Decreased ribosomal DNA transcription in dorsal raphe nucleus neurons differentiates between suicidal and non-suicidal death

Brainstems of suicide victims with unknown psychiatric comorbidity (typical for most of the suicide cases autopsied in our Department of Forensic Medicine) and sudden death controls were obtained during routine forensic autopsies in accordance with existing EU law regulations. The study has been approved by the local ethics committee of the Medical University of Gdańsk and performed in accordance with the ethical standards laid down in the Declaration of Helsinki of 1989.

Detailed diagnostic, demographic and toxicological data together with raw data of AgNOR measurements are presented in Supplementary Table. Violent suicide methods prevailed in the suicide cohort (23 out of 27), which are representative for our autopsy material. All brains were free of gross neuropathology suggestive of vascular, traumatic, inflammatory, neoplastic and neurodegenerative processes. A toxicology screening on blood and urine for ethanol was performed at each autopsy. The majority of investigated cases (17 suicide victims and 28 controls, see Supplementary Table) revealed the blood alcohol concentration (BAC) below the limit of quantification (LOQ), i.e. <0.2 g/l according to internationally accepted analytical guidelines. Other substances of abuse, antidepressant and antipsychotic drugs, as well as their metabolites were investigated when an intoxication was suggested by the scene inspection and/or other available information sources prior to the autopsy, i.e. in three cases. These cases together with a helium inhalation victim constituted the non-violent suicide subgroup (see Supplementary Table).

Brainstems were isolated from the brains and fixed in toto in 10 % phosphate-buffered formaldehyde for 1 week. After being fixated, tissue blocks containing the entire DRN were isolated from the brainstems and embedded in paraffin. Subsequently, serial 5-µm-thick transverse sections were cut along the entire rostrocaudal axis of the DRN. Every 200th section was mounted and stained for AgNOR.

Silver staining was carried out as previously described [24]. Briefly, 5-µm paraffin sections were dewaxed and rehydrated through graded alcohols. The silver staining was freshly prepared by dissolving 2 g/dl gelatin in 1 ml/dl aqueous silver nitrate solution at a 1:2 ratio. The sections were incubated with this solution in a dark moist chamber at room temperature for 45 min and subsequently washed with deionised water. Following this protocol, the AgNOR area—containing AgNORs (that are clustered, undistinguishable from each other) and representing the nucleolus—appears as an intranuclear, clearly delineated black or dark brown small spot, and the nuclear border is clearly visible in the majority of large DRN neurons (Fig. 1), which are most probably serotonergic output neurons [48]. Glia cells were distinguished from neurons according to the established cytomorphological criteria [49].

In each of the DRN subnuclei (ventral, ventrolateral, dorsal, interfascicular and caudal [48]), AgNOR parameters were determined in 40 neurons with clearly visible borders of the nucleus and AgNOR areas (Fig. 1), which were selected throughout the available AgNOR-stained sections, i.e. in seven sections for a case, on average. Thus, the AgNOR parameters of 200 DRN neurons were investigated in each case. The number of investigated neurons was established arbitrarily in accordance with the guidelines on quantitative evaluation and diagnostic and research studies employing the AgNOR method. This method does not require an estimation of the number of cells and/or nuclei [24]. The neurons were sampled using a 400× magnification. The AgNOR areas (composed of clustered AgNORs and representing the nucleoli), their number and the nuclear area within a single-sampled neuron were determined using a light microscope attached to a computer image analysis system (cellSens^®, Olympus, Japan). In this system, each of the neurons sampled by 400× magnification was visualised digitally and focused, and the sharpest and longest profiles of the nucleus and AgNOR areas were traced by the mouse cursor on the screen. As a result, the numerical values of AgNOR and the nuclear areas and the numbers of AgNOR areas were calculated automatically. Subsequently, the AgNOR ratio (relative AgNOR area) was derived by dividing the total AgNOR area by the nuclear area, taking into account all the AgNOR areas present per neuronal nucleus. This procedure was performed separately for each of the sampled neurons. The sampled measures were averaged to derive a single set of values for each DRN subnucleus, for the rostral and caudal subdivisions of the DRN and for the entire DRN as a single anatomical structure in each of the investigated cases.

Statistical analyses were performed with the data analysis software system STATISTICA version 10 (StatSoft^®, Inc. 2011, www.​statsoft.​com). As normal distribution was not given for all analysed AgNOR parameters, nonparametric statistical procedures were used. Unadjusted two-way comparisons with the Mann–Whitney U test were carried out to detect between-group differences.

The χ ² test and the U test were used to detect the possible differences between the study groups with respect to sex, age, season of the year (month of death in spring/summer vs. autumn/winter), brain weight, post-mortem delay and BAC (whose values below LOQ were accounted null values in statistical analysis). All statistical tests were two tailed.

STATISTICA Automated Neuronal Networks (SANN) module containing receiver operating characteristic (ROC) procedure was applied for the evaluation of potential diagnostic value of the method, i.e. its accuracy (represented by the area under ROC curve), sensitivity and specificity.

After AgNOR staining of the DRN neurons, borders of the AgNOR area (containing AgNORs that were clustered together and indistinguishable from each other) were clearly visible (Fig. 1) in line with the staining patterns in DRN neurons presented previously [24]. Most of the neurons contained one AgNOR area (representing the nucleolus in this staining method). Two or more AgNOR areas were observed very rarely, which explains why the AgNOR number was near 1.

The differences in AgNOR parameters were beyond qualitative evaluation, and they could only be captured by means of quantitative measurements.

The statistical analysis revealed highly significant differences by means of the cumulative analysis of all DRN subnuclei. Both the AgNOR area and the relative AgNOR area were decreased in suicide victims compared to controls (Table 1). No significant differences existed in any of the AgNOR parameters between violent (n = 23) and non-violent (n = 4) suicide victims. As revealed by the ROC curve for the AgNOR area, the method accuracy (represented by the area under ROC curve, AUC) is 83 %; its sensitivity according to the suicide diagnosis is 67 %, whereas the specificity of suicide exclusion (i.e. the diagnosis of non-suicidal case) is 90 %. The respective values for the relative AgNOR area are: 85 % (AUC), 78 % (sensitivity) and 90 % (specificity), and for both parameters when tested simultaneously they are: 89 % (AUC), 85 % (sensitivity) and 87 % (specificity).

Variables which could influence the AgNOR parameters in the DRN neurons, such as post-mortem interval (PMI), sex, age at death, season of the year (month of death in spring/summer vs. autumn/winter) and brain weight were not associated with the values of the AgNOR parameters in the compared groups. No significant differences in these variables between suicides and controls were revealed. As the PMI is a very important confounder in post-mortem research, the hypothetical influence of this factor was excluded by an additional analysis of data. The cases with longest PMI (i.e. ≥70 h, 6 cases in suicidal and 3 in control group, see Supplementary Table) were separated, and AgNOR parameters were compared between them and remaining cases in respective groups, which revealed non-significant U test P values. Moreover, intergroup comparisons were performed after exclusion of these cases, which supported the highly significant differences between suicidal and control groups in both the AgNOR area (U test P value 0.000046) and the relative AgNOR area (U test P value 0.000007; for a comparison see Table 2).

Although males unequivocally prevailed in suicidal versus control group (χ ² test P value 0.0002, Table 2), no intragroup differences related to sex were found in any of AgNOR parameters, among them the AgNOR area and the relative AgNOR area (non-significant U test P values).

The present study has certain limitations that have to be considered: (1) A relatively small number of cases have been analysed, especially for the evaluation of potential diagnostic value of the method; therefore, results have to be confirmed in a larger sample. (2) The psychiatric diagnoses (also according to substance use disorders) and the data on possible psychotropic medication preceding death were not available. The levels of psychotropic drugs were established only in three suicide victims where medication overdose constituted a cause of death. However, our current study did not aim at the relation between suicide and other mental disorders, and our previous studies did not suggest that the decreased AgNOR area in DRN neurons may be related to the medication in the last three months of life [24]. (3) The application of paraffin-embedded tissue is a limitation of our method compared to frozen brain samples, which would allow the application of a wider set of approaches. (4) Although serotonergic neurons comprise about 70 % of DRN neurons [7] and nearly all the large cells in this nucleus are serotonergic [5, 48], we could not determine the type of selected neurons by AgNOR staining. Since AgNOR areas correspond almost 100 % to the nucleoli in DRN neurons, immunohistochemistry for tryptophan hydroxylase with a Nissl counterstain could bring to light, whether the cells with reduced AgNOR areas are really serotonergic. Furthermore, confocal immunofluorescence microscopy using antibodies against nucleolin and tryptophan hydroxylase could be another approach to resolve this issue.