Neurofibromatosis 1 (NF1) is the most common autosomal dominant single-gene neurodevelopmental disorder with incidence of 1:2700, [1] caused by loss of function mutations in the NF1 gene on chromosome 17q11.2 encoding for neurofibromin. Although identified by neurocutaneous manifestations, morbidity in childhood NF1 usually relates to cognitive, social and behavioural difficulties, with moderate cognitive impairment and academic underachievement in about 80% [2] and attention-deficit/hyperactivity disorder (ADHD) in 38–50% [2, 3]. Recent evidence of autism spectrum disorder (ASD) prevalence of ~ 25% with partial traits in a further 20% [4, 5] support NF1 as a promising single-gene syndromic model for understanding ASD pathology [6].

The neurobiology of the social and learning deficits in NF1 has been studied in Nf1^+/− mouse models and recent in-human studies [7]. Neurofibromin is a negative regulator of rat-sarcoma viral oncogene homologue (Ras); loss of neurofibromin causes disinhibition of the RasMAPK pathway with consequent GABA/glutamate disequilibrium, impairment in long-term potentiation (LTP) and synaptic plasticity [8]. Upregulation of the Ras pathway can also directly affect myelin formation and axonal integrity [9] and dysregulate nitric oxide signalling pathways in oligodendrocytes [10]. Recent diffusion tensor imaging (DTI) study in human NF1 [11] demonstrated increased apparent diffusion coefficient (ADC) values localised in the caudate and other deep grey nuclei, diencephalon and frontal white matter in NF1 children compared to controls, consistent with decreased neuronal density or myelin sheath disorganisation; the extent of these effects was associated with neurological symptoms. Other imaging studies in human NF1 have identified reduced cortical GABA [12, 13], reduced cerebral perfusion [14], alteration in diffusion-weighted imaging [15] and abnormal network connectivity on resting state fMRI [16, 17].

This emerging understanding of NF1 neural system pathophysiology from animal and human studies has provided a rationale for experimental intervention trials. Compensatory downregulation of Ras activation can be achieved by blocking its farnesylisation, using 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors (statins). Attenuation of Ras activity in Nf1 ^+/− mouse models using lovastatin [18] or, alternatively, through genetic co-deletion of the Pak1 gene (Nf1 ^+/− , Pak1 ^+/− ) [7] rescues the biochemical, electrophysiological and behavioural deficits, including normalisation of social memory and autism-like behavioural phenotypes. Convergence in effect of these two methods supports the specificity of the target mechanism. Further, the Pak1 gene co-deletion experiment illustrated the potential of such studies to illuminate causal pathogenic pathways, by suggesting functional localisation of the primary Ras-related pathology in the amygdala and other parts of the social brain network, and causal involvement of specific synaptic proteins [7].

Translational statin intervention studies in human NF1, based on the Ras downregulation hypothesis, have had mixed results. Improvements in verbal and non-verbal memory were reported within a 12-week phase 1 single-arm study examining the safety and tolerability of lovastatin in 23 children aged 10–17 years [19] and in a 14-week randomised controlled trial (RCT) of lovastatin in 44 10–50-year-olds [20]. Normalisation of pseudo-resting state functional connectivity in areas of the default mode network (DMN) following lovastatin treatment was found in a case series of 7 children from the prior child cohort [21]. A case-control study of trans-cranial magnetic stimulation in 11 adults with NF1 showed impaired synaptic plasticity and deficits in phasic alertness at baseline compared to controls, which improved after 4 days of high-dose (200 mg) lovastatin [22]. However, larger statin trials have found little effect. A 12-week double-blind, placebo-controlled RCT of simvastatin in 62 children with NF1 aged 8–16 found no group differences on primary behavioural outcome measures and minimal improvements in cognitive aspects of visual synthesis in the simvastatin group; [23] another RCT of simvastatin (84 children aged 8–16 years) found no improvements in cognitive deficits or parent-reported behavioural problems [24]. A 16-week RCT of lovastatin in 146 8–15-year-olds with NF1 and visuospatial learning/attention deficits found no improvements on a paired associate learning task [25].

This human intervention work has been on mid-childhood or older cohorts rather than in earlier development. Trials have also not specifically targeted NF1-autism behavioural outcomes or used multiparametric imaging techniques. We report here therefore on the first experimental trial of a statin in young children with NF1 with co-occurring autism, using detailed multilevel measurements designed to assess the pathogenic pathway identified in NF1 animal models from gene disruption to cognitive and behavioural pathology. These included (i) statin effects at a cellular level on Ras activation, using peripheral MAPKinase assay; (ii) multiparametric imaging to reflect different related aspects of neural system structure, neurophysiology and in vivo function; and (iii) NF1-relevant cognitive and autism-related behavioural outcomes. By investigating statin effects on these different levels, we aim to illuminate the dynamics and possible causal relationships of this pathogenic pathway in humans. Hypotheses were that (i) statin treatment in young children with NF1-autism would be feasible, safe and acceptable to families; (ii) peripheral MAPKinase assay and awake multiparametric imaging could be acquired; and (iii) that although the study was not powered for definitive treatment effect estimation, signals of change in MAPK and multimodal imaging parameters would be detectable, along with change in autism and other cognitive and behavioural symptoms. Specific imaging parameters for testing and hypothesised parameter changes were selected on the basis of the existing imaging literature in NF1, especially those linked to known abnormalities in idiopathic autism. Thus, we expected normalisation of the reduced GABA and perfusion metrics, and reduction in the abnormalities in DTI and connectivity metrics found in NF1 (see “Methods” section).

Previous statin trials in older children and adults have shown mixed effects, but most used lovastatin and measured outcomes at just cognitive or behavioural levels. This trial used simvastatin, considered the most effective neuroprotective statin [44]. It is also the first trial that has looked in detail at statin effects on ‘upstream’ process at cell and neural system levels, reflecting a pathogenic pathway between gene disruption and the autism-related behavioural psychopathological outcomes known in NF1. We interpret the outcomes therefore in relation to each of these levels, while acknowledging that the restricted sample size in this data-rich trial, and the variable amounts of data available for different analyses, limits precision of estimation.

At the cellular level, the moderate between-group point estimate showing peripheral lymphocyte reduction of MAPK function was in the hypothesised direction, consistent with a statin effect at the cell level on the Ras pathway activation; the wide 95% CI values ranged from a large decrease to a small increase. Preparation, international transport and storage of the samples may have introduced increased variance in the assay results.

At a neural system level, neuroimaging shows evidence of specific statin effects in regions of interest of the brain including frontal white matter, deep grey nuclei (lentiform, caudate and thalamic nuclei), cingulate gyrus, ventral diencephalon and occipital/occipito-parietal cortex. Detection of GABA in white matter has been reported in other studies, albeit at lower levels than in grey matter [45]. The effects of the statin on the multiparametric data are in a direction consistent with normalising many aspects of the underlying NF1-related neuropathology identified in previous studies. Thus, the suggested increased absolute GABA levels in frontal white matter is consistent with reversing the reduced cortical GABA found in previous studies in children and young adults with NF1 [12, 13] (a contrast to the increase found in animal experiments [7, 18]). The variation in GABA findings by brain region in our study is echoed in a recent NF1 animal study, [46] reporting differential localization of GABA between prefrontal cortex and hippocampus and speculating that this may relate to differential effects on pre- and post-synaptic receptors. In the future, it would be possible to study this important variability further in humans by measuring GABA type A receptor binding using [¹¹C]-flumazenil PET alongside GABA concentration with MRS, as in [12].

Interpretation of our evidence suggesting reduced Glx concentration in deep grey nuclei in relation to the existing NF1 literature is uncertain since findings on Glx concentration in NF1 have been previously conflicting. However, in young children with idiopathic autism, elevated deep brain Glx has been found in the anterior cingulate cortex in one large sample study [47] and reported to correlate with quality of social interaction in another [48]. The finding in this current study therefore can be interpreted within this context as positive in relation to autism symptoms.

The reduction in ADC found within the cingulate gyrus, and the significant ADC finding within the machine learning analysis, needs to be interpreted in the context of other work, which has shown increased ADC and decreased FA values in NF1 including in the cingulate [15]. Such findings suggest reduction in cellular packing and intra-myelinic oedema and have been associated with NF1 neurological symptom status [11]. The effects found in this current study therefore are consistent with reduced extra-cellular water free diffusion in NF1, and a positive simvastatin effect to reduce intra-myelinic oedema and improve cellular packing. The presence of microstructural abnormalities, reflected in increased ADC values, have also been described beyond NF1 in idiopathic autism [49–51] and potentially give these findings wider relevance in relation to this NF1-autism cohort.

The increased perfusion in the ventral diencephalon can be understood in the context of diminished perfusion in cingulate gyrus, medial frontal cortex, centrum semiovale, thalamus and temporo-occipital cortex found in NF1 children (n = 14, mean age = 10.2 years) [14] and related hypo-metabolism predominantly within the thalamus in FDG PET studies [52–54]. Statins may increase cerebral blood flow by improving cerebral vasomotor reactivity through increased NO bioavailability, promotion of microvascular reperfusion, and enhanced eNOS in the thalamus, as well as cerebellum, visual cortex and posterior cingulate [55].

No statistically robust difference in the DMN was identified between treatment and placebo groups, but findings at the 10% level raise the possibility of a trend that might be detected in a larger study. Diminished functional connectivity has been found in the posterior cingulate in human NF1 [16], and there is evidence from a small case series with children that statin treatment can induce improvements in functional connectivity in posterior cingulate cortex [21]. Here, simvastatin could potentially be acting in a focal manner on microstructural and vascular changes resulting in better regulation of function through a regional improvement in myelination and resultant neuronal function.

For behavioural outcomes, while the sample was too small for definitive estimation, we found that 25% of the simvastatin sample, compared to none of the placebo group, showed a clinical response using standard criteria measured using independently triangulated parent-report with clinician judgement.

This study demonstrates the acceptability and safety of simvastatin treatment for young children with NF1 and autism; feasibility of awake scanning, data acquisition and peripheral biomarker assay in such children given the right preparation; and the value of such a multiparametric approach in capturing the likely complexity of pathogenic mechanisms.

The trial findings are suggestive of specific simvastatin effects in brain areas that have been shown to be part of NF1 neural pathology in previous studies. Furthermore, many of these areas have functional significance as part of the ‘social brain network’, highly associated with social impairment and autism psychopathology [56]. This functional localisation may thus be relevant both to the high autism prevalence in NF1 and to how simvastatin could have specific remedial effects on NF1-autism at the level of brain structure and function.

In terms of pathophysiological mechanism, the initial rationale for statin intervention was its action in NF1 animal models to downregulate the Ras pathway with consequent effect to reduce GABA, improve synaptic long-term potentiation and rescue the behavioural phenotype [7, 8, 18]. This trial in young children gives evidence consistent with that model operating in humans through its evidence of a simvastatin treatment effect (albeit with wide CI) towards reduced cellular pMAPK activation on peripheral assay, and associated biologically plausible effects found on GABA/glutamate balance in FWM and DGN. However, the results also suggest simvastatin action through additional mechanisms, such as direct effects on myelin formation and regional axonal and astrocyte integrity in NF1. Pleiotropic effects of this kind from statins in the CNS are well recognised [57–59]. Our findings further suggest that treatment may affect such mechanisms in relevant functional brain areas in NF1 autism. This has future potential for insights into causal pathogenesis in autism and NF1 as well as suggesting more focused treatment targets. Larger studies will be necessary to further test these possibilities and to link them to any confirmed effect on behavioural symptom outcomes. While the initial results are encouraging and suggest specific hypotheses for further testing, this preliminary study was not powered to provide evidence to support clinical use of simvastatin in the disorder in children at this time.

In a wider context, the SANTA trial is, to our knowledge, the first RCT in syndromic autism, or indeed in clinical neuroscience generally, to have successfully tested effects simultaneously on relevant cellular activity markers, neural system multiparametric imaging and behavioural outcomes. As such, it provides a model of a new cohort of experimental intervention designs to link brain process and behavioural outcomes in the context of an experimental intervention trial. This has the eventual goal of treatment discovery in autism, plus the illumination of pathogenic pathways from gene effect to behavioural outcome in neuropsychiatric disorder; in terms of  both regional brain localization and underlying pathogenic mechanisms.