This is the first systematic review of all studies investigating the interplay between components of the stress response and the cannabinoid system in modulating psychosis risk and outcome in humans. Previous reviews have mainly focused on the two systems independently, summarizing the role of the hypothalamic-pituitary-adrenocortical (HPA) axis [13] or the stress response system [67] as well as of exogenous cannabinoids [6, 15] or the endocannabinoid (eCB) system [20] in increasing the risk of psychosis onset [6, 13, 20] or modulating psychosis outcome [15, 20, 67]. It is noteworthy that the evidence for an alteration and its direction in either the HPA axis/stress response system or in eCB components is not always unequivocal, possibly because of substantial methodological heterogeneity across studies [8].

Evidence from this review offers some insight into the stress-vulnerability model of psychosis, indicating HPA axis anomalies in the psychosis spectrum, both at baseline [26, 32] and following experimental [27] but not lifetime [31] stress exposure, with the latter being relevant for brain [28] and dopamine [33, 34] function as well as symptom manifestation [33]. Only one study did not find evidence of HPA axis abnormalities in schizophrenia [30]. In terms of outcome, a higher risk of incident clinical psychosis as a function of SLEs was found [58]. Also, apart from a few studies not supporting an association between CT and psychosis-related measures [40, 43, 45, 51, 60], CT, as a proxy of significant stress exposure, was found to be associated with schizotypy [35] and independently increase the risk of psychosis diagnosis and symptoms [38, 41, 42, 44, 46-48, 53, 54, 57, 61, 63] as well as earlier onset [39] and poorer outcome [36]. Further, female subjects exposed to CT [59], and earlier [47] and more severe [41] CT were found to carry a greater risk of psychosis-like experiences (PLE) [47, 59] and recent-onset psychosis (ROP) [41].

Different lines of research reviewed here also corroborate the cannabinoid-hypothesis of psychosis. Direct evidence points towards a psychosis-inducing effect of THC in otherwise healthy subjects [29], which is possibly accounted for by cognitive fusion [35]. Indirect evidence suggests that cannabis use may alter cortical [33] and striatal [34] dopamine signaling, with relevance for the onset of attenuated psychotic symptoms [33], in line with the long-lasting schizophrenia dopamine-hypothesis [68]. Also, cannabis use was found to increase the risk of PLE [46, 47, 53, 57, 61, 63], psychosis diagnosis [53] including ROP [42], incident clinical psychosis [58], psychotic symptoms in bipolar disorder [40], and poorer schizophrenia outcome [36]. However, a few studies have disconfirmed these findings [43, 45, 48, 51, 60], possibly because of a higher risk mainly in the context of recent/ current use [44, 54] as well as regular and more severe use [38, 41]. Earlier psychosis onset as a function of CU was also suggested [39]. Finally, cognitive intermediate phenotypes of psychosis were found to differ depending on CNR1 genetic variation and its interaction with cannabis use [66]. The CNR1 gene encodes for the CB1 receptor, whose localization on different neuronal subpopulations specifically modulates emotional and social behavior [69-72], with potential implications for the psychosis phenotype.

However, both preclinical and clinical studies have suggested a crosstalk between the eCB system and HPA axis and stress response, urging a deep reappraisal of the inter-relationship between these two systems, especially in the context of psychosis [8, 21]. Overall, this review indicates that the HPA axis and the eCB system, both at baseline and following an exogenous perturbation (e.g., stress exposure, cannabis use), interact with each other in health and disease, with relevance for the pathophysiology of psychosis. The most interesting evidence from this review regarding the interplay between the two models of psychosis is discussed in the following section.

First, THC and CBD, two cannabis ingredients with potentially opposite effects in psychosis [73], may exert specular effects also on the HPA axis. In fact, THC seems to amplify pre-existing HPA abnormalities among psychosis patients [26] and induce transient psychotic symptoms in otherwise healthy individuals through HPA disruption [29]. Conversely, CBD is suggested to counterbalance the detrimental effects of experimental stress on the HPA axis and behavior among individuals at risk of developing psychosis [27], despite not restoring the underlying dysfunctional brain activity [28]. In line with evidence from THC studies, cannabis use was found to disrupt the HPA axis among patients at different stages of illness, including clinical high-risk (CHR) [32], ROP [31], and schizophrenia patients [30]. Interestingly, such disrupting effect of cannabis on the HPA axis was evident also among control subjects (Fig. 2) [31].

Second, lifetime stress exposure (e.g., stressful life events, SLEs; childhood trauma, CT) was not found to modulate the HPA axis when the detrimental effects of cannabis use were taken into account [31]. Also, experimental stress exposure impacted patients’ HPA axis [33, 34], with relevance for the psychosis-related dopaminergic state [33, 34] and symptom severity [33], but such pattern of dopamine-cortisol response to experimental stress was reported to be blunted [33] or decoupled [34] in patients who also used cannabis (Fig. 2).

Third, THC was found to induce more severe symptoms of psychosis in those subjects with a history of CT [35]. In line with such evidence, some [39, 40, 46, 48, 51, 53, 61, 63] but not all [36, 42-44, 49, 52, 59, 60] studies report a greater risk of PLE [46, 48, 53, 61, 63], psychotic symptoms in bipolar disorder [40], and psychosis diagnosis [51, 53] and earlier onset [39] as a function of being exposed to both cannabis use and childhood trauma (CT), with the highest risk among individuals with earlier CT [47] as well as early [45], recent [54], and regular [38] CU. Further, additional known risk factors are suggested to play a role in the interaction between CU and CT on the risk of PLE, such as the COMT gene [60, 61], white matter integrity [62], urbanicity [50, 55, 56], threat-induced cognitive biases [57], cognitive alteration [56], cognitive fusion [37], and affective comorbidity [55, 56], with SLEs being relevant for ROP [41]. Noteworthy, additional evidence supports a direct role of SLEs in conferring a higher risk of incident clinical psychosis in the context of CU (Fig. 2) [58].

Fourth, an independent line of research indicates that either stress [64] and CT [65] affect endocannabinoid (eCB) signaling in both schizophrenia [64] and CHR [65] individuals. Also, a differential effect of cannabis use [64] as well as a dose-dependent effect of CT and psychosis risk intensification [65] on disrupting the eCB signaling have been found. Further, single evidence supports the notion that, in the context of specific CNR1 genetic vulnerability, first-episode psychosis (FEP) patients would present with higher stress and poorer memory and attention performance, with the highest perceived stress and the worst longer-term cognitive improvement in those with also a history of cannabis use (Fig. 2) [66].

The findings of this systematic review must be seen considering some strengths and limitations. While substantially converging, evidence regarding the association between aberrancies in the stress response system and the endocannabinoid system in psychosis is limited and very heterogeneous. It is currently unclear whether the two systems affecting one other are specific to psychosis or would be commonly reported across different clinical conditions. Also, considering evidence that earlier disruption of the two biological systems may result in more severe clinical phenotypes of psychosis, their interplay needs to be further addressed by tracking illness progression in longitudinal studies. Further, potential gender-driven vulnerabilities remain to be investigated. Treatment-wise, while the therapeutic effects of CBD in psychosis seem to be reasonably mediated by its action on both systems, such line of research is still in its infancy and no other treatment strategies to restore the two systems’ function have been identified. In fact, while CBD treatment has been reported to modulate eCB signaling [74], its direct effect on CB receptors has been disconfirmed in favor of serotoninergic [75] and dopaminergic [76, 77] activities, with the modulation of the crosstalk between the dopamine and cannabinoid systems [78] being particularly relevant for CBD antipsychotic effectuates as an early treatment [79, 80].