Address correspondence and reprint requests to Dr Francesc Artigas, Department of Neurochemistry, Institut d' Investigacions Biomèdiques de Barcelona (CSIC), IDIBAPS, Rosselló, 161, 6th floor, 08036 Barcelona, Spain. E-mail: email@example.com
Pyramidal neurons of the medial prefrontal cortex (mPFC) project to midbrain serotonergic neurons and control their activity. The stimulation of prefrontal 5-HT2A and AMPA receptors increases pyramidal and serotonergic cell firing, and 5-hydroxytryptamine (5-HT) release in mPFC. As the mPFC contains abundant α1-adrenoceptors whose activation increases the excitability of pyramidal neurons, we examined the effects of their stimulation on local 5-HT release, using microdialysis. The application of the α1-adrenoceptor agonist cirazoline by reverse dialysis increased the prefrontal 5-HT release in a concentration-dependent manner, an effect antagonized by coperfusion of TTX, prazosin (α1-adrenoceptor antagonist), BAY × 3702 (5-HT1A agonist), NBQX (AMPA/KA antagonist) and 1S,3S-ACPD (mGluR II/III agonist), but not by MK-801 (NMDA antagonist). Cirazoline also enhanced the increase in 5-HT release induced by DOI (5-HT2A/2C agonist) and AMPA. In addition, M100907 (5-HT2A antagonist) but not SB-242084 (5-HT2C antagonist) reversed the cirazoline- and AMPA-induced 5-HT release. These results suggest that the stimulation of prefrontal α1-adrenoceptors activates pyramidal afferents to ascending serotonergic neurons. The effect of cirazoline was also reversed by coperfusion of classical (chlorpromazine, haloperidol) and atypical (clozapine, olanzapine) antipsychotics, which suggests that a functional antagonism of the α1-adrenoceptor-mediated activation of prefrontal neurons may partly underlie their therapeutic action.
Male Wistar rats (Iffa Credo, Lyon, France) weighing 280–320 g at the time of the experiments were used. The animals were housed in groups of four per cage until the onset of the experiments and kept under a controlled temperature of 22 ± 2°C and a 12-h lighting cycle (lights on at 07 : 00 h). After surgery, rats were housed individually. Food and water were always freely available throughout the experiments. All experimental procedures were in strict compliance with the Spanish legislation and the European Communities Council Directive on ‘Protection of Animals Used in Experimental and Other Scientific Purposes’ of 24 November 1986 (86/609/EEC).
Drugs and reagents
5-HT oxalate, (S)-AMPA (alpha-amino-3-hydroxy-5-methyl-4-isoxazole-4-propionate), chlorpromazine, cirazoline, DOI (1-[2,5-dimethoxy-4-iodophenyl-2-aminopropane]), (+)-MK-801 (dizocilpine), NBQX (2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo(f)quinoxaline), SB 242084 (6-chloro-5-methyl-1-[6-(2-methylpyridin-3-yloxy)pyridin-3-ylcarbamoyl]indoline), prazosin and tetrodotoxin (TTX), were from Sigma/RBI (Natick, MA, USA). 1S,3S-ACPD (1S,3S-aminecyclopentane dicarboxylic acid), haloperidol and clozapine were from Tocris (Bristol, UK). BAY × 3702 (R-(–)-2-[4-[(chroman-2-ylmethyl)-amino]-butyl]-1,1-dioxo-benzo[d]isothiazolone·HCl), citalopram·HBr, M100907 (R-(+)-alpha-(2,3-dimethoxyphenyl)-1-[4-fluorophenylethyl]-4-piperidinemethanol; Lilly code LY 368675) and olanzapine were from Bayer AG, Lundbeck A/S and Eli Lilly & Co, respectively. Other materials and reagents were from local commercial sources. Drugs were dissolved in the perfusion fluid or water (except clozapine, dissolved in acetic acid, and olanzapine, dissolved in HCl). Concentrated solutions (1 mm; pH adjusted to 6.5–7 with NaHCO3 when necessary) were stored at − 80°C and working solutions were prepared daily by dilution in artificial CSF. Concentrations are expressed as free bases. Control rats were perfused for the entire experiment with artificial CSF. The bars in the figures show the period of drug application (corrected for the void volume of the system).
Surgery and microdialysis procedures
An updated description of the microdialysis procedures used can be found in Adell and Artigas (1998). Briefly, anesthetized rats (sodium pentobarbital, 60 mg/kg i.p.) were stereotaxically implanted with concentric microdialysis probes equipped with a Cuprophan membrane. In most experiments rats were implanted with one probe in mPFC at the following coordinates (in mm): AP +3.2, L −0.8, DV −6.0 (probe tip 4 mm) taken from bregma and dura mater (Paxinos and Watson 1986). To determine whether the perfusion of cirazoline in the mPFC increased the activity of the ascending serotonergic system, as we hypothesized, an additional experiment was conducted, in which rats were implanted with two probes, i.e. one in the mPFC, as above, and the other in the dorsal raphe nucleus (DR) at AP −7.4, L −3.1, DV − 7.5 with a lateral angle of 30° (probe tip 1.5 mm) taken also from bregma and dura mater (Paxinos and Watson 1986). All microdialysis experiments were performed in freely moving rats on the day following implants. The probes were perfused at 1.5 µL/min with artificial CSF (125 mm NaCl, 2.5 mm KCl, 1.26 mm CaCl2 and 1.18 mm MgCl2) containing 1 µm citalopram. After 1 h stabilization period, four fractions were collected to obtain basal values before local administration of drugs by reverse dialysis. Successive 20-min (30 µL) dialysate samples were collected. At the end of the experiments, rats were killed by an overdose of anesthetic. The placement of the dialysis probes was examined by perfusion of fast green dye and visual inspection of the probe track after cutting the brain at the appropriate level.
The concentrations of cirazoline and prazosin were determined in pilot experiments. Those of atypical antipsychotics were from Bortolozzi et al. (2003) whereas the rest of the drugs were used at concentrations known to reverse the increase in prefrontal 5-HT release induced by DOI (Martín-Ruiz et al. 2001). Given the in vitro nanomolar affinity of cirazoline and prazosin for α1-adrenoceptors, the use of micromolar concentrations used may appear non-selective. However, effective concentrations applied by reverse microdialysis to stimulate/block brain receptors or transporters differ typically 3–4 orders of magnitude from in vitro affinities (see for instance Hervás et al. 2000; Tao et al. 2000; Sakai and Crochet 2001; West and Grace 2002). This difference is due mainly to the low application rates used together with the continuous clearance of applied drugs via the brain capillaries and the CSF so that only a very small drug fraction reaches the target receptors. This factor is particularly important in the present study as the effect of cirazoline on 5-HT release requires the stimulation of a substantial receptor population in projection neurons to the DR in order to elicit a measurable increase in terminal 5-HT release.
The concentration of 5-HT in dialysate samples was determined by HPLC, as described by Adell and Artigas (1998). 5-HT was separated using a Beckman (San Ramon, CA, USA) 3-µm particle size column and detected with a Hewlett-Packard 1049 electrochemical detector at + 0.6 V. Retention time was between 3.5 and 4 min and the limit of detection was typically 1 fmol/sample.
Data and statistical analysis
Data (mean ± SEM) are expressed as fmol/fraction (uncorrected for recovery) and shown in figures as percentages of basal values, averaged from four predrug fractions. Statistical analysis of drug effects on dialysate 5-HT was performed using analysis of variance (anova) for repeated measures with time as repeated factor and drug concentration as independent factor. Average values of selected time periods were also calculated and compared using paired t-test. Statistical significance was set at the 95% confidence level (two tailed).
Baseline 5-HT values were 26.2 ± 0.7 fmol/fraction in mPFC and 44.4 ± 7.9 in DR (n = 196 and 8, respectively). The perfusion of artificial CSF for 4 h did not alter the 5-HT release in mPFC (Fig. 1). The local application of cirazoline (30, 100 and 300 µm) by reverse dialysis increased dialysate 5-HT in a concentration-dependent manner compared with controls receiving artificial CSF (F3,212 = 19.06, p < 0.00001, group effect; F15,315 = 32.2, p < 0.00001, time effect; F45,315 = 6.1, p < 0.00001, time–group interaction). The mean elevation once the effect of cirazoline had stabilized was 110 ± 6%, 171 ± 9% and 223 ± 14% for 30, 100 and 300 µm, respectively (Fig. 1). In fact, the two groups of eight rats correspond to two different experiments with four animals each carried out 10 months apart (Fig. 1b). Both experiments yielded the same results and the data was therefore pooled. In pilot experiments, the perfusion of increasing concentrations of cirazoline (100 and 300 µm, 2 h each) also elicited a concentration-dependent increase in 5-HT (141 ± 18% at 100 µm and 194 ± 28% at 300 µm; data not shown). The coperfusion of 1 µm TTX completely canceled the increase in 5-HT release induced by cirazoline and reduced 5-HT levels to below baseline (F9,36 = 24.9; p < 0.00001) (Fig. 2).
In double probe microdialysis experiments, the perfusion of cirazoline 300 µm in the mPFC elevated significantly the 5-HT release in both areas, although the effects was more marked in mPFC (F15,105 = 31.6; p < 0.001) than in the DR (F15,105 = 6.9; p < 0.000001) (Fig. 3a,b). The increase in dialysate 5-HT produced by the perfusion of 300 µm cirazoline in these animals was the same as that observed in animals implanted with a single probe. On the other hand, a previous dual-probe study showed that the perfusion of a CSF in the mPFC did not alter the release of 5-HT in the DR (Celada et al. 2001).
The coperfusion of the selective α1-adrenoceptor antagonist prazosin (100 and 300 µm) reversed the 5-HT increase elicited by cirazoline 300 µm (F9,36 = 6.2, p < 0.0001 at 100 µm; F9,36 = 26.4, p < 0.00001 at 300 µm). Both concentrations of prazosin were equally effective and produced a slow decline in 5-HT which nearly reached baseline values at the end of the prazosin perfusion (Fig. 4a). The infusion of 100 µm prazosin rapidly and completely reversed the 5-HT increase induced by the application of 100 µm cirazoline (F9,36 = 17.6, p < 0.00001; Fig. 4b).
Previous observations indicate that the coperfusion of the selective 5-HT1A receptor agonist BAY × 3702 reverses the increase in 5-HT release induced by the local application of DOI and AMPA in mPFC (Martín-Ruiz et al. 2001; Bortolozzi et al. 2003). This led us to examine the effect of BAY × 3702 on the effect of cirazoline. The coperfusion of 30 µm BAY × 3702 significantly reversed the increase in 5-HT release induced by 300 µm cirazoline (F9,45 = 3.6, p < 0.002; Fig. 5a). A higher concentration of BAY × 3702 (100 µm) elicited a similar antagonism (data not shown). However, 30 µm BAY × 3702 rapidly and completely reversed the effect of 100 µm cirazoline, and reduced 5-HT release to slightly below baseline (F9,27 = 7.1, p < 0.00005; Fig. 5b).
The coperfusion of cirazoline 300 µm enhanced the 5-HT elevation induced by the perfusion of DOI 100 µm (F9,63 = 9.7, p < 0.00001; Fig. 6). The stimulatory effect of DOI on 5-HT release in mPFC depends on glutamatergic transmission through AMPA receptors (Martín-Ruiz et al. 2001). We therefore examined whether the effect of cirazoline was also dependent on glutamatergic inputs in mPFC. The increase in 5-HT release elicited by 300 µm cirazoline was reversed by the coperfusion of the AMPA/KA receptor antagonist NBQX (300 µm) (F9,27 = 9.8, p < 0.00001; Fig. 7a) but not by the NMDA receptor antagonist MK-801 (Fig. 7b). Also, the non-selective mGluR II/III agonist 1S,3S-ACPD partially reversed the cirazoline-induced 5-HT increase at 3 but not at 1 mm (F9,36 = 12.0, p < 0.00001; Fig. 7c). Also, as previously shown (Martín-Ruiz et al. 2001), the local perfusion of AMPA 300 µm increased the 5-HT release (Fig. 7d). This effect was potentiated by the coperfusion of cirazoline 300 µm, which elevated 5-HT to 438 ± 34% of baseline (F9,36 = 26.8, p < 0.00001; Fig. 7d).
The close relationship between the AMPA-mediated transmission, α1-adrenoceptors and 5-HT2A receptors was also illustrated by the functional antagonism of the S-AMPA-induced 5-HT release exerted by prazosin (100 µm) and M100907 (300 µm). The coperfusion of either antagonist reversed the increase in 5-HT release produced by 300 µmS-AMPA (F9,36 = 12.0, p < 0.00001 prazosin effect; F9,27 = 10.6, p < 0.0001, M100907 effect; Fig. 8a). The perfusion of prazosin 100 µm totally reversed the 5-HT elevation induced by the application of DOI 100 µm (F9,81 = 40.2, p < 0.00001; Fig. 8b). Likewise, the coperfusion of the selective 5-HT2A receptor antagonist M100907 (300 µm) antagonized the 5-HT increase induced by cirazoline. This antagonism was partial at 300 µm cirazoline (F9,36 = 9.8, p < 0.00001; Fig. 8c) and total at 100 µm cirazoline (F9,36 = 11.2, p < 0.00001; Fig. 8d). However, the selective 5-HT2C receptor antagonist SB 242084 (100 µm) failed to significantly alter the effect of cirazoline (Fig. 8c).
The increase in 5-HT release induced by cirazoline 100 µm was also reversed by the coperfusion of 300 µm of the classical antipsychotics haloperidol and chlorpromazine (F9,36 = 14.9, p < 0.00001 and F9,45 = 14.8, p < 0.00001 for haloperidol and chlorpromazine, respectively; Fig. 9a,b). Likewise, the atypical antipsychotics clozapine and olanzapine (300 µm each) significantly reduced 5-HT levels to or below baseline (F9,36 = 10.2, p < 0.00001 and F9,27 = 23.2, p < 0.00001 for clozapine and olanzapine, respectively; Fig. 9c,d).
In additional experiments we determined the effects of the administration of the different compounds that reduced dialysate 5-HT when perfused in combination with cirazoline. For this purpose BAY × 3702 (30 µm), M100907 (300 µm), NBQX (300 µm), prazosin (100 µm), haloperidol (300 µm), chlorpromazine (300 µm) and clozapine (300 µm) were perfused alone. The response of dialysate 5-HT was averaged over the last four samples, once the maximal effect was stabilized, and expressed as the percentage change from the corresponding basal (predrug) values. Paired t-test revealed that each of these compounds, except NBQX, reduced significantly (p < 0.01) the release of 5-HT (Fig. 10).
Three main findings derive from the present study. First, the activation of α1-adrenoceptors in mPFC increases the local release of 5-HT by an impulse-dependent mechanism. Second, this effect is dependent on AMPA-mediated inputs. Finally, antipsychotic drugs reduce the basal 5-HT release and reverse the effect of α1-adrenoceptor activation, an observation possibly related to their therapeutic actions.
We would like to stress two different points relevant to the discussion of the data of this study. First, the fact that antipsychotics reverse the cirazoline-induced increase in prefrontal 5-HT release does not imply that psychotic states are necessarily associated to an increase in cortical serotonergic transmission. We used the stimulation of α1-adrenoceptors in mPFC as a mean to activate the mPFC-raphe circuit in order to explore drug interactions in vivo. Second, because the mPFC has essentially an associative role, these drug interactions need to be interpreted at cellular (pyramidal) and not at receptor level, because several drugs used to reverse the effect of cirazoline (M100907, BAY × 3702, NBQX) are not expected to interact with α1-adrenoceptors in the experimental conditions used.
The effect of cirazoline likely involves the activation of α1-adrenoceptors on pyramidal neurons projecting to the DR, as previously observed for 5-HT2A receptors (Fig. 11). This assumption is based on (i) the common signal transduction mechanisms activated by 5-HT2A and α1-adrenoceptors (see Introduction); (ii) the great abundance of both receptors in the prelimbic and infralimbic areas of the mPFC (Pazos et al. 1985; Palacios et al. 1987) which project to the DR (Hajós et al. 1998; Peyron et al. 1998); (iii) the increase in the DR 5-HT release produced by cirazoline application in mPFC; and (iv) the reversal of the effect of cirazoline by agents acting on pyramidal neurons (see below).
Cirazoline is not entirely selective for α1-adrenoceptors and displays affinity for imidazoline receptors and α2-adrenoceptors, where it behaves as an antagonist (Ruffolo and Waddell 1982). However, the comparatively lower affinity for these sites (Molderings et al. 1998) suggests that its effects are mediated by α1-adrenoceptors. Moreover, the blockade of its effect by prazosin suggests that cirazoline acts via α1-adrenoceptors although its similar affinity for the various subtypes does not allow to clarify which one(s) were involved. Two areas projecting to the mPFC (thalamus and midbrain raphe) express abundant α1B-adrenoceptor mRNA. The good correspondence between receptor protein and mRNA suggests a somatodendritic location (Palacios et al. 1987; McCune et al. 1993; Pieribone et al. 1994; Day et al. 1997; Domyancic and Morilak 1997) and appears to exclude the possibility that putative terminal α1B-adrenoceptors mediate the effect of cirazoline. Terminal 5-HT2A receptors (Jakab and Goldman-Rakic 1998) in thalamocortical afferents to the mPFC have been suggested to mediate the 5-HT2A receptor-dependent increase in the spontaneous excitability of pyramidal neurons in mPFC (Aghajanian and Marek 1999; Marek et al. 2001). However, terminal 5-HT2A receptors in mPFC do not seem to be located in glutamatergic axons and extensive thalamic lesions left unaltered the effect of DOI on pyramidal cell firing (Miner et al. 2003; Puig et al. 2003), which suggests that postsynaptic 5-HT2A receptors are involved in the excitatory effect of 5-HT2A receptor stimulation. The analogy of effects of 5-HT and noradrenaline on pyramidal excitability (Marek and Aghajanian 1999) suggests a similar location for α1-adrenoceptors. Moreover, the effect of cirazoline was canceled by the coapplication of NBQX, BAY × 3702 and antipsychotic drugs, which act on receptors located on intrinsic neurons of the prefrontal cortex (Petralia and Wenthold 1992; Kia et al. 1996; Vysokanov et al. 1998; De Felipe et al. 2001). Given the complex pharmacological profile of the mGluR II/III agonist 1S,3S-ACPD, it is unclear whether this agent may act presynaptically (i.e. by reducing glutamate release) and/or postsynaptically, by activating postsynaptic inhibitory mGluRs.
As observed with the action of DOI (Martín-Ruiz et al. 2001), the 5-HT-increasing action of cirazoline depends on glutamatergic transmission in mPFC as it was reversed by AMPA/KA (but not NMDA) receptor blockade and mGluR II/III activation, and was mimicked by the local application of S-AMPA. Indeed, the 5-HT- and noradrenaline-induced increase in pyramidal excitability was also abolished by AMPA receptor blockade (Marek and Aghajanian 1999), suggesting a dependence on glutamatergic inputs onto mPFC.
The activation of 5-HT1A receptors by the pre- and postsynaptic 5-HT1A agonist BAY × 3702 (De Vry et al. 1998; Casanovas et al. 1999, 2000) counteracted the effect of DOI and cirazoline on 5-HT release (Martín-Ruiz et al. 2001; this study). 5-HT1A receptors have been reported to occur in the somatodendritic compartment and axon hillock of pyramidal neurons (Kia et al. 1996; De Felipe et al. 2001) and their activation results in neuronal hyperpolarization and reduction of firing rate (Araneda and Andrade 1991; Ashby et al. 1994). Hence, BAY × 3702 may oppose to the increase in excitability produced by the activation of α1-adrenoceptors, thus reducing the excitatory input onto midbrain 5-HT neurons and, hence, 5-HT release (see scheme in Fig. 11). The specificity of BAY × 3702 is supported by its total lack of action in the mPFC of 5-HT1A receptor knockout mice at the concentration used herein (Amargós-Bosch et al., unpublished results).
The reciprocal antagonism between 5-HT2A and α1-adrenoceptors (M100907 of cirazoline's effect and prazosin of DOI's effect) appeared surprising. These neurochemical results parallel behavioral data showing that the 5-HT2A-mediated, DOI-induced head shakes in rodents were suppressed by prazosin and a number of ligands acting at cortical receptors, such as 5-HT1A agonists, 5-HT2A/2C antagonists or classical antipsychotics such as haloperidol, among others (Schreiber et al. 1995; Dursun and Handley 1996), an observation for which no clear neurobiological basis has been provided so far. The present data suggest that pyramidal neurons may play an integrative role for these actions to modulate motor output. Indeed, our observations suggest a close association between 5-HT2A, α1-adrenoceptors and AMPA receptors to regulate the activity of projection neurons in mPFC, which is the driving force of the observed changes in 5-HT release (Fig. 11). Also, cortical 5-HT2A, α1-adrenoceptors (but not AMPA receptors) appear to tonically control basal 5-HT release, given the reduction in 5-HT release produced by their local perfusion (α1-adrenoceptors in the raphe also control tonically the activity of 5-HT cells and its local and terminal release; Baraban and Aghajanian 1980; Rouquier et al. 1994; Adell and Artigas 1999; Bortolozzi and Artigas 2003).
The effects of 5-HT2A receptor and α1-adrenoceptor activation on pyramidal cell excitability are consistent with a postsynaptic location. Indeed, most cortical 5-HT2A receptors in mPFC are located postsynaptically (Miner et al. 2003). The tonic activation of both receptors can elicit a phospholipase C-mediated increase in Ca2+ signaling (see Introduction), which may facilitate AMPA-mediated transmission. This is also consistent with the α1-adrenoceptor-mediated facilitation of the excitatory action of glutamate on cortical neurons (Mouradian et al. 1991; McCormick et al. 1993). The removal of the tone on either receptor by the respective antagonist may result in a loss of synergism and a subsequent reduction of pyramidal activity and of the descending excitatory input onto 5-HT neurons, which might explain the effect on basal and cirazoline-stimulated 5-HT release. Interestingly, prazosin and M100907 application completely reversed the 5-HT-increasing action of S-AMPA, an observation, which further supports the interaction between these receptors. However, we cannot clarify whether M100907 and prazosin act as pure antagonists in vivo as at least prazosin has been reported to be an inverse agonist in artificial cell systems (Zhu et al. 2000; Hein et al. 2001).
Interestingly, the basal and cirazoline-stimulated 5-HT release was also reversed by classical (chlorpromazine, haloperidol) and atypical antipsychotics (clozapine, olanzapine). All these agents display high in vitro affinity for α1-adrenoceptors (in the low nanomolar range), whereas the only the atypical drugs have such high affinity for 5-HT2A receptors (Arnt and Skarsfeldt 1998; Bymaster et al. 1999; Sebban et al. 1999). Both prazosin and M100907 reversed the elevation in mPFC 5-HT release produced by cirazoline (this study) and DOI (Bortolozzi et al. 2003). Similarly, chlorpromazine, haloperidol, clozapine and olanzapine also counteracted the increase in 5-HT produced by cirazoline (this study) and DOI (Bortolozzi et al. 2003). Based on the relative affinities of the four antipsychotic drugs tested, we postulate that only α1-adrenoceptor blockade participates in the reversal of the effect of cirazoline by classical antipsychotics whereas both 5-HT2A receptors and α1-adrenoceptors may be involved in the action of atypical antipsychotics. Given the complex pharmacological profile of these drugs, it is likely that only the use of murine knockout models can clarify which receptor is involved in this reversal.
Atypical antipsychotics are 5-HT2A receptor antagonists (Meltzer 1999). Likewise, the blockade of α1-adrenoceptors by prazosin potentiated the antipsychotic-like effect of dopamine D2 receptor antagonists (Wadenberg et al. 2000) and there is increasing interest in the role played by 5-HT1A receptors in the activity of atypical antipsychotics (Millan 2000; Ichikawa et al. 2001). It is noteworthy that these three properties (5-HT2A receptor and α1-adrenoceptor blockade, stimulation of 5-HT1A receptors) converge in the same effect in mPFC, i.e. a reduction of the 5-HT release, which likely parallels the change in activity of pyramidal neurons. This suggests that, in addition to their antidopaminergic action, antipsychotics may partly exert their palliative effect by reducing the activity of prefrontal pyramidal neurons by any of these mechanisms. This would agree with the key role of the frontal lobe in the pathophysiology of schizophrenia and its treatment (for review, see Weinberger et al. 1994; Arnt and Skarsfeldt 1998; Lidow et al. 1998; Lewis and Lieberman 2000). Further work is required to examine the neuronal distribution of these three receptors in mPFC in order to clarify the cellular site(s) of interaction.
Work supported by grants SAF2001-2133 and FIS 2001–1147. Financial support from Bayer S.A and Eli Lilly & Co is also acknowledged. The authors thank Leticia Campa for help in HPLC analyses and the pharmaceutical companies for generous drug supply. AB is recipient of a postdoctoral fellowship from the Fundación Carolina. MA-B is recipient of a predoctoral fellowship from IDIBAPS.