This study aimed to elucidate mechanisms underlying side effects associated with antipsychotic use in dementia. To our knowledge we provide the first evidence mechanistically linking antipsychotics with specific cardiovascular and infectious diseases which are common side effects of their use in dementia. Supporting our hypothesis, the initial high throughput screen identified three conditions related to known side-effects which were associated with the antipsychotics; atherosclerosis cases were enriched for positive scores with all three antipsychotics, venous thromboembolism cases were enriched with positive scores for risperidone and volinanserin, and influenza cases were enriched with positive scores for amisulpride. Supplementing these drug-disease associations, a number of biological pathways related to cardiovascular biology, infectious disease and inflammation/immune system were enriched across antipsychotic signatures. These findings suggest specific cardiovascular and immune processes may underlie some harmful effects of antipsychotics and for the first time provide a number of candidates which can now be prioritised for further investigation.
Notable pathways enriched in risperidone include BDNF, PDGFR-beta, TNF and TGF-beta signalling. Findings from previous in-vitro and in-vivo studies strongly implicate PDGFR-beta in atherosclerosis and cardiovascular disease, providing a possible mechanism to explain the positive association between the three antipsychotics and atherosclerosis and VTE observed in this study [23]. Similarly, BDNF also plays a role in the cardiovascular disease (as well as neuroplasticity and development) [24, 25] and is expressed in a variety of blood cells, the heart and vasculature [26]. It is also noteworthy that previous studies have demonstrated that part of risperidone’s pro-cognitive therapeutic mechanism of action may be via BDNF [27]. It is evident from our findings that more work must be done to untangle this complex element of antipsychotic MoA, where BDNF is plausibly related to both beneficial and detrimental effects of antipsychotics, which is highly relevant to dementia where the margin between clinical benefit and harm is so narrow. Two pathways linked to the essential amino acid selenium were enriched in amisulpride and volinanserin. Selenium plays a role in preventing oxidative stress and has been widely linked in observational studies to cardiovascular disease and atherosclerosis [28]. Moreover, one study in patients with schizophrenia implicated selenium deficiency in the adverse cardiac effects of clozapine, though it was not clear whether the deficiency was caused by the drug or the schizophrenia itself [29]. Our findings bring greater clarity to this previous work by providing evidence that antipsychotics directly act on selenium pathways. This has particular relevance to neurodegeneration where selenium deficiency in Alzheimer’s disease brain tissue has been observed and is hypothesised to play a role in cardiovascular side effects in Parkinson’s disease [30, 31]. Our findings provide a clear indication for prioritising study of selenium deficiency and its interaction with antipsychotics in people with neurodegenerative disease in order to understand if it may be a clinically useful marker.
Infectious disease and immune pathways were also enriched across all three antipsychotics. These included a range of viral and influenza-linked GO terms in amisulpride and volinanserin, and TNF and TGF-beta in risperidone. Consistent with this, a recent study showed a considerable global suppression of immune response in mice treated with risperidone, indicated by reduction in a number of cytokines during treatment [32]. Our findings suggest that this impact extends to other antipsychotics and so underscore the need to prioritise investigation of immune response in people with dementia. They also suggest that susceptibility to infection associated with antipsychotics is not solely secondary to more general effects of antipsychotics like sedation-induced inactivity or failure to clear the chest.
Although more work needs to be done to build on the candidate mechanisms highlighted in this study, their initial identification is an important step which could ultimately have important implications for clinical decision making. For example, the incorporation of more formal cardiovascular history screening, with a particular focus on thrombosis risk or selenium deficiency, into clinical decision making could result in greater harm reduction.
We note that there were differences in the pathways enriched between antipsychotics however it would not be appropriate to draw direct comparison between them at the specific pathway level or interpret differences as clinically relevant. This is because these experiments were conducted in-vitro, so cellular responses will be affected by dosing and duration of exposure to each compound, similarly, equivalent doses and bioavailability of drugs in humans will differ. At a broader level however, it is worth noting that associations between antipsychotics and side-effects, and enrichment of relevant biological pathways were observed across all compounds, despite their differing MoAs. Further comparison in different biological models, including those where ageing and frailty can be incorporated, and epidemiological studies is now warranted [33]. This line of investigation could have important implications for Alzheimer’s disease, Parkinson’s disease, and elderly people with schizophrenia where clinical trials of amisulpride and pimavanserin (a highly selective 5-HT2A inverse agonist) have recently been published and more antipsychotic-like drugs are in development [2, 34–36].
The overall trend towards downregulation of genes in this experiment is also worth comment. This pattern was particularly notable in risperidone, where 53 genes were upregulated and 756 were downregulated. However, although notable this is not without precedent. One study, with a similar design, which treated SK-N-SH neuroblastoma cell lines with risperidone for 24 hours showed 80% of genes were downregulated in analysis of microarray data [12].
With regard to limitations, the design and analysis of this study follows the same principles as Cmap and therefore the same caveats apply. These include the comparison between cell line-derived signatures and human studies, specifically that it would be premature to draw concrete conclusions on the clinical profile of compounds based on these data alone. However, as with Cmap, the trade-off is an experimental design which provides a high throughput low cost screen, analogous to a drug repurposing experiment where thousands of licensed compounds are triaged against a single disease signature. Similarly, in this study, screening three antipsychotic signatures against thousands of diseases showed that mechanisms underlying venous thromboembolism, atherosclerosis and infection may be relevant to the side effect profiles of antipsychotics, providing a clear rationale for prioritising their investigation in different biological models and epidemiological studies. In doing so, this study also represents an important step towards safety screening for compounds in development of neuropsychiatric symptoms in Alzheimer’s disease.