An independent committee of experienced clinicians made consensus DSM-IV diagnoses for all subjects using information obtained from clinical records and structured interviews with surviving relatives. The same approach was used to exclude psychiatric illness in the unaffected comparison subjects. All procedures were approved by the University of Pittsburgh’s Institutional Review Board for Biomedical Research and Committee for Oversight of Research and Clinical Trials Involving the Dead.
Tissue processing and immunohistochemistry
For fluorescence immunohistochemistry, tissue sections were incubated for 72 hours at 4oC in a combination of a rabbit PV antibody (Swant PV-28; 1:1,000 dilution) and a mouse aggrecan antibody (Acris SM1353P; 1:50 dilution), previously shown to label PNNs in human postmortem tissue and to recognize an aggrecan epitope that is not glycosylated . In immunohistochemical run 2, guinea pig anti-NeuN (Millipore ABN90; 1:1,000 dilution) was included to enhance the identification of pyramidal cells based on their characteristic morphology. The specificity of the PV antibody has been recently verified by our laboratory and the NeuN antibody has been extensively characterized by others (for initial characterization see ). Following washes to remove the primary antibodies, non-specific lectin binding sites were blocked using CarboFree blocking buffer (Vector SP-5040) and sections were then incubated in biotinylated-WFA (Sigma L1516; 1:400 dilution) for 1 hour at room temperature. Labeling with WFA has been used by multiple laboratories to label PNNs in human brain . Sections were then incubated in the following secondary antibodies for 18 hours at 4oC to detect the primary antibodies and biotinylated-WFA: donkey anti-rabbit Alexa 568 to detect anti-PV, donkey streptavidin Alexa 647 to detect WFA, donkey anti-mouse Alexa 488 to detect anti-aggrecan and donkey anti-guinea pig CF405 to detect NeuN. Following the secondary antibodies, sections were then incubated for 30 min in NeuroTrace 435 (Invitrogen) to label Nissl substance, mounted on gelatin-coated slides, coverslipped with Vectashield HardSet mounting medium (Vector Laboratories), and coded as to subject number and diagnosis.
The mean (+SD) dose of haloperidol decanoate (16.0±2.1 mg/kg) produced mean trough serum level of 4.3±1.1 ng/mL, which is known to be in the therapeutic range for the treatment of schizophrenia . After an overdose of pentobarbital, the brains were removed, subjected to a 45 minute postmortem interval and then processed in a fashion identical to the human brain tissue. Sites from both the dorsal and ventral banks of the principal sulcus were obtained during imaging. All procedures were approved by the University of Pittsburgh’s Institutional Animal Care and Use Committee.
Stereological methods and image capture
Sections from both members of a subject pair were imaged consecutively on an Olympus BX51 upright microscope equipped with an Olympus DSU spinning disk confocal, a Hamamatsu C4742-98 CCD camera (Hamamatsu, Bridgewater, NJ), Olympus mercury light source, excitation/emission filter wheels, a 89000 Sedat Quad ET filter set (Chroma Technology Corp, Bellows Falls, VT), and high precision Prior Scientific motorized XY stage (Prior Scientific, Inc., Rockland, MD) equipped with a linear XYZ encoder (Ludl Electronic Products, Ltd., Hawthorne, NY). Image collection was controlled using SlideBook version 5.0 software (Intelligent Imaging Innovations, Denver, CO). All imaging and processing were performed using SlideBook 5.0.
Because the highest densities of PV neurons and PNNs are found in layer 3 of the DLPFC, labeled structures were assessed in randomly-selected sampling sites as follows: In locations of the tissue section cut perpendicular to the pial surface, layer 3 was identified using Nissl/NeuN labeling and a sampling traverse ~4 mm wide was imaged under epifluorescence conditions using a 2x wide field objective. For low magnification (20x objective, 0.75 NA) imaging, the SlideBook stereology package was used to subdivide the traverse into 150 µm X 150 µm sampling sites spaced apart by 161 µm. Adjacent sites were not imaged to avoid any photo-bleaching between sites. Twelve sites were imaged by first determining the top and bottom positions of the tissue section at a given site and then collecting an entire Z-stack using a step size of 0.75 µm. The mean (±SD) number of PV cells sampled was 119.9±16.8 per healthy subject and 108.8±21.3 per schizophrenia subject.
For high magnification (super-corrected 60x oil objective, 1.4 NA) imaging, a similar sampling scheme was applied except that the sampling sites were 100 µm X 100 µm with 250 µm spacing, and the first 18-25 sites containing PV-positive cells were imaged using a step size of 0.5 µm. The mean (±SD) number of PV cells sampled was 42.1±8.5 per healthy subject and 42.3±9.7 per schizophrenia subject. The average number of cells sampled was based upon a power analysis using the low magnification findings for the density of PV cells and proportion of PV cells with PNNs.
Image stacks collected using the 20x objective were subjected to NoNeighbors deconvolution (for low magnification cell counting analyses of PV neurons). Counts of PV-positive cells were made from the stacks using a 290 µm X 290 µm counting frame that included the entire z-stack except for the top and bottom 3 images. All PV cells within the counting frame were initially counted with the rater blinded to whether a putative PNN was visible. All PV cells were then scored for the presence of a PNN under identical conditions.
For quantitative fluorescence analysis of individual PV neurons and their associated PNNs, image stacks (following exposure correction) taken using the 60x objective were subjected to 20 iterations of the AutoQuant blind deconvolution algorithm. Following deconvolution the intensity histogram settings for each fluorescence channel were made equivalent across subjects (for a given channel) and each PV cell subjectively scored as PNN positive or negative (using aggrecan labeling as the PNN marker). An object mask was manually drawn in a single z-plane to cover the entire soma at the largest cross-sectional area of all PV cells, this mask was dilated by 10 pixels and the original soma mask subtracted, resulting in a ring around each soma encompassing the putative PNN. To determine individual mask volumes and mean intensity values for each fluorescent channel SlideBook 5.0 was used.
In the paired ANCOVA model, the measured value was the dependent variable, diagnostic group the main effect, subject pair a blocking factor and storage time and immunohistochemical run covariates. Subject pairing is an attempt balance groups for sex, age and PMI and to account for the parallel processing of the tissue, and thus does not represent a true statistical paired design. Consequently, the unpaired ANCOVA model included the measured value as the dependent variable, diagnostic group as the main effect, and PMI, age, sex, storage time and immunohistochemical run as covariates. Degrees of freedom related to the ANCOVA statistics vary for the following reasons: a) only significant covariates (p<0.05) were included in the reported model, b) outliers, as determined using Tukey’s hinges test (using ±3x the inter-quartile range) , were excluded or, c) some measures were not available for all subjects. All reported values are from the paired ANCOVA since both models yielded similar results (except where noted in the text). The results from all ANCOVA models are presented in Supplemental Table 2.
Since a number of comorbid factors could potentially affect the levels of the dependent variables, each factor (sex, death by suicide, diagnosis of schizoaffective disorder, tobacco use at time of death, and use of antidepressants or benzodiazepines and/or sodium valproate at time of death) was assessed in schizophrenia subjects. ANCOVA models included the comorbid factor of interest as the main effect and age, sex, storage time, immunohistochemical run, and PMI as covariates. Only significant covariates (p<0.05) were included in the reported ANCOVA analyses and results from all analyses are provided in Supplemental Table 3. Analyses of the antipsychotic-treated monkeys were performed using two-tailed unpaired Student’s t-Tests.
Hilbig H, Bidmon HJ, Blohm U, Zilles K (2001). Wisteria floribunda agglutinin labeling patterns in the human cortex: a tool for revealing areal borders and subdivisions in parallel with immunocytochemistry. Anat Embryol (Berl)203(1): 45-52.
Mauney SA, Athanas KM, Pantazopoulos H, Shaskan N, Passeri E, Berretta S, et al (2013). Developmental pattern of perineuronal nets in the human prefrontal cortex and their deficit in schizophrenia. Biol Psychiatry74(6): 427-435.
Mullen RJ, Buck CR, Smith AM (1992). NeuN, a neuronal specific nuclear protein in vertebrates. Development116(1): 201-211.
Pantazopoulos H, Woo TU, Lim MP, Lange N, Berretta S (2010). Extracellular matrix-glial abnormalities in the amygdala and entorhinal cortex of subjects diagnosed with schizophrenia. Arch Gen Psychiatry67(2): 155-166.
Pierri JN, Chaudry AS, Woo TU, Lewis DA (1999). Alterations in chandelier neuron axon terminals in the prefrontal cortex of schizophrenic subjects. Am J Psychiatry156(11): 1709-1719.
Rocco BR, Sweet RA, Lewis DA, Fish KN (2015). GABA-synthesizing enzymes in calbindin and calretinin neurons in monkey prefrontal cortex. Cereb Cortex. e-pub ahead of print 30 March 2015. doi: 10.1093/cercor/bhv051.
Tukey JW (1977). Exploratory data analysis Addison-Wesley Pub. Co.: Reading, Mass., xvi, 688 p.pp.
Virgintino D, Perissinotto D, Girolamo F, Mucignat MT, Montanini L, Errede M, et al (2009). Differential distribution of aggrecan isoforms in perineuronal nets of the human cerebral cortex. J Cell Mol Med13(9B): 3151-3173.