Physiology of synaptic pruning

Synaptic pruning plays a crucial role in brain maturation by reducing excessive amounts of synapses developed in utero (Chechik Reference Chechik, Meilijson and Ruppin1998). The process of pruning is activity dependent, using the principle of ‘use it or lose it’. Pruning followed by myelination to create more efficient circuits usually takes place during three age-related phases between birth and early adulthood (Box 1). Serial magnetic resonance imaging (MRI) has revealed that these phases appear to be associated with grey matter volume loss, which can be considered a proxy measure of the extent of pruning (Sowell Reference Sowell, Thompson and Tessner2001).

Pruning is carried out by microglia performing phagocytosis of unwanted synapses (Trapp Reference Trapp, Wujek and Criste2007). However, to preserve other active circuits, synapses which are to be eliminated need to be ‘tagged’ with a protein marker. Studies in mice show that rapidly replicating neurons produce a signalling protein called fractalkine, which binds to the receptor CX3CR1, which is found exclusively in microglia, causing activation (Jones Reference Jones, Beamer and Ahmed2010).

Animal studies have shown that microglia can also be activated in the prefrontal cortex by chronic stress (Wohleb Reference Wohleb, Fenn and Pacenta2012). The relationship between stress and microglial activation is based on the release of cytokines such as interleukin-6 (IL-6), induced by stressors. IL-6 penetrates the blood–brain barrier and activates microglia in the central nervous system (Liu Reference Liu, Ho and Mak2012). Furthermore, IL-6 plays a key role in pathogenesis of mood disorders.

Microglial activity can be measured via positron emission tomography (PET) using radioactive ligands binding to benzodiazepine receptors (Banati Reference Banati2002). More recently, quantitative analysis of microglial activation has been achieved using the ligand PK11195, another proxy measure of pruning (van Berckel Reference van Berckel, Bossong and Boellaard2008).

Cells supporting neurons – astrocytes – also assist synaptic pruning by identifying synapses primed for elimination (Chung Reference Chung, Clarke and Gordon2013). Astrocytes secrete a cytokine named transforming growth factor beta (TGF-β), which activates the complement cascade C1q, C4, C2 and finally C3 (Fishelson Reference Fishelson, Attali and Mevorach2001; Banati Reference Banati2002), again leading to microglial activation.

It is unclear why there are recurrent spells of pruning: whether onset is a threshold effect mediated by synaptic load, or a response to social, hormonal or genetic signals. There appear to be variable amounts of pruning in different parts of individual brains; again the reason is unclear, but it is possibly associated with differential brain usage by individuals.

Early-onset psychiatric disorders: three syndromes

Persisting, recurrent or brief psychotic illnesses in the absence of gross brain abnormality are called ‘functional’ psychoses (Murray Reference Murray, Hill and McGuffin1997). Across the world it is estimated that 1% of the adult population experience these syndromes. These include schizophrenia, bipolar disorder and a ‘hybrid’ psychosis called schizoaffective disorders, which is a mixture of symptoms seen in schizophrenia and bipolar disorder. These disorders usually show ‘prodromal’ features in adolescence such as attentional and memory difficulties, as well as sequencing problems and affective changes or apathy. The onset of acute symptoms such as mania, delusions and hallucinations generally occur in early adulthood, often in the context of major life changes or substance misuse. It is also not uncommon for individuals with psychoses to have experienced severe stress (for example sexual abuse) in childhood, which is known to heighten immune responsivity (Ayaydin Reference Ayaydin, Abali and Okumus2016).

Evidence of pruning in psychoses

Thompson and colleagues (Reference Thompson, Vidal and Giedd2001) serially scanned the brains of young adolescents with schizophrenia, alongside their healthy siblings and matched controls, over a 5-year period. They found a fourfold excess of permanent grey matter loss in the children with schizophrenia compared with controls. There was also evidence of excess grey matter loss in siblings, which subsequently ameliorated (Mattai Reference Mattai, Weisinger and Greenstein2011). Furthermore, a recent PET study found excess microglial activity in people with schizophrenia and, to a lesser extent, in people at high risk of the disease compared with controls (Bloomfield Reference Bloomfield, Selvaraj and Veronese2015).

Immunological research in schizophrenia has shown that the complement cascade is altered, with C1, C3 and C4 showing increased mean activity in people with schizophrenia, whereas C2 shows reduced activity (Hakobyan Reference Hakobyan, Boyaivan and Sim2005). A further study showed increase in interleukin 1 beta (IL-1β, an inflammatory marker) in patients with schizophrenia (Soderland Reference Soderland, Schroder and Nordin2009). The latest finding is an allele of the C4 gene (C4A) that is more likely to be present in patients with schizophrenia (Sekar Reference Sekar, Bialas and de Rivera2016).

In bipolar disorder, serial MRI scanning of adolescents (Gogtay Reference Gogtay, Ordonez and Herman2007) found that those proceeding to a bipolar illness or psychosis with mood lability (schizoaffective disorder) appeared to show excessive grey matter loss in the bilateral anterior and subgenual cingulate cortex rather than generalised grey matter loss. Major depressive disorder has been associated with increased microglial activation in the cingulate gyrus, a finding that is also being reported in bipolar depression (Steiner Reference Steiner, Walter and Gos2011).

Evidence in autism

Evidence points to an acceleration of brain growth during the early years in autism. Courchesne and colleagues (Reference Courchesne, Carper and Akshoomoff2003) have demonstrated that children with autism have lower than expected brain weight at birth, thereafter changing to excess brain weight during the next 3 years; both grey and white matter in all brain regions except occipital grey matter are involved, with the most growth demonstrated in the frontal, temporal and cingulate areas (Schumann Reference Schumann, Bloss and Barnes2010). Diffusion tensor imaging of young children with autism (Walker Reference Walker, Gozzi and Lenroot2012) suggests increased axons and myelination between neighbouring areas of the brain compared with more distal connections. Furthermore, there has been evidence of increased density of activated microglia in the dorsolateral prefrontal cortex, but not in other brain regions (Walker Reference Walker, Gozzi and Lenroot2012).

Attention-deficit hyperactivity disorder

Rapoport and colleagues (Reference Rapoport, Castellanos and Gogate2001) compared serial MRI findings from children with ADHD with those from children with childhood-onset schizophrenia. They found that the degree of grey and white matter loss in ADHD was less than that in schizophrenia. Grey and white matter losses were in the dorsolateral prefrontal cortex, caudate, pallidum, corpus callosum and cerebellum (Durston Reference Durston2003).

Current explanations

At present, the only well-evidenced causative mechanism to explain ADHD, bipolar disorder and autism is that family history increases risk of these disorders via multiple genetic routes. There is no evidence that immune reactions to vaccinations play a role in any of these conditions.

Schizophrenia is the only condition of the three for which where there has been an evidenced-based theory, that of dopamine and glutamate dysregulation, i.e. there is excess dopamine supply to the striatal areas due to reduced inhibition of this process caused by prefrontal inactivity and subsequent reduction in glutamate supply to the striatum (Stone Reference Stone, Morrison and Pilowsky2007). The reason for frontal inactivity has not yet been elucidated. From a genetic perspective, Harrison & Weinberger (Reference Harrison and Weinberger2005) have provided an overview of how the various putative schizophrenia susceptibility genes (such as DISC1, NRG1 and DTNBP1) could interact to produce neuropathology by way of synaptic pruning.

Potential common pathogenic pathway

A parsimonious explanation of the pathogenesis of all these disorders based on the above evidence is increasingly discussed by researchers in the field, albeit without peer-reviewed publications apart from a single paper on schizophrenia (Keshavan Reference Keshavan, Anderson and Pettegrew1994). The idea is that there are differing patterns of synaptic pruning: it is excessive in all brain regions in schizophrenia and also excessive in some brain regions in bipolar disorder (consistent with the spectrum of clinical presentation shared by schizophrenia and bipolar disorder); and it is inadequate at all phases in autism, and low grade but persistent across all of childhood in ADHD (again accounting for a shared spectrum of clinical presentation and/or comorbidity of these two conditions). It is also being considered that unipolar major depression is due to stress-induced ‘reactive’ synaptic pruning in frontal and cingulate areas potentially at any age.

Data interpretation problems and future strategy

The main problem with the available evidence is the small samples in each study, typically fewer than 20 in each arm, and the lack of replication of positive findings. It is likely that publication bias has limited the publication of negative findings. It also has to be noted that the positive findings are simply group mean differences. Furthermore, serial scanning of children and adolescents, especially using radioactive ligands, is a limiting factor. However, imaging of children in the early stages of illness carries less likelihood of confounding due to prolonged psychotropic use, both illicit and prescribed. Replication of the positive findings mentioned in this article would be helpful using the same inclusion criteria and controlling for psychotropic use.

Brain volume assessment and measurement of regional ligand-based perfusion tend to vary between different imaging centres; there is a need for multicentre studies with a single imaging protocol and central evaluation of findings. Similar problems exist in biochemical analysis of complement activity.

Thus, future strategy should involve a multicentre longitudinal MRI study of adolescents at risk for schizophrenia, bipolar disorder, ADHD or autism compared with siblings and age- and gender- matched controls, with annual scanning over 2 years. Furthermore, a study of activated microglia following diagnosis, looking at effects of time, psychosocial intervention and medication needs to be carried out with subsequent reporting controlled centrally.

Potential for biomarkers and novel interventions

Immunological findings could be utilised as biomarkers of active disease and prognostic indicators of chronicity. In particular, direction of travel of inflammatory markers (especially complement parameters) could help judge the duration of maintenance treatment required to avoid relapse in psychosis and ADHD. This is the main area of investigation currently, involving microglial and complement activity.

A combination of serial grey matter volume studies and studies of the extent of microglial or complement activation might predict outcome in first-episode psychosis and assist decisions on antipsychotic treatment duration. Similarly, continuing asymmetry of grey matter volume in bipolar disorder and lack of grey matter growth in ADHD could inform maintenance therapy regimes.

As regards immunotherapy, it is possible to develop drugs that increase neuronal pruning by activating the complement system and drugs that limit pruning by reducing glial activity (Box 3). However, immunotherapy can cause side-effects, including leukopenia and infection (Mak Reference Mak, Cheak and Tan2009b). Therefore, the prescription of immunotherapy should be based on a risk–benefit analysis and patient- and disease-related factors.

Rapamycin, used in immunosuppression after transplantation, enhances neuronal pruning, as evidenced in rats (Sato Reference Sato, Kasai and Kobayashi2012). This drug is toxic to humans, but variants of the molecule might be useful in autism, even after emergence of symptoms, as found in animal studies.

Minocycline, an antibiotic and anti-inflammatory agent, has been shown to reduce microglial activation, and has shown early promise as an augmenter of antipsychotics to treat negative and cognitive symptoms in schizophrenia (Levkovitz Reference Levkovitz, Mendlovich and Riwkes2010). In vitro studies of some antipsychotic agents also show attenuation of microglial activation via cytokine production (Bian Reference Bian, Kato and Monji2008). Lithium also appears to have an effect in reducing microglial activation via the P13K/Akt intracellular signalling pathway (Dong Reference Dong, Zhang and Dai2014), suggesting a strong rationale for its continued use in treatment.

The potential of increasing activated microglia via bone marrow transplantation has been confirmed using mouse models of Rett syndrome (Derecki Reference Derecki, Cronk and Lu2012), which is similar to autism. Analogous to the treatment of blood cancers (Davila Reference Davila, Riviere and Wang2014), microglia at an early stage of development could be removed from marrow regardless of a person's age, grown and activated in vitro before being returned to help the pruning process. Peripheral infusion of activated microglia is possible, as microglia migrate to the brain (Grossmann Reference Grossmann, Stence and Carr2002).

Finally, the finding of an excess of a C4 allele in schizophrenia could suggest gene silencing of this protein (Liu Reference Liu, Hammerlindl and Keller2011).