Reduced anterior insular cortex volume in male heroin addicts: a postmortem study

All brains were obtained from the Magdeburg Brain Bank. Sampling and preservation of the human brain material were done in accordance with the Declaration of Helsinki, German law and the local institutional review board at the University of Magdeburg. The analysis included 14 chronic male heroin addicts who died from drug overdose and 13 male controls (Table 1). Patients and controls were matched as closely as possible but differed in terms of age and duration of autolysis (although the latter did not reach significance; p = 0.097). Information on clinical characteristics was extracted from the clinical records and via structured interviews with people closely related to the subjects using a psychological autopsy [26]. In addition to heroin, all but one of the addicts had a history of abusing other legal and/or illegal substances, including morphine, cannabis, alcohol, cocaine, barbiturates, benzodiazepines and hallucinogens. However, the tested patients fulfilled the diagnostic criteria of an addiction only for heroin. An experienced neuropathologist (CM) ruled out qualitative neuropathological changes due to neurodegenerative disorders (such as Alzheimer’s disease, Parkinson’s disease, Pick’s disease), tumors, or inflammatory, vascular or traumatic processes, using samples with Nissl myelin staining, as well as HLA-DR-, beta-amyloid-, and tau-immunostainings. None of the heroin addicts were HIV-positive. A toxicology screen of blood and urine for ethanol and other substances of abuse was performed at each medico-legal autopsy and evaluated by forensic pathologists (KT and TG).

Tissue preparation was performed as previously described [27]. Brains were removed and fixed in toto in 8% phosphate-buffered formaldehyde for at least two months. Frontal and occipital poles were separated by coronal cuts anterior to the genu and posterior to the splenium of the corpus callosum. After embedding all parts of the brains in paraffin wax, serial, whole brain coronal sections of the middle block were cut using a large-scale microtome (Balzers, Liechtenstein) at a thickness of 20 μm and mounted. Volume shrinkage was determined for each brain before and after dehydration and embedding of the tissue using the formula: VSF = (A1/A2)^3/2 (VSF = volume shrinkage factor; A1 = cross-sectional area before processing of tissue; A2 = cross-sectional area after processing of tissue).

For anatomical orientation and morphometric investigations, every 50th serial coronal whole brain section was treated with a combined cell and fiber solution of Nissl (cresyl violet) and Heidenhain-Woelcke myelin stains and sampled [28, 29], resulting in an intersectional distance of 1 mm. All subsequent examinations were performed using an Olympus BX60 microscope with an associated DP22 camera and the CellSens image analysis program (OLYMPUS K.K., Tokyo, Japan), with the operators blinded to diagnosis. Overview images of the whole area of the insular complex were captured at 12.5 × magnification with a linear resolution 2.990 μm/px and an automatic exposure time. All available serial coronal brain sections (average of 31 sections per brain) with uni- or bilaterally recognizable insular cortex were evaluated. On the scanned overview images, delineation of the insular cortex was determined as described by Ding et al. [30] and as illustrated in Fig. 1. Subdivision of the insular cortex into the anterior and posterior regions was performed as described by Senatorov et al. based on the location of the central insular sulcus [10]. Using the CellSens function “freehand polygon”, the cross-sectional areas of the insula were outlined and measured by planimetry in both hemispheres. Insular cortex volumes were calculated by summing the measured areas multiplied by the distance to the next measured section, as previously described [31, 32]. These were multiplied by the VSF (related to the dehydration and embedding process) to estimate the insular volumes in fresh brain tissue.

Neuron counting was performed only in the anterior insula as the main region of interest. The images were captured at 200× magnification with a linear resolution 0.18343 μm/px and a constant exposure of 9.092 ms (Fig. 1). In the acquired images, both pyramidal cells and interneurons with clearly visible nucleoli were counted manually within the delineated cross-sectional area of the anterior insular cortex. We applied 30 counting boxes for each hemisphere per section, with three sections per case, resulting in 180 counting boxes per case. Each box corresponded to a size of 0.0016742884 mm³ = 352.2 μm × 264.1 μm × 18 μm edge length (20 µm section thickness minus guard zones of 2 × 1 µm). The observed coefficients of error (OCE) were calculated as described by Gundersen and Jensen [29]. The mean OCE values were 0.078. The Pearson interrater correlation showed good reliability with a value of r = 0.872. A counting grid was used to define a three-dimensional box within the thickness of the section as described previously [27], allowing 2 × 1 μm guard zones at the top and bottom of the section for the application of a direct, three-dimensional counting method.

Cortical layer-specific evaluation was performed by subdividing cortical gray matter regions into superficial and deep layers, as proposed by Katsel et al. [33]. As summarized in Table 2, five counting boxes were placed in layers III and IV each. Layers I and II as well as V and VI were combined due to difficulties in exact delineation (10 counting boxes per double layer). Care was taken that the counting boxes did not overlap. Application of the optical disector made it necessary to measure movements in the z-axis using a microcator as an integral part of the microscope. The determined numerical densities of neurons per counting box were averaged per layer and summed. Estimated total neuron numbers of the left and right anterior insular cortex were extrapolated by multiplying the averaged numerical densities of neurons from all cortical layers with the determined respective total anterior insular cortex volumes in mm³ divided by 0.0016742884 mm³ (i.e., size of one counting box).

Data analyses were performed using the statistical software package R (http://​www.​r-project.​org). Demographic data were compared using Student’s t tests. Volume data were normally distributed as indicated by Shapiro–Wilk-tests. As summarized in Table 1, total brain volume was larger in the heroin group compared to controls (1534 ± 65 versus 1390 ± 104 cm³; T = 4.287; p < 0.001). In determinations of diagnosis-dependent insular cortex volumes, we accounted for whole brain size as a potential confounding factor using relative insular volumes (corrected by VSF) normalized to whole brain volume (relative volume = volume of the respective insular structure divided by the whole brain volume). For analysis of relative insular volume data, a repeated measures analysis of variance (rmANOVA) was performed using “hemisphere” and “region” (anterior/posterior insula) as within-subject factors and “diagnosis” as a between-subject factor. As there was a “region*diagnosis” interaction, an rmANOVA was performed for each region. To account for multiple comparisons, false discovery rate (FDR)-corrections were performed for p values of post-hoc t tests for each rmANOVA [34]. Since the number of neurons was determined only in the anterior insula, an overall rmANOVA for the within-subject factor “region” could not be used, but all other steps were identical as in the relative volume analysis. Apart from the main effects of “diagnosis” or “hemisphere”, the interaction of these factors was determined by rmANOVA. Cohen’s d was used to assess effect size (ES), with d ≥ 0.2,  ≥ 0.5 and ≥ 0.8 considered as small, medium and large ES, respectively.

To identify potential confounding factors, rmANCOVA were performed with the testwise inclusion of “age” and “duration of autolysis” as covariates. In addition, Pearson correlation coefficients were used to calculate correlations of anterior insula volumes with respective neuronal cell densities, age with anterior/posterior insula volumes, and “duration of autolysis” with the shrinkage factor. Finally, Pearson correlation coefficients were calculated to assess the potential association of anterior insula volume reduction with previously published brain volumetric data from patients with heroin addiction [5‐8].

The relative insular volume showed an interaction between diagnosis and region (anterior/posterior insula) [F (1,25) = 8.993, p = 0.006**]. However, there was no cerebral lateralization (i.e., no effect of “hemisphere”) and the factors “diagnosis” and “hemisphere” did not interact significantly. Because of the significant diagnosis*region interaction, a separate analysis of the relative volume of the anterior and posterior insula was carried out. The anterior insula showed a smaller relative volume in heroin addicts versus controls (3010 ± 614 *10^–6 versus 3970 ± 1306 *10^–6; F(1,25) = 6.112, p = 0.021*, Cohen’s d = − 0.989; Table 1 and Fig. 2). Consistent with this finding, the proportion of the anterior to total insula was also found to be lower in heroin addicts versus controls (67.50 ± 5.47% versus 71.44 ± 2.89%; F(1,25) = 5.345, p = 0.029*, Cohen’s d = − 0.925). No significant diagnosis-related differences were observed regarding the relative volume of the posterior insula (Table 1 and Fig. 2).

Due to the above-mentioned volume reduction of the anterior insula, neuronal counting was performed to check whether this volume reduction might be associated with neuronal loss. However, relative volumes of the anterior insula showed no significant correlation with respective neuronal cell densities. Moreover, there was no main effect of “diagnosis” on estimated neuron numbers [179.07 ± 36.90 versus 200.3 ± 81.4; F(1,25) = 0.779, p = 0.386; Table 1].

Testwise rmANCOVAs employing potential confounding variables did not reveal a significant effect of age [anterior insula: F(1,24) = 0.654, p = 0.427; posterior insula: F(1,24) = 0.001, p = 0.972] or duration of autolysis [anterior insula: F(1,24) = 0.615, p = 0.441; posterior insula: F(1,24) = 0.040, p = 0.843] on diagnosis-dependent relative insular volume differences. In addition, the percentage of anterior to whole insula volume was not affected by age [F(1,24) = 0.419, p = 0.524) or autolysis [F(1,24) = 2.022, p = 0.168]. With neuronal cell count data available only for the anterior insula, testwise rmANCOVAs employing potential confounding variables did not reveal a significant influence of age [F(1,24) = 1.361, p = 0.255] or duration of autolysis [F(1,24) = 0.049, p = 0.827] on diagnosis-dependent neuronal cell count differences. In addition, no significant correlations were found between age and anterior insula volume (r = 0.280, p = 0.160) or between the duration of autolysis and shrinkage factor (r = − 0.140, p = 0.490).

Comparison of the current findings with those from our previous studies [5‐8] showed that volumes of the right anterior insula and right nucleus accumbens were correlated in heroin-addicted subjects, although this did not pass FDR correction. No further significant correlations were observed with previous volumetric findings of our workgroup on heroin addiction (Supplementary Table 1).