Elevated brain-derived cell-free DNA among patients with 1 first psychotic episode-a proof-of-concept study 2 3

3 Asael Lubotzky , Ilana Pelov, Ronen Teplitz, Daniel Neiman, Adama Smadja, Hai 4 Zemmour, Sheina Piyanzin, Bracha-Lea Ochana, Kirsty L. Spalding, Benjamin Glaser, Ruth 5 Shemer, Yuval Dor, Yoav Kohn 6 Department of Developmental Biology and Cancer Research, Institute for Medical Research 7 Israel-Canada, the Hebrew University-Hadassah Medical School, Jerusalem, Israel 8 Neuropediatric Unit, Shaare Zedek Medical Center, Jerusalem, Israel 9 Jerusalem Mental Health Center, Eitanim Psychiatric Hospital, Jerusalem, Israel. 10 Hebrew University-Hadassah School of Medicine, Jerusalem, Israel. 11 Karolinska Institute, Department of Cell and Molecular Biology Stockholm, Stockholm, 12 Sweden. 13 Department of Endocrinology and Metabolism Service, Hadassah Medical Organization and 14 Faculty of Medicine, Hebrew University of Jerusalem, Israel. 15 These authors contributed equally to this work 16 Co-senior authors. 17 18


30
Schizophrenia is a common, severe and debilitating psychiatric disorder. Despite extensive 31 research there is as yet no biological marker that can aid in its diagnosis and course prediction.

32
This precludes early detection and intervention. Imaging studies suggest brain volume loss 33 around the onset and over the first few years of schizophrenia, and apoptosis has been proposed 34 as the underlying mechanism. Cell-free DNA fragments (cfDNA) are released into the 35 bloodstream following cell death. Tissue-specific methylation patterns allow the identification of 36 the tissue origins of cfDNA. We developed a cocktail of brain specific DNA methylation 37 markers, and used it to assess the presence of brain-derived cfDNA in the plasma of patients with 38 a first psychotic episode. We detected significantly elevated neuron-(p=0.0013), astrocyte-39 (p=0.0016), oligodendrocyte-(p=0.0129) and whole brain-derived (p=0.0012) cfDNA in the 40 plasma of patients during their first psychotic episode (n=29), compared with healthy controls 41 (n=31). Increased cfDNA levels were not correlated with psychotropic medications use. Area

Introduction Early diagnosis in schizophrenia
Schizophrenia is a complex and chronic psychiatric disorder affecting around 1% of the global 50 population. It is associated with increased disability, mortality and high private and social 51 economic burden 1-3 .

52
Psychosis is the first presentation of schizophrenia that usually leads to the diagnosis. Although 53 psychosis typically onsets during adolescence and young adulthood, there is accumulating data 54 that demonstrates underlying biological changes, beginning years prior to the psychotic 55 symptoms 4,5 . Thus, identifying biomarkers that will allow early diagnosis and therapeutic 56 interventions is of the highest importance.

57
The etiology of schizophrenia is multifactorial, with a complex interaction of polygenic risk and 58 environmental factors 3,6 . Although over 100 genetic variants were identified contributing to the 59 risk of schizophrenia, their broad distribution in the general population precludes clinical use 7 .

60
Other potential biomarkers were studied, including neurotransmitters and their metabolites 8 , 61 associated endophenotypes such as smooth pursuit eye movement (SPEM) 9 and sensory gating 62 defects (P150) 10 . Results of those studies are somewhat inconsistent and are not specific enough 63 to allow their use as a diagnostic tool. Brain imaging has demonstrated notable structural 64 changes mostly in late and more chronic stages of the disease, too late for useful intervention 11 . 5 cfDNA molecules are rapidly cleared from the circulation with a half-life of ~15 minutes, thus 70 reflecting events taking place at the time of blood draw 14,15 . Although the mechanism underlying 71 the release and rapid clearance of the cfDNA is still unclear, many researchers have been trying 72 to develop methods that will enable the use of these short fragments for the diagnosis and follow-73 up of various medical conditions. For example, cfDNA-based methods of prenatal diagnosis 74 have been developed 16,17 , and ongoing work is advancing approaches for cancer liquid biopsy 18-75 20 and for detection of rejection in organ transplant recipients 15,21,22 .

76
The possibility of using cfDNA as a biomarker in the medical conditions detailed so far is based 77 on the fundamental genetic difference between the DNA of a host and that of a tumor, fetus or 78 the graft 21,23-26 . In comparison it is more complicated to assess changes in the levels of cfDNA in 79 plasma of healthy individuals or individuals without mutated DNA in their blood. It is known 80 that plasma levels of cfDNA vary over time depending on age, increased physical activity, and 81 the existence of various medical conditions such as infectious diseases 15 . Therefore, it has not 82 been possible to use the amount of cfDNA in the blood as a marker for a specific pathology.

83
Moreover, although the cfDNA that circulates in the blood originates from different tissues, it 84 has the same genome and therefore cannot be associated with a specific source tissue using DNA 85 sequencing 23,27,28 .

86
We and others have recently described an approach to identify the tissue origins of cfDNA, 87 based on tissue-specific methylation patterns. Such cell-type specific markers allow the inference 88 of cell death in multiple settings, for example cardiomyocyte cell death following myocardial 89 infarction or hepatocyte death in patients with liver metastasis 21,23-26 .

91
Neuro-anatomical changes are known in schizophrenia. Numerous longitudinal and cross-92 sectional imaging studies recruited individuals at high and ultra-high risk for developing a 93 psychotic illness and demonstrated dynamic brain changes emerging around onset and over the 94 years of schizophrenia. The studies generally demonstrate gray matter reduction in the frontal 95 and temporo-limbic regions 29-31 . Structural brain abnormalities in psychosis occur prior to full 96 blown symptoms and progressively worsen as psychosis develops. Although the underlying 97 causes and the exact timing of the morphologic changes in the brain remain obscure, apoptosis 98 has been proposed as a potential mechanism that could contribute to this progressive pathology.

99
This idea is also supported by post-mortem evidence of regional reduction of neuronal and glia  (Figure 1-figure supplement 1A). To determine assay sensitivity, we serially 126 diluted brain DNA into leukocyte DNA, and found that the brain markers allowed the detection 127 of as little as 0.1% brain DNA in a mixture, or just one brain genome in a mixture of a thousand 128 genomes (Figure 1-figure supplement 1B).

8
Total levels of cfDNA circulating in plasma were higher in patients with first psychotic episode 134 compared with healthy controls (Mann-Whitney test for controls vs. patients, p-value =0.017, 135 Figure 1A). cfDNA levels from all brain cell combined were higher in patients with first 136 psychotic episode compared with healthy ( Figure 1B,C). cfDNA levels from each brain cell type 137 separately or from whole-brain markers were also significantly elevated in patients (Mann-138 Whitney test for patients vs controls, neurons p-value =0.0013, astrocytes p-value =0.0016, 139 oligodendrocytes p-value =0.0129, whole brain markers p-value = 0.0012). Elevated brain-140 derived cfDNA was seen when measuring either its absolute concentration ( Figure 1D), or its 141 fraction (Figure 1-figure supplement 2A-D). This is consistent with elevated release of brain-142 derived cfDNA in patients with first psychotic episode, reflecting brain pathology, rather than a 143 non-specific effect.

144
To determine how well brain-derived cfDNA markers can distinguish patients with first 145 psychotic episode from healthy controls, we generated a receiver operating characteristic (ROC) 146 curve for the combined signal from all brain cell types. The brain cfDNA score was able to 147 identify the plasma of patients with a first psychotic episode with an area under the curve (AUC) 148 of 0.77 (p value=0.0003). At 90% specificity, the sensitivity for identifying patients with first 149 psychotic episode was 65% ( Figure 1E). The markers of each brain cell type were able to

152
There was no significant correlation between brain markers and clinical parameters including the 153 duration of the current episode, duration of hospitalization or the number of psychotic episodes 154 (see Source data 1).

156
Since previous studies found a correlation between the use of antipsychotic drugs and anatomic 157 changes in the brain 37 , we examined whether the increase of brain-derived cfDNA levels in 158 patients is related to their psychiatric drug intake. To this effect, we considered (1) the amount of 159 antipsychotic drugs taken on the day of the patient's blood test for cfDNA, and (2) the total 160 amount of antipsychotic drugs taken throughout the hospitalization until the blood test. These 161 data were taken from medical records. In order to generate a single score for all patients,

170
After correcting for multiple testing, we did not find any statistically significant correlation 171 between drug intake and cfDNA. Also, all statistical tests apart from 3 were non-significant on a 172 nominal level (p>0.05). We found nominally statistically significant correlations between the 173 levels of brain-derived cfDNA and drug intake for only one type of astrocyte marker (PRDM2) 174 and one type of neuron marker (ITF) (Spearman correlation coefficient for astrocyte marker 175 PRDM2: p-value=0.024/drugs on blood test day, and p-value=0.025/overall amount drugs; for 176 neuron ITF: p-value=0.016/overall amount of drugs).

179
We have demonstrated increased levels of brain-derived cfDNA in adult patients with a first 180 psychotic episode. This is a proof-of-concept study, demonstrating the feasibility of using 181 cfDNA as a biomarker in psychiatric patients, and the results call for replication in a larger 182 number of well characterized patient samples, requiring better control for diagnosis and drug 183 treatment. Further research is needed to examine the association of increased brain cfDNA levels 184 with specific diagnoses, mainly schizophrenia, and with other factors such as genetic 185 predisposition and course of disease.

186
We note that our study detected elevated brain cfDNA but did not reveal the cause of release of 187 these molecules to circulation. We hypothesize that our findings reflect brain cell apoptosis, 188 which is consistent with the literature indicating brain damage in such patients. However 189 elevated brain cfDNA could also result from a disruption of the blood-brain barrier, from a 190 defect in the local removal of debris from dying brain cells, or a combination of these biological 191 phenomena. Regardless, elevated brain cfDNA provides an intriguing, measurable link between 192 a physical process in the brain and a psychotic episode. Another alternative driver of cfDNA, 193 which we could not robustly rule out, was psychiatric drug treatment or illicit drug abuse both of 194 which could theoretically induce brain cell death, although we partially controlled for that.

195
Future exploration should aim at distinguishing between these possibilities. We also note that 196 samples were obtained after the initial stabilization and partial remission of the psychotic 197 symptoms, suggesting the possibility that a more dramatic elevation of brain cfDNA takes place 198 during the acute phase of the psychotic episode. Finally, we observed an elevation in the total 199 concentration of cfDNA among psychotic patients. Since brain cfDNA comprises a small 200 fraction of cfDNA even among patients (<1% of total cfDNA), this elevation must originate in other tissues. Elevated total cfDNA can be related to drug toxicity, stress mechanisms or more 202 likely derive from immune or inflammatory cells, perhaps related to the proposed link between 203 schizophrenia and the immune system 39,40 . In a recent paper 41 we characterized cfDNA released 204 from specific immune cell types. We suspect that the total cfDNA elevation originates from 205 these cells and hope to explore that in future research. our results suggest that pathological brain processes occur for a longer period than the acute 214 presentation. This is also supported by the literature as brain changes emerging around onset of 215 the schizophrenia and accumulate over the years of the disease. It would be very interesting to 216 sample patients in the earlier stage of psychosis, but indeed will be difficult due to ethical 217 (consent) issues. Our results may also aid in monitoring treatment response. Increased brain-218 derived cfDNA can be a marker of psychotic exacerbation in chronic schizophrenia. This study 219 was not designed to prove the above-mentioned hypotheses, and this could be a goal for further 220 research.

Materials and Methods
We recruited men and women over