Association of serum retinoic acid with depression in patients with acute ischemic stroke

Retinoic acid (RA), produced by the metabolism of vitamin A, makes effects on depression and stroke. This study was aimed to evaluate the relationship between RA levels in serum and post-stroke depression (PSD). A single-center (Chengdu, China) prospective cohort study was conducted on patients with acute ischemic stroke. The RA serum level was measured at admission. The PSD was assessed in the 3-month follow-up. The RA-PSD relationship was evaluated with conditional logistic regression. In total, 239 ischemic stroke cases and 100 healthy controls were included. The median RA serum level in patients with ischemic stroke was 2.45 ng/ml (interquartile range [IQR], 0.72-4.33), lower(P<0.001) than 3.89 ng/ml of those in control cases ([IQR]: 2.62-5.39). The crude and adjusted odds ratios [OR] (and 95% confidence intervals [CI]) of PSD associated with an IQR increase for RA were 0.54 (0.44, 0.67) and 0.66 (0.52, 0.79), respectively. Higher ORs of PSD associated with reduced RA levels (<cut-off=2.8ng/ml) were observed (OR=3.01 [95% CI, 2.34-4.98]; P<0.001). This study revealed that, in patients with ischemic stroke, reduced RA serum level was related to higher risk of PSD at 3 months, which may be applied as a predictive indicator.

AGING Nucleus accumbens shell is crucial for depression and one study reported that RA signaling pathway was enhanced in this region [13], which influenced on the emotional behavior, such as depression and anxiety [14]. Accumulating studies have showed an association between RA and depression development [15][16][17]. However, the relationship between RA and PSD has not elucidated. We conducted a prospective single-center cohort study on patients with acute ischemic stroke (AIS) for evaluating the relationship between RA serum levels and PSD, and the clinical significance of RA was also compared with other known PSD indicators.

Main results
For the 363 patients with AIS, 53 patients were diagnosed as depression at 3-month follow-up, showing the incidence rate was 22.2% (95%CI: 16.9%-27.4%). The characteristics of the patients with or without depression were presented (Table 1). Compared with those without PSD, PSD patients showed higher age, ratio of female and widowhood, stroke severity and 3month functional disability, as well as higher serum levels of Hs-CRP and HCY.  Figure 2). The serum RA level was negatively correlated with the severity of depression (r[spearman]=-0.249, P<0.001), which was defined with HAM-D score. Furthermore, negative correlations were observed between RA serum level and the NIHSS score (r=-0.175, P=0.007), Hs-CRP serum levels (r =-0.209, P=0.001) HCY serum levels (r=-0.132, P=0.041) and age (r =-0.184, P=0.005). The RA serum level was not fluctuated with the blood collection time from stroke onset (P=0.17).
The ORs and 95% CIs of PSD associated with an IQR increase in RA were compared with other risk indicators including the NIHSS score (Table 2). Higher serum levels of RA significantly decreased the risk of PSD. The crude OR (and 95% CI) was 0.54(0.44, 0.67). The  OR was adjusted for potential confounders and the adjusted OR (95% CI) was 0.66(0.52, 0.79). In addition, age, sex, widowhood, higher initial stroke severity, 3-month functional disability and serum levels of Hs-CRP and HCY were significant indicators for depression, different from others factors (Table 2). Furthermore, multivariate analysis was performed to obtain adjusted OR (95% CI) of PSD associated with RA quartiles (with Q1 as reference) are shown in Table 3. PSD distribution for the RA quartiles ranged from 36.7% (Q1) to 8.5% (Q4) (P <0.001). RA serum levels in Q3 and Q4 were related to less risk of PSD, which was decreased by 48% (OR=0.52; 95%CI: 0.36-0.92; P=0.039) and 67% (OR=0.36; 95%CI: 0.18-0.63; P<0.001).

AGING
The cut-off value of RA for predicting PSD was calculated from ROC curves. When the RA serum level was 2.8ng/ml, the AUC was 0.69 (95% CI, 0.61-0.76), providing a best sensitivity of 77.4% and specificity of 51.6% ( Figure 3). RA presented a better discriminatory ability than that of age (AUC, 0.58; 95%CI, 0.50-0. 64

DISCUSSION
Previous studies have confirmed that PSD was caused by multiple factors, conformed to the biopsychosocial model of mental illness [18]. RA functioned in the development of the central nervous system [19]. In our prospective cohort study, the RA was proved as a new and independent predictor for indicating 3-month PSD in ischemic stroke patients. It was compatible with the existing indicator of clinical NIHSS score, while superior to that of other indicators.
Consistent with this study, Duan et al. [20] reported that decreased serum levels of RA at baseline were related to the risk of 3-month depression in AIS patients and Hou et al. [21] showed that reduced RA serum level was associated with 3-month poststroke cognitive impairment (PSCI) in AIS patients. PSD might occur through a pathophysiological pathway other than classical depression [22]. PSD may greatly influence the prognosis of stroke [23], both depression symptoms and cognitive impairment were independent predictors of impaired higher-level functioning and community reintegration 2 to 3 years after stroke [24]. A study showed that PSD made negative effects on functional outcome, despite of pharmacological therapy [25].
Various parameter would be tested on admission, some of which were probably associated with PSD, for. instance, vitamin D [26], vitamin B [27], insulin-like growth factor 1 [28], interleukin-6 [29]; hs-CRP [30], neopterin [31], tumor Necrosis Factor-α [32], and irisin [33]. Timely and accurate risk assessment was useful for depression prevention and care of stroke patients [34]. Our study illustrated that serum RA level was additionally helpful for predicting the 3-month PSD in AIS patients.  Significant association of RA level and age in PSD patients had been found. This is not surprising that the absorption of nutrients such as vitamin A decrease among older patients. Thus, the age would contribute to the association between RA and PSD. However, this relationship persists after adjusting for age in multivariate analysis. Furthermore, patients with depression was accompanied with higher infarct volumes and severer functional disability. RA levels were negatively correlated with the functional disability degree. The association between RA and PSD might be resulted from the stroke severity and disability. Again, correlation still persists after adjusting for these factors in multivariate analysis. Thus, the role of RA in pathophysiology of PSD might be functioned through the following pathways. First, RA played role in both the innate and adaptive immune systems [8], thus regulating the brain function at the cellular and systemic levels, finally functioned in pathology of depression [35]. Second, RA promoted the induction of proinflammatory cytokines and the maintenance of tissue homeostasis after the brain ischemia, finally contributing to the pathogenesis of mood disorders [36,37]. In mice with chronic inflammation in the ileum, oral supplementation with RA could alleviate inflammation by regulating the balance between the Th17 and Treg populations. In addition, the number of CD103+ DCs and RALDH2 expression would be enhanced via a positive feedback mechanism [38,39]. Third, neurovascular unit dysfunction with blood-brain barrier (BBB) hyperpermeability contributed to depressive disorder [40]. In an in vitro oxygen-glucose deprivation (OGD) treatment, RA made protective effects on BBB, depending on RA receptor α [41]. Fourth, increased oxidative stress in cerebral tissues during ischemia had been proposed to implicate in the pathogenesis of depressive-like symptoms following stroke [42]. One study showed that RA reduced staurosporine-induced oxidative stress and apoptosis and supported the antioxidant defense system [43]. One study reported that all-trans-RA (at-RA) increased nitric oxide (NO) synthesis in endothelial cells [44]. NO mechanism in the protective effect of naringin against PSD in mice had been recommended [45]. Fifth, bone morphogenetic protein (BMP) signaling in the hippocampus regulated depression, suppressed BMP signaling involved in the effects of some antidepressants [46]. 9c-RA had protective effects AGING against ischemic brain injury, and these effects involve BMP [47]. Lastly, RA and RA receptor (RAR) and their complex play roles in innate and adaptive immunity [48]. In addition, RA showed neuroprotective activity, which could increase barrier tightness in humaninduced pluripotent stem cell-derived brain endothelial cells by activating RARα, RARγ, and RXRα [49]. RAR plays both sides of homeostatic plasticity [50], and Chiang et al., [51] showed that RAR contributed to long-term depression, revealing a novel and unexpected role for vitamin A in higher cognitive functions. RARα was implicated in the activation of HPA axis thus participating in the etiology of depression [52].
Observative research design could not clarify the relationship between the RA, depression and outcomes in stroke patients. Machado-Pereira et al. [53] suggested that systemic administration of RA-loaded nanoparticles protected neurovascular integrity and improved the inflammatory environment from ischemia in the immature brain, demonstrating the effects of RA in the regulation of neuroinflammation. Plane et al. [54] showed that the RA could improve the survival of adultgenerated subventricular zone-derived neurons and stimulate post-stroke function recovery. 9-cis-RA had neuroprotective effects against ischemia-related brain injury [55]. At-RA showed potent therapeutic efficacy in ischemic stroke by attenuating neural inflammation through STAT1 signaling [56]. In cerebral ischemia rats, the infarct volume could be reduced with RA, and the neural functional recovery would be enhanced [19]. The role of serum RA during stroke, and whether supplemented RA could prevent depression to improve stroke prognosis need further experimental validation.

Study strengths and limitations
This study included first-ever AIS patients admitted 48 h within the onset of symptoms. These samples were more representative of the acute state. Furthermore, a range of confounding risk factors had been collected and adjusted. Lastly, we used the method of interquartile range in statistical analysis. There were still some notable limitations. First, this is a preliminary study with a small sample size (N=239) from one single center. The conclusion could not be over-interpreted. Second, patients with a lifespan of less than three months and aphasia were excluded, and these patients might be suffered from depressive symptoms. This might cause a selection bias, resulting in a lower rate of PSD than actual data. Third, the PSD patients' quality of life was not assessed in this study. Fourth, serum level of RA was tested once at admission, thus the changes in the disease course could not be monitored. RA levels were assessed without previous vitamin A and/or base treatment, which may affect RA determination. Thus, the results should be verified in the future. Furthermore, we tested serum rather than cerebral spinal fluid RA levels, thus RA levels in the central nervous system could not been determined. Lastly, this was an observational cohort study which could not draw any causal relationship. Whether maintaining normal RA serum level favored for PSD prevention should be further investigated.
In conclusion, reduced RA serum level was correlated to higher risk of 3-month depression in AIS patients, suggesting that RA may be a promising predictive biomarker for PSD. This may be used for improving prediction of depression and assignment of effective treatment for stroke in the future.

Ethical statement
This study was reviewed and approved by the Patient Research Ethics Committee of the Sichuan Provincial People's Hospital (Permission Numbers: 2016-03-18A) and conducted strictly adhered to the principles of the Declaration of Helsinki. Written informed consents were obtained from all participants or from their relatives (e.g., paralysis or blindness).

Included patients
This single-center prospective cohort study was conducted in Chengdu, China. From September 1, 2017 to October 30, 2018, all patients were first-ever AIS who were admitted to the Department of Emergency of our hospital. AIS was confirmed according to the criteria of WHO Multinational Monitoring of Trends and Determinants in Cardiovascular Disease (WHO-MONICA) [57]. The inclusion criteria included: (1) stroke was diagnosed based on magnetic resonance imaging (MRI) and/or computerized tomography (CT).
At the same time, 100 healthy volunteers with matched age and sex were recruited in control group from hospital medical examination center. Exclusion criteria for patients, was also applied to the control group. In addition, the psychiatric or cerebrovascular diseases AGING were excluded by a neurological doctor (not the author). The Hamilton Scale in those subjects should less than 7.

Clinical variables at admission
At admission, the basic information (age, sex, education and BMI), vascular risk factors (hypertension, diabetes mellitus, coronary heart disease and family history of stroke), family history of psychosis and marital status were recorded. Stroke severity was analyzed by skilled neurologists according to the NIHSS score [ranged from 0 to 42]; higher score indicated severer disease) [58]. MRI was used to evaluate the site and cause of stroke within 24hours after admission. Stroke etiology was evaluated with the TOAST (Trial of Org 10172 in Acute Stroke Treatment) criteria [59]. In addition, functional outcome was achieved at discharge based on the modified Rankin Scale (mRS) blinded to RA serum levels [60].

Assessment of depression
The depression was the primary end point of AIS patients in the 3-month follow-up. The depression was defined by the Diagnostic and Statistical Manual of Mental Disorders IV (DSM-IV) criteria. Patients with a diagnosis of PSD must show depressed mood or loss of interest or pleasure, accompanied with 2~5 symptoms of major depression lasted for at least 2 weeks [61]. In addition, the validated Chinese translation Hamilton Depression Rating Scale 17 items (HDRS, also abbreviated HAMD) was also used to quantify the severity of depressive symptoms [62]. Cut-off threshold for depression were chosen as >7 points for each of the HDRS [63]. Good internal consistency of HDRS was observed in the patients included in our study (Cronbach's α for HDRS-was 0.82). During the followup, antidepressant medication will be given immediately if the patients were diagnosed as PSD. The all-cause mortality within the 3-month follow-up was the secondary end point. Outcome assessment was performed by two skilled physician who were blind to the patients' information. A structured telephone interview was performed on the patient or their relatives.

Blood sampling and biomarkers testing
Fasting venous blood of patients was collected on admission, within 72 h after the onset of symptom (within 0-3[n=25], 3-12 [n=94], 12-24 [n=52], and 24-72[n=68]). The serum was obtained by centrifuge at 3000 g and then frozen at −70 °C until testing. The RA serum level was quantified with ELISA kit (catalog no. MBS705877; MyBio-Source). Previous studies had suggested the high sensitivity and specificity of this method in the determination of human RA [12,64]. This kit showed slight cross-reactivity with vitamin B-12(within 10%), vitamin D (within 8%) and vitamin A (within 10%). The coefficient of variance for the interassay and intra-assay repeatability was less than 10% and 8%, respectively. In addition, the serum levels of other parameters were also tested with standard methods, including homocysteine (HCY), hypersensitive C-reactive protein (Hs-CRP) and fasting blood glucose (FBG).

Statistical analysis
The normal distribution of data was evaluated with Kolmogorov-Smirnov test. The categorical variables and continuous variables were expressed as numbers (percentages) or edians (interquartile range, IQR), respectively. The comparison among groups was performed with Student's t-test or Mann-Whitney's U test. Correlation was assessed with Spearman's correlation coefficient.
The correlation between PSD and RA serum level was evaluated with conditional logistic regression models. A range of risk factors of PSD and potential confounders were confirmed according to univariate analysis. We adjusted the confounders in the model, including age, sex, family history of psychosis (yes vs. no), widowhood or divorced (yes vs. no), NIHSS at admission, time from stroke onset to blood collection, mRS at discharge, FBS, Hs-CRP, HCY and RA. The association between RA and PSD were presented as odds ratio (OR) and 95% confidence intervals (95% CI) associated with an interquartile range (IQR) increase. Furthermore, multivariate analysis was also used to obtain adjusted OR (95% CI) of PSD associated with RA quartiles (with Q1 as reference).
Accuracy of RA serum level and other factors to predict PSD was evaluated with ROC. Optimal RA serum level in the ROC was confirmed with the level with the highest Yuden index, for predicting depression symptoms after stroke. The effects of reduced levels of RA (<cut-off value, defined by ROC analysis) on PSD was also assessed. Statistics was performed with SPSS 23.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 8.0(GraphPad Software., San Diego, CA 92108, USA). P< 0.05 was considered as statistically significant.

Ethics
Written informed consents were obtained from all patients; and, this study conformed to the principles of the Declaration of Helsinki were approved by the investigational review board of the Sichuan Provincial People's Hospital.

AUTHOR CONTRIBUTIONS
CM had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: YC, CM, LW, ZD; Acquisition of data: YC, CM, ZD; Analysis and interpretation of data: YC, CM, LW; Drafting of the manuscript: YC, LW, ZD; Critical revision of the manuscript for important intellectual content: CM; Administrative, technical, or material support: YC, CM, LW, ZDJ; Study supervision: CM.

ACKNOWLEDGMENTS
We are grateful to the staff in the Department of Emergency from Sichuan Provincial People's Hospital for their support with patient recruitment. We also grateful to the patients who participate in the study.

CONFLICTS OF INTEREST
The authors declare that there is no conflicts of interest.