OUP user menu

GABA deficit in the visual cortex of patients with neurofibromatosis type 1: genotype–phenotype correlations and functional impact

Inês R. Violante, Maria J. Ribeiro, Richard A. E. Edden, Pedro Guimarães, Inês Bernardino, José Rebola, Gil Cunha, Eduardo Silva, Miguel Castelo-Branco
DOI: http://dx.doi.org/10.1093/brain/aws368 918-925 First published online: 12 February 2013


Alterations in the balance between excitatory and inhibitory neurotransmission have been implicated in several neurodevelopmental disorders. Neurofibromatosis type 1 is one of the most common monogenic disorders causing cognitive deficits for which studies on a mouse model (Nfl+/−) proposed increased γ-aminobutyric acid-mediated inhibitory neurotransmission as the neural mechanism underlying these deficits. To test whether a similar mechanism translates to the human disorder, we used magnetic resonance spectroscopy to measure γ-aminobutyric acid levels in the visual cortex of children and adolescents with neurofibromatosis type 1 (n = 20) and matched control subjects (n = 26). We found that patients with neurofibromatosis type 1 have significantly lower γ-aminobutyric acid levels than control subjects, and that neurofibromatosis type 1 mutation type significantly predicted cortical γ-aminobutyric acid. Moreover, functional imaging of the visual cortex indicated that blood oxygen level-dependent signal was correlated with γ-aminobutyric acid levels both in patients and control subjects. Our results provide in vivo evidence of γ-aminobutyric acidergic dysfunction in neurofibromatosis type 1 by showing a reduction in γ-aminobutyric acid levels in human patients. This finding is relevant to understand the physiological profile of the disorder and has implications for the identification of targets for therapeutic strategies.

  • neurofibromatosis type 1
  • GABA
  • neurodevelopmental disorders
  • visual cortex
  • spectroscopy


The balance between excitatory and inhibitory neurotransmission is tightly regulated in the human brain and largely dependent on the levels of glutamate, the major excitatory neurotransmitter, and γ-aminobutyric acid (GABA), the major inhibitory neurotransmitter. Alterations in this push–pull mechanism might underlie the cognitive deficits found in several neurodevelopmental disorders (Ramamoorthi and Lin, 2011).

Neurofibromatosis type 1 is an autosomal dominant developmental disorder with a prevalence of 1 in 3500 (Kayl and Moore, 2000). Clinical features include café-au-lait spots, benign tumours and cognitive deficits, namely, learning and visuospatial impairments, which occur even in the absence of visible neural pathology (Jett and Friedman, 2010). The disease is caused by mutations in the NF1 gene that encodes neurofibromin, a negative regulator of the RAS signalling cascade (Ballester et al., 1990). Members of the RAS family are critical for controlling cellular growth and differentiation while also being involved in learning and memory (Fernandez-Medarde and Santos, 2011). Previous studies using a mouse model of neurofibromatosis type 1 (Nf1+/) indicated that neurofibromin also modulates GABAergic neurotransmission leading to enhanced inhibitory activity during periods of high-frequency neural stimulation. This was reflected by increased inhibitory postsynaptic potentials in the hippocampus (Costa et al., 2002; Cui et al., 2008), medial prefrontal cortex and striatum (Shilyansky et al., 2010). However, the hypothesis of altered inhibition has never been tested in humans.

To test this hypothesis, we applied magnetic resonance spectroscopy to measure GABA levels in patients with neurofibromatosis type 1. This is the only in vivo tool capable of non-invasively measuring brain metabolites. Magnetic resonance spectroscopy measurements were performed in early visual cortex, an area where the role of inhibition has been widely studied and where we previously found evidence for functional alterations (Ribeiro et al., 2012; Violante et al., 2012). Moreover, it has recently been shown that GABA concentration in the human visual cortex is significantly related to performance levels in visual perceptual tasks assessing orientation-specific surround suppression (Yoon et al., 2010) and orientation discrimination performance (Edden et al., 2009).

When studying the human disorder, the variability of mutations that give rise to the disease (Thomson et al., 2002) should be taken into account. Therefore, it is noteworthy to investigate whether genotypic differences lead to changes in GABA levels. We addressed this by testing the hypothesis that different types of mutations would impact neurofibromin function differently and consequently manifest in cortical GABA levels. Our rationale was based on the fact that different types of mutation can result in different levels of phenotypic severity. For instance, missense or splice-site mutations can act as dominant negatives disrupting the function of the healthy allele and resulting in a more severe phenotype than nonsense mutations (Khajavi et al., 2006).

To assess the functional correlates of the magnetic resonance spectroscopy signal, participants underwent functional MRI while performing a visual task. Importantly, the blood oxygen level-dependent (BOLD) signal is sensitive to the balance between excitation and inhibition (Logothetis, 2008) and therefore sensitive to variations in GABA levels (Chen et al., 2005; Licata et al., 2011). Consequently, it is physiologically relevant to determine whether the previously observed relationships between BOLD and GABA (Northoff et al., 2007; Muthukumaraswamy et al., 2009, 2012) are preserved in patients with neurofibromatosis type 1.

Materials and methods


We studied 20 patients with neurofibromatosis type 1 [mean age ± standard deviation (SD) 13.0 ± 3.1 years, age range 7.76–19.49 years, 15 females] and 26 healthy control subjects (mean age ± SD 13.1 ± 2.9 years, age range 7.42–19.67 years, 16 females), matched for age and gender. Patients were diagnosed by NIH criteria (National Institutes of Health Consensus Development Conference, 1988). Exclusion criteria for all participants were as follows: psychiatric disorder, neurological illness affecting brain function other than neurofibromatosis type 1, epilepsy or a clinically significant intracranial abnormality detected on MRI. Unidentified bright objects (T2-hyperintensities commonly found in patients with neurofibromatosis type 1) located in the occipital cortex were considered exclusion criteria. Additionally, we excluded patients with IQ <70. The control group was recruited from local schools, and their academic records were in accordance with chronological age, indicating adequate intellectual functioning. Children prescribed with stimulant medication (methylphenidate) were not medicated on the day of testing (n = 4). None of the participants were taking any other type of medication.

The study was conducted in accordance with the Declaration of Helsinki; all procedures were reviewed and approved by the Ethics Commission of the Faculty of Medicine of the University of Coimbra. Written informed consent was obtained from participants aged >18 years or from the children’s legal representatives.

All participants had normal or corrected-to-normal visual acuity. To rule out eye disorders in the neurofibromatosis type 1 group, patients underwent an ophthalmological assessment including best-corrected visual acuity, stereopsis evaluation using Randot, slit lamp examination of anterior chamber structures and fundus examination. No anomalies that could affect vision were found. We assessed handedness using the Edinburgh Inventory (Oldfield, 1971); in the neurofibromatosis type 1 group, 15 subjects were right-handed and five had mixed-handedness. In the control group, 20 subjects were right-handed and six had mixed-handedness. There were no between-group differences in age (t-test, not significant), gender or handedness (chi-square test, not significant).

Two participants with neurofibromatosis type 1 failed to perform the functional MRI task as instructed, thus analysis involving functional MRI data was carried out on 18 patients with neurofibromatosis type 1 and 26 control subjects. Participants underwent all imaging acquisitions on the same day.

Genetic characterization

Neurofibromatosis type 1 mutational profile was obtained by whole NF1 gene sequencing. The mutation analysis of 19 of the 20 patients has been reported previously (Ribeiro et al., 2012). For the patient missing genetic testing, DNA was extracted from peripheral blood and standard procedures were used to perform whole-gene sequencing and multiplex ligation-dependent probe amplification analysis (Upadhyaya et al., 2009).

Functional magnetic resonance imaging paradigm

Visual stimuli were similar to those used by Muthukumaraswamy et al. (2009) and consisted of a circular moving grating (80% contrast, spatial frequency 2 cycles/°, 4° diameter, velocity 1°/s), equiluminant to the background. Stimuli were presented in the lower left visual field, subtended 4° horizontally and vertically, with the centre of the stimulus located 3.3° from a central fixation point. Stimulus duration was chosen randomly in an interval between 1.5–2 s followed by 10 s of fixation point only; 30 events were presented. Participants were instructed to maintain fixation on the central point for the entire experiment and to press a button, as fast as possible, when the grating disappeared. The aim of this task was to keep subjects engaged throughout the functional MRI acquisition. Participants could make two types of errors: (i) pressing the button before the disappearance of the grating or (ii) failure to report the disappearance of the grating. All participants performed the task with a high-level of accuracy, >95% correct responses. However, patients with neurofibromatosis type 1 were slower than control subjects [neurofibromatosis type 1: 521.7 ± 178.1 ms; control subjects: 445.6 ± 201.7 ms; U(42) = 134.5, Z = −2.375, P = 0.018].

Magnetic resonance imaging and spectroscopy acquisitions

Scanning was performed on a 3 T Siemens scanner, using a 12-channel birdcage head coil. For each participant we acquired the following: (i) a T1-weighted MP-RAGE sequence, 1 mm3 isotropic voxel, repetition time 2.3 s, echo time 2.98 ms, inversion time 900 ms, flip angle 9°, field of view 256 × 256 mm, 256 × 256 matrix, 160 slices, GRAPPA acceleration factor = 2; (ii) a T2-weighted FLAIR sequence used to identify unidentified bright objects, 1 mm3 isotropic voxel, repetition time 5 s, echo time 388 ms, inversion time 1.8 s, field of view 250 × 250 mm, 256 × 256 matrix, 160 slices, GRAPPA acceleration factor = 2; (iii) a single-shot echo-planar image with interleaved acquisition for the functional MRI acquisition, 3 mm3 isotropic voxel, repetition time 2 s, echo time 39 ms, flip angle 90°, field of view 256 × 256 mm, 86 × 86 matrix, 28 slices, GRAPPA acceleration factor = 2; and (iv) a GABA-edited magnetic resonance spectra using the MEGA-PRESS method (Mescher et al., 1998; Edden and Barker, 2007), 3 cm3 isotropic voxel, echo time 68 ms, repetition time 1.5 s, 196 averages, 1024 data points. During odd number acquisitions, a frequency-selective inversion pulse was applied to the GABA-C3 resonance at 1.9 ppm (‘on resonance’). During even number acquisitions, the pulse was applied at 7.5 ppm (‘off resonance’). The voxel was positioned within the occipital cortex with its lower face aligned with the cerebellar tentorium. The sagittal sinus was avoided ensuring that the volume remained inside the occipital lobe (Fig. 1A). During structural and spectroscopic acquisitions, participants watched cartoons. This helped children to stay motionless during acquisitions. To ensure that head position was maintained and no movements occurred during acquisitions, participants’ eyes were monitored in real-time with a camera placed on the mirror system (Avotec Real Eye 5721). If a movement occurred, the acquisition was stopped and reinitiated. Moreover, participants’ eyes were monitored to ensure that during the spectroscopic acquisition, none of the participants fell asleep, and that fixation was maintained during the functional MRI acquisition.

Figure 1

GABA measurements. (A) Localization of the magnetic resonance spectroscopy voxel (white square) in the visual cortex of a representative participant. (B) Edited magnetic resonance spectroscopy spectrum from a representative participant showing clearly resolved peaks for GABA and glutamine+glutamate (Glx). (C and D) Cortical GABA/total creatine (tCr) levels and glutamine/total creatine levels for patients with neurofibromatosis type 1 (NF1) (black, n = 20) and control subjects (grey, n = 26), respectively. Graphs depict individual values, mean and standard deviation. **P = 0.001.

Magnetic resonance imaging data analysis

T1-weighted images were used for magnetic resonance spectroscopy voxel placement and image segmentation. Segmentation was performed using in-house software written in Matlab7 (The MathWorks Inc) and the VBM8 toolbox in SPM8 (http://www.fil.ion.ucl.ac.uk/spm) and applied to determine the relative proportions of grey matter, white matter and CSF in the voxel. FLAIR images were used to identify T2 hyperintensities, a common neuroradiological finding in patients with neurofibromatosis type 1. None of the control participants had unidentified bright objects, and none of the unidentified bright objects found in patients were located in the occipital cortex.

Functional magnetic resonance imaging data analysis

Image processing and analysis were conducted using BrainVoyager QX2.3 (Brain Innovation). We applied slice scan-time correction, linear trend removal, temporal high-pass filtering (two cycles per run), spatial smoothing (full-width at half-maximum 5 mm) and motion correction. None of the participants had within-run movements >3 mm. General linear model predictors were built by convolving a 2/10 s boxcar time course with a two-gamma function. Statistical thresholding was performed using false discovery rate, q(FDR) < 0.05. BOLD signal peak amplitude (% signal change) at the peak voxel was extracted for each participant, as this variable had been shown previously to be highly correlated with GABA levels (Muthukumaraswamy et al., 2012).

Magnetic resonance spectroscopy data analysis

Spectra from all subjects were inspected for movement artefacts. A difference spectrum was generated for each participant (on resonance-off resonance). Using in-house software written in Matlab, we applied 4 Hz exponential line broadening to all spectra. Peak integration was used to quantify GABA (∼3 ppm) and the combined glutamate plus glutamine peaks (∼3.75 ppm) in the difference spectra and total creatine peak (3 ppm) in the summed spectra. Integrals of GABA, glutamine and total creatine peaks were automatically calculated using a linear fit of the baseline and Gaussian fits to the peaks.


Statistical analyses were performed with PAWS Statistics 18 (SPSS Inc). First, we verified the normality assumption for the different parameters using the Shapiro Wilk’s test. All data were normally distributed, except in the case of reaction times in the functional MRI task. For analysis, we used independent samples t-tests, Mann–Whitney U test and Pearson’s correlation analyses.

To assess the effect of genotype on GABA levels, we first transformed GABA/total creatine measurements of all participants to Z-scores and performed a multiple regression analysis using a forced entry model, in which all predictors are forced into the model simultaneously. To construct the model, we took into account the fact that age (Marenco et al., 2010), gender (O’Gorman et al., 2011) and grey matter content (Jensen et al., 2005) may affect GABA concentration. Additional predictors were constructed from categorical variables relative to the type of neurofibromatosis type 1 mutation (nonsense, missense, splice-site and mutation not found in the NF1 gene).


Magnetic resonance spectroscopy GABA levels

Magnetic resonance spectroscopy measurements were performed in the visual cortex (Fig. 1) of 20 children and adolescents with neurofibromatosis type 1 and 26 matched control subjects. GABA and glutamate + glutamine peaks were well edited for all participants. Ratios of GABA/total creatine and glutamine/total creatine were calculated for each subject. GABA/total creatine provides reliable GABA concentration estimates and reduces inter-subject variance attributable to differences in signal-to-noise ratio and CSF fraction within the voxel (Bogner et al., 2010).

Patients with neurofibromatosis type 1 displayed significantly reduced GABA/total creatine levels compared with control subjects, [t(44) = −3.61, P = 0.001], whereas glutamine/total creatine levels were not significantly different [t(44) = −0.835] (Fig. 1C and D).

The percentages of grey matter, white matter and CSF in the spectroscopy voxel were assessed for each participant and were not significantly different between patients and control subjects (P = 0.196, P = 0.074 and P = 0.604, respectively). This ensured that any difference in metabolite levels did not arise as a consequence of differences in tissue content between groups. Moreover, GABA/total creatine was not correlated with grey matter (patients r = −0.066, P = 0.782; control subjects r = 0.203, P = 0.321), white matter (patients r = −0.037, P = 0.877; control subjects r = −0.302, P = 0.134) or CSF (patients r = 0.023, P = 0.924; control subjects r = 0.056, P = 0.785). Additionally, differences in GABA levels cannot be explained by a different impact of age or gender in patients and control groups because neither age nor gender were associated with GABA/total creatine (age: patients r = 0.267, P = 0.255; control subjects r = −0.095, P = 0.643) (gender: patients r = 0.266, P = 0.257; control subjects r = 0.142, P = 0.490).

Effect of genotype on GABA levels

The NF1 mutational profile of the patients is shown in Table 1. The mutational analysis of 19 of the 20 patients has been reported previously (Ribeiro et al., 2012). For the new patient, we found a novel missense mutation, which represents a novel change within the gene that has not been identified in >1000 normal chromosomes from unaffected individuals studied for the entire NF1 gene mutations. Disease-causing mutations were identified in 65% of the patients, in accordance with previous studies (Griffiths et al., 2007). Cases where the mutation was not identified might have a mutation missed by whole-gene sequencing but affecting gene expression, given that the criteria for diagnosis were fulfilled. A multiple regression model was applied to establish whether the type of mutation (nonsense, missense, splice site or mutation not found in the NF1 gene) might explain the GABA levels in the occipital cortex. We used as predictors, the type of mutation, age, gender and percentage of grey matter in the voxel. Given that only one frameshift mutation was present in our cohort, this type of mutation was not included as a predictor. The model obtained was significant for the predictor variables [F(7,37) = 2.414, P = 0.039, R2 = 0.31].

View this table:
Table 1

Summary of the NF1 mutations identified in the patients with neurofibromatosis type 1

Mutation typeMutation at complementary DNA levelMutation at protein level
Nonsensec.2041C > Tp. Arg681X
c.3318C > Gp.Tyr1109X
c.3942G > Ap.Trp1314X
c.5458C > Tp.Gln1828X
Frameshiftc.2500-2501 ins C
Missensec.2048T > Cp.Leu695Pro
c.2786T > Cp.Leu929Pro
c.5501T > Gp.Leu1834Arga
Splice-sitec.730+1G > AIVS5
c.1720+3A > GIVS11b
c.5944-6A > GIVS31
c.8097+1G > AIVS47
  • a Genetic characterization reported for the first time.

  • b Mutation found in two related patients.

We found that splice-site or missense mutations are significant contributors to predict reduced GABA levels in the occipital cortex (beta = −0.423, t = −3.005, P = 0.005 for splice-site mutations and beta = −0.324, t = −2.104, P = 0.042 for missense mutations). However, nonsense mutations or cases where the mutation is not found are insufficient to predict GABA levels below those of control subjects (Fig. 2). Additionally, age, gender and grey matter content did not add predictive power to the model. Furthermore, there was no significant interaction between the type of mutation and age.

Figure 2

GABA/total creatine (tCr) Z-scores for each participant grouped by genotype. GABA/total creatine measurements from each participant were transformed to Z-scores and plotted as groups according to the type of mutation found in our cohort of patients with neurofibromatosis type 1.

Correlation between GABA and blood oxygen level-dependent signal

We measured functional MRI BOLD signal while participants performed a low-level visual task (Fig. 3A). GABA/total creatine levels were significantly correlated with peak BOLD amplitude in patients (n = 18, r = −0.502, P = 0.034) and control subjects (n = 26, r = −0.458, P = 0.019) (Fig. 3B). Neither the correlation coefficients (z = 0.172, P = 0.863) nor the regression slopes [t(40) = 0.830, P = 0.412] were significantly different between groups. During functional MRI acquisition, participants were required to press a button to report the disappearance of the visual stimulus. Reaction times in this visual task were not correlated with BOLD signal or occipital GABA concentration in either group. In spite of the reduced levels of GABA found in patients and the correlation between GABA and BOLD, peak haemodynamic responses were not significantly different between groups (Fig. 3C). This results from the fact that for the different average GABA/total creatine in each group, the BOLD signal presents the same mean activation. In contrast, no correlations were observed for the glutamine/total creatine levels in patients or control subjects, in agreement with previous findings (Muthukumaraswamy et al., 2009; Stagg et al., 2011)

Figure 3

Functional MRI results. (A) BOLD activation in the visual cortex of a representative participant (P<0.05, corrected for multiple comparisons using false discovery rate). A black crosshair indicates the peak voxel. (B) GABA in the occipital cortex correlates with BOLD levels in patients (black, n = 18) and control subjects (grey, n = 26). (C) BOLD response time courses to the visual stimulation at the peak voxel for patients (orange) and control subjects (blue), error bars depict SEM. tCr = total creatine.


In this study, we report for the first time, measurements of GABA levels in patients with neurofibromatosis type 1, which proved to be crucial to translate the hypothesis of altered inhibition in this condition. Our results point to lower GABA levels, whereas no alterations were observed in glutamine/total creatine levels, indicating an alteration in the balance between excitation and inhibition in the patients’ visual cortex. An imbalance in the excitatory/inhibitory push–pull mechanism is in agreement with our previous findings of impaired visual contrast sensitivity in patients with neurofibromatosis type 1 (Ribeiro et al., 2012). Impaired contrast sensitivity was also found in pharmacological studies using drugs that alter GABAergic transmission (Blin et al., 1993; Haris and Phillipson, 1995). Furthermore, decreased GABA levels are in line with the evidence that lower GABA concentration is related to poorer visual orientation discrimination performance (Edden et al., 2009), a hallmark deficit of the cognitive impairments found in patients with neurofibromatosis type 1 (Schrimsher et al., 2003).

Previous studies conducted in animal models applied electrophysiological recordings in brain slices and provided evidence of increased inhibitory postsynaptic potentials, an indication of augmented inhibitory neurotransmission. In an elegant study, Cui et al. (2008) used Cre-loxP mice to limit neurofibromin heterozygosity to specific cell types. They found that increased inhibition was related to a rise in miniature inhibitory postsynaptic potential frequency as a result of NF1 heterozygosity in inhibitory neurons, suggesting a role for neurofibromin in GABAergic neurotransmission. Accordingly, increased frequency of miniature inhibitory post-synaptic potentials in Nf1+/ mice was attributable to higher levels of GABA release owing to increased phosphorylation of synapsin 1, a protein with a critical role in the regulation of neurotransmitter storage and release (Hilfiker et al., 1999). Mutations in the neurofibromin gene result in hyperactivation of Ras (Xu et al., 1990) and consequently of its downstream effectors, such as the mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK) (Li et al., 2005; Guilding et al., 2007). Findings by Cui et al. (2008) indicate that higher extracellular signal-regulated kinase activation in inhibitory neurons owing to reduced neurofibromin activity results in higher synapsin 1 phosphorylation, and consequently greater GABA release.

Our results in human patients point to reduced GABA levels, whereas no alterations were observed in glutamine/total creatine levels, in accordance with the findings from the animal model showing that neurofibromin alterations do not affect excitatory transmission (Cui et al., 2008).

Reduced GABA levels in patients are not necessarily inconsistent with increased inhibitory activity observed in the (Nf1+/) mouse model. It is important to highlight that there is a distinction between total GABA concentration measured by magnetic resonance spectroscopy and GABA neurotransmission measured in the animal model. Nonetheless, although magnetic resonance spectroscopy provides information about the overall concentration of GABA, it is mainly the cytosolic, extracellular and vesicular pools that are measured, as GABA bound to macromolecules (e.g. GABA transporters, GABA receptors) is not detected by the method. Moreover, GABAergic inhibition is modulated by alterations in GABA metabolism, which determines the cytosolic concentration of the neurotransmitter (Golan et al., 1996), and animal studies have suggested that the concentrations of vesicular and non-vesicular pools of GABA appear to be in equilibrium (Waagepetersen et al., 1999). This implies that the GABA levels measured by magnetic resonance spectroscopy are in close relationship with the concentration at the vesicular pool. In fact, magnetic resonance spectroscopy measurements of healthy individuals under the influence of benzodiazepines (Goddard et al., 2004) revealed decreased GABA levels in the visual cortex, whereas studies in animals reported increased inhibitory postsynaptic potentials (Perrais and Ropert, 1999). A similar pattern was observed in studies using alcohol (Roberto et al., 2003; Gomez et al., 2012). However, these are more difficult to interpret because of the pleiotropic effects of alcohol on other neurotransmitter systems, in particular the glutamatergic system (Thoma et al., 2011).

Mechanistically, it is possible that in the human brain, the activation of synapsin 1 is also increased, leading to a larger pool of releasable synaptic vesicles. The low GABA levels measured here might reflect a compensatory mechanism. Increased GABAergic neurotransmission can modulate GABA metabolism by downregulating GABA synthesizing enzymes, limiting the GABA available for packaging and release (Sheikh and Martin, 1998). Further studies are needed to evaluate these possibilities. In particular, studies in both patients and animal models employing 13C-magnetic resonance spectroscopy and molecular imaging using PET would deepen our understanding of the disease mechanism.

Interestingly, other neurodevelopmental disorders have been related with GABAergic dysfunction, namely, autism, Down syndrome, schizophrenia and Fragile X (Ramamoorthi and Lin, 2011). Similar to the alterations proposed by the neurofibromatosis type 1 mouse model, a mouse model of Down syndrome (Ts65Dn) showed increased inhibition (Kleschevnikov et al., 2004), whereas decreased GABA levels have been reported in patients in vivo (Smigielska-Kuzia et al., 2010) as well as in biochemical measurements in foetuses (Whittle et al., 2007). To develop potential treatments, it is critical to understand how enhanced inhibitory neurotransmission in mice models might be reflected in decreased GABA levels in humans. For that, it will be crucial to apply to the mouse models the analytical tools available to measure GABA levels in humans.

In this study, we also investigated the genotypic impact on GABA levels. The study of genotype–phenotype correlations in neurofibromatosis type 1 is hampered by the high mutation rate of the NF1 gene, resulting in >1000 different disease-causing mutations reported (HGMD, http://www.hgmd.cf.ac.uk). Nonetheless, by grouping mutations by type and performing linear regression analysis, we found that individuals with missense and splice-site mutations have significantly lower GABA/total creatine levels than control subjects.

Our findings are consistent with the idea that abnormal neurofibromin proteins resulting from missense or splice-site mutations can act as dominant negatives disrupting the function of the healthy allele, given that patients with neurofibromatosis type 1 are heterozygous for the mutation (Jett and Friedman, 2010), and resulting in a more severe phenotype than nonsense mutations (Khajavi et al., 2006). These results should, however, be interpreted cautiously in terms of mutation type, owing to the relatively reduced number of patients.

Finally, we confirmed that BOLD and GABA levels are negatively correlated in both patients with neurofibromatosis type 1 and control children, in agreement with previous studies (Muthukumaraswamy et al., 2009, 2012; Stagg et al., 2011) and with the notion that BOLD functional MRI response is sensitive to the excitation-inhibition balance and therefore to GABA concentration (Chen et al., 2005). The regression analysis for each group indicated that we can predict equally well in both groups the effect of GABA on BOLD (similar correlation coefficients, showing similarly explained variance). Peak haemodynamic responses were not significantly different between groups, implying that other factors besides the physiological coupling between BOLD and inhibition also contribute to the BOLD signal variance (Logothetis, 2008).

Regarding methodological considerations, several studies have addressed the issue of sensitivity and validity of GABA measurements using MEGA-PRESS (Puts and Edden, 2012). Still, it is relevant that the GABA peak may be contaminated by macromolecules (Rothman et al., 1993; Kegeles et al., 2007). However, it seems unlikely that differences between patients with neurofibromatosis type 1 and control subjects were explained by macromolecules, particularly taking into account the correlation found between GABA measurements and BOLD signal. Besides, there is mounting evidence for the biological relevance of GABA measurements using magnetic resonance spectroscopy and its correlation with behavioural processes (Edden et al., 2009; Sumner et al., 2010; Stagg et al., 2011).

In summary, we have reported GABA changes in patients with neurofibromatosis type 1. Our results show that the excitation/inhibition balance is altered in the visual cortex of patients, as a consequence of reduced GABA levels. Although our conclusions are derived from measurements of GABA in the visual cortex, it is likely that analogous abnormalities generalize to other cortical areas, in agreement with the widespread GABAergic alterations observed in mice (Shilyansky et al., 2010). GABA concentration was related to the type of mutation, providing a link between neurofibromin function and regulation of GABA metabolism in humans. Importantly, our experimental and methodological approach provides an example that could be applied to monitor GABA changes as a consequence of therapeutic approaches in neurofibromatosis type 1 and other neurodevelopmental disorders where alterations in GABAergic neurotransmission were proposed to underlie cognitive deficits, namely, autism, Fragile X syndrome and Down's syndrome (Ramamoorthi and Lin, 2011).


University of Coimbra (Grant number III/14/2008) and the Portuguese Foundation for Science and Technology (Grant numbers PIC/IC/83155/2007, PIC/IC/82986/2007, COMPETE PTDC/SAU-ORG/118380, individual fellowships SFRH/BD/41348/2007 to I.R.V., SFRH/BPD/34392/2006 to M.J.R and SFRH/BD/41401/2007 to I.B.).


The authors thank all the participants and their families for their participation in this study. The authors wish to express their gratitude to Dr. Graeme Mason for his insightful comments on the manuscript. The authors also wish to acknowledge the contributions of Carlos Ferreira and João Marques in technical assistance with MR scanning.

blood oxygen level-dependent
γ-aminobutyric acid


View Abstract