A Nomothetic Network Model Disclosing the Comorbidity of Depression and Unstable Angina: Effects of Atherogenicity, Insulin Resistance, Immune Activation, Antioxidants, the Endogenous Opioid System, Trace Elements, and Macrominerals

a Faculty of Veterinary Medicine, University of Kerbala, Iraq. E-mail: ranafadhil78@yahoo.com. b Department of Chemistry, College of Science, University of Kufa, Iraq. E-mail: hassann.hassan@uokufa.edu.iq. c Department of Ecology, College of Science, University of Kufa, Iraq. E-mail: hasana.albuthabhak@uokufa.edu.iq. d Department of Chemistry, College of Science, University of Kufa, Iraq. E-mail: headm2010@yahoo.com. e Corresponding Author: Department of Psychiatry, Faculty of Medicine, Chulalongkorn University, Bangkok, Thailand. f Department of Psychiatry, Medical University of Plovdiv, Plovdiv, Bulgaria. g School of Medicine, IMPACT Strategic Research Centre, Deakin University, PO Box 281, Geelong, VIC, 3220, Australia. E-mail: dr.michaelmaes@hotmail.com.


Introduction
The incidence of cardiovascular disease (CVD) is strongly predicted by atherosclerosis (ATS) (Rahman and Woollard, 2017) and CVD is the top-1 cause of mortality worldwide (Mc Namara et al., 2019). As long as atherosclerotic plaques do not obstruct more than 40% of the blood vessel lumen, ATS may remain asymptomatic, but when the obstruction proceeds, myocardial ischemia may develop with or without significant myocardial necrosis and the latter condition is named unstable angina (UA) (Wiviott and Braunwald, 2004). CVD/ATS patients show an increased incidence of depressive symptoms (Hare et al., 2014) and depression is associated with increased risk to coronary heart disease (Moise et al., 2016;Raič, 2017) and UA (Lespé rance et al., 2000). Moreover, depression due to ATS/UA is accompanied by an increased risk of cardiac morbidity and mortality (de Melo et al., 2017;Maes et al., 2011;Strik et al., 2001).
The main pathways causing ATS comprise increased atherogenicity, immune cell dysfunctions, increased oxidative stress, and insulin resistance (Frasca et al., 2017;Hansson, 2011). Increased oxidative stress toxicity and especially increased oxidation of low-density lipoprotein cholesterol (LDLc) with consequent IgG-mediated autoimmune responses and lowered antioxidant defenses are important pathways leading to ATS (Leopold and Loscalzo, 2009;McMurray et al., 1992;Sinha et al., 2013). There is now also evidence that major depression is a neuro-immune disorder characterized by immune activation, a chronic low-grade inflammatory state, increased oxidative stress and lowered antioxidant levels (Maes, 1995;Maes et al., 2011).
Comprehensive reviews showed that shared metabolic, immune-inflammatory and nitro-oxidative pathways may explain the comorbidity between depression and ATS/CVD (de Melo et al., 2017;Maes et al., 2011).
First, increased atherogenicity is not only a hallmark of ATS/UA, but also occurs in major depression (Bortolasci et al., 2015;Nunes et al., 2015;Vargas et al., 2014) as indicated by increased Castelli risk indices 1 and 2 and an increased atherogenic index of plasma (AIP) with lowered levels of high-density lipoprotein cholesterol (HDLc) and elevated triglyceride levels (Maes et al., 2011;Oliveira et al., 2017). Second, increased IR is not only a hallmark of ATS/UA, but also major depression is accompanied by increased IR as measured with the homeostatic model assessment of IR (HOMA-IR). Thus, a recent review and meta-analysis found a small but significant association between IR and depression, although there are also negative findings (de Melo et al., 2017;Kan et al., 2013;Landucci Bonifá cio et al., 2017;Silva et al., 2012).
Third, ATS/CVD and major depression share a chronically activated immuneinflammatory response system (IRS), as indicated by increased levels of various pro-inflammatory cytokines, including interleukin (IL-6), increased production of reactive oxygen and nitrogen species (RONS), and nitro-oxidative stress toxicity (NOSTOX) as indicated by higher levels of oxidized LDLc, malondialdehyde (MDA), and autoimmunity directed against oxidative specific epitopes, and lowered levels of antioxidants with immune regulatory activities, e.g. zinc (Maes et al., 2011). As such, the onset of depression in ATS/CVD may be explained by the combined activities of IRS, RONS and NOSTOX, and depression may increase risk towards ATS/CVD via increased atherogenicity, RONS and ROSTOX (de Melo et al., 2017;Maes et al., 2011).
Recently, we found that ATS may be discriminated from the healthy state using a combination of different biomarkers in machine learning models, namely an increased HOMA-IR index, IL-6, IL-10 (a negative immune regulatory cytokine), β-endorphin, copper and magnesium, and lower zinc (Qazmooz et al., 2020). Moreover, ATS with UA was significantly discriminated from ATS without UA by increased levels of triglycerides, IR, IL-10, β-endorphin, and mu-opioid receptors (MOR), and lowered levels of vitamin D3, another antioxidant (Qazmooz et al., 2020).
In this respect, it is interesting to note that increased IL-10 may play a protective role in ATS and in major depression (Maes and Carvalho, 2018;Mallat et al., 1999). IL-10 is a major component of the compensatory immune regulatory system (CIRS), which attenuates the IRS in immune disorders including ATS and major depression (Maes and Carvalho, 2018;Mallat et al., 1999).
Lower vitamin D3 is not only detected in CVD/ATS but also in major depression and perinatal depression (Sotoudeh et al., 2020;Trujillo et al., 2018), whilst vitamin D supplementation has a protective effect in depression (Vellekkatt and Menon, 2019). Increased levels of β-endorphin and Mu opioid receptor (MOR) are not only detected in CVD/UA, but also in major depression (Al-Fadhel et al., 2019;Browne and Lucki, 2019;Callaghan et al., 2018). These endogenous opioid system (EOS) peptides have immune-regulatory effects and, therefore, may exert CIRS functions in both CVD/ATS and major depression (Al-Fadhel et al., 2019;Qazmooz et al., 2020). Finally, a recent meta-analysis showed that elevated copper levels are associated with major depression and are involved in the pathophysiology of that illness (Ni et al., 2018).
Nevertheless, no studies have examined whether UA is accompanied by increased depressive symptoms and whether the association among UA and depression may be explained by shared pathways including increased atherogenicity (as assessed with AIP and Castelli risk 1 indices) and insulin resistance (as assessed with an increased HOMA-IR index), IRS (elevated IL-6) and CIRS (elevated IL-10) activation, lowered levels of antioxidants such as zinc and vitamin D, aberrations in the EOS as indicated by increased β-endorphin and MOR levels, and increases in copper. Hence this study was conducted to examine whether increased depressive rating occur in ATS or UA and whether the above-mentioned biomarkers of ATS/UA are also associated with 8 8 severity of depression in ATS/UA thereby suggesting that shared pathways may underpin bot ATS/UA and depression due to that condition.

Participants
This study recruited 120 ATS patients and 58 normal volunteers. All participants were admitted to the Sadr-Teaching Hospital, Najaf Governorate, Iraq from November 2019 to January 2020. We made the diagnosis of ATS employing the ICD-10 criteria (CM-170) based on a full medical history, physical examination, blood pressure estimations, echocardiography, Doppler sonography, and electrocardiogram (ECG). UA was diagnosed using the guidelines of the American College of Cardiology Foundation/American Heart Association Task Force (Anderson et al., 2013) and patients were classified according into class III (n=34) or IV (n=26) according to Canadian Cardiovascular Society (CCS) criteria (Campeau, 1976). As such, the ATS study sample was divided into three subgroups, namely ATS without UA (ATS), UA class III and UA class IV.
The European Society of Hypertension (ESH) and the European Society of Cardiology (ESC) guidelines were used to make the diagnosis of hypertension (HT) (Mancia et al., 2013). HT patients had blood pressure measurements of > 140 / 95 mmHg using a conventional sphygmomanometer in a seated posture and with the arm in the horizontal position after fifteen minutes of quiet sitting. The World Health Organization criteria (WHO, 2006;WHO, 2011) were used to make the diagnosis of type 2 diabetes mellitus (T2DM) when fasting plasma glucose (FPG) ≥ 7.0 mM, and glycated hemoglobin (HbA1c) > 6.5 %.
The 58 apparently healthy controls were age and sex-matched to the ATS groups and were free from ATS/CVD/UA. Exclusion criteria for patients and controls were renal or hepatic 9 9 diseases, myocardial infarction, stroke, cancer, (auto)immune disorders, neuroinflammatory and neurodegenerative disorders, and subjects with an albumin/creatinine ratio > 30 mg/g and serum CRP > 6 mg/dL, thereby excluding subjects with overt inflammation. We excluded subjects with psychiatric axis-I diagnosis according to DSM-5 criteria (APA, 2013), except subjects with a "mood disorder due to a general medical condition, depressed mood or diminished interest or pleasure in all or almost all activities". All patients had serum FBG <25 mmol/L and fasting insulin <400 pmol/liter to comply with the criteria of the Homeostatic Model Assessment (HOMA) calculator program and triglycerides < 4.5 mmol/L to comply with Friedewald's formula requirements. Written consent was obtained from all subjects before participating in the study, which was approved by the IRB of the University of Kufa (487/2019) in compliance with the International Guidelines for Human Research protection as required by the Declaration of Helsinki.

Methods
Depression was measured using the 1996 revision of the Beck Depression inventory (BDI-II) (Beck et al., 1996). This self-rating depression scale has high internal consistency, high testretest reliability and shows a strong correlation with the Hamilton Depression rating Scale (Beck et al., 1996). The BDI-II score can be used as a cost-effective instrument to detect depression including in patients with medical conditions (Wang and Gorenstein, 2013). The items of the BDI-II scale cover all symptoms of the DSM-IV diagnostic criteria for major depression. We computed the total score of all items as an overall index of severity of illness. The total BDI-II score was dichotomized using two different cut-off values namely ≥19 and ≥ 29. The first threshold value delineates moderate depression (versus no + minimal + mild depression) and the second threshold values delineates severe depression (versus no + minimal + mild + moderate depression) (Beck et al., 1996). We also computed BDI-II subdomain scores, namely a) key depressive symptoms (sum of the items sadness, pessimism, loss of pleasure, loss of interest), b) lowered self-esteem (LSE) (sum of the items past failure, guilty feelings, punishment feelings, self-dislike, self-criticalness, worthlessness), and 3) physiosomatic symptoms (sum of the items loss of energy, changes in sleeping pattern, changes in appetite, concentration difficulties, tiredness or fatigue, loss of interest in sex).
Fasting blood samples (10 mL) were drawn in the morning from all subjects after an overnight fast. After complete clotting, the blood was separated by centrifugation at 3000 rpm for 10 minutes and then stored at -80 o C until analyzed. Total cholesterol (TC), triglycerides (TG), calcium, magnesium, glucose, albumin, urea, and creatinine, were measured spectrophotometrically by kits supplied by Biolabo ® (Maizy, France). Serum high-density lipoprotein cholesterol (HDLc) was measured after precipitation of all other lipoproteins by a reagent consisting of sodium phosphotungstate and magnesium chloride and the cholesterol levels in the supernatant were measured spectrophotometrically. Low-density lipoprotein cholesterol (LDLc) was computed from Friedewald's formula: LDLc = TC − HDLc -TG/2.19. Serum CRP was measured by latex agglutination principles using a kit supplied by Spinreact ® (Barcelona, Spain). Copper and zinc were measured in serum spectrophotometrically by kits supplied by Giesse Diagnostics (Rome, Italy). Commercial ELISA sandwich kits were used to measure serum IL-6 and β-endorphin (Melsin Medical Co, Jilin, China), IL-10 (Elabscience ® , Hubei, China), insulin (DRG ® International Inc., NJ, USA), and MOR (Mybiosource ® Inc. San Diego, California, USA). The sensitivities of the kits were for IL-6: 0.1 pg/mL, IL-10: 4.69 pg/mL, β-endorphin: 0.1 pg/mL, insulin: 12.22 pmol/L, and MOR: 7.18 pg/mL. The intra-assay CVs (precision withinassay) of all assays were less than 10%.
As explained previously (Bonifá cio et al., 2020;Flauzino et al., 2019) we computed different atherogenic and insulin resistance indices using z unit-weighted composite scores, namely z TCz HDLc (zTC-zHDLc), reflecting the Castelli risk index; z TGz HDLc (zTG-zHDLc), reflecting AIP; and z glucose + z insulin (IR), reflecting insulin resistance (IR). These composite scores are strongly correlated with the indices they are intended to reflect (Bonifá cio et al., 2020;Flauzino et al., 2019) Statistical analysis Continuous variables are displayed as mean ± standard deviation (or standard error) and normality of distribution of continuous data was tested using the Kolmogorov-Smirnov test.
Comparisons of continuous variables between study groups were performed using analysis of variance (ANOVA) and differences between nominal variables using analysis of contingency tables (using χ 2 test). Univariate generalized linear model (GLM) analysis was employed to examine the relationships among the biomarkers and diagnostic classes while controlling for covariates including age, BMI, tobacco use disorder (TUD), and sex. Effect sizes were computed using partial η 2 values, and GLM-generated estimated marginal mean (SE) values were computed.
We conducted protected pairwise comparisons among treatment means and used p-correction for false discovery rate (FDR) to adjust for type 1 errors due to multiple comparisons (Benjamini and Hochberg, 1995). Binary logistic regression analysis was used to delineate the essential explanatory variables that predict depression (with no depression as reference group). We used multiple regression analysis to predict the BDI-II and subdomain scores using atherogenicity and IR indices, trace elements, immune and EOS biomarkers as explanatory variables while allowing for the effects of UA, age, sex, BMI and smoking.
We also conducted multilayer perceptron (MLP) Neural Network (NN) analysis to assess the prediction of increased BDI-II scores (either ≥ 19 or ≥ 29) as input variables and the biomarkers as input variables. In this study we employed an automated feedforward model and employed 2 hidden layers with up to 6 nodes and up to 250 epochs with mini-batch training and gradient descent. One consecutive step with no further decrease in the error term was used as the stopping criterion. The study sample was divided into training (46.7%), testing (20%) and holdout (33.3%) samples. Error, relative error, the area under the ROC curve, the confusion matrices, and the importance of the explanatory variables (displayed in an importance chart) were computed. Tests were 2-tailed, and a p-value of 0.05 was used for statistical significance. All statistical analyses were performed using IBM SPSS windows version 25, 2017.
Partial Least Squares (PLS) path analysis (SmartPLS) (Ringle et al., 2015) was employed to assess the causal association between biomarkers, atherogenicity, IR, and the phenome of ATS/UA including the BDI-II score. All variables were entered as single indicators or as latent vectors (LVs) extracted from their reflective manifestations. Complete PLS analysis was conducted using 5000 bootstrap samples when the outer and inner models complied with prespecified quality data, namely a) the model fit SRMR is < 0.080, b) LVs have good composite reliability (> 0.7), Cronbach's alpha (> 0.7), and rho_A (> 0.8), with an average variance extracted (AVE) > 0.500, c) LV loadings are > 0.666 at p<0.001, d) Blindfolding shows adequate construct cross-validated redundancies and communalities, e) Monotrait-Heterotrait analysis indicates adequate discriminatory validity. Moreover, Confirmatory Tetrad Analysis (CTA) was used to assess whether the reflective model of the LVs is not mis-specified. Differences in the PLS model and pathways between men and women were assessed using permutation and Multi-Group Analysis. Table 1 shows the socio-demographic data of ATS patients with (AST+MDD) and without depression (AST-MDD) and healthy controls. There were no significant differences in age, sex ratio, body weight, BMI, and TUD between the three study groups. There was a significantly higher incidence of UA and class IV (versus class III) according to the Canadian classification score in the ATS+MDD versus the ATS-MDD group. The systolic blood pressure was significantly different between the three diagnostic groups and increased from controls → ATS-MDD → ATS+MDD. Diastolic blood pressure was significantly higher in ATS patients than in controls. The frequency of T2DM were not significantly different between both ATS groups. This table also lists the differences in the drug state among the groups and shows that patients with AST+MDD were more frequently treated with aspirin, atorvastatin, bisoprolol, clopidogrel, isosorbide dinitrate, and trimetazidine than the AST-MDD group. Table 2 shows the measurements of the biomarkers in the study groups divided according to BDI-II scores. Total cholesterol, TG, LDLc, AIP, Castelli-1, FPG, insulin, IR, IL-10, βendorphin, and copper were significantly higher in ATS patients than in controls. Serum zinc and HDLc levels were significantly lower in the ATS patients in comparison with the control group.

Associations depression groups and biomarkers
Total serum magnesium was not different between the three groups. Serum vitamin D3 was significantly lower in ATS-MDD patients than the control group. MOR and IL-6 were significantly higher in the ATS+MDD group than in the other 2 study groups. We found that p-correction did not change the significant differences of all biomarkers listed in Table 2 and covarying for the drug state did not change the results. Table 3 shows the outcome of binary logistic regression analyses with AST+MDD as dependent variable (and ATS-MDD as the reference group). The first regression discriminated patients with BDI-II score ≥19 versus BDI-II < 19 and showed that 4 input variables significantly discriminated both groups (χ 2 =66.306, df=4, p=<0.001), namely IL-6, MOR, age and UA (all positively associated). The Nagelkerke effect size was 0.585, and the accuracy of the classification was 82.9% with a sensitivity of 84.3% and a specificity of 80.9%. The second regression separated ATS patients with BDI-II ≥ 29 from ATS patients with BDI-II < 29 and showed that IL-6 and UA were significant (χ 2 =43.02, df=2, p=<0.001) explanatory variables with a pseudo-R 2 Nagelkerke value of 0.442 and an overall accuracy of 80.3% and sensitivity=57.6% and specificity=89.3%.

Best predictions of AST+MDD versus ATS-MDD
The forced entry of the drug state in this regression did not change the significance of these two predictors. Table 4 shows the results of the most accurate neural network (NN#1) discriminating ATS+MDD from ATS-MDD patients (using BDI ≥ 29 as a cut-off value). This feedforward network was trained with 2 hidden layers, with 4 units in layer 1 and 3 units in layer 2 and hyperbolic tangent as the activation function in the hidden layers and identity in the output layer.
The sum of squares error term was much lower in the testing than in the training sample, and the percentage of incorrect classifications was reasonably constant in the three samples, indicating that the model learned to generalize from the trend and is not overtrained. The AUC ROC was 0.831 with a sensitivity of 71.4%, and specificity of 90.0%. Figure 1 displays the importance chart and shows that IL-6 has the highest predictive power of the model, followed at a distance by MOR, UA, zinc, β-endorphin, and total calcium, and again at a distance by magnesium and AIP. Table 4 also shows the network information and model summaries of NN#2 discriminating ATS+MDD from AST-MDD using BDI-II ≥ 19 as a cut-off value. This model was trained using 2 hidden layers, with 5 units in layer 1 and 4 units in layer 2, again with hyperbolic tangent and identity as activation functions in the hidden layers and output layer, respectively. The percentage of incorrect classifications was fairly constant in the three sets while the AUC ROC curve was 0.931 with a sensitivity of 88.0%, and specificity of 81.0%. Figure 2 shows that IL-6 has the highest predictive power of the model, followed by β-endorphins, MOR, AIP, zinc, insulin resistance, total calcium, and vitamin D3. Table 5 shows the intercorrelations between the total score and subdomain scores on the BDI-II and the biomarkers. In the whole study group, the BDI-II scores were significantly correlated with most biomarkers, except albumin, magnesium, and vitamin D3. In ATS patients, the BDI-II total score was significantly correlated with total calcium (r=0.303, p<0.01), IL-6 (r=0.530, p=<0.001), IL-10 (r=0.187, p=<0.01), and MOR (r=0.344, p<0.001). Table 6 shows the results of different multiple regression analyses with the total BDI-II score and symptom domain scores as dependent variables and the biomarkers as explanatory variables while allowing for the effects of age, sex, BMI, UA and the drug state of the subjects.

Relationships depression scores and biomarkers
We found that 53.9% of the variance in total BDI score (regression #10 , and 39.8% of the variance in BDI Key symptom score (regression #2) could be explained by age, IL-6, MOR, and UA. Figure  3 shows the partial regression of the BDI-II physiosomatic score on IL-6 after adjusting for the variables listed in Table 6, regression #1. Regression #3 shows that 45.5% of the variance in selfesteem could be explained by IL-6, UA, age, atorvastatin (all positively) and male sex. We found that 42.2% of the variance in BDI-II physiosomatic sub-score (regression # 4) could be explained by IL-6, UA, and atorvastatin (all positively) and FPG (inversely).
We re-ran all multiple regression analyses but now without UA, thereby allowing for other biomarkers to enter in the models. We found that 56.2% of the variance in total BDI score (regression #5) could be explained by IL-6, IR, MOR, UA, hypertension, and age. Up to 45.0% of the variance in BDI key symptoms can be explained by IL-6, IR, MOR, triglycerides, and age (regression #6). Regression #7 found that 48.8% of the variance in self-esteem could be explained by IL-6, IR, AIP, age (all positively), zinc (inversely) and male sex. Finally, we found that 39.7% of the variance in the physiosomatic subdomain score could be explained by IL-6, IR, MOR, age, and hypertension (Regression #8).  communalities and redundancies were more than adequate. Confirmatory Tetrad Analysis showed that the model was not mis-specified as a reflective model. We found that 66.2% of the variance in the LV was explained by the regression on six biomarkers and age. Permutations and Multi-Group Analysis disclosed no significant differences between men and women in this model. Figure 6 shows a clustered bar graphs with the summaries of the separate BDI-II and subdomain scores in 4 study groups, namely controls, ATS, UA class III, and UA class IV.

Association BDI-II scores and ATS/UA/CCS classes III and IV
Univariate GLM analysis (adjusted for age and sex) showed that the total BDI-II score (F=86.60, df=3/169, p<0.001), key depressive (F=47.86, p<0.001), LSE (F=40.60, p<0.001), and physiosomatic (F=40.17, p<0.001) subdomains were significantly different among the four study groups and increased from controls → AST →UA class III → UA class IV, except LSE which was not different between ATS and UA class III. Figure 7 shows the latent variable scores computed using the first PLS analysis comprising latent scores reflecting atherogenicity, IR, ATS+UA+BDI score and a combination of these LVs (all_LS). We also added IL-6, IL-10 (as most strongly associated with the BDI-II score), and the total BDI score in this clustered bar graph. The total BDI-II, and ATS+UA+BDI and all_LS latent variable scores were significantly different between the four groups. Increased atherogenicity was a hallmark of ATS, increased IR for ATS and UA class IV, increased IL-6 for UA class IV, and IL-10 for ATS and UA class IV.

Clinical aspects of the comorbidity of depression and ATS/UA
The first major finding of this study is that there is a strong association between depressive ratings and ATS/UA. As described in the Introduction, there is now evidence that both depression and ATS co-occur with bidirectional associations (de Melo et al., 2017;Hare et al., 2014;Maes et al., 2011;Moise et al., 2016;Raič, 2017;Strik et al., 2001). Moreover, the present study found significantly increased total and subdomain BDI-II scores in UA patients as compared with patients without UA. These findings extend those of Lesperance et al. (2000), who described an association between depression and UA (Lespé rance et al., 2000). The same authors showed that the BDI identified depression in 41.4% of UA patients and that depressed patients showed an increased risk of mortality due to nonfatal myocardial infarction or cardiac death (Lespé rance et al., 2000). The Coronary Psychosocial Evaluation Studies (COPES) reported that in UA patients, a high BDI score was associated with a significantly increased mortality, 42 months later as compared with patients without depression (Whang et al., 2010). The American Heart Association, in a recent publication, states that depression predicts a poor prognosis in patients with the acute coronary syndrome (Lichtman et al., 2014). Whang et al. (2010) established that UA patients with depression have an increased risk of all-cause mortality at 42-months even after adjusting for the Charlson comorbidity index, the GRACE risk score and left ventricular ejection fraction (Whang et al., 2010). As such, comorbid depression in ATS/UA appears to increase risk towards cardiac and non-cardiac death.
Furthermore, in the present study, we found that the BDI-II score and all subdomain scores were significantly higher in UA class IV as compared with class III, indicating that along the ATS spectrum from normal controls → ATS without UA → UA class III → UA class IV there is a gradual increase in the BDI-II scores, key depressive and physiosomatic symptoms, and a gradual decrease in self-esteem. In the present study, we included only depressed patients with a mood disorder due to a general medical condition, namely ATS/UA, and as such the findings indicate that the ATS spectrum is accompanied by increased severity of "secondary" depression. In this respect, depression due to medical illness is generally accompanied by more severe symptoms and more medical costs as compared with patients without comorbid depression (Katon et al., 2007;Maes et al., 2011;Wang and Gorenstein, 2013).

Depressive symptoms, atherogenicity, and IR in ATS/UA.
We found that the BDI-II scores, the ATS features, atherogenicity and IR indices were reflective manifestations of a single latent trait, namely "severity of the atherogenicity-IRphenome of ATS". Moreover, this new construct showed accurate "psychometric" properties, including internal consistency, predictive relevance, construct replicability, and convergent validity. As such, the latent variable scores extracted from these data constitute a replicable and reliable score reflecting the severity of ATS/UA, its atherogenic and IR-associated pathophysiology, and severity of comorbid depression as well. Moreover, these results show that the BDI-II score and its subdomains are manifestations of a same trait comprising atherogenicity and IR.
These results extend those of previous reports that major depression is accompanied by increased AIP, Castelli 1 risk index, and triglyceride levels, and lowered HDLc (Bortolasci et al., 2015;de Melo et al., 2017;Maes et al., 2011;Nunes et al., 2015;Vargas et al., 2014). Liang et al. (2014) observed that depressive symptoms are positively correlated with lipid levels including total cholesterol, triglycerides, and LDLc, and negatively with HDLc, and that these associations are partially mediated by heart disease (Liang et al., 2014). Lowered levels of HDLc in major depression are not only a consequence of activated immune-inflammatory pathways (Maes et al., 1997), but also of lowered levels of paraoxonase 1 (PON1) arylesterase activity Moreira et al., 2019a;Moreira et al., 2019b). Upon secretion of the PON1 enzyme from the liver into the blood, PON1 is integrated into HDL to form an HDL-PON1 complex (Moreira et al., 2019a). The anchored PON1 protects HDL (and LDL) from oxidation and activates HDL to optimize cholesterol efflux from macrophages to the liver (Moreira et al., 2019a). As such, the lowered PON1 activity (in part determined by PON1 Q192R single nucleotide polymorphisms) in major depression may be causally related to increased oxidation of HDL and LDL and, therefore may underpin ATS (Billecke et al., 2000;Chistiakov et al., 2017;Costa et al., 2005;Kowalska et al., 2015;Mackness et al., 2002) and the comorbidity between ATS and major depression (Moreira et al., 2019a).
A recent review and meta-analysis reported a significant association, albeit with small effect size, between major depression and IR, although some papers reported negative findings (Bonifá cio et al., 2020;de Melo et al., 2017;Kan et al., 2013;Silva et al., 2012). The IR index, insulin and glucose levels are associated with major depression, which is, in turn, is closely related to an increased atherogenic potential (Di Pino and DeFronzo, 2019). All in all, our results show that increased atherogenicity and IR are pathways leading to ATS (Frasca et al., 2017;Hansson, 2011) and to depression due to ATS/UA. Moreover, it appears that depression due to ATS/UA is more strongly related to IR than major depression.

Depressive symptoms, immune and oxidative stress biomarkers, and ATS/UA.
This study found that IL-6 and MOR are important biomarkers of comorbid depression due to ATS/UA and that the BDI-II scores are significantly and positively associated with IL-6, IL-10, MOR, β-endorphins, and copper, and negatively with zinc, magnesium and vitamin D. Increased levels of IL-6 and IL-10 are now established biomarkers of major depression (Maes and Carvalho, 2018) and significant correlations between depression scores and IL-6 and IL-10 are frequently observed (Al-Fadhel et al., 2019;Wiener et al., 2019). Inflammatory cytokines, including IL-6, not only increase the risk to develop ATS via activation of macrophages leading to endothelial dysfunction and atherothrombosis (Caruso et al., 2020), but also major depression . Increased IL-6 is also observed in patients with depression due to multiple sclerosis (Kallaur et al., 2016). The increased levels of IL-10 indicate that activation of the CIRS is another shared pathway among major depression (Maes and Carvalho, 2018) and depression due to ATS/UA (this study).
Moreover, lowered zinc is another shared biomarker of depression due to ATS/UA (this study) and major depression (Maes et al., 1997;Roomruangwong et al., 2017;Twayej et al., 2019). Zinc is an acute phase reactant which displays not only anti-inflammatory and antioxidant effects but also anti-atherogenic effects (Bao et al., 2010). Moreover, zinc deficiency may cause endothelial cell dysfunctions and supplementation with zinc may improve LDL oxidation, vascular endothelial cell functions, and inflammation (Bao et al., 2010).
Increased levels of MOR and β-endorphins are observed in major depression  and ATS (Wilbert-Lampen et al., 2007) and also in depression due to ATS/UA (this study). Both EOS biomarkers play a role in monocytic and endothelial endothelin secretion, thereby regulating the balance between vasoactive products, which mediate vasoconstriction (Wilbert-Lampen et al., 2007). β-endorphins and/or MOR agonists may contribute to ATS, plaque instability, endothelial dysfunctions, and IR (Okano et al., 2020). Moreover, these and other EOS system peptides are secreted during immune activation and may function as part of the CIRS (Al-Fadhel et al., 2019;Moustafa et al., 2020).
Nevertheless, two biomarkers of depression due to ATS/UA, namely increased copper and calcium (this study) were not always increased in major depression. For example, Narang et al.

Limitations
The results of our study should be interpreted with regard to its limitations. First, we used a case-control design and consequently, we may not establish firm causal interpretations. Second, it would have been even more interesting if we had measured other biomarkers of oxidative stress, including oxidized LDL and IgG responses to oxidized LDL, as well as Q192R PON1 genotypes and PON1 arylesterase activity.

Conclusions
Current views are that the functional impairments that accompany medical illnesses are a cause of depression (Wang and Gorenstein, 2013) and that depression in patients with ATS may partly be attributed to increased psychological stress (Glassman et al., 2009). Nevertheless, our results show that depression due to ATS/UA is a reflective manifestation of the atherogenicity and IR pathophysiology and the phenome of ATS/UA and that this single latent trait is strongly associated with immune activation, lowered antioxidant levels, the EOS, trace elements, and macrominerals. As such, atherogenicity, IR, immune activation, lowered zinc, and vitamin D3 as well as increased copper and calcium are new drug targets to treat depression in patients with ATS/UA. These new treatments of depression due to ATS/UA should consider a combination of anti-inflammatory agents targeting IL-6 trans-signaling, zinc, vitamin D, and magnesium, and maybe curcumin and resveratrol (de Melo et al., 2017).

Funding
There was no specific funding for this specific study.

Authorships.
All authors contributed significantly to the paper and approved the final version.          and the LS extracted from atherosclerosis, unstable angina and depression (ATS+UA+BDI) data, and the LS extracted from the latter and the atherogenic index of plasma, Castelli risk index 1, and the IR index (ALL_LS). In this graph we also show the measurements