Introduction

Obsessive–compulsive disorder (OCD) is a chronic disorder, characterized by recurrent intrusive thoughts and compulsive time-consuming behaviors. Worldwide, OCD prevalence is 2.5% in the general population and 1.5% in noninstitutionalized older adults (Teachman 2007). The estimated direct and indirect costs of OCD are around 8.4 billion dollars per year in the US (DuPont et al. 1995). OCD will be the sixth most incapacitating disease worldwide by 2020 (Murray and AD 1996). Despite its undeniable burden, the pathophysiological basis of OCD is yet to be clarified.

Both structural and functional neuroimaging studies have indicated disturbances in cortico-striato-thalamo-cortical circuitry (CSTC) as a possible mechanism underlying OCD symptoms (Rauch et al. 1994a; Subira et al. 2016). Among the structures involved in this circuitry, abnormalities in the orbitofrontal cortex (OFC) have been one of the most consistent findings in neuroimaging studies of OCD patients (Hoexter et al. 2015). However, current evidence is controversial. Structural imaging studies comparing OCD patients and controls reported greater OFC gray matter volume (Christian et al. 2008; Kim et al. 2001; Valente et al. 2005), while others showed a reduction (Hoexter et al. 2012; Pujol et al. 2004; Rotge et al. 2010; Szeszko et al. 1999; Togao et al. 2010; Yoo et al. 2008). Functional imaging studies have shown similarly conflicting results, since both hyperactivation (Gillan et al. 2015; Harrison et al. 2012; Maltby et al. 2005) and hypoactivation (Chamberlain et al. 2008) of the OFC areas were reported. Several factors could have contributed to these uncertainties, such as different imaging techniques and heterogeneity in OCD clinical presentation.

The human OFC is subdivided into areas or regions with distinct cytoarchitecture and hodology (Beck 1949; Brodmann 1909; Economo and Koskinas 1925; Hof et al. 1995; Mackey and Petrides 2010; Sanides 1962; Uylings et al. 2010; Vogt and Vogt 1919; Ongur et al. 2003). In addition, the main regions of the OFC, particularly the medial orbitofrontal (MO), anterolateral (AL) and anteromedial (AM), may have distinct roles (Chamberlain et al. 2008; Milad and Rauch 2007, 2012). For this reason, we focused onto these three important regions in our study.

Neuropathological examination of post-mortem brains is the gold standard method to assess the clinicopathological basis of diseases, including dementia, movement disorders and stroke. For example, in Alzheimer’s disease, specific protein accumulation, such as tau phosphorylation and amyloid beta, were detected and are used as diagnostic evidence (Braak and Braak 1991). In the case of OCD, to date, no specific neuropathological lesion has been recognized in post-mortem brain examinations.

Stereology is the method of choice for assessing neuronal number and volume. It requires the availability of the whole regions of interest, as well as the unified random selection of fields under stereologic analysis (Schmitz and Hof 2005). These requirements pose limitations because entire brains of well-characterized OCD patients are rarely available. For this reason, very few neuropathological or morphometric post-mortem studies of psychiatric disorders have been performed, and none of them has been conducted on OCD (Beasley et al. 2005; Cotter et al. 2005; Law and Harrison 2003; Pakkenberg 1990; Rajkowska et al. 2001; Stockmeier et al. 2004; Todtenkopf et al. 2005). Therefore, we applied unbiased stereological tools to estimate post-mortem neuronal density and volume, within OFC areas to achieve a clearer understanding of the neurobiological basis of OCD.

Materials and methods

Ethics statements

The brains studied herein were part of the psychiatric disorders collection of the Brain Bank of the Brazilian Aging Brain Study Group (Psy-BBBABSG). They were sourced from the Autopsy Service of the University of São Paulo between 2009 and 2011 (de Oliveira et al. 2012). All protocols concerning the acquisition of data and tissue processing were approved by the Research Ethics Committees of the University of São Paulo Medical School (Process number: 740/09), and by the Brazilian National Commission of Research Ethics (CONEP, Process number: 540/09) and are in accordance of The Code of Ethics of the World Medical Association (Declaration of Helsinki). The informed consent was obtained from all individual participants included in this research.

Participants

We analyzed 14 hemispheres (7 with OCD and 7 controls) from subjects who were ≥ 50 years old and without signs of clinical dementia at the time of death. Control cases were selected to match to the OCD group, regarding post-mortem interval (PMI), cerebrospinal fluid pH (CSF pH), age, sex, and hemisphere analyzed. Further detailed information concerning our collection is discussed elsewhere (de Oliveira et al. 2012).

Clinical assessment

After signing a term of free informed consent, the deceaseds’ next of kin (NOK) answered a comprehensive interview on clinical, functional, cognitive and psychiatric information (Ferretti et al. 2010). In participants with a presumable diagnosis of OCD during the screening interview, we invited the NOK to return for a second interview, with an experienced psychiatrist for confirmation of OCD diagnosis. This complete evaluation comprised a Structured Clinical Interview for DSM IV Axis I disorder (SCID) (Del-Ben et al. 1996; First et al. 2002; Spitzer et al. 1992), the Yale–Brow Obsessive–Compulsive Scale (Y-BOCS) (Goodman et al. 1989), and a short version of DY-BOCS (Rosario-Campos et al. 2006). We considered a cutoff ≥ 1 for the Neuropsychiatry Inventory score.

It is relevant to note that none of the subjects diagnosed with OCD post-mortem has received a diagnosis of OCD during their lives.

Histological processing

After removal from the skull, the brains were subdivided by a medio-sagittal cut. One hemisphere was dissected fresh into 46 blocks and frozen for future studies. The contralateral hemisphere was perfusion-fixed with formalin at 20% and post fixed for at least 3 months. After fixation, the brain tissue was dehydrated and embedded in celloidin. The blocks were cut along the coronal axis using a sliding microtome (Polycut, Cambridge Instruments, UK), producing serial sections of 440 µm thickness.

Photograph image acquisition

A monopod with a digital camera (Canon EOS 5D Mark II 21.1 Megapixel®, Tokyo, Japan) was firmly attached to the microtome knife holder, allowing each section to be photographed while the block was cut. These images were used for computer-assisted three-dimensional (3D) reconstruction of all frontal lobes, using Amira software (FEI, ThermoFisher Scientific) 3D reconstruction was mandatory for architectonic evaluation of the regions of interest.

Odd-numbered sections were stored in 70% alcohol and representative planes of sections were used to perform the neuropathological examination for Alzheimer’s disease, according to The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) (Mirra et al. 1991) criteria and Braak and Braak staging (Braak and Braak 1991). All even-numbered serial sections were stained with gallocyanin (Einarson 1932) and photographed at low magnification, using a digital single-lens reflex camera (Canon EOS 5D Mark II 21.1 Megapixel®, Tokyo, Japan) mounted with a 1:2:8/90 mm macro lens (Vivitar®, Tokio, Japan). Moreover, a wide-field stereomicroscope (Olympus®, Tokyo, Japan) with zoom lenses ranging in magnification from 5 to 120 × was used to delineate the cytoarchitectonic fields on the Nissl-stained sections and to draw the boundaries of the orbitofrontal areas AM, MO and AL (Heinsen et al. 1999).

Cytoarchitectonic delineation

The criteria for cytoarchitectonic delineation included: absolute cortical thickness and relative thickness of the cortical layers I through VI in corresponding locations in the gyri, relative thickness of the combined layers I through IV versus layers V + VI, size, staining properties, and cell densities of pyramidal cells in layers IIIc and V, subdivision of layer V into Va and Vb ; presence, thickness, and granule cell density of layer IV, and thickness, and sharpness of cortical to subcortical border of layer VI. For parcellation of the orbitofrontal cortex, we were using the terminology of Hof et al. (1995).

The anterior borders of AM, MO, and AL are formed by the frontopolar region, characterized by a high thickness of layer IV and high cell number of granule cells (Fig. 1a, b). The caudal border of AL was set at the transition of the granular orbitofrontal area AL to the dys- or agranular part of the lateral orbital gyrus, the continuation of the inferior frontal gyrus (Fig 1a, d, e, F3o) (Eberstaller 1890; Economo and Koskinas 1925). Sagittal sections represent the standard plane of section for anterior–posterior delineation of rostro-caudally arranged cytoarchitectonic fields. However, neither the anterior nor the caudal borders were sharply delimited at low-power magnification (Fig. 1a). Furthermore, we have excluded region PM from our study, because the delineation of granular- and dysgranular areas in coronal sections was subject to considerable interindividual variability between researchers involved in delineation. The same was true for the frontopolar delineation. There was an area of transition (Fig. 1a) between the frontopolar regions and the ensuing caudal orbitofrontal fields and particularly the increasing number of layer IV granule cells made the delineation of MO and AL uncertain in coronal sections. According to Hof et al. (1995) AM covers the depth of the olfactory sulcus and also the ventral part of the rectus gyrus. We had set the medial border of AM into the olfactory sulcus, because in 5 out of our 14 cases the rectus gyrus was artifactually lesioned and a cytoarchitectonic delineation of the complete AM was not possible. The medio-lateral delineation of fields AM, MO, and AL was less ambiguous because the borders were always in the same plane of our coronal sections.

Fig. 1
figure 1

Sagittal section through the right orbitofrontal cortex of a 68-year-old female control case. The plane of section is indicated by the red line in f. The low power overview clearly demonstrates postero-anterior changes in laminar width and staining properties of layers III through VI. Layer V is prominent in the posterior orbital part of the superior frontal gyrus (F1) and sublayer Va is characterized by is strong staining (chromophilia) with gallocyanin and Vb by its pallor due to a relatively low overall cell density. This chromophilia is decreasing in the anterior (rostral) direction and the superior and inferior borders of layer Va become blurred whereas the cell density in increasing in Vb to reach a maximum in the frontopolar region. The granule cells of layer IV are small, round, and less-well stained with gallocyanin, layer IV is interspersed between layers III and V. It is missing in posterior (agranular) fields of F1, barely recognizable in the dysgranular fields, and it becomes more prominent and reaches a maximum in width and cell density in the frontopolar region or area 10 of Brodmann. Black arrow marks an emergent layer IV, black bar in lower-right, 100 µm, same magnification in all microscopic pictures (b–e); white arrows and letters indicate the position of figures b through e at higher microscopic magnification. f Computer-assisted reconstruction of the orbitofrontal regions AM (anteromedial), MO (medial orbitofrontal), and AL (anterolateral) according to Hof et al. (1995). F1o, F2o, F3o continuation of the superior (F1), medial (F2), and inferior frontal gyri (F3) into the orbital region based on the terminology of Economo and Koskinas (1925). The lateral parts of F1o correspond to the medial orbital gyrus, the rostral parts of F2o to the anterior, the posterior parts of F2o to the posterior orbital gyrus, and F3o to the lateral orbital gyrus of Chiavaras and Petrides (2000) respectively

Full size image

Stereological estimations

Neuronal density is not homogeneous but subject to gradual caudo-rostral increase in layers II through VI (Fig. 1a). After 3D-reconstruction of the cytoarchitectonic fields AM, MO, and AL, we selected coronal planes of section midway between the presumptive rostral and caudal borders of the fields. We placed a transparent folia printed with a 1 × 1 mm grid over the serial sections and selected those crossing points hitting representative and undistorted sections through the dome (crown), wall, fundus or sulcal floor (Triarhou 2013) of AM, MO, and AL. To eliminate bias caused by tangential sections, only crossing points hitting perpendicular sections through complete gyral profiles were selected. Neuron densities were determined along a virtual perpendicular line through the crossing point within layers IIIa, IIIb, IIIc, IV, Va, Vb, VIa, and VIb in the dome, wall, and floor regions, summarized and divided by three (mean of each layer of the three representative regions of a gyrus).

We estimated neuron density using the unbiased optical disector. Each cortical layer of the area was counted separately. In each cortical layer, the disector height was set to 30 µm and the guard zone to 15 µm (Heinsen et al. 1999). We took serial images of optical stacks comprising 11 pictures, focusing within the z-axis of the microscope by turning the fine adjustment knob (within a range of 30 µm). The microscope’s lens was a PlanApo40/1.0 oil and the photographic ocular a FK 5 ×. The stack of virtual images was imported to the cell counter plugin of the Image J 1.45I software (National Institutes of Mental Health, EUA). The final magnification on the computer screen was 1500-fold. Scrolling through the optical stack is like virtual focusing through gallocyanin stained neurons. All profiles of neuronal nucleoli that came anew onto the computer screen were counted. Nucleoli in granule cell nuclei could not always be identified or could be confused with larger clumps of heterochromatin. In this case, profiles of granule cell nuclei coming anew into the field of focus were counted. We have never observed neurons with more than one nucleolus per cell. The area of the counting frame was 80 × 80 µm or 0.0064 mm2. We set the left and the lower margin as forbidden lines. The neuronal density (N) was calculated according to the formula (1):

$$Nv=N/A \times t,$$
(1)

where N is the number of nucleoli counted, A is the area of the frame (0.0064 mm2) and t is the optical thickness (30 µm or 0.03 mm). For volume estimation, digital photos of all slices of the OFC were imported into Image J. A virtual grid with d = 0.7 mm was randomly overlaid on each slice. All crossing points that hit the region of interest were counted. Then volume (V) was calculated using the Cavalieri principle (West 1999), according to the formula (2):

$$V=k \times {d^2} \times t \times {P_T},$$
(2)

where k was the distance between sections (2, since we had stained every second section), d2 was the area of the grid associated with each crossing point (P) (Weibel 1979) of the counting grid (0.7 × 0.7 = 0,49 mm2) PT is the sum of the crossing points counted for all sections, and t was the mean thickness of the hemispheric sections (≈ 0.4 mm).

The precision and effectiveness of counting is a matter of debate (Schmitz 1997, 1998, 2000; Schmitz and Hof 2000, 2005; Schmitz et al. 1999a, b). In analogy with Schmitz (1998) recommendations we have counted between 671 and 2239 neurons per case to reduce the methodological error associated with sampling and counting procedures to ≈ 5%.

Statistical analysis

We described the variables of interest using means and standard deviations for continuous variables, and absolute and relative frequencies for categorical variables. We used Fisher’s exact test to compare categorical variables between control and OCD groups, and the Mann–Whitney test to compare continuous variables. For all tests, the significance level was set at 0.05. We used Stata 13.0 software (College Station, TX: StataCorp LP) for statistical analysis.

Results

Participants

Seven subjects who had OCD and seven matched controls were selected. The inspection of each brain showed no significant difference between OCD and control subjects regarding macroscopic neuropathological lesions, such as infarcts. However, we observed of tau and a amyloid-beta pathology in two candidates (one case and one control), based on CERAD criteria (Table S1).

OCD clinical presentation of cases, as reported by the NOK, is presented in Table 1. As expected, the groups were similar regarding variables used to paired participants (i.e. age, sex, hemisphere analyzed, PMI, and cerebrospinal fluid pH) (Table S2). Groups were also similar for other important variables, such as education level, smoking and alcohol use. The Neuropsychiatry Inventory (NPI) score and SCID were negative in the control group.

Table 1 Description of obsessive–compulsive symptoms of the OCD patients
Full size table

Qualitative differences between control and OCD cases

In general, perfusion fixation of hemispheres from controls and cases with short post-mortem time and with controlled CSF-pH resulted in excellent preservation of neuronal size, shape, and Nissl substance. Preservation of subcortical gray matter was not always optimal and most likely the consequence of intravascular blood clotting and limited flow of formalin. With low power magnification, the caudo-rostral and medio-lateral gradation in layer width, number and intensity of staining did not show pathognomonic differences between controls and the OCD cases (Figs. 1a, 2a). Applying the criteria described in “Material and methods”.

Fig. 2
figure 2

a Coronal section through the right orbitofrontal cortex of a 71-year-old male control case. From the medial (left) to lateral (right) margin of the orbitofrontal cortex a similar layer-wise gradation concerning chromophilia, increase in neuron density, and changes in absolute and relative width of layers III through VI as in Fig. 1a can be observed. Latero-medial increase in depth and increase in granule cell number parallels an increase in width of layer III and particular increase in the number and size of deep IIIc pyramidal cells are main characteristics of this medio-lateral gradation, whereas the size and staining intensity of layer V pyramidal cell decreases. Black arrows indicate the position of figures b through d at higher microscopic magnification. b anteromedial region; c medial orbitofrontal region; d anterolateral region. Black bar lower right of c 100 µm. Stippled lines from left to right cytoarchitectonic borders between AM/MO and MO/AL

Full size image

AM, MO, and AL could be unequivocally delineated from each other medio-laterally. However, the anterior borders with the frontopolar field and the posterior borders with dysgranular fields were less sharp and barely visible in coronal sections. These central parts of AM, MO, and Al comprised a minimal number of 10 and a maximal number of 17 serially cut slices with uncertain rostral and caudal boundaries.

With higher magnification, we became aware of local or extended lower cell densities in cases with OCD (Figs. 2b–d, 3b–d). These figures should be viewed from a considerable distance to blur neuronal details. Irregularly shaped paler regions will become visible in different layers after this maneuver. Uniform laminar differences in neuron density were more difficult to diagnose macroscopically (Figs. 2a, 3a). Laminar or focal pallor in the cortical gray matter can be the result of neuron loss that can be quantified and subjected to statistical analyses.

Fig. 3
figure 3

a Coronal section through the left orbitofrontal cortex of a 66-year-old female OCD case. The overall cell density in all three regions (b anteromedial; c medial orbitofrontal; d anterolateral) is lower than in the control case, the presence of hyperchromic and large IIIc pyramidal cells and the columnar (vertical) arrangement of neurons is striking. The latter phenomenon is masked in Fig. 2 by higher overall cell density. Black arrows indicate the position of figures b through d at higher microscopic magnification. Black bar lower right of c, 100 µm. Stippled lines from left to right cytoarchitectonic borders between AM/MO and MO/AL

Full size image

Neuron density and areal cortical volume in OCD and controls

Our quantitative estimation of neuron densities confirmed the qualitative impression of reduced total neuronal densities (Table 2) ranging from − 25% in AM and AL and − 21% in MO. However, there was no consistent trend of a layer-specific decrease in all three OFC fields analyzed. Numerically, the infragranular layers were most heavily affected in all three fields, the supragranular layer IIIb particularly in AM and MO of the OCD cases. The mean volume of the combined areas was slightly but not significantly lower in our OCD cases (Table 3).

Table 2 Comparison of the neuron density (N/mm3) in the cortical layers between OCS and controls
Full size table
Table 3 Comparison of the volume, between OCD subjects and controls
Full size table

Discussion

To our knowledge, this is the first cytoarchitectonic study that includes estimation of neuronal density and volume of defined OFC-regions in post-mortem brains of subjects with OCD. Cytoarchitectonic delineation of the human orbitofrontal cortex was subject of numerous investigators and the results were not always uniform [for a recent summary and review see (Henssen et al. 2016)]. Nevertheless, we succeeded in subdividing the central orbitofrontal regions into three medio-lateral fields that correspond to subdivisions described previously by authors using combined Nissl and immunohistochemical stains (Ongur et al. 2003; Uylings et al. 2010) or sophisticated image analysis (Henssen et al. 2016). Our field AL did not extend into polar regions like the one depicted by Hof et al. (1995). Subtle architectonic differences between AL and FP after SMI-32 immunohistochemical staining could explain this discrepancy. We observed a puzzling non-uniform decrease of laminar neuron densities in OCD versus age- and sex-matched control cases. The greatest quantitative differences were more frequently encountered in the infragranular layers, whereas layer IIIb appeared particularly vulnerable in AM. Granule cell density in layer IV seemed to be unaffected.

Differences in neuron density can be artefactual and simply reflect less shrinkage in individual brains during fixation, embedding, staining, and coverslipping (Heinsen et al. 1994; Oorschot 1994). Multiplication of neuron density with the volume of a cytoarchitectonic field yields total neuron number, a dimension that is not biased by histological artifacts. However, we had exclusively estimated laminar neuron density in the sulcal depth, the wall and the crown in central regions of the orbitofrontal cortex. This procedure does not reflect the relative volumes of cortical layers and the high neuron density of the thin layers IV or Va will automatically bias calculations of mean cortical neuron densities. Consequently, multiplication of our mean cortical neuron densities with the volume of the cortical fields estimated by the Cavalieri principle will necessarily yield biased total areal neuron numbers with little biological significance. Our primary focus was on the analysis of laminar neuron densities. This is only possible in strictly perpendicular sections through the cortex, whereas unbiased total neuron number must also be estimated in grazing or tangential sections of a cortex as well, depending on the random location of a hitting point (see Fig. 1a where laminar diagnosis in several profiles of the cortex dorsal to the more or less perpendicularly cut ventral cortex is illusive). On the other hand, selective regional differences in laminar neuron densities provide a strong argument against artifact-induced swelling associated with decreased neuron density and the impression of neuron loss. Artifact-associated histological procedures were reported to affect cortex and white matter in a more comprehensive or global way (Kretschmann et al. 1982). Finally, enhanced or reduced neuron density does not necessarily imply higher total or reduced neuron number respectively. We have recently reported a circumscribed laminar neuronal loss and overall decreased neuron number in Huntington’s disease in the presence of higher average neuron density associated with global cerebral atrophy and a reduced neuron number with massive gliosis and unchanged average cortical volume in chorea-acanthocytosis, an extremely rare disease with clinical symptoms similar to Huntington’s disease (Liu et al. 2018).

Nevertheless, the average neuron density was by 21–25% lower in the OCD cases. Comparing b–d in Figs. 2 and 3 reveals a selective deficit or loss of smaller pyramidal neurons in OCD cases with preservation of bigger, particularly layer IIIc and chromophilic Vb neurons. Neuron loss by degeneration, neuronal deficits due to maldevelopment, or age-related proliferation with hypertrophic plasticity of the neuropil in the orbitofrontal cortex need to be verified as a pathogenetic mechanism behind diffuse and/or localized laminar pallor in Nissl-stained sections. Indeed, whereas we found a non-significant slight volume decrease in our sample (not corrected with individual shrinkage factors) (Heinsen et al. 1994; Oorschot 1994). de Wit et al (2014) describe complex local group-by-age interactions with a relative preservation of orbitofrontal cortical volumes comprising area 13 and 47 (parts of our MO) in aged OCD. Their data are based on an international voxel-based morphometric in-vivo 1.5 T T1-weighted MRI study on 412 OCD patients and 368 control cases. Given decreased neuron density reflects decreased neuron number, a relative preservation of orbitofrontal cortical volume would imply a proliferation of neurites and neuropil increase. In this respect, the neuropathology of the orbitofrontal cortex in OCD would be intermediate between pathognomonic laminar neurodegeneration in Huntington’s disease and probable neuropil hypertrophy with massive gliosis in chorea-acanthocytosis. Scanning of additional hemispheres from new cases with high-resolution MRI scanners (Grinberg et al. 2008) and calculation of post-mortem in cranio volumes of cortical and subcortical regions with stereological volume estimation of the same regions after histological processing and calculation of shrinkage factors and transformation of serial coronal sections into high-resolution virtual sagittal sections for better areal delineation (Alegro et al. 2016; Alho et al. 2018) could be a strategy to solve the question of artifactual or biological volume changes, neuron density, and neuron number differences in controls and age- and sex-matched OCD cases.

The orbitofrontal cortex integrates several sensory modalities including taste, olfaction, pain, auditory, and visual stimuli. These stimuli are associated with learning, memory, emotional, and hedonic aspects. Higher cognitive capacities that depend on these inputs including reward, moral, social judgement, and conduct are obviously arranged in a topographic manner from more medial to more lateral and posterior to anterior aspects (Henssen et al. 2016), The majority of the orbitofrontal areas are covered by a clear-cut six-layered isocortex with increasing width and cell-density of layer IV (Sanides 1962). Pyramidal cells represent about 80% the major cell type of the isocortex. Their layer-wise arrangement reflects connectional/functional aspects of association, commissural, and projection neurons to cortical and subcortical targets (Nieuwenhuys et al. 2008). Layer IV granule cells and layer III pyramidal cells are the main targets of intracortical feed-forward circuits (Pandya and Yeterian 1985). Heavy neuron deficits/loss in orbitofrontal infragranular layers V and VI could be associated with disconnections to accumbens, ventral striatum, pallidum, ventral tegmental area, and extended regions of the substantia nigra pars compacta (Haber et al. 1995). Combined loss of layer III and layer V pyramidal cells could disconnect caudally respectively rostrally directed intraorbital streams whereas intact granule neurons still receive unimpaired cortical somatosensory and visual afferents.

Presently, it is illusive to correlate our findings with resting metabolic state, symptom provocation paradigms, and pre-post treatment functional neuroimaging studies implicating a hyper-activated OFC pattern in the pathophysiology of OCD (Rauch et al. 1994b; Saxena and Rauch 2000). Findings from structural neuroimaging research have reported OFC volumetric alterations, including both increments and decrements, in OCD patients (Hoexter et al. 2012; Kim et al. 2001; Radua and Mataix-Cols 2009; Rotge et al. 2010; Togao et al. 2010; Valente et al. 2005). Our quantitative data concerning volume differences of OFC areas were subject to histological artifacts and, therefore, are equivocal concerning pathological volume increments of decrements in OCD cases. The majority of cortical pyramidal cells are glutamatergic (Meldrum 2000). Correlation of laminar and region-specific neuron deficits with elevated glutamate and glutamine (Glx/Cr) in the OFC white matter of OCD patients (van de Vondervoort et al. 2016), candidate gene association studies (Mattheisen et al. 2015; Stewart et al. 2013) and the differential expression of the glutamate transporter genes SLC1A1, SLITRK, DLGAP1 and GRIN2B in OCD remain to be established (van de Vondervoort et al. 2016).

The study has some limitations (1) the small sample size, and stereological sampling strategies most likely limited its statistical power to determine significant differences in the total number of neurons. (2) Given the case–control design of post-mortem investigations, it is not possible to determine whether microscopic alterations in the OFC represent a primary site of the disorder, the result of a compensatory mechanism, or an epiphenomenon (3). Our findings are exploratory and, as such, were not adjusted to multiple testing. (4) The subjects were not diagnosed with OCD while alive. To overcome this limitation, we did an extensive post-mortem neuropsychiatric evaluation with the individuals’ families to characterize their OCD symptoms. It is worth emphasizing that the post-mortem diagnostic ability of our initial interview was validated (Ferretti et al. 2010). Several of our cases (6) had difficulties in discarding things and a tendency to accumulate possessions; however, all these patients had also one additional obsessive–compulsive symptom. It is worth highlighting that more than 50% of OCD patients had comorbid hoarding symptoms (Torres et al. 2012). Therefore, obtaining well-characterized and diligently processed post-mortem brains from individuals with OCD is a challenge. Future studies should include a higher number and, if possible younger cases with probable OCD.