-2-positive cells (Fig. 2). In addition, IL-6 levels increased in rat CSF 6 h after GTN, but not vehicle infusion (n = 6 in each group). IL-6 levels were 66 ± 12 pg/ml and 153 ± 43 pg/ml at 4 and 6 h after GTN infusion, versus 23 ± 5 pg/ml 6 h after vehicle infusion (P < 0.05 at 6 h).

Fig. 1

IL-1β protein expression after GTN infusion. IL-1β protein expression in rat dura mater homogenates was analysed using Western blots. A representative image is shown in A. Densitometric measurements of IL-1β protein expression (B) show a biphasic increase after GTN infusion (2 μg/kg/min for 30 min) with a peak protein expression at 1 and 6 h (n = 3 per time point). IL-1β protein expression 6 h after LPS injection (5 mg i.p.) was used as positive control. Values are given as mean (± standard deviation).

Fig. 2

IL-6 expression in rat dura mater after GTN infusion. Rats were infused with GTN (2 μg/kg/min) (AC) or vehicle (DF) for 30 min. Macrophage staining (red in A and D) was performed with a specific antibody against the ED-2 antigen, which was recognized by a Cy3-labelled secondary antibody. A FITC-labelled secondary antibody was used to visualize IL-6 staining (green in B and E). Double-stained elements appear yellow in C and F. The arrow in B points to a branch of the middle meningeal artery. Scale bar = 100 μm.

Inducible nitric oxide synthase (iNOS) after GTN infusion in rat dura mater

iNOS protein was not constitutively expressed nor was its mRNA detected in vehicle-treated animals. However, beginning 4, 6 and 10 h after GTN infusion, a 130 kDa band corresponding to iNOS protein was detected (Fig. 3A). Consistent with these findings, iNOS mRNA increased 2 and 4 h after GTN infusion (but not at 30 min), anticipating a rise in iNOS protein. By 6 h, the mRNA signal was lost (Fig. 3B). INOS protein expression was not observed in brainstem and cortex, but was found in the spleen (data not shown).

Fig. 3

Expression of inducible nitric oxide synthase (iNOS) in rat dura mater after GTN infusion. (A) No iNOS protein was detected in vehicle-treated rats (6 h after infusion), or in rats receiving GTN up to 2 h after infusion. iNOS protein appeared at 4 h and peaked at 6 h, decreasing at 10 h post-infusion. LPS injection (5 mg i.p.) was used as positive control. β-Actin (a house-keeping protein) levels did not change. iNOS mRNA preceded protein expression by ~2 h as indicated by reverse transcriptase–PCR. A representative of five similar experiments is shown in B. (C) Dose–response curve for iNOS protein expression (6 h) in dural extracts as detected by Western blots (representative of three experiments performed with tissues pooled from two rats for each GTN dose). β-Actin levels were not affected by the treatment. (D and E) Densitometric measurements of iNOS protein levels on Western blots time course (n = 3 per time point) and dose response (n = 3 per dose), corresponding typical blots are shown in A and C.

A dose response for iNOS was established at the 6 h time point (0.2–20 μg/kg/min i.v. for 30 min). GTN at 1 μg/kg/min was the lowest dose producing a detectable increase in iNOS protein (Fig. 3C). The level of iNOS expression at 20 μg/kg/min was only marginally higher than at 2 μg/kg/min. In all subsequent experiments a dose of 2 μg/kg/min was used as previously reported (Read et al., 1999). Higher doses (20 μg/kg/min) caused a significant decrease in blood pressure (data not shown) and drops in blood pressure increase immediate early gene expression (Graham et al., 1995).

Cellular source for iNOS protein

iNOS was found almost entirely in ED-2-positive cells within dura mater using fluorescence microscopy (Fig. 4). The number of macrophages did not appear to differ between groups, but ED-2-staining was more intense 4 and 6 h after GTN (two observers naive to the treatment group). To determine if leucocytes infiltrate into dura mater after GTN, myeloperoxidase staining was performed on whole mount preparations 0–6 h after GTN. Only rare myeloperoxidase-positive cells were detected in each sample, which did not change over time. This suggests that myeloperoxidase-positive leucocytes do not play a major role after GTN infusion. Following topical GTN administration, strong iNOS immunoreactivity was observed in perivascular ED-2-positive macrophages along the middle menigeal artery and its branches, which differed significantly from the vehicle-treated group (where some iNOS immunoreactivity was restricted to the burr hole area).

Fig. 4

iNOS protein co-localizes in macrophages after GTN treatment. Rats were infused with GTN (2 μg/kg/min) (AC) or vehicle (DF) for 30 min. Macrophages were stained with the specific antibody ED-2 which was recognized by a Cy3-labelled secondary antibody (red in A and D). The iNOS antibody was recognized by a FITC-labelled secondary antibody (green in B and E). Double-stained cells appear yellow in C and F. While the vast majority of iNOS-positive cells appear to be macrophages, some iNOS-stained cells were not ED-2 positive (arrow heads), indicating that iNOS may also be expressed by another cell type. The arrow in B points to a branch of the middle meningeal artery. Scale bar = 100 μm.

GTN infusion causes plasma protein leakage in dura mater

Plasma protein leakage from blood vessels may be caused by increased NO production and oedema formation is an important component of inflammation. Extravasation of [125I]BSA in dura mater 4 h after GTN infusion increased from 6.2 ± 0.7 c.p.m./mg wet weight (n = 12) in the vehicle group to 11.7 ± 1.4 c.p.m./mg wet weight after GTN (n = 14, P < 0.05). L-NMMA administration attenuated protein leakage by 47% (n = 6; P < 0.05). A selective iNOS inhibitor, L-NIL, significantly reduced GTN induced protein extravasation by 77% to 7.4 ± 1.4 c.p.m. (n = 6; P < 0.05). These data establish the importance of the iNOS isoform (NOS Type II) for GTN induced delayed events in meninges (Fig. 5).

Fig. 5

NOS inhibition attenuates GTN-induced dural plasma protein extravasation. GTN (2 μg/kg/min) was infused 4 h before [125I]BSA injection into the femoral vein. Rats were perfused with saline 15 min later and radioactivity was counted in the dura mater as described in Material and methods. The non-selective NOS inhibitor L-NMMA (administered 90 and 210 min after GTN) and the iNOS selective inhibitor L-NIL (administered 90 min after GTN) both significantly attenuated GTN-induced plasma leakage (P < 0.05 and P < 0.01, respectively).

Mast-cell immunohistochemistry

Mast-cell activation is associated with protein extravasation and inflammation in dura mater. Therefore we investigated whether GTN infusion leads to dura mast cell activation. A significant number of mast cells showed signs of degranulation, as determined by the presence of methylene blue-stained granular content outside the perimeter of mast cells, along the middle meningeal artery 6 h after GTN infusion. The degree of activation increased from 17 ± 5% in vehicle-treated animals (2 ± 1 out of 16 ± 4 mast cells per field, n = 4) to 48 ± 5% after GTN (6 ± 2 cells per field out of 12 ± 3 mast cells per field, n = 4; P < 0.01). Because completely degranulated mast cells do not exhibit metachromatic staining with methylene blue, the percentage of activated mast cells could be underestimated by this technique.

These findings were confirmed by electron microscopy, which showed that mast cells lost filopodia and had fewer electron-dense granules 4 h and 6 h after GTN treatment (n = 2 per group and time point; Fig. 6). It should be noted that many secretory granules had altered shapes and content, not recognizable by light microscopy, indicative of secretion. Mast cells investigated immediately after GTN infusion had very few such changes (results not shown).

Fig. 6

Ultrastructural changes in rat dural mast cells 4 h after GTN infusion (B and B′) compared with vehicle (A and A′). (A and A′) Normal mast cells exhibit filopodia (arrowhead in A) and numerous electron-dense granules (G). (B and B′) Notice the disappearance of plasma membrane folds, the presence of electron-lucent secretory granules (g) showing a progressive dissolution (secretion) of their contents. In B′, some granule contents are also visible lying outside the cell (arrow). Scale bar = 2 μm.


Our study shows that nitroglycerin infusion causes a delayed inflammatory response within the rat dura mater, increases expression of Type II or inducible NOS within resident macrophages, and upregulates the pro-inflammatory cytokines IL-1β and IL-6. Four hours after infusion, we detected augmented plasma protein extravasation in dura mater, a marker of inflammation, which was attenuated by a specific Type II NOS inhibitor (L-NIL). At this time, GTN partially degranulated dural mast cells, consistent with the genesis of a delayed inflammatory response. Taken together, these data suggest that GTN, either directly or indirectly, produces a delayed inflammation within rat dura mater. We postulate that inflammation within dura mater provides an important substrate for the development of headache in man.

We believe that NO is the mediator of the delayed changes induced by GTN in the present study. Read and colleagues demonstrated previously that following administration of GTN (2 μg/kg/min) according to an identical protocol, NO concentrations increase within physiological range in rat cortex during infusion and remain elevated for at least another 30 min post-infusion (Read et al., 1999). Although GTN can accumulate and reach toxic levels in adipose tissue and lipid-rich organs such as brain, this seems unlikely to occur in the fibrous dura mater (Torfgard et al., 1991). Moreover, the GTN dosage used by Torfgard was 1000-fold higher than the dose we used. NO reacts with superoxide anions to form a powerful oxidant, peroxynitrite, which in turn can promote tissue injury by oxidation and nitration of proteins and other macromolecules. Interestingly, levels of superoxide anion decrease following GTN infusion in rat cortex (Read et al., 1999). This decrease could be indicative of peroxynitrite anion formation or may be unrelated to NO generation. Additional studies will be needed to clarify this point.

The brain may not be an important player in this model of meningeal inflammation because iNOS upregulation was found in other tissues with a high concentration of macrophages (e.g. spleen), and because iNOS was not upregulated in brain following GTN intravenous infusion. In addition, topical GTN caused iNOS upregulation in dural macrophages, suggesting that NO produced locally from GTN in dura mater contributes to inflammation within this tissue. In our view, a similar mechanism accounts for inflammation in dura mater after systemic GTN infusion.

The response to GTN resembles what has been reported after trigeminal nerve stimulation, notably oedema, leakage of plasma proteins, vasodilation and mast cell degranulation (Markowitz et al., 1987; Dimitriadou et al., 1991; Goadsby, 1993; Ebersberger et al., 1999). Underlying mechanisms in the two models appear to overlap but are not identical. For example, CGRP (calcitonin gene-related peptide) is released after trigeminal nerve stimulation (Limmroth et al., 2001) as well as during migraine attacks, presumably from meningeal blood vessels (Goadsby et al., 1990) and topical application of GTN to meningeal vessels promotes relaxation, in part, by releasing CGRP from trigeminal axons (Wei et al., 1992). Since both an NO donor (Masini et al., 1994) and CGRP/substance P trigger mast cell degranulation and CGRP potentiates IL-6 release from isolated peritoneal macrophages after LPS stimulation in an NO-dependent manner (Tang et al., 1999), we speculate that CGRP is a likely mediator of GTN-induced response in dura mater.

Type II NOS expression and NO generation play a key role in the development of inflammation in other tissues. Macrophages cultured from numerous tissues express cytokines and Type II NOS after LPS/IFN-γ challenge or exposure to IL-1β or TNF (tumour necrosis factor)-α (for review, see MacMicking et al., 1997). Type II NOS is usually expressed in leucocytes or antigen-presenting cells after injury, trauma and after administration of pro-inflammatory stimuli. This enzyme is also upregulated in models of inflammation such as endotoxin shock, adjuvant-induced arthritis, carrageenin-induced pleuritis and paw oedema (Cannon et al., 1996; Handy and Moore, 1998; Wolkow, 1998; Cuzzocrea et al., 2000; Petrov et al., 2000). Type II NOS expression leads to tissue oedema in models of inflammatory pain and after LPS challenge (Iuvone et al., 1998). In carrageenin-induced inflammation, pretreatment with the selective iNOS inhibitor L-NIL also reduces delayed oedema formed in the hind-paw, as does L-NMMA (Handy and Moore, 1998). In the present study, we showed that iNOS activity is important for development of delayed plasma protein extravasation in meninges following GTN infusion. Hence, both models may be dependent upon iNOS activity.

Like NO, IL-6 and IL-1β can be pro-inflammatory (Cuzzocrea et al., 1999; Lopez-Figueroa et al., 2000). After GTN infusion, IL-1β expression increased at 1 h and again at 6 h. The first burst of IL-1β may induce iNOS expression in dural macrophages in a way similar to IL-1β-induced Type II NOS expression in rat heart and brain (Balligand et al., 1995; Lopez-Figueroa et al., 2000). Possibly, the second IL-1β peak develops in response to NO generated by induced iNOS, which is consistent with findings in hypothalamus following administration of the NO-donor sodium nitroprusside (Tringali et al., 1999). Although the above is most consistent with our formulation, other reports imply an inhibitory effect of NO on IL-1β expression (Schroeder et al., 1999), pointing to important cell type specific differences.

The expression of IL-6 is transcriptionally regulated by the nuclear factor kappa B and so is the expression of iNOS, as shown in stimulated macrophages and other cell types (Xie et al., 1994; Grigoriadis et al., 1996; Lin et al., 1997). Depending upon the model, NO may either increase or decrease iNOS expression (Chen et al., 1995; Togashi et al., 1997; Peng et al., 1998). Our study shows that GTN-derived NO induces iNOS expression, in agreement with the pro-inflammatory effects of the NO donor DETA (diethylenetriamine)/NO. Low dose application of this agent causes nuclear factor kappa B translocation and increases iNOS mRNA in macrophages following IFN-γ plus LPS administration (Sheffler et al., 1995; Connelly et al., 2001). In the heart, NO donors cause nuclear translocation of nuclear factor kappa B to increase iNOS expression and further NO generation (Xuan et al., 1999). Further studies with inhibitors of the nuclear factor kappa B transcription pathway will establish its role in the induction of iNOS and cytokines after GTN administration.

Clinical implications

In addition to cited clinical data and the occurrence of headache in workers manufacturing explosives, NO has also been implicated in migraine pathophysiology based upon its activity in experimental migraine models. Some models use c-fos expression within trigeminal nucleus caudalis (TNC) as a surrogate for neuronal activation. Induction of c-fos in laminae I, II neurons of TNC follows noxious tissue stimulation and depends in part upon small diameter primary sensory inputs including those projecting from the meninges (Nozaki et al., 1992). Sagittal sinus stimulation induces c-fos expression within ipsilateral TNC (Kaube et al., 1993) and the response is attenuated by administering L-NAME (n(G)-nitro-l-arginine methyl ester), a non-selective NOS inhibitor (Hoskin et al., 1999). Pardutz et al. (2000) and Tassorelli and Joseph (1995) found that delayed c-fos and neuronal NOS expression (4 h) develop after bolus nitroglycerin injection (10 mg/kg; s.c.) in a similar distribution within TNC as observed following direct trigeminal nerve stimulation (Kaube et al., 1993). However, it would be important to determine whether the c-fos response reflects direct effects of GTN and NO on TNC or indirect effects via inflammation and trigeminal nerve activation.

We believe these experiments in rodents are relevant to the GTN response in man. At least two hypotheses may account for the development of migraine headache after GTN. GTN and its derivative NO may upregulate cytokines and iNOS, and cause delayed inflammation. The time for transcription and translation-dependent mechanisms explain the latency of this delayed response in dura mater. Upregulation of cytokines and NO generated from iNOS sensitize (or less likely, discharge) small unmyelinated trigeminal fibres to thereby generate pain and headache. Activation of trigeminal afferents, in turn, promotes neuropeptide release and neurogenic oedema within the meninges. Alternatively, or in addition, GTN may directly activate trigeminal afferents innervating the dura mater, causing early headache followed by macrophage activation, inflammation and delayed headache. The two explanations are not mutually exclusive and are testable.

The induction of delayed headache selectively in migraineurs could be genetically determined and may relate either to an augmented response to GTN (iNOS, cytokines, mast cells), a higher sensitivity to cytokines together with iNOS-derived NO and mast cell mediators, and/or sensitization of primary afferents of the trigeminovascular system. Future studies will be needed to clarify this point. Additionally, we suggest that meningeal inflammation and subsequent headache may not be unique to GTN infusion; rather oedema formation, macrophage activation, gene upregulation and mast-cell degranulation may reflect a prototypical response of human meninges to a host of drugs which provoke headache, and might represent novel targets for treatment approaches which prevent or abort headache. In susceptible individuals, this tissue response may explain the genesis of delayed migraine headache following the ingestion of certain foods and beverages, and, as postulated previously, may be caused by events generated within cortical grey matter adjacent to meninges, such as cortical spreading depression (Moskowitz et al., 1993).


The authors wish to thank Andrew A. Parsons, Neuroscience Research, SmithKline Beecham Pharmaceuticals, New Frontiers Science Park, Harlow, Essex, UK for his helpful suggestions to choose the GTN dose. This work was supported by grants from the NIH 5P01 NS35611 (to M.A.M. and C.W.), the International Headache Society Research Fellowship 1999 (to U.R.) and the Deutsche Forschungsgemeinschaft (Re-1316/1–1).


View Abstract