Poststroke Depression Biomarkers: A Narrative Review

Poststroke depression (PSD) is the most prevalent psychiatric disorder after stroke, which is independently correlated with negative clinical outcome. The identification of specific biomarkers could help to increase the sensitivity of PSD diagnosis and elucidate its pathophysiological mechanisms. The aim of current study was to review and summarize literature exploring potential biomarkers for PSD diagnosis. The PubMed database was searched for papers published in English from October 1977 to December 2017, 90 of which met inclusion criteria for clinical studies related to PSD biomarkers. PSD biomarkers were subdivided into neuroimaging, molecular, and neurophysiological. Some of them could be recommended to support PSD diagnosing. According to the data, lesions affecting the frontal-subcortical circles of mood regulation (prefrontal cortex, basal nuclei, and thalamus) predominantly in the left hemisphere can be considered as neuroimaging markers and predictors for PSD for at least 1 year after stroke. Additional pontine and lobar cerebral microbleeds in acute stroke patients, as well as severe microvascular lesions of the brain, increase the likelihood of PSD. The following molecular candidates can help to differentiate PSD patients from non-depressed stroke subjects: decreased serum BDNF concentrations; increased early markers of inflammation (high-sensitivity C-reactive protein, ferritin, neopterin, and glutamate), serum pro-inflammatory cytokines (TNF-α, IL-1β, IL-6, IL-18, IFN-γ), as well as pro-inflammatory/anti-inflammatory ratios (TNF-α/IL-10, IL-1β/IL-10, IL-6/IL-10, IL-18/IL-10, IFN-γ/IL-10); lowered complement expression; decreased serum vitamin D levels; hypercortisolemia and blunted cortisol awakening response; S/S 5-HTTLPR, STin2 9/12, and 12/12 genotypes of the serotonin transporter gene SLC6A4, 5-HTR2a 1438 A/A, and BDNF met/met genotypes; higher SLC6A4 promoter and BDNF promoter methylation status. Neurophysiological markers of PSD, that reflect a violation of perception and cognitive processing, are the elongation of the latency of N200, P300, and N400, as well as the decrease in the P300 and N400 amplitude of the event-related potentials. The selected panel of biomarkers may be useful for paraclinical underpinning of PSD diagnosis, clarifying various aspects of its multifactorial pathogenesis, optimizing therapeutic interventions, and assessing treatment effectiveness.


INTRODUCTION
Poststroke depression (PSD) is the most prevalent psychiatric disorder after stroke, which affects nearly one-third of the survivors during first 5 years after disease onset (1)(2)(3). The diagnosis of PSD includes the following characteristics: (1) presence of major/minor depressive episode according to DSM-III-IV-5 or other valid approaches; (2) evidence of stroke from history, physical examination, and/or neuroimaging data; and (3) onset of PSD is temporally related to the stroke (3). Several epidemiological findings have demonstrated that PSD is independently linked to negative clinical outcomes, such as significantly longer hospitalization; more severe functional disability (3)(4)(5)(6); profound diminutions in physical, psycho-social, cognitive, and eco-social domains of quality of life (3,7); unsatisfactory results of poststroke rehabilitation (8); elevated rates of mortality (3,(9)(10)(11); higher risks of recurrent stroke at 1 year (12); as well as considerable strain for caregivers (13). Data mentioned above highlight the importance of identifying PSD among stroke survivors.
The detection of depressive symptoms at early stroke stages and recognition subjects at risk for PSD diagnosis remains challenging. Clinical measures currently used to assess PSD, especially in the acute poststroke patients, may lack the specificity necessary to detect symptoms (14,15). From this point of view, the identification of specific biomarkers might help to increase the sensitivity of PSD diagnosis. Moreover, it could be helpful for elucidating the pathophysiological mechanisms of PSD and ultimately lead to choosing specific targeted treatment (16).
Thus, we aimed to review and summarize the literature exploring potential biomarkers for PSD diagnosis.

METHODS
We searched PubMed database for studies published in English using keywords: "poststroke depression" and "depression after stroke." The search covered a period from October 1977 to December 2017. We also reviewed the reference lists of obtained articles for additional information. Further, human clinical articles related to biomarkers of PSD were subjected to a comprehensive analysis. The inclusion criteria were: (1) peerreviewed original studies with case-control design and all types of reviews, where the relationship between PSD and possible biomarkers was studied; (2) age of participants ≥ 18 years; (3) patients had ischemic or hemorrhage stroke at the time of entry; (4) valid instruments for PSD assessment; (5) standardized measurements for biomarkers. We excluded duplicate articles with the same data set and studies without sufficient data. All articles were reviewed and analyzed by the first author (O.L.). Received results were checked for accuracy by the second investigator (A.T.). Discrepancies if occurred were resolved by discussion and consensus.

RESULTS AND DISCUSSION
At the first stage, 764 clinical and experimental articles were identified, 90 of which met inclusion criteria and underwent a detailed analysis at the second stage (Figure 1). The results of included studies were subdivided according to the type of investigated biomarkers into neuroimaging, molecular, and neurophysiological. In turn, molecular PSD biomarkers were categorized into monoamines, growth factors, markers of neuroinflammation, markers of the hypothalamic-pituitaryadrenal axis, markers of oxidative damage, other metabolites, and genetic markers for convenient structural representation of the data obtained.

NEUROIMAGING BIOMARKERS OF PSD
In spite of being one of the straight roots of depression, a question if a stroke-determined neuroanatomical substrate is actually contributing to the development of PSD remains a matter of debate (17,18). The results of neuroimaging research in PSD patients are presented in Table 1. Methodological peculiarities of the main studies allow having an idea of the importance of clinical-anatomical relationships in this cohort of patients.
The results of some well-organized studies using different MRI/CT techniques do not yield evidence that lesions in a distinct neuroanatomical region induce PSD (19)(20)(21)(22). In contrast, a large number of clinical-neuroimaging investigations found significant associations between stroke location and PSD. Murakami et al. revealed that PSD was associated with infarcts located in the left frontal cortical region, left basal ganglia, and brainstem (23). These data are partially in line with the earlier results, which pointed to the importance of closeness of the damage focus to the left frontal pole (24), involving prefrontal or basal ganglia structures (25)(26)(27), for the PSD appearance.
A resting-state functional MRI study (28) identified changes in the affective network in PSD subjects and revealed that the stroke of the left orbital part of the inferior frontal gyrus was tightly associated with PSD severity. Yang et al. studied the neuroanatomical basis of PSD in relation to white matter connectivity (17). The researchers separated 17 nodes to construct a PSD-related subnetwork. They demonstrated that local efficiency of the subnetwork was significantly declined and this functional decrease was a predictor of PSD (RR:0.84, 95% CI:0.72-0.98). Damages in the left putamen, right insular cortex, and right superior longitudinal fasciculus were found to be correlated with PSD. According to Nys et al., PSD was connected to lesion size but not to lesion location (29).
The discrepancies among studies could be explained by the fact that the relationship between the stroke focus and the likelihood of developing depression may depend on the time since the onset of an acute cerebrovascular accident (30,31). A large amount of data confirming this point of view has been accumulated. Left-sided stroke lesion was a factor contributing to early onset of PSD (in 2 weeks) (32). Thalamic lesions were significantly associated with PSD in the acute stage of stroke (8). Left lenticulocapsular infarcts were an independent predictor of depressive symptoms at 1 month after stroke onset (33). These data are in agreement with one of the earliest CT-investigations, which revealed the importance of the left hemisphere and basal ganglia lesions for PSD existence at 2 months of the disease (34). The results of Lassalle-Lagadec et al. study demonstrated that a deterioration of default mode network, which play a key role in mood control, was correlated with PSD severity (35). The investigators revealed that the PSD score at 3-month follow-up was associated with changes in functional connectivity of the left middle temporal cortex and precuneus at 10 days after a first stroke. Ischemic lesions in the frontal-subcortical paths were significantly correlated with depression in contrast to non-depressed subjects at 3 months after stroke onset (36). Moreover, in further logistic regression analysis after adjusting for relevant confounders the relationship (OR:2.6, 95% CI:1.3-5.1) was still significant. Similar data about the crucial role of frontal-subcortical lesions with left-sided predominance (37)(38)(39) or without it (40)(41)(42) for PSD were also received earlier. Furthermore, Vataja et al. and Lauterbach et al. pointed to the special role of the left pallidum lesions for the onset of PSD (43,44).
According to a meta-analysis of 31 reports involving 5,309 subjects (18), patients with left hemisphere infarcts might be more vulnerable to PSD during first 6 months after disease onset (OR:1.5, 95% CI:1.2-1.9). Lesions in the left dorsolateral prefrontal cortex are linked to more severe PSD in chronic poststroke patients (45).
It was demonstrated that additional pontine microbleeds in acute ischemic stroke patients significantly increased the possibility of PSD (46), whereas lobar microbleeds decreased the remission rate of PSD (47). Tang et al. also received results, which suggested that lobar microbleeds might be crucial for the PSD onset in subjects with lacunar infarcts (48).
The results of some studies revealed that the type of ischemic stroke might be a risk factor for PSD. Subjects with total anterior cerebral infarcts had a greater prevalence of depression (OR:1.76, 95% CI:1.1-2.7) (49). At the same time, Arba et al., exploring the association between different subcortical ischemic stroke lesions and occurrence of PSD at 1 year after disease onset, established that the lacunar subtype was least associated with PSD (OR:0.71, 95% CI:0.55-0.93) compared to other stroke variants (50). Nevertheless, accumulation of subcortical lacunar lesions in the deep white matter, basal ganglia, and thalamus might be a more significant predictor of PSD than solitary lacunas (51,52).
In some studies, the influence of different types of cerebral vasculopathy for PSD in ischemic stroke patients was assessed. Enlarged perivascular spaces (markers of cerebral small vessel disease) in the centrum semiovale on axial T2 weighted MRI independently predicted PSD occurrence at a 3-month period after stroke, according to the Liang et al. data (53). Severe white matter hyperintensities, another cerebral sign that reflects cerebral microvasculopathy, were identified as an independent factor of PSD at 3 months after stroke onset (54). Moreover, intracranial atherosclerosis on MRI scans might be essential for prediction of PSD in ischemic stroke subjects (55). In general, mentioned data are in accordance with the meta-analytical evidence that the reduction of overall brain perfusion has an impact on PSD (56).
Few studies have examined clinical and neuroimaging correlations of PSD after hemorrhagic stroke. Stern-Nezer et al. after investigating 89 patients with spontaneous intracerebral hemorrhage concluded that PSD was not associated with hematoma volumes and presence of intraventricular hemorrhage (57).
Summarizing obtained neuroimaging data in PSD patients, the conclusion can be made that clinical-anatomical correlations were found only for ischemic stroke lesions. Localization of the focus, the volume of ischemia, and additional burdening anatomical factors were contributing to PSD occurrence. For Review n/a n/a n/a In acute stage and 3-6 months after stroke, PSD is associated with left-hemisphere lesion severity and proximity of the lesion to the frontal pole and related to the dysfunction of (cortico-) Frontiers in Neurology | www.frontiersin.org Neuropathological PSD Non-PSD 20 21 DSM-IV Basal ganglia and thalamic lacunes were the best neuropathological correlates of PSD; the higher the lacunar scores in these areas, the greater the risk of PSD (Mann-Whitney U test, z = −3.129; P = 0.002). Semiquantitative severity scores for white matter lacunes were also related to the occurrence of PSD (Mann-Whitney U-test, z = −2.211; P = 0.03). This was not the case for microinfarcts nor for diffuse or periventricular white matter demyelination (Mann-Whitney U-test, z = −0.425, probability values of 0.67, z = −0.705, 0.48, and z = 0.087, 0.93, respectively). The combined brain lacune score Frontiers in Neurology | www.frontiersin.org 1 year after the stroke onset, lesions affecting the frontalsubcortical affective network (prefrontal cortex, basal nuclei, and thalamus) predominantly in the left hemisphere can be considered as neuroimaging markers for PSD and as predictors of PSD development. Total anterior cerebral infarcts lead to a higher PSD occurrence, whereas lacunar lesions less often cause depression symptoms. Supplementary pontine and lobar cerebral microbleeds in acute stroke patients, as well as severe brain microvasculopathy, increase the likelihood of PSD.

Monoamines
Molecular markers of PSD are presented in Table 2. It was revealed that PSD patients had significantly lower liquor concentrations of 5-hydroxyindoleacetic acid (a 5-HT metabolite) compared to non-depressed subjects with acute stroke lesions and non-depressed patients without stroke lesions (92). The results demonstrate that serotonergic mechanisms are Cross-sectional Poststroke 29 n/a DST's sensitivity was 15%, its specificity was 67%, and its positive predictive value was 48%. DST yields no more information than would be gained from random assignment of the diagnosis of depression     implicated in PSD pathogenesis. Nevertheless, these data need to be further confirmed.

Growth Factors
Accumulating evidence shows that expression of brainderived neurotrophic factor (BDNF) is involved in the pathophysiological mechanisms of depression (93) and PSD (94).
A meta-analysis of four studies including 499 stroke patients (58) revealed that significant decrease in serum BDNF concentrations in the early period after stroke predisposed to the development of depression. The data correspond to the results of an earlier meta-analysis Noonan et al. (56). A growing amount of evidence indicates the pathogenic influence of insulin-like growth factor 1 (IGF-1) on a major depressive disorder (MDD) (95). Most of the studies showed that increased peripheral IGF-1 levels might predict the occurrence of MDD, whereas decreased IGF-1 levels might reflect the treatment effectiveness (96). Nevertheless, Yue et al. found no differences in serum IGF-1 concentrations in PSD patients as opposed to non-depressed poststroke patients and persons with MDD (97). On the other hand, the authors revealed significantly greater serum IGF-1 mRNA concentrations in PSD group compared to depressed subjects without stroke (97).
Therefore, the decrease of serum BDNF concentrations after stroke can be used as a PSD predictor.

Markers of Neuroinflammation
Despite serious methodological issues, current research found that immune dysfunction is crucial for the pathophysiology of PSD (98). Immunological mechanisms can initiate the inflammation-bound cell death in mood-related cerebral areas (99). Therefore, markers of neuroinflammation could be helpful for PSD diagnosing.
It was shown that increased early markers of inflammation predicted further PSD development. Tang et al. found that elevated serum concentrations of high-sensitivity C-reactive protein (Hs-CRP) in the acute stroke phase independently predisposed to PSD occurrence at 6 months after its onset (OR:1.3, 95% CI:1.2-1.5, and AUC value of 0.765, 95% CI:0.701-0.983) (59). Similar results were published by Yang et al. (60). The authors also established that higher risk of PSD is related to serum Hs-CRP concentrations ≥ 0.85 mg/dL (OR:7.8, 95% CI:4.2-14.6) (60).
Serum level of ferritin (an inflammatory acute phase protein) ≥ 130.15 µg/L was independently related to depression after a 2-month period of stroke onset (OR:5.4, 95% CI:1.7-16.8) in accordance with Zhu et al. (61). Patients with PSD showed higher levels of serum neopterin (a marker of cellular immune system activation) at admission compared with non-depressive poststroke subjects (21.6 vs. 14.6 nmol/L) (62). Furthermore, serum neopterin independently predicted PSD occurrence after 6 months of stroke onset (OR:1.95, 95% CI:1.36-2.81) and demonstrated a prominently higher discriminatory ability when compared to Hs-CRP.
Interestingly, IL-17 serum levels did not distinguish poststroke patients with and without PSD in Swardfager et al. study (66). Nevertheless, IL-17 was correlated with lower cognitive functioning in PSD patients. In those depressive individuals, IL-17 was related to increased lipid hydroperoxide, a measure of oxidative stress (ρ = 0.52), and decreased IL-10 (ρ = −0.48), in contrast to subjects without PSD (66). Authors concluded that cognitive PSD symptoms might be linked to IL-17 related signaling, including pro-inflammatory and anti-inflammatory imbalance and hyperoxidation.
Cytokines can realize their pathogenic impact by driving intrinsic apoptotic pathways, involving intracellular calcium, glutamate excitotoxicity, and reactive oxygen species, which significantly elevates the risk of depression (99). It was also hypothesized that the increased production of pro-inflammatory cytokines due to stroke may amplify the pro-inflammatory processes, predominantly in limbic regions, extensively activate indoleamine 2, 3-dioxygenase, and, consequently, reduce serotonin production in paralimbic areas, including ventral lateral frontal and polar temporal cortex, as well as basal ganglia. The sequential physiological dysregulation might result in PSD (101).
The complement system is usually considered to be an essential part of the innate immunity and a linkage to the acquired immunity throughout pro-inflammatory cytokine transmission. Nguyen et al. revealed that lowered complement expression in serum was associated with PSD symptoms at 3 months after stroke (67).
Some studies revealed that decreased concentrations of vitamin D, which is essential for immunoregulation, were associated with depression in poststroke patients (68,102). PSD subjects had significantly decreased serum concentrations of vitamin D within 24 h after admission compared to nondepressed individuals. Serum vitamin D lower than 37.1 nmol/L were independently related to the onset of PSD (OR:8.8, 95% CI:2.0-38.7) (68).

Markers of Hypothalamic-Pituitary-Adrenal Axis (HPA)
According to review (69), persistent HPA dysregulation occurs in up to 40% of stroke patients. The level of hypercortisolemia is moderately determined by the volume and location of the infarct. In 1-month to 1-year term after stroke onset hypercortisolemia is correlated with PSD (69). Kwon et al. compared the cortisol awakening response (measuring saliva cortisol directly, 15, 30, and 45 min after wakening) in PSD patients (2 months after a stroke or longer) with age-matched controls (70). In PSD group, salivary cortisol concentrations did not increase considerably at any measured time, demonstrating the blunted cortisol awakening response. Furthermore, a prominent adverse association between the cortisol awakening response and the severity of depression in PSD group was shown (70).
It was detected in the meta-analysis Noonan et al. that dexamethasone moderately suppressed elevated cortisol concentrations (OR:3.3, 95% CI:1.3-8.4) in PSD patients (56). Abnormal dexamethasone suppression test (DST) results at 3 weeks after stroke might be a potential PSD predictor (71). Nevertheless, some investigators declared low utility of the test. Thus, Lipsey et al. found that DST sensitivity and specificity was only 67 and 70%, respectively (72). Grober et al. indicated that the DST sensitivity and specificity was only 15 and 67%, respectively, whereas, the positive predictive value was 48% (73). Therefore, it was concluded that the DST provides no more information for PSD diagnosis (73).
Thereafter, it can be suggested that among HPA axis dysfunction markers, hypercortisolemia and the blunted cortisol awakening response are the most prominent in PSD patients.

Markers of Oxidative Damage
A great amount of evidence suggests that stroke is accompanied by oxidative stress. Some studies investigated the links between oxidative stress and PSD. Cichon et al. evaluated a possible relationship between plasma protein oxidative damage and the likelihood of PSD (103). The research showed that oxidative proteins damage was associated with the severity of PSD. Liu et al. found a positive correlation between serum malondialdehyde (oxidative stress biomarker) levels and PSD severity (r = 0.54) during 1 month follow up after stroke onset (104). According to the ROC-analysis, the optimal cutoff value of serum malondialdehyde concentrations as an indicator to support a PSD diagnosis was 2.898 nmol/mL, which yielded a sensitivity and a specificity of 77.9 and 81.1%, respectively, with AUC of 0.883 (95% CI:0.836-0.929). Elevated malondialdehyde higher than 2.898 nmol/mL was an independent predictive marker of PSD (OR:24.3, 95% CI:9.5-62.4).
Nevertheless, specificity of mentioned oxidative markers for delineating PSD and non-PSD poststroke patients should be supported in larger sample studies. Therefore, they couldn't be recommended for routine clinical practice.

Metabolites
Evidence suggests that elevated acute serum glucose concentrations might be a predictor of PSD after ischemic stroke. The PSD score (according to Beck Depression Inventory) at 12 months after a stroke had a positive association with the serum glucose concentration at admission (r = 0.32) (74). The authors established that the acute glucose concentrations higher than 126 mg/dL could be a predictor of PSD occurrence.
Gu et al. examined a possible association between serum uric acid levels within 24 h after stroke onset and the development of PSD at a 3-month poststroke period (75). They demonstrated that uric acid concentrations lower than 239.0 and higher than 328.1 µmol/L were independently related to the onset of PSD (OR:7.76, 95% CI:2.56-23.47, and OR:0.05, 95% CI:0.01-0.43, respectively). Summarizing previously obtained data, authors concluded that possible antidepressant effects of uric acid could be explained by its multiple antioxidant (scavenging of free radicals and reactive oxygen species, chelation of transition metals, prevention of lipid peroxidation) and antiinflammatory roles, decreasing blood-brain barrier permeability, and, consequently, diminishing central nervous tissue damage and neuronal death (75).
Substantial differences were established between the PSD and non-PSD patients with acute stroke regarding bilirubin concentrations (76). In post-hoc comparisons, the percentage of subjects with bilirubin ≥ 14.1 µmol/L was significantly greater among PSD patients (37.7 vs. 19.7%). After the final regression analysis, bilirubin concentration ≥ 14.1 µmol/L still independently predicted PSD. According to the authors' view, high bilirubin concentrations in PSD patients may reflect the intensity of initial oxidative stress, as well as indicate a higher level of perceived psychological stress (76).
ApoE plays a key role in lipid metabolism regulation. It was indicated that PSD individuals are more likely to demonstrate dyslipidemia and abnormal serum ApoE levels (105). Zhang et al. established that compared to non-depressed poststroke patients, PSD subjects had decreased peripheral ApoE microRNA expression and increased serum ApoE (77). Higher serum leptin (a hormone predominantly made by adipose cells) concentrations were found in PSD group in comparison with non-PSD poststroke individuals (38.5 vs. 8.2 ng/mL) (78). Increased serum levels of leptin at 7th and 30th day of poststroke also showed a correlation with later onset of PSD (79).
Xiao et al. used nuclear magnetic resonance spectroscopybased metabonomic analysis to determine urine metabolites that are significantly altered in PSD patients (80). This approach could differentiate PSD patients from healthy control and non-depressed poststroke patients with high accuracy. Authors identified the panel of urinary metabolites, which included arabinitol, formate, lactate, phenylalanine, and αhydroxybutyrate. They found that PSD patients had higher urine concentrations of lactate and α-hydroxybutyrate and lower urine concentrations of phenylalanine, formate, and arabinitol compared to healthy controls. The satisfactory predictability of the panel (AUC of 0.946) demonstrated that it could be a "good" classifier for PSD diagnosing (80). Zhang and Zhang (81) proposed a combined panel of six urinary biomarkers (azelaic acid, glyceric acid, phenylalanine, pseudouridine, tyrosine, and 5-hydroxyhexanoic acid) that might separate PSD patients from non-depressed poststroke individuals with AUC of 0.961 in a training set and 0.954 in discriminating blinded test samples.
Overall, considerable amount of limitations including restricted ethnical groups, a single metabolomics platform, inability to differentiate PSD from stroke patients with other neuropsychiatric disorders, and the necessity to collect cerebrospinal fluid from PSD patients to ensure that mentioned above serum/plasma/urinary markers are relevant to the focus of disease pathogenesis require future multinational investigations with appropriate methodology.

Genetic Markers
A number of studies consider that PSD might be caused by genetic susceptibility. Much attention has been paid to the serotonin transporter gene SLC6A4 polymorphisms, especially 5-HTTLPR, STin2 VNTR, and rs25531. Summarizing data about them were presented in Kohen et al. review (82). The SERT gene is found on chromosome 17q11.1-17q12 and includes 14 exons. SERT most often studied variant 5-HTTLPR, which is found in the promoter region, is divided into a long (L) and short (S) allele, based on the absence or presence of a 43 bp deletion/insertion polymorphism. rs25531 is a single nucleotide polymorphism (SNP), existing in a common (A) or rare (G) variant, which location is immediately upstream of 5-HTTLPR in the SERT gene. A STin2 VNTR SERT polymorphism is situated in intron 2 and comprises a variable number of nearly equal 17 bp segments (usually 9, 10, or 12).
Several studies were devoted to the serotonin transporter gene-linked polymorphic region (5-HTTLPR) genotype in PSD subjects. Most studies showed that S/S (short allele) 5-HTTLPR genotype was significantly more frequent in PSD patients compared to non-depressed poststroke subjects (82)(83)(84)(85)(86)(87). In contrast, LL (long allele) genotype was more prevalent in nondepressed poststroke subjects compared to PSD patients (87). This regularity can be explained by the fact that the transcription capacity of the S allele is lower to that of the L allele, leading to poor serotonin expression in the areas of action (106).
In addition to the 5-HTTLPR polymorphism, expression of SLC6A4 is influenced by DNA methylation status. Kim et al. found that hypermethylation of SLC6A4 promoter was independently related to PSD at 2 weeks and 1 year after ischemic brain incident. Moreover, it was significantly linked to the increase of depression severity during a 1-year poststroke period (88).
Few studies assessed the association between some serotonin receptors (5-HTR) and growth factors genotypes and PSD. 5-HTR2a (serotonin 2a receptor) 1438 A/A genotype was linked to major PSD, whereas BDNF met/met genotype was linked to major as well as minor PSD (86). The authors found a substantial association between 5-HTR2a 1438A/G and BDNF val66met polymorphisms for major PSD and a marginally significant association between BDNF val66met polymorphisms for both (major and minor) PSD (86). Furthermore, Tang et al. established considerable interactions between the HTR2c gene and PSD presence in the male Chinese poststroke individuals (90). They showed that rs12837651 T allele and rs2192371 G allele significantly correlated to PSD status with OR of 4.02, 95% CI:1.16-13.93, and 2.87, 95% CI:1.06-7.75, respectively (90).
Along with genetic profiles, BDNF secretion is influenced by epigenetic factors. In this regard, Kim et al. (91) demonstrated that BDNF promoter hypermethylation independently correlated with the prevalence, persistence, and incidence of PSD, as well as with aggravating of depression severity over a 1-year period after stroke (91).
A potential role of microRNAs in PSD pathogenesis was observed in the review of experimental studies by Yan et al. (107). Implementation of those results into clinical practice may be helpful for the diagnosis and prognosis of PSD.
On the whole, genetic markers, namely S/S 5-HTTLPR, STin2 9/12, and 12/12 genotypes of the serotonin transporter gene SLC6A4, 5-HTR2a 1438 A/A, and BDNF met/met genotypes, can reflect the hereditary predisposition of PSD. To epigenetic factors of PSD, higher SLC6A4 promoter and BDNF promoter methylation status can be referred.

NEUROPHYSIOLOGICAL MARKERS EEG
To assess abnormalities in EEG complexity in PSD subjects Zhang et al. used Lempel-Ziv Complexity (108). It was shown that PSD individuals had lower neural complexity in whole cerebral areas compared with poststroke non-depressed persons and healthy controls. As screening indicators for PSD, Lempel-Ziv Complexity parameters demonstrated more than 85% in specificity, sensitivity, and accuracy ( Table 3). The lack of severe PSD patients in this study and absence of correlations between a stroke location and neurophysiological data do not allow expanding obtained data on the entire PSD population currently.
Wang et al. examined quantitative EEG changes in PSD subjects with basal ganglia infarcts (109). Left-hemisphere PSD patients showed increased beta2 power in frontal and central areas, whereas right-hemisphere PSD ones showed increased theta and alpha power mainly in occipital and temporal regions. Additionally, for left-hemisphere lesions, beta2 power in central and right parietal regions provided high discrimination between PSD and poststroke non-depressed subjects, and for righthemisphere lesions, theta power was similarly discriminative in PSD patients showed lower neural complexity compared with non-PSD and HC in the whole brain regions. In stroke patients, significant correlation was between HDRS and LZC in the whole brain regions, especially in frontal and temporal. LZC parameters used for PSD recognition possessed more than 85% in specificity, sensitivity and accuracy  (77). The average incubation period of N200 and P300 ERPs waves was prolonged, and the P300 amplitude was decreased in PSD subjects in comparison with non-PSD stroke patients and healthy individuals (P < 0.01). Wenzhen et al. also revealed that the incubation period of N400 was significantly prolonged and the average amplitude of the ERPs component was reduced in PSD group in comparison with those in the non-PSD group (110). These findings demonstrate that PSD subjects are prominently worse at recognizing target stimuli, indicating lower perceptual abilities and/or cognitive processing (77,110). Therefore, neurophysiological markers of PSD, reflecting a violation of perception and cognitive processing, are the elongation of the latency of N200, P300, and N400, as well as a decrease in the P300 and N400 amplitude of the ERPs, could be used in the diagnosis process.

CONCLUSION
Summarizing obtained data, we can highlight that revealed biomarkers reflect complicated neurobiological mechanisms of PSD (111). They are caused by neuroanatomical substrates and involve different molecular signal pathways including serotonergic dysfunction, growth factors failure, neuroinflammation, HPA dysregulation, oxidative stress, and metabolic abnormalities. These acquired pathogenic mechanisms proceed against the background of a hereditary vulnerability, which links mainly to the serotonergic system of the brain and the mechanisms of neurotrophic support. Together, they lead to violations of emotional and cognitive processing. Some of them could be recommended to support the PSD diagnosis, while others need to be clarified before routine clinical use.
Concerning neuroimaging data, we can conclude that ischemic stroke lesion localization, its size, and additional burdening anatomical factors are pathogenically related to PSD. Lesions affecting the frontal-subcortical circles of mood regulation (prefrontal cortex, basal nuclei, and thalamus) predominantly in the left hemisphere can be considered as imaging markers for PSD and also as predictors of PSD development for at least 1 year after the stroke onset. Total anterior cerebral ischemia leads to a higher PSD occurrence; lacunar infarcts less often cause depression symptoms. Additional pontine and lobar cerebral microbleeds in acute stroke patients, as well as severe microvascular lesions of the brain, increase the likelihood of PSD occurrence. Considering that the maximal recovery after stroke is reached within the first year from its onset, it may be recommended to take into account the indicated localizations of stroke lesions in an acute period for preventive and therapeutic strategies of depression.
Resuming the studies on molecular markers, we can distinguish the following candidates, with the help of which PSD patients can be differentiated from non-depressed stroke patients. A significant decrease in serum BDNF concentrations at the early stage of stroke predisposes to the development of PSD. Increased early markers of inflammation (Hs-CRP, ferritin, neopterin, and glutamate), as well as serum proinflammatory cytokines (TNF-α, IL-1β, IL-6, IL-18, IFN-γ) and pro-inflammatory/anti-inflammatory ratios (TNF-α/IL-10, IL-1β/IL-10, IL-6/IL-10, IL-18/IL-10, IFN-γ/IL-10) are associated with PSD development. PSD is also characterized by lowered complement expression and decreased serum vitamin D levels. Hypercortisolemia and the blunted cortisol awakening response are the most prominent features of HPA axis dysfunction in PSD. In our view, studies of treatment effects directed on the above mentioned immunological and neuroendocrine mechanisms can validate these markers to be incorporated in routine clinical practice.
Genetic markers, namely S/S 5-HTTLPR, STin2 9/12 and 12/12 genotypes of the serotonin transporter gene SLC6A4, 5-HTR2a 1438 A/A and BDNF met/met genotypes, can reflect the genetic basis for the hereditary predisposition of PSD. To epigenetic markers of PSD, higher SLC6A4 promoter and BDNF promoter methylation status can be assigned.
Neurophysiological markers of PSD, reflecting a violation of perception and cognitive processing, are the elongation of the latency of N200, P300, and N400, as well as a decrease in the P300 and N400 amplitude of the ERPs. The validity of this biological marker should be additionally checked after antidepressant treatment.
In our opinion, taking into account previous remarks, the selected panel of biological markers may be useful for paraclinical underpinning of PSD diagnosis, clarifying various aspects of its multifactorial pathogenesis, optimizing therapeutic interventions, and assessing treatment effectiveness.
The validity of mentioned bellow markers for clinical practice has to be confirmed in further research. Among them are lower 5-hydroxyindoleacetic acid levels in cerebrospinal fluid, increased IL-17 serum levels, elevated serum concentrations of malondialdehyde and oxidative damage of proteins, high serum levels of glucose, uric acid, bilirubin, ApoE, and leptin together with changes in urine concentrations of arabinitol, azelaic acid, formate, glyceric acid, lactate, phenylalanine, pseudouridine, tyrosine, α-hydroxybutyrate, and 5-hydroxyhexanoic acid.

AUTHOR CONTRIBUTIONS
OL formulated the main concept, searched the literature, and wrote the manuscript. AT provided critical review and revision of the article. Both authors prepared the final version of the manuscript.