Proteomic analysis of post mortem brain tissue from autism patients: evidence for opposite changes in prefrontal cortex and cerebellum in synaptic connectivity-related proteins

Post mortem prefrontal cortex (PFC) (Brodmann area 10; autism n = 10, controls n = 10) and cerebellum (CB) (lateral posterior and anterior lobe, autism n = 16, controls n = 17) were provided by the National Institute of Child Health and Human Development (NICHD) Brain and Tissue Bank for Developmental Disorders (University of Maryland School of Medicine, Baltimore, MD, USA). Patients were matched to controls with respect to age of death (PFC: P = 0.99; CB: P = 0.80), gender (PFC: P = 1.00; CB: P = 0.70), post mortem interval (PMI) (PFC: P = 0.08; CB: P = 0.18) and brain pH (PFC: P = 0.51; CB: P = 0.66), which is indicated in the additional files (see Additional file 1 for PFC and Additional file 2 for CB). The analysis of prefrontal cortex included samples from 10 autism individuals (age = 21.3 ± 4.3 years, PMI = 22.1 ± 3.9 hours) and 10 matched controls (age = 21.4 ± 4.3 years, PMI = 14.7 ± 0.8 hours), and cerebellum tissue included 14 individuals with a diagnosis of autism (age = 18.6 ± 3.4 years, PMI = 22.7 ± 3.5 hours) and 17 matched controls (age =17.5 ± 2.9 years, PMI = 16.7 ± 1.6 hours) (Table 1). The study was approved by the Columbia University Medical School Institutional review board, consent was obtained from next of kin and all samples were de-identified and personal information anonymised. Diagnosis of autism was confirmed by the structured Autism Diagnostic Interview-Revised (ADI-R) carried out with the parents, the childhood autism rating scale (CARS), family history documentation, gene analysis to exclude genetic disorders and mutations, and neurological examination (Table 1). Subjects with a diagnosis of autism were included in the study. Local ethical approval for use of this tissue was granted by the Cambridgeshire Local Research Ethics Committee.

All biochemicals and reagents were obtained from Sigma-Aldrich (Poole, UK) unless specified otherwise. Brain tissues (approximately 30 mg) were sectioned using a Leica Cryostat (Milton Keynes, UK), collected into pre-chilled lysing matrix D tubes (MP Biomedicals; Cambridge, UK) and stored at -80°C until use. Protein extraction was performed by addition of fractionation buffer (7 M urea, 2 M thiourea, 4% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate, 2% ASB14 and 70 mM dithiotreitol (DTT), followed by sonication for 10 seconds using a Branson Sonifier 150 (Thistle Scientific; Glasgow, UK) and vortexing for 30 minutes at 4°C. The homogenates were centrifuged for 3 minutes at 17,000 g and the supernatants collected for precipitation of the proteins using 4:1 volumes ice-cold acetone. The resulting pellets were suspended in 100 μL of 50 mM NH₄HCO₃ (pH 8.0). Sulfhydryl groups on proteins were reduced by incubation with 100 mM DTT for 30 minutes at 60°C and alkylated with 200 mM iodacetamide for 30 minutes at 37°C. Proteins were cleaved into peptides by incubation with 1:50 (trypsin:protein) porcine trypsin (Promega; Madison, WI, USA) for 17 hours at 37°C and stopped after 16 hours by addition of 0.80 μL of 8.8 M HCl. Samples were stored at -80°C. Prior to mass spectrometry analyses, 0.1% formic acid was added to a final concentration of 0.12 μg/μL protein.

Digested prefrontal cortex and cerebellum proteomes were analysed using SRM-MS on a Xevo TQ-S mass spectrometer coupled to a nanoAcquity UPLC system (Waters Corporation, Wilmslow, UK), as described previously [23]. Protein candidates were selected for this analysis given their association with pathways which have been implicated previously in autism. Physiochemical criteria for selecting tryptic peptides were based on peptide count, uniqueness and quality of transitions. All SRM functions had a 6 minute window of the predicted retention time and scan times were 0.02 seconds. Two peptides were selected for each target protein and isotopically-labelled peptides were synthesized at JPT Peptide Technologies GmbH (Berlin, Germany). The samples and labelled peptides were mixed together and separated using the following 42 minutes gradient: 97/3% (A/B) to 70/30% in 20 minutes; 70/30% to 15/85% in 5 minutes; and in 3 minutes to 97/3%. Peptide spectra were acquired in SRM mode using a capillary voltage of 2.35 kV and a cone voltage of 33 V. At least three transitions were measured for each peptide. The candidate proteins were identified with the SRM method as described previously [23].

Statistical analyses of SRM data were conducted in the R statistical programming language (version 2.15.3; http://​www.​r-project.​org), using the package MSstats [26‐28]. Data processing consisted of log₂ transformation to stabilize the variance and normalize peak intensities. The data were then analysed using a linear mixed model to detect proteins or peptides that were present at significantly different levels in cases compared to controls (autism/control). In the model, the experimental characteristics were linked to the data in terms of biological condition, labelling state, technical and biological replication and potentially interfering transitions [28]. The scope of inference for both biological replicates and technical replicates were set to ‘restricted’, as conclusions about the data are confined to the study itself. Furthermore, additional model interactions were performed for interference transitions. These analyses resulted in P-values, Q-values (allowing for multiple testing) and ratio changes (autism/control).

SRM-MS analysis of prefrontal cortex protein extracts from autism and control patients showed no batch effects as indicated by principal component analysis (PCA), of which an additional plot shows this in more detail (Additional file 3). Conversely, analysis of the cerebellum protein extracts led to the identification of one outlying sample. After removal of this sample, no further outliers were detected, as is visualised in the additional PCA plot (Additional file 4). All further statistical analyses were carried out using only the remaining samples.

SRM-MS was used for targeted analysis of proteins associated with energy metabolism, astrocyte function, myelination and synaptic vesicles in accordance with literature evidence [4‐21]. The proteins which were detectable using SRM-MS analysis are indicated in Table 2. GFAP, vimentin (VIME), creatine kinase B-type (CKB), MAG, MBP, myelin-oligodendrocyte glycoprotein (MOG), myelin proteolipid protein (PLP1), dynamin-2 (DNM2), syntaxin-1A (STX1A), syntaxin-binding protein 1 (STXBP1), synapsin-2 (SYN2), synaptotagmin-1 (SYT1), protein kinase C casein kinase substrate 1 (PACSIN1) were detected in both prefrontal cortex and cerebellum from all patients and controls. However, reelin (RLN) was only identified in the cerebellum and was therefore not used in the comparative analysis.

Prefrontal cortex and cerebellum showed a similar decrease in the levels of the immature astrocyte marker VIME, whereas the mature astrocyte marker GFAP was increased and this occurred only in the cerebellum (Figure 1). The energy-related CKB was increased by more than 70% in the prefrontal cortex and showed a non-significant decrease (P = 0.2605) in the cerebellum. The myelin-related proteins MAG and PLP1 were decreased in the prefrontal cortex with a concomitant increase in the synaptic proteins SYN2 and STBP1, although the latter was not significant (P = 0.1002). Conversely, the myelin related proteins MAG, MBP, MOG and PLP1 were all increased by more than 30% and the synaptic proteins STX1, SYT1 and PACSIN1 were decreased in the cerebellum. DNM2 was also decreased in the cerebellum although this was not significant (P = 0.2910).