Abnormal medial prefrontal cortex connectivity and defective fear extinction in the presymptomatic G93A SOD1 mouse model of ALS


  • C. Sgobio,

    1. CNR Institute for Neuroscience and Department of Experimental Neurology, IRCCS S. Lucia Foundation, and §Department of Neuroscience, University of Tor Vergata, Rome, Italy
    Search for more papers by this author
  • , A. Trabalza,

    1. CNR Institute for Neuroscience and Department of Experimental Neurology, IRCCS S. Lucia Foundation, and §Department of Neuroscience, University of Tor Vergata, Rome, Italy
    Search for more papers by this author
  • , A. Spalloni,

    1. CNR Institute for Neuroscience and Department of Experimental Neurology, IRCCS S. Lucia Foundation, and §Department of Neuroscience, University of Tor Vergata, Rome, Italy
    Search for more papers by this author
  • C. Zona,

    1. CNR Institute for Neuroscience and Department of Experimental Neurology, IRCCS S. Lucia Foundation, and §Department of Neuroscience, University of Tor Vergata, Rome, Italy
    Search for more papers by this author
  • , I. Carunchio,

    1. CNR Institute for Neuroscience and Department of Experimental Neurology, IRCCS S. Lucia Foundation, and §Department of Neuroscience, University of Tor Vergata, Rome, Italy
    Search for more papers by this author
  • , P. Longone,

    1. CNR Institute for Neuroscience and Department of Experimental Neurology, IRCCS S. Lucia Foundation, and §Department of Neuroscience, University of Tor Vergata, Rome, Italy
    Search for more papers by this author
  • and ,, M. Ammassari-Teule

    Corresponding author
    1. CNR Institute for Neuroscience and Department of Experimental Neurology, IRCCS S. Lucia Foundation, and §Department of Neuroscience, University of Tor Vergata, Rome, Italy
    Search for more papers by this author

*M. Ammassari-Teule, CNR Institute of Neuroscience, IRCCS S. Lucia Foundation, Via Fosso di Fiorano 64-65, 00143 Rome, Italy. E-mail: martine.teule@ipsifar.rm.cnr.it


Amyotrophic lateral sclerosis (ALS) is a fatal progressive neuropathy associated with the degeneration of spinal and brainstem motor neurons. Although ALS is essentially considered as a lower motor neuron disease, prefrontal cortex atrophy underlying executive function deficits have been extensively reported in ALS patients. Here, we examine whether prefrontal cortex neuronal abnormalities and related cognitive impairments are present in presymptomatic G93A Cu/Zn superoxide dismutase mice, a mouse model for familial ALS. Structural characteristics of prelimbic/infralimbic (PL/IL) medial prefrontal cortex (mPFC) neurons were studied in 3-month-old G93A and wild-type mice with the Golgi–Cox method, while mPFC-related cognitive operations were assessed using the conditioned fear extinction paradigm. Sholl analysis performed on the dendritic material showed a reduction in dendrite length and branch nodes on basal dendrites of PL/IL neurons in G93A mice. Spine density was also decreased on basal dendrite segments of branch order five. Consistent with the altered morphology of PL/IL cortical regions, G93A mice showed impaired extinction of conditioned fear. Our findings indicate that abnormal prefrontal cortex connectivity and function are appreciable before the onset of motor disturbances in this model.

Amyotrophic lateral sclerosis (ALS) is a fatal progressive motor neuropathy with an incidence estimated around 1.5/100 000. Although degeneration of spinal (Frey et al. 2000; Jablonka et al. 2004; Pun et al. 2006; Zang & Cheema 2002; Zhang et al. 1997) and cortical (Maekawa et al. 2004; Nihei et al. 1993; Schiffer et al. 1994) motor neurons leading to a severe cortical atrophy (Grosskreutz et al. 2006; Kiernan & Hudson 1994) is well documented in ALS patients, the link between peripheral and central motor neuron abnormalities remains largely unknown. In nondemented patients, investigations on the functional consequences of cortical atrophy have shown consistent deficits in executive functions (Abrahams et al. 1996; Dary-Auriol et al. 1997; Kilani et al. 2004; Neary et al. 2000), that is, in operations depending on prefrontal cortex areas (Ragozzino et al. 1999, 2003). Executive function deficits were prominently represented by impairments in verbal/nonverbal fluency (Kew et al. 1993), concept formation and retrieval (Strong et al. 1999), working memory/sustained attention (Hanagasi et al. 2002; Mantovan et al. 2003), emotional memory (Papps et al. 2005) and response flexibility (Schreiber et al. 2005). In addition, the observation that patients with unclear speech showed considerable reduction in movement planning suggested an extended degeneration across wide regions of the frontal lobe (Santhosh et al. 2004).

Mutations in the Cu/Zn superoxide dismutase (SOD1) gene are implicated in about 20% of the familial form of ALS, and transgenic mice overexpressing the human SOD1 (GLY93 → ALA) mutation (G93A) show an ALS-like phenotype (Gurney et al. 1994). These mice exhibit a progressive muscle atrophy leading to abnormal gait and four-limb paralysis occurring at about 4 months of age and then die soon thereafter. Longitudinal evaluation of their motor performance in tasks assessing different aspects of motor function has shown some abnormalities in muscle strength and co-ordination around 8 weeks of age, although spontaneous activity was not impaired until 15 weeks of age (Barneoud et al. 1997). The presence of subtle motor deficits early in development prompted us to examine whether mutation-related changes in cognition might also precede the onset of severe neuropathic symptoms. We first reported defective striatal-dependent learning accompanied by impaired frontostriatal long-term depression in 12-week-old mice (Geracitano et al. 2003). However, we also observed that, at the same age, these mice exhibited a paradoxical enhancement in hippocampus-dependent memory, an increased expression of hippocampal α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) subunit GluR1 protein levels and a lower threshold to CA3–CA1 long-term potentiation induction (Spalloni et al. 2006). Indirect support to our findings comes from data showing increased levels of phosphorylated Erk, an extracellular signal-regulated kinase involved in neural plasticity and memory (Villarreal & Barea-Rodriguez 2006), in symptomatic G93A mice (Chung et al. 2005). Thus, cognitive changes associated with anatomically specific neural modifications occur in G93A mice before clinical symptoms become evident. Because (1) prefrontal cortex abnormalities mediating executive function deficits seem to be a constant feature of ALS patients and (2) to our knowledge, structural alterations of cortical neurons have not yet been described in ALS mouse models, here we examine the structure and function of prefrontal cortex neurons in G93A mice, with a particular attention devoted to the prelimbic/infralimbic (PL/IL) areas mediating one aspect of executive function, that is, response flexibility.

Materials and methods


B6SJL-TgN(SOD1-G93A)1Gur mice expressing the human G93A SOD1 mutation were obtained from Jackson Laboratories (Bar Harbor, ME, USA). Selective breeding maintained the transgene in the hemizygous state in a hybrid C57BL/6J × SJL genetic background (Gurney et al. 1994). Screening for the presence of the human transgene was performed on tail tips homogenized in phosphate-buffered saline, freeze/thawed twice and centrifuged at 18 000 gfor 15 min at 4°C. The pellet extracts were analyzed by staining for SOD1 activity after Laemmli’s polyacrylamide gel electrophoresis (10% resolving and 4% stacking). Mice were divided into two groups: transgenic G93A SOD1 mice (G93A) and their nonmutant (wild type, WT) littermates, which were considered as controls. They were housed in groups of three or four in a temperature-controlled room (22°C) with a 12-h light:dark cycle (light on: 0700–1900 h). Food and water were given ad libitum. Experiments were conducted in compliance with the European Council Directive (86/609/EEC) for the use and care of laboratory animals. All experiments were carried out in male mice between postnatal days P85 and P90 and performed blind. Independent groups of mice were used for Golgi–Cox staining and behavioral experiments.

Golgi–Cox impregnation of brain tissue

Mice (G93A: n = 4; WT: n = 4) were anesthetized with chloral hydrate (400 mg/kg) and perfused intracardially with 0.9% saline. The brains were dissected and impregnated using a standard Golgi–Cox solution (1% potassium dichromate/1% mercuric chloride/0.8% potassium chromate) according to the method described by Glaser and Van der Loos (1981). The brains immersed in the Golgi–Cox solution were stored at room temperature for 6 days, transferred to a sucrose solution (30%) for 5 days and then sectioned coronally (150 mm) using a vibratome. Sections were mounted on gelatinized slides, stained according to the Gibb and Kolb (1998) method and covered with Permount.

Morphological analysis

Dendrite morphology and dendritic spine density

Four brains of each genotype were processed for morphological analyses. Measurements were performed on impregnated neurons identified under low magnification (× 20/0.5NA). Within each hemisphere, three pyramidal cortex neurons with the soma in layer V and apical dendrites reaching layers II and IV were selected in the PL/IL regions of the medial prefrontal cortex (mPFC) (Bregma 1.98–1.78 mm, Franklin & Paxinos 1997, Fig. 1a) and subsequently analyzed under higher magnification (× 63/0.75NA). Only fully impregnated pyramidal neurons displaying dendritic trees without obvious truncations and isolated from neighboring impregnated neurons were retained for the analysis (Vyas et al. 2002). Morphological measurements were made by an experimenter blind to the genotype of the animal. Because no interhemispheric difference was detected, the data were pooled so that six neurons per animal were considered in each analysis. Measurements were carried out using a microscope (DMLB, Leica) equipped with a motorized stage and a camera connected to a software for morphological analyses allowing quantitative three-dimensional analysis of complete dendritic arborization (Neurolucida 7.5, MicroBrightField, Inc., Williston, VT, USA). The length and the number of branch nodes of the dendritic trees were quantified tracing the entire apical and three basal dendrites and performing Sholl analyses. The basal dendritic trees were required to be well discernible from neighboring impregnated dendrites and to have no breaks in staining and at least one four-order branch to maximize the possibility of counting spines until the most distal branch order present in this genotype (Table 1). Briefly, using the center of the soma as reference point, dendritic length and branch points were measured as a function of their radial distance from the soma by adding up all values in each successive concentric segment (segment radius: 25 μm). On each dendrite category, dendrite diameter and spine density were estimated as a function of the branch order under higher magnification (× 100/0.75N). All protrusions with or without bulbous expansions were counted as spines if they were in direct continuity with the dendritic shaft. Spine density was expressed as the number of spines per micrometer of dendritic length. It was calculated for the entire arborization of selected dendrites by dividing the number of spines by the length of the dendrite branch. Densities of spines on the apical and on the three basal dendritic trees were averaged for a neuron mean, and the six neurons from each mouse were averaged for an animal mean.

Figure 1.

Morphological measurements. (a) Coronal sections of mPFC (adapted from Franklin & Paxinos 1997). The black areas depict the cortical regions (PL/IL) of interest. (b) Typical pyramidal neuron from IL/PL cortices with apical (white arrows) and basal (black arrows) dendrites. Measurements were performed on apical dendrites in layers I–IV and on basal dendrites in layer V that show the most extended ramifications. Differences in extension of the dendritic arbor were assessed by means of a Sholl analysis based on measurements performed on successive concentric segments (segment radius: 25 μm). Dendrite diameter and spine density were estimated as a function of the branch order.

Table 1.  Spine density and dendrite diameter across branch orders on apical and basal dendrites
OrderNumber of miceNeurons per mouseSpine density (spines/μm)Average diameter (μm)
  1. Measurements of spine density performed on apical dendrites did not show genotypic variations in the number of spines per branch. In the basal compartment, the minor length of dendrites in the mutants prevented spine density and dendrite diameter to be estimated at branch orders more distal than four. A two-factor anova (genotype × branch order) performed on these data showed a significant effect of the branch order factor (F3,18 = 79.05, P < 0.001) and of the genotype × branch order interaction (F3,18 = 13.81, P < 0.001). Post hoc comparisons showed that the number of spines on branch order four was significantly lower in G93A mice than in WT mice (P < 0.05). The diameter of apical branches did not vary between genotypes. The diameter of order one basal branches was significantly lower in G93A than in WT mice [significant effect of the branch order (F3,18 = 131.32, P < 0.001) and of the genotype × branch order interaction (F3,18 = 4.96, P < 0.05); Tukey test for branch order one, P < 0.01).

Apical dendrites
1446 ± 06 ± 00.069 ± 0.0040.035 ± 0.033.08 ± 0.273.05 ± 0.0
2446 ± 06 ± 00.265 ± 0.0060.262 ± 00071.72 ± 0.19182 ± 0.13
3446 ± 06 ± 00.464 ± 0.0090.466 ± 0.011.31 ± 0.111.34 ± 0.05
4446 ± 06 ± 00.492 ± 0.0070.533 ± 0.0111.19 ± 0.091.26 ± 0.01
5446 ± 05.5 ± 0.50.587 ± 00090.525 ± 0.0061.19 ± 0.091.20 ± 0.11
6445.5 ± 0.295.5 ± 0.50.582 ± 0.0080.689 ± 0.0131.21 ± 0.081.16 ± 0.03
7445.5 ± 0.295.5 ± 0.50.693 ± 0.0080.628 ± 0.0081.19 ± 0.061.20 ± 0.05
8445.25 ± 0.485.25 ± 0.480.698 ± 0.0030.618 ± 0.0101.05 ± 0.101.14 ± 0.09
9443 ± 0.715.25 ± 0.480.786 ± 0.0070.744 ± 0.0101.23 ± 0.111.09 ± 0.03
10442.25 ± 0.754.25 ± 0.630.781 ± 0.0080.591 ± 0.0071.23 ± 0.081.12 ± 0.04
11442.25 ± 0.754 ± 0.410.609 ± 0.0060.619 ± 0.0060.97 ± 0.081.06 ± 0.07
12442 ± 0.713 ± 0.410.620 ± 0.0070.635 ± 0.0141.06 ± 0.090.99 ± 0.05
13241.5 ± 0.961.5 ± 0.290.676 ± 0.020.469 ± 0.0150.96 ± 0.010.87 ± 0.08
14130.75 ± 0.250.75 ± 0.250.405 ± 0.0150.89 ± 0.10
15110.5 ± 0.50.25 ± 0.25
16110.5 ± 0.50.25 ± 0.25
1710.25 ± 0.25
Basal dendrites
1446 ± 06 ± 00.144 ± 0.0280.086 ± 0.0301.22 ± 0.051.40 ± 0.03*
2446 ± 06 ± 00.225 ± 0.0590.264 ± 0.0491.02 ± 0.061.09 ± 0.06
3445.75 ± 0.256 ± 00.313 ± 0.0490.379 ± 0.0540.87 ± 0.040.92 ± 0.05
4443.25 ± 0.255.25 ± 0.480.312 ± 0.0110.506 ± 0.070*0.85 ± 0.050.81 ± 0.02
5230.5 ± 0.292.25 ± 0.350.305 ± 0.0640.508 ± 0.1080.85 ± 0.030.82 ± 0.19
6120.5 ± 0.250.75 ± 0.480.513 ± 0.0080.95 ± 0.07
720.5 ± 0.290.188 ± 0.0011.02 ± 0.15
820.5 ± 0.290.454 ± 0.0430.85 ± 0.07

Behavioral analysis

Contextual fear conditioning and extinction of fear responses

The apparatus was a conditioning chamber (28 × 28 × 10 cm) made of transparent plastic material with a metal grid floor wired to a shock generator. The top of the cage was a transparent cover with drilled holes. Tones were delivered through a loudspeaker hanging from the ceiling (60 cm above the cage) of a soundproof room. On day 1, mice (G93A: n = 7; WT: n = 7) were individually pre-exposed to the conditioning chamber for a 2-min period, and the distance traveled (in cm) during this period was retained as a motor activity score. On day 2, mice were given one conditioning trial. They were returned to the conditioning chamber for a 12-min period and exposed to five tone-footshock pairings starting after 2 min of free exploration (intervals between tones: 99 seconds). Each tone (80 dB, 2000 Hz, 20 seconds) terminated at the time a scrambled footshock (0.7 mA, 1 second) was delivered (trace conditioning). Fear extinction was assessed on days 3, 4 and 5. On each day, mice were exposed to three tone-alone presentations at intervals of 99 seconds in the same apparatus located in a different chamber. This was performed to avoid context-footshock associations interfering with the extinction of tone-shock associations. A camera connected to a personal computer and mounted above the top of the cage was used to record the distance traveled during pre-exposure, conditioning and extinction and, only during the two latter episodes, the time spent freezing (absence of visible movements except those required for respiration). These behaviors were scored during 60 seconds following the end of each tone using a software for behavioral analysis (ethovision 3.0; Noldus, Wageningen, the Netherlands).

Statistical analyses

Differences in dendrite length and in branch nodes in each dendrite compartment were estimated using two-factor analysis of variances (anovas), with ‘genotype’ as main factor and ‘radial distance from soma’ as repeated factor. Differences in spine density and in dendrite diameter were estimated for each branch order in each dendrite compartment by means of two-way anovas, with ‘genotype’ as main factor and ‘branch order’ as repeated factor. Differences in activity level during pre-exposition of mice to the conditioning chamber in each context were estimated by means of the Student’s t-test. Differences in fear-conditioning performances (time spent freezing and distance traveled) were estimated by means of two-factor anovas, with ‘genotype’ as main factor and conditioned stimulus-unconditioned stimulus pairing (CS-US) (training) or ‘tone-alone’ (extinction) presentations as the repeated factor. The anovas were followed by honest significant difference Tukey post hoc tests.


Reduced basal dendrite arborization and spine density on PL/IL neurons in G93A mice

Figure 2a shows Golgi-stained pyramidal PL/IL neurons in G93A and WT mice, while Fig. 2b,c depicts the results of the Sholl analysis for dendrite length and branch nodes on apical (left) and basal (right) dendrites, respectively. Statistical analyses performed on apical dendrites data did not show any difference between genotypes. In the basal dendrite compartment, a main effect of genotype was found for dendrite length (F1,6 = 13.03, P < 0.05) and branch nodes (F1,6 = 9.49, P < 0.05) with significant ‘genotype × radial distance from the soma’ interactions (dendrite length: F13,76 = 8.01, P < 0.05; branch nodes: F6,36 = 2.46, P < 0.05), indicating that morphological measurements performed at increasing distance from the soma evolved differently according to the genotype. Post hoc comparisons then showed that, in the mutants, dendrites lying 50–100 μm from the soma were shorter and those lying 25–50 μm from the soma had fewer nodes (P < 0.05 for all comparisons). Spine density was then estimated as a function of the branch order on apical and basal dendrites. In the latter case, the minor length of mutant basal dendrites did not allow spine density to be compared between genotypes in all the mice and/or in six neurons per mice at branch orders more distal than four. The data are shown in Fig. 2d, while spine density values ± standard error of the mean are reported in Table 1. Statistical analyses of spine density estimated at branch order 1–12 of apical dendrites showed no effect of the genotype (F1,6 = 54.84, P < 0.001) but an effect of the branch order (F11,66 = 54.84, P < 0.001). Conversely, spine density estimated at branch order 1–4 of basal dendrites indicated a significant effect of the branch order (F3,18 = 79.05, P < 0.001) and of the genotype × branch order (F3,18 = 13.81, P < 0.001) interactions. Post hoc pair comparisons then showed that the number of spines on branch order four was significantly lower in G93A mice than in WT mice (P < 0.05). Also, there was no major variation in the diameter of dendrites between genotypes except the smaller diameter of basal dendrites on branch order one in G93A mice [significant effect of the branch order (F3,18 = 131.32, P < 0.001) and of the genotype  × branch order (F3,18 = 4.96, P < 0.05) interactions; Tukey test for branch order one, P < 0.01].

Figure 2.

Structural characteristics of prefrontal cortex neurons in G93A and WT mice. (a) Representative Golgi-stained pyramidal neurons in PL/IL cortices in G93A (top) and WT (bottom) mice. (b–d) The curves depict measurements of dendrite length (b) branch nodes (c) and spine density (d) on apical (left) and basal (right) dendrites in each genotype. Statistical comparison showed a reduction in the extension of the basal dendritic arborization, the number of dendritic nodes and spine density at branch order four in G93A mice compared with WT (*P < 0.05). No genotypic difference was found in the apical dendrite compartment.

Pretraining motor activity was similar in G93A and WT mice

On day 1, both G93A and WT mice traveled the same distance when pre-exposed to the conditioning chamber for a 2-min period [t (12) = 0.135, P > 0.1] (data not shown).

Fear conditioning developed similarly in G93A and WT mice

As shown in Fig. 3a,b, conditioning developed similarly in G93A and WT mice. On day 1, the amount of freezing (Fig. 3a) was equivalent (no significant effect of genotype, F1,12 = 2.38, P > 0.1) and increased progressively across CS-US presentations (significant effect of the repeated factor, F2,48 = 31.92, P < 0.001) in both groups, showing a similar reaction to tone-shock associations. In the same fashion, the distance traveled in the conditioning chamber during CS-US presentations (Fig. 3b) did not vary between genotypes (F1,12 = 0.06, P > 0.1) but decreased in both groups across tone-shock presentations (F12,48 = 8.312, P > 0.01).

Figure 3.

Fear conditioning and extinction of conditioned fear responses. Fear conditioning developed similarly in G93A and WT mice. During training, (a) the time spent freezing increased (P < 0.001) and (b) the distance traveled in the conditioning chamber decreased (P < 0.01) to the same extent across CS-US presentations in G93A and WT mice. Conversely, extinction of fear responses followed a different trend according to the genotype. Wild-type mice showed (c) a reduction of time spent freezing (P < 0.05) and (d) an increase in the distance traveled (F8,96 = 3.21, P < 0.01) across tone-alone presentations that were not found in G93A mice. Post hoc comparisons, *P < 0.05.

G93A mice failed to extinguish conditioned fear responses

Extinction of conditioned fear responses developed differently according to the genotype. Wild-type mice showed a progressive reduction of freezing responses (Fig. 3c) and an enhancement of locomotion across exposures to the tone-alone presentation (Fig. 3d), while these changes were not found in G93A mice. Analysis of variances performed on these data showed significant effects of genotype (freezing: F1,12 = 6.22, P < 0.05; distance traveled: F1,12 = 10.38, P < 0.01), of presentations (freezing: F8,96 = 3.29, P < 0.01; distance traveled: F8,96 = 3.98, P < 0.01) and of significant genotype × presentation interactions (freezing: F8,96 = 2.37, P < 0.05; distance traveled: F8,96 = 3.21, P < 0.01). Post hoc comparisons showed that, on day 3, G93A mice exhibited a stronger amount of freezing and ran a shorter distance in comparison with their WT controls following the three tone-alone presentations (P < 0.05 for all comparisons).


The main result of the present study is that structural alterations in mPFC neurons associated with fear extinction deficits are present in G93A mice before the onset of the motor symptoms. Specifically, these alterations consisted in a dramatic reduction of the dendritic arbor accompanied by a mild decrease in spine density on basal dendrites of pyramidal neurons lying in the PL/IL regions.

We first observed a minor complexity of basal dendrite arborization in PL/IL neurons in G93A mice. Sholl analysis showed that the length of basal dendrites and the number of nodes were reduced by 37% and 14%, respectively, and that this reduction was mainly evident on medium-order branches lying 50–100 μm from the soma for the dendrite length and more proximally (25–50 μm) for the dendrite nodes. On the other hand, apical dendrites were not affected. Because basal dendrites receive the majority of synapses innervating neocortical pyramidal neurons (Gordon et al. 2006), the present pattern of mPFC dendritic reorganization is therefore consistent with a robust decrease in neuronal connectivity adversely affecting intracortical excitatory inputs.

In addition to the reduction in dendrite length and branching, we also observed a decrease in the number of spines specifically found on branch order five of basal dendrite segments. Considering that dendritic spines are postsynaptic elements involved in excitatory transmission (Harms & Dunaevsky 2007) and that the density of spines predicts the number of excitatory synapses (Wallace & Bear 2004), the finding that G93A mice show a decrease in spines on a reduced basal dendritic arborization in PL/IL neurons further points to a reduction in intracortical excitatory neurotransmission. This reduction might be viewed as an early cellular adaptation resulting from a decrease in peripheral inputs knowing that, in this model, the gradual loss of synaptic connections begins before the onset of clinical symptoms (Frey et al. 2000) and that peripheral synapses at neuromuscular junctions are lost long before axons degenerate.

Assessment of mPFC-related behaviors in G93A mice showed that their abnormal PL/IL cortical morphology interfered with their capability of extinguishing conditioned fear responses. The persistent freezing behavior observed in the mutants cannot be ascribed to genotypic differences in motor activity as these mice moved to the same extent as the WT controls during pre-exposure to the conditioning chamber. It is now well established that distinct subregions of the prefrontal cortex are implicated in a variety of high-level cognitive and executive operations (see Dalley et al. 2004 for a review). Within this framework, demonstration that mPFC regions govern response flexibility, that is, the capability of forming new associations inhibiting those formerly established, largely derives from evidence implicating PL/IL neurons in extinction of conditioned fear (Milad & Quirk 2002; Morgan et al. 1993; Quirk et al. 2006). Data showing that acquisition and extinction of conditioned fear involve similar neurochemical (Prado-Alcala et al. 1994) and molecular (Lin et al. 2003; Santini et al. 2001) mechanisms underline the view that defective fear extinction actually consists in a new learning. Also in humans, response flexibility has extensively been examined in relation to posttraumatic stress disorder, and the observation that subjects with difficulties in forgetting a traumatic experience showed a marked reduction in mPFC activity largely contributed to demonstrate the role of mPFC in adapting behavior to the situation’s demand (Bremner et al. 2005; Williams et al. 2006). Thus, the observation that G93A mice with structural alterations on PL/IL neurons show defective fear extinction is consistent with a mutation-related executive function deficit reminiscent of the human ALS prefrontal syndrome.

Interestingly, recent investigations on the functional impact of stress-induced alterations in cortical morphology have shown the existence of a direct relationship between retraction of dendrite branching in mPFC regions and executive function deficits. For example, reduction of dendritic arbors in mPFC neurons underlying impairments in attention set-shifting was found to occur in rats subjected to chronic restraint stress (Liston et al. 2006). Closer to our findings, dendritic retraction in IL/PL cortices causing increased resistance to fear extinction was reported in rats (Miracle et al. 2006) and mice (Izquierdo et al. 2006) exposed to chronic or brief uncontrollable restraint stress. In these experiments, however, stress-induced dendrite retraction specifically affected apical dendrites. In addition, apical dendrite arbors returned to the prestress levels several weeks after the stressor was suppressed (Radley et al. 2004), thus indicating that the retraction was only transient. Differently, we found a robust decrease in length and nodes as well as a decrease in the number of spines on basal dendrites of PL/IL neurons in G93A mice reared in standard condition, suggesting a direct effect of the mutation on the morphology of IL/PL neurons and on the behaviors they subserve.

Peripheral and central neuronal loss at both presymptomatic (Jaarsma et al. 2000) and symptomatic (Gurney et al. 1994) stages have been extensively reported in G93A mice. However, no description of dendritic material is, to our knowledge, available in this genotype, despite the early unbalance between excitatory and inhibitory synaptic inputs in G93A motor neuron (Schutz 2005), suggesting that the neuropathy could be the consequence of abnormal neural networks properties (Durand et al. 2006). Altogether, the results of this study provide evidence that overexpression of human SOD1 with a G93A mutation decreases connectivity in mPFC networks and alters mPFC-related cognitive operations in presymptomatic G93A mutants compared with age-matched WT mice. Hence, they suggest that assessment of mPFC function by noninvasive brain imaging and neuropsychological techniques might be exploited for early diagnosis of familial ALS associated with the SOD1 mutation.


This work was supported by a Fondi di Investimento per la Ricerca di Base (FIRB) project N° RBAU01A7T4-002 from the Italian Ministry for Education, University and Research.