We have previously shown that levels of the endocannabinoid anandamide in cerebrospinal fluid (CSF) in antipsychotic-naive patients with schizophrenia ^{Reference Leweke, Giuffrida, Wurster, Emrich and Piomelli1–Reference Leweke, Giuffrida, Koethe, Schreiber, Nolden and Neatby3} are both markedly elevated and negatively correlated with psychotic symptoms. A model of dopamine/endocannabinoid interaction in acute schizophrenia was proposed where (over-)activation of dopamine D₂ receptors is associated with an increased release of anandamide counter-balancing dopamine-mediated psychotic symptoms by strengthening an endogenous feedback loop. This model suggests an adaptation of endocannabinoidergic function over a long period to a slow, gradual increase in dopaminergic neurotransmission. ^{Reference Leweke and Koethe4} We investigated whether changes in the endocannabinoid system are already noticeable in initial prodromal states of psychosis and test the hypothesis that an elevation of anandamide in CSF is apparent in this early stage of the disease.

Method

The ethics committee of the University of Cologne approved the study. All patients and healthy volunteers provided written informed consent.

We consecutively recruited 108 participants, comprising 81 healthy controls with no family history of psychiatric or neurological disorders (56% male, median age 27.3 years (inter-quartile range (IQR) 23.9 to 29.9)) and 27 patients with initial prodromal states of psychosis (70% male, age 24.3 years (IQR 19.9 to 29.1)). Controls were recruited by word of mouth and were paid for their expenses. Out of the 27 patients, 8 (30%) received antipsychotic medication.

Controls were screened for current or previous medical and psychiatric disorders. No cannabis use within 6 weeks prior to the study was accepted. Overall, 55 controls (68%) had taken cannabis <20 times/lifetime and 26 (32%) had used cannabis >20 but <50 times/lifetime.

Patients were assessed at the Cologne Early Recognition Centre using the Positive and Negative Symptom Scale (PANSS), ^{Reference Kay, Fiszbein and Opler5} the Scale of Prodromal Symptoms (SOPS), ^{Reference Ruhrmann and Klosterkötter6} and the Schizophrenia Prediction Instrument for Adults (SPI–A). ^{Reference Miller, McGlashan, Woods, Stein, Driesen and Corcoran7} An initial prodromal state was assumed if any of ten cognitive–perceptive basic symptom or any attenuated or transient psychotic symptom was present. After examination, 11 patients (41%) developed schizophrenia; 36% had already been treated with antipsychotics before transition. The 75th percentile of time from baseline to transition was 13 months; median follow-up was 61 months (IQR 56 to 80). Overall, 16 patients used cannabis <20 times/lifetime and 11 used cannabis >20 times/lifetime. All participants showed a negative urine drug screening for illicit drugs.

Cerebrospinal fluid and serum were collected around mid-day using a procedure described elsewhere. ^{Reference Giuffrida, Leweke, Gerth, Schreiber, Koethe and Faulhaber2} No participants revealed pathognomonic CSF findings according to a standard examination. ^{Reference Andersson, Alvarez-Cermeno, Bernardi, Cogato, Fredman and Frederiksen8} We measured CSF and serum levels of anandamide along with oleoylethanolamide, another endogenous fatty acid ethanolamide. These were purified and quantified by isotope dilution high-performance liquid chromatography/mass spectrometry as previously described. ^{Reference Schreiber, Harlfinger, Nolden, Gerth, Schömig and Klosterkötter9}

Skewed empirical distributions of continuous variables were summarised by median and IQR (25th to 75th percentile). Independent groups were compared by exact Wilcoxon rank sum test (corrected for ties). Spearman's rho was calculated to assess pairwise correlation of variables. The cumulative probability of transition to psychosis was estimated by the Kaplan–Meier method and compared overall by the log rank test (hazard ratio). Statistical analyses were performed using SPSS and R for Windows.

Results

In controls, the median concentration of anandamide in CSF was <0.001 pmol/ml (IQR<0.001 to 0.005) and <0.001 pmol/ml (<0.001 to 0.700) in serum. We found no significant correlation between anandamide concentrations in CSF and serum (r _S=0.033, P=0.768). Median CSF anandamide levels of patients were 0.006 pmol/ml (IQR<0.001 to 0.073, P=0.004; online Fig. DS1a) and 0.216 pmol/ml (IQR<0.001 to 0.515, P=0.838) in serum. Oleoylethanolamide levels did not differ in CSF or serum between controls (CSF: 0.680 pmol/ml (IQR 0.280 to 1.100); serum: 7.700 pmol/ml (IQR 4.885 to 11.000)) and patients (CSF: 0.600 pmol/ml (IQR 0.315 to 1.026), P=0.896, online Fig. DS1b; serum: 6.641 pmol/ml (IQR 2.867 to 9.891), P=0.139).

A significant inverse correlation between anandamide in CSF and symptoms in patients was observed for PANSS factor 3 ‘cognitive syndrome’ (r _S=–0.533, P=0.005). No other psychopathological measures correlated with ethanolamide levels.

In this sample, no relevant differences for CSF and serum levels of anandamide and oleoylethonalamide were found between patients and controls regarding smoking, cannabis use, medication or body mass index. No significant difference was found for all ethanolamide levels between medicated and antipsychotic-naive patients. Interestingly, patients with cannabis use <20 times/lifetime showed no significantly different ethanolamide levels compared with those with >20 times/lifetime use.

By splitting patients at the median CSF anandamide value (0.006 pmol/ml) into low and high groups, patients with higher anandamide levels in CSF were found to be associated with an extended transition time into frank psychosis (P=0.095; online Fig. DS1c). The relative risk of transition of individuals in the high v. low anandamide groups was 0.33 (95% CI 0.09–1.29). A similar analysis for lifetime cannabis use was not associated with the likelihood of transition to psychosis.

Discussion

This is the first study investigating levels of anandamide and oleoylethanolamide in CSF and serum of patients with initial prodromal states of psychosis. We found levels of anandamide in CSF significantly elevated in patients. The anandamide elevation was comparable to that described for first-onset, antipsychotic-naive schizophrenia ^{Reference Giuffrida, Leweke, Gerth, Schreiber, Koethe and Faulhaber2} and was found in CSF exclusively, making its central origin most likely. As oleoylethanolamide concentrations did not differ significantly between patients and controls, it is unlikely that changes in anandamide levels were caused by a generalised alteration in ethanolamide signalling.

Interestingly, we found a trend for an association of anandamide elevation in CSF with a longer time to psychosis transition in patients; this trend, however, did not reach statistical significance mainly owing to lack of power. This is in line with our hypothesis of an adaptive role for anandamide in schizophrenia, indicating that those patients expressing higher CSF anandamide are less likely to have psychotic symptoms. Four patients who transited to schizophrenia were treated with dopamine D₂-antagonists at examination and showed no significantly different ethanolamide levels compared with antipsychotic-naive patients. However, the observation that some patients did not show significant changes in CSF anandamide levels might imply a fluctuation or state-dependency of the alterations of the endocannabinoid system during the initial prodromal period. This has been recently suggested for other markers in CSF. ^{Reference Huang, Leweke, Tsang, Koethe, Kranaster and Gerth10} Alternatively, this may be related to the particular condition investigated, as not all individuals diagnosed with an initial prodromal state develop schizophrenia but may either return to a normal psychiatric status or develop another psychiatric disorder. ^{Reference Corcoran, Malaspina and Hercher11} It is interesting to note that in this sample a history of cannabis use did not have an impact on levels of CSF anandamide as previously reported in acute antipsychotic-naive schizophrenia. ^{Reference Leweke, Giuffrida, Koethe, Schreiber, Nolden and Kranaster12} Further, no relation between history of cannabis use and risk of transition to psychosis was observed. Both aspects may be related to the relationship between Δ⁹-tetrahydrocannabinol and cannabidiol in street cannabis consumed, which was not further evaluated. ^{Reference Morgan and Curran13}

Our findings suggest that endogenous anandamide mobilisation may play a protective role in at least a subgroup of patients with early stage schizophrenia. ^{Reference Leweke and Koethe4} Based on this observation, it may be hypothesised that assaying the endocannabinoids in CSF may provide useful biomarkers in early schizophrenia.