Large-scale analysis of DNA methylation identifies cellular alterations in blood from psychosis patients and molecular biomarkers of treatment-resistant schizophrenia

  1. Eilis Hannon
  2. Emma L Dempster
  3. Georgina Mansell
  4. Joe Burrage
  5. Nick Bass
  6. Marc M Bohlken
  7. Aiden Corvin
  8. Charles J Curtis
  9. David Dempster
  10. Marta Di Forta
  11. Timothy G Dinan
  12. Gary Donohoe
  13. Fiona Gaughran
  14. Michael Gill
  15. Amy Gillespie
  16. Cerisse Gunasinghe
  17. Hilleke E Hulshoff
  18. Christina M Hultman
  19. Viktoria Johansson
  20. Rene S Kahn
  21. Jaakko Kaprio
  22. Gunter Kenis
  23. Kaarina Kowalec
  24. James MacCabe
  25. Colm McDonald
  26. Andew McQuillin
  27. Derek W Morris
  28. Kieran C Murphy
  29. Collette Mustard
  30. Igor Nenadic
  31. Michael C O’Donovan
  32. Diego Quattrone
  33. Alexander L Richards
  34. Bart PF Rutten
  35. David St Clair
  36. Sebastian Therman
  37. Timothea Toulopoulou
  38. Jim Van Os
  39. John L Waddington
  40. Wellcome Trust Case Control Consortium 2
  41. CREeTable AR consortium
  42. Patrick Sullivan
  43. Evangelos Vassos
  44. Gerome Breen
  45. David Andrew Collier
  46. Robin Murray
  47. Leonard S Schalkwyk
  48. Jonathan Mill

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Objective

Psychosis - a complex and heterogeneous neuropsychiatric condition characterized by hallucinations and delusions - is a common feature of schizophrenia. There is evidence for altered DNA methylation (DNAm) associated with schizophrenia in both brain and peripheral tissues. We aimed to undertake a systematic analysis of variable DNAm associated with psychosis, schizophrenia, and treatment-resistant schizophrenia, also exploring measures of biological ageing, smoking, and blood cell composition derived from DNAm data to identify molecular biomarkers of disease.

Methods

We quantified DNAm across the genome in blood samples from 4,483 participants from seven case-control cohorts including patients with schizophrenia or first-episode psychosis. Measures of biological age, cellular composition and smoking status were derived from DNAm data using established algorithms. DNAm and derived measures were analyzed within each cohort and the results combined by meta-analysis.

Results

Psychosis cases were characterized by significant differences in measures of blood cell proportions and elevated smoking exposure derived from the DNAm data, with the largest differences seen in treatment-resistant schizophrenia patients. DNAm at 95 CpG sites was significantly different between psychosis cases and controls, with 1,048 differentially methylated positions (DMPs) identified between schizophrenia cases and controls. Schizophrenia-associated DMPs colocalize to regions identified in genetic association studies, with genes annotated to these sites enriched for pathways relevant to disease. Finally, a number of the schizophrenia associated differences were only present in the treatment-resistant schizophrenia subgroup.

Conclusions

We show that DNAm data can be leveraged to derive measures of blood cell counts and smoking that are strongly associated with psychosis. Our DNAm meta-analysis identified multiple DMPs associated with both psychosis and a more refined diagnosis of schizophrenia, with evidence for differential methylation associated with treatment-resistant schizophrenia that potentially reflects exposure to clozapine.

Article activity feed

  1. eLife

    ###Reviewer #3:

    This is the largest study of DNA methylation differences in the blood of controls and patients with psychosis, performed in a sample of 4,483 participants. As is predictable, the authors found significant differences in measures of blood cell proportions and smoking exposure in patients with psychosis compared with controls, and in patients with schizophrenia with clozapine treatment compared with other patients. They also detected differentially methylated positions in such comparisons. The authors have employed an appropriate methodology to search for schizophrenia- and psychosis- associated methylation changes, and the manuscript is interesting and well-written. However, I think a more extensive analysis may increase our insight about DNA methylation differences in schizophrenia, and is therefore necessary.

    1. An important question is whether the methylation differences are pre-existing the disorder or a consequence, an epiphenomenon of the disorder. The fact that the authors detect a higher number of DMPs when they exclude individuals with first episode psychosis from their analysis could suggest that the methylation differences are not present before the onset of the disorder. However, the authors have the resources and the ability to better answer this question. For example:

    1a) I think they should report in a separate section the results in the two samples of FEP individuals compared with age-matched controls. Can they identify any FEP-specific DMP?

    1b) Also, I think they could try to integrate their data with other blood methylation datasets, to see whether the DMPs associated with psychosis/schizophrenia have been associated with environmental risk factors associated with schizophrenia. For example, the authors could check the overlap of the DMPs with blood methylation changes associated with gestational age (PMID: 32114984; this work contains references to other studies that may be useful too). Data on methylation and cannabis or other environmental factors, if available, may be useful too.

    1c) The authors could also explore, in patients and controls, the relationship between age and methylation of the DMPs. An increase of the differences between patients and controls in older ages would suggest that the methylation differences are related to factors that are secondary to the disorders, while the presence of methylation differences at younger ages could suggest the opposite. Analyzing the interaction between methylation and age on case-control status could be an alternative way to answer this question.

    1. Sex is an important biological variable that the authors could analyze more extensively, considering that being male is a risk factor for schizophrenia, and is associated with a different epigenetic regulation. The authors have already the statistics to analyze whether the psychosis/schizophrenia-associated DMPs are also associated with sex. Moreover, they could analyze the interaction between methylation and sex on case-control status and/or perform analyses stratified by sex.

    2. The authors did not find association of schizophrenia with age acceleration. However, a recent study has performed a comprehensive analysis of 14 epigenetic clocks categorized according to what they were trained to predict: chronological age, mortality, mitotic divisions, or telomere length. I think it is relevant that the authors try to validate and perhaps extend the findings of Higgis-Chen and coll. ("Schizophrenia and Epigenetic Aging Biomarkers: Increased Mortality, Reduced Cancer Risk, and Unique Clozapine Effects", PMID: 32199607).

    3. Adjustment: I have not found any clear information about ethnicity/race. I assume the samples were mainly composed by white Caucasians. Did the authors perform any adjustment for ethnicity/race or population stratification? Also, were principal components of negative control probes included as covariates?

    4. Replication: was there any replication at the level of DMP in the data from Montano et al.? Also, if many DMPs are under genetic control, we should expect an overlap between DMPs in blood and brain of patients with schizophrenia. Have the authors analyzed such overlap?

    5. I think the authors should be more cautious in interpreting the clozapine data. They write: "Studies have also shown that higher neutrophil counts in schizophrenia patients correlate with a greater burden of positive symptoms (Núñez et al., 2019) suggesting that variations in the number of neutrophils is a potential marker of disease severity(Steiner et al., 2019). Our sub-analysis of treatment-resistant schizophrenia, which is associated with a higher number of positive symptoms (Bachmann et al., 2017), found that the increase in granulocytes was primary driven by those with the more severe phenotype, supporting this hypothesis." Actually, the fact that TRS cases are characterized by a significantly higher proportion of granulocytes could be related a "recruitment bias": because clozapine administration is associated with a risk of agranulocytosis, clozapine is usually not prescribed to patients with low number of granulocytes. I think this possibility needs to be mentioned, unless the authors can exclude it.

  2. eLife

    ###Reviewer #2:

    This is an important piece of work conducted to the highest standards of methodological rigour. By drawing together most case-control DNAm studies of schizophrenia in a single meta-analysis, this work will provide the most up-to-date information for some time, and is likely to generate a lot of interest.

    I think there are no critical methodological problems with the manuscript. Points for consideration include:

    1. The abstract details the (unsurprising) smoking results but lacks other findings, such as the GO analysis and the localisation of findings to previously associated GWAS loci.

    2. The authors could consider providing a DNAm-based predictor of SCZ/SCZ-resistance based on their dataset - to be tested in a series of leave-one-out analyses. In my opinion, this would provide further interest in the results, provide evidence of replication somewhat lacking from the current version, and could be used by others to test for SCZ/TRS prediction in their cohorts or for the purpose of PheWAS.

    3. There are a large number of findings reported with only a p-value given, and no effect size. In many cases, I think there's no reason that additional info couldn't be added.

    4. It's not sufficiently clear in the text how the effects of SCZ were disambiguated from TRS - when the latter group is nested within the first.

    5. Whether DNAm is a cause or consequence of liability to SCZ could be further examined in the paper - and I'm not sure why the authors have stopped short of further MR-based tests of this question.

    6. The correction for smoking is somewhat heterogeneous across studies ('smoking status'). If they were current non-smokers, was this recent? Further examination of whether reporting findings attenuate after inclusion of AHRR CpGs would provide greater confidence that some are not due to residual confounding. Alcohol and BMI are also likely to give rise to similar issues.

  3. eLife

    ###Reviewer #1:

    This is a large study of multiple cohorts of individuals with schizophrenia and controls and comparing DNA methylation in blood samples. The main findings are replications of smaller studies. The purported goal is identification of a biomarker but the impact of medication effects on blood cell composition cannot be ruled out and therefore confounds any conclusions about future utility. The confirmation of heavier smoking in individuals with schizophrenia also seems of limited use.

  4. eLife

    ##Preprint Review

    This preprint was reviewed using eLife’s Preprint Review service, which provides public peer reviews of manuscripts posted on bioRxiv for the benefit of the authors, readers, potential readers, and others interested in our assessment of the work. This review applies only to version 2 of the manuscript.

    ###Summary:

    This is the largest study of DNA methylation differences in the blood of controls and patients with psychosis, performed in a sample of 4,483 participants. This is an important piece of work conducted to the highest standards of methodological rigour. By drawing together most case-control DNAm studies of schizophrenia in a single meta-analysis, this work will provide the most up-to-date information for some time, and is likely to generate a lot of interest.

    As predictable, the authors found significant differences in measures of blood cell proportions and smoking exposure in patients with psychosis compared with controls, and in patients with schizophrenia with clozapine treatment compared with other patients. They also detected differentially methylated positions in such comparisons. The authors have employed an appropriate methodology to search for schizophrenia- and psychosis- associated methylation changes, and the manuscript is interesting and well-written.

  5. Version published to 10.1101/2020.04.05.026211v2 on bioRxiv
  6. Version published to 10.1101/2020.04.05.026211v1 on bioRxiv