Sage Journals: Discover world-class research

Abstract

Objective: In order to conduct postmortem human brain research into the neuropatho-logical basis of schizophrenia, it is critical to establish cohorts that are well-characterized and well-matched. The aim of the present study was therefore to determine if specimen characteristics including: diagnosis, age, postmortem interval (PMI), brain acidity (pH), and/or the agonal state of the subject at death related to RNA quality, and to determine the most appropriate reference gene mRNAs.

Methods: A matched cohort was selected of 74 subjects (schizophrenia/schizoaffective disorder, n = 37; controls, n = 37). Middle frontal gyrus tissue was pulverized, tissue pH was measured, RNA isolated for cDNA from each case, and RNA integrity number (RIN) measurements were assessed. Using quantitative reverse transcription–polymerase chain reaction, nine housekeeper genes were measured and a geomean calculated per case in each diagnostic group.

Results: The RINs were very good (mean = 7.3) and all nine housekeeper control genes were significantly correlated with RIN. Seven of nine housekeeper genes were also correlated with pH; two clinical variables, agonal state and duration of illness, did have an effect on some control mRNAs. No major impact of PMI or freezer time on housekeeper mRNAs was detected. The results show that people with schizophrenia had significantly less PPIA and SDHA mRNA and tended to have less GUSB and B2M mRNA, suggesting that these control genes may not be good candidates for normalization.

Conclusions: In the present cohort <10% variability in RINs was detected and the diagnostic groups were well matched overall. The cohort was adequately powered (0.80–0.90) to detect mRNA differences (25%) due to disease. The study suggests that multiple factors should be considered in mRNA expression studies of human brain tissues. When schizophrenia cases are adequately matched to control cases subtle differences in gene expression can be reliably detected.

Keywords

housekeeping genes pH postmortem brain postmortem interval RNA integrity number

Schizophrenia research on postmortem human tissue has matured in the past decade, largely due to two factors: (i) the widespread availability of larger cohorts that are more extensively screened; and (ii) the readily available quantitative molecular tools especially for gene expression studies. In the past, lack of replication of basic biological findings in brains of people with schizophrenia across laboratories hampered progress. The renewed research effort on postmortem brains is finally yielding numerous replicable findings across independent laboratories and independent cohorts, which is a critical advance for postmortem brain research on schizophrenia. Studies of postmortem human brain provide a valuable avenue for elucidating the underlying cellular and molecular mechanisms of schizophrenia that are not accessible through live imaging of people or through animal models. Thus, there is a need for direct studying of postmortem human brains from people with schizophrenia, although collecting and characterizing human brains is methodologically challenging [1–3], and standards regarding inclusion criteria necessary for rigorous postmortem research are not universally agreed upon and continue to change [4–11]. Numerous premortem, antemortem and postmortem factors can influence the quality of human brain tissue and need to be considered, and the relative importance of confounding factors for specific molecules are often not known.

Despite these limitations, postmortem human studies of psychiatric illnesses such as schizophrenia remain an extremely valuable tool because they have already begun to shed light on a number of molecular mechanisms that contribute to the aetiology and development of schizophrenia [12–17]. The establishment of multiple large cohorts of well-characterized brains of people with schizophrenia and controls from around the world contribute to our understanding of the biological basis for schizophrenia and will be critical to advancing the field. Therefore, given our experience, and with access to a fairly large collection of postmortem tissue samples, we set out to establish a new cohort of schizophrenia and control brains from Sydney, Australia.

The assembly and characterization of this cohort has been driven by the Schizophrenia Research Institute (SRI), a virtual institute supporting the collaboration of Australian scientists and clinicians who bring a variety of knowledge, skills and techniques together to find ways to prevent and cure schizophrenia [18]. In 2007, through the Institute's Developmental Neurobiology Panel, SRI commenced a process of actively facilitating the development of a large-scale collaborative research programme based on the use of a common collection of postmortem brains for schizophrenia research from the Tissue Resource Centre at the University of Sydney. The results from this multifaceted investigation will be collected and collated, thereby allowing for testing of relationships among a variety of measures. This report represents the first step in this collaborative process whereby the collection and quality control processes for this new SRI tissue cohort are described and the selection of reference genes is considered.

Methods

Subjects

All research was approved and conducted under the guidelines of the Human Research Ethics Committee at the University of New South Wales (HREC 07261). The study cohort included 37 schizophrenia/ schizoaffective disorder patients and 37 matched controls (Table 1). Controls were selected prior to RNA extraction based on the following parameters (in descending order of importance for matching), brain pH (±1), age at death (within 10 years), and postmortem interval (PMI; within 10 h). Other factors such as hemisphere, gender, time in freezer and agonal state were matched on a stratum (groupwise) basis. Details of brain collection and determination of clinical and tissue factors can be found in Supplementary Materials and Methods, and in Tables 1S, 2S. Supplementary material to be found online, at http//www.informaworld.com/10.1080/00048670903393662.

Table 1.

Subject characteristics

Demographics	Controls	Schizophrenia subjects
Age (years)	51.1 ± 2.40	51.3 ± 2.32
pH (DLPFC)	6.66 ± 0.29	6.61 ± 0.30
PMI (hours)	24.8 ± 10.97	28.8 ± 14.07
RIN	7.3 ± 0.57	7.3 ± 0.58
Gender	7F, 30M	13F, 24M
Hemisphere	23R, 14L	17R, 20L

DLPFC, dorsolateral prefrontal cortex; PMI, postmortem interval; RIN, RNA integrity number; F, female; M, male; R, right; L, left

Table 1S

Detailed cohort demographic information.

Case	Dx	Gender	Age at death	PH	PMI (Hrs)	RIN	Hemi	Race	Cause of death	Manner of death	Toxicology at postmortem
1	scz	F	51	5.69	12	6.2	R	C	Congestive cardiac failure,	N	Lithium 20mg/L; Midazolam 0.02mg/L
2	scz	M	56	6.16	15	6.7	L	C	COPD	N	Nil detected
3	scz	M	52	6.35	46	7.1	R	c	Cardiomegaly	N	TOX not performed
4	scz	F	68	6.24	32	7.5	L	c	Acute pancreatitis	N	Alcohol 0.055g/100ml
5	scz	M	54	6.19	27.5	6.3	R	c	Coronary artery thrombosis	N	Chlorpromazine: 0.7mg/L; Diazepam: <0.1 mg/L; Nordiazepam: 0.1mg/L; Insulin: 2uU/mL
6	scz	F	58	6.34	19	6.8	R	c	Sepsis and chronic renal failure	N	Morphine: 0.06mg/L, Codeine: 0.05mg/L, Carbamazepine: 7mg/L, Pethi dine: 0.1 mg/L, Paracetamol: 6mg/L, Metoclopramide 0.1 mg/L, Diazepam: <0.1mg/L
7	scz	F	66	6.33	12.5	6.6	R	c	Faecaloid peritonitis	N	Nil detected
8	scz	F	55	6.30	33.5	7.6	L	c	ASCVD	N	Amisulpride 18mg/L; Clozapine 1.4mg/L
9	scz	M	55	6.37	72	6.6	R	c	Ischaemic heart disease	N	Metoprolol: 0.2mg/L
10	scz	F	54	6.41	29	7.2	R	c	Asthma	N	Citalopram 0.6mg/L
11	SCZ-A	M	57	6.41	28	7.4	R	c	COPD	N	Paracetamol 3mg/L; lithium 4.6mg /L
12	SCZ-A	F	61	6.43	17	6.8	R	c	Myocarditis	N	Clozapine: 0.4mg/L; Lithium: 0.1 mg/L
13	scz	F	67	6.44	27	6.3	L	c	Emphzyma, right lung	N	Benztropine 0.2mg/L; Mesoridazine 0. 9mg/L; Thiori dazine 0.6mg/L; Paracetamol < 3.0mg/L
14	scz	M	75	6.56	36	6.9	L	c	Ischaemic heart disease	N	Olanzapine −0.2mg/L; Fluvoxamine −0.7mg/L
15	scz	M	40	6.50	21.5	7.6	L	c	Obstructive sleep apnea, dihydrocodeine toxicity	N	Valproic acid 20mg/L; dihydrocodeine 0.7mg/L; quetiapine 0.3mg/l; sertraline
16	scz	M	44	6.52	29.5	7.3	L	c	Hanging, suicide	S	Urine THC detected
17	scz	M	51	6.53	21	6.9	L	c	Ischaemic heart disease	N	Thioridazine 2.2mg/L; Mesoridazine
18	scz	M	27	6.56	10	7.0	L	c	Clozapine toxicity	U	Clozapine 8.6mg/L
19	scz	M	51	6.68	18	7.9	L	c	Ischaemic heart disease	N	Nil detected
20	scz	M	52	6.70	8.4	7.7	R	c	Ischaemic heart disease	N	Temazepam <0.1mg/L
21	scz	M	57	6.72	48	6.5	R	c	ASCVD	N	Carbamazepine 10mg/L, citalopram 0.2mg/L, quetia pine <0.1mg/L
22	scz	M	33	6.75	48	8.3	L	c	Hanging suicide	S	Doxylamine: 0.9mg/L; Olanzapine: 0.2mg/L; Paracetamol: 3mg/L
23	scz	M	30	6.77	24	7.3	L	c	CO poisoning, suicide	S	Carbon Monoxide 74% saturation; Clozapine 0.7mg/L
24	SCZ-A	M	73	6.79	14	7.0	L	c	Asphyxia, choked on food	A	Fluoxetine: 0.2mg/L; 7-amino nitrazepam: 0.1 mg/L; diazepan: <0.1mg/L; paracetamol: <3mg/L
25	scz	F	56	6.81	34	7.7	L	c	Pulmonary thrombo-	N	Paracetamol 4.5mg/L
26	scz	M	27	6.82	38.5	8.0	L	c	Myocarditis	N	Clozapine 0.9mg/L
27	SCZ-A	M	34	6.95	26	7.3	R	A	Hanging, suicide	S	Carbamazepine 1mg/L)
28	SCZ-A	F	33	6.93	50	8.1	R	c	Hanging, suicide	S	Nil detected
29	SCZ	M	57	6.98	38	7.5	L	C	Probable myocardial scarring, cardiac arrythmia	N	Thioridazine 0.6mg/L; Seraline <0.1mg/L
30	SCZ	M	27	6.84	33	8.2	R	C	Hanging, suicide	S	Nil detected
31	SCZ	M	59	6.87	26.5	7.6	R	c	Ischaemic heart disease	N	TOX not performed
32	SCZ-A	F	73	6.9	20	7.2	L	c	Right ventricular dysplasia (Uhl’ s anomaly)	N	Doxepin 0.1mg/L, chlorpheniramine 0.1, trifluoperazine 0.1, codeine 0.05, pseudoephedrine 0.1, paracetamol 10mg/L
33	SCZ-A	M	30	7.01	26	7.2	L	c	Hanging, suicide	S	Venlafaxine 0.9mg/l
34	SCZ	F	61	6.91	42	6.8	R	c	ASCVD	N	Clozapine 1.1mg/L; diazepam 0.2mg/L; Laudanosine 0.4mg/L; Nordiazepame 0.4mg/L; Olanzapine
35	SCZ	M	32	6.99	26	7.9	L	c	Hanging, suicide	S	Priadel 400 mg, carbamazepine 400mg-600mg, pimozine 4mg, lithium carbonate, chlorpromazine, fluphenazine decanoate, pimozide, thioridazine
36	SCZ	M	67	6.82	5	8.4	L	c	Ischaemic heart disease	N	Nil detected
37	SCZ	F	56	7.09	39	7.6	R	c	Undetermined	N	Thioridazine (1mg/L); Mesoridazine (0.6mg/L)
38	CON	M	46	5.84	29	6.7	L	c	Acute Ml	N	Nil detected
39	CON	M	60	6.02	25	7.1	R	c	Bacterial Peritonitis stomach tumor,	N	Nil detected
40	CON	M	37	6.23	11	6.9	R	c	Pulmonary Embolism	N	Nil detected
41	CON	M	78	6.29	6.5	7.5	L	c	Adenocarcinoma (lung,	N	TOX not performed
42	CON	M	61	6.31	27.5	6.6	R	c	Ischaemic heart disease	N	TOX not performed
43	CON	M	74	6.31	10	7.4	L	c	Respiratory arrest	N	Nil detected
44	CON	F	78	6.37	11	6.1	R	c	Respiratory failure due to pulmonary fibrosis	N	Control less 20g
45	CON	F	56	6.47	23	7.1	R	c	ulmonary thrombo-embolus	N	Nil detected
46	CON	M	60	6.55	21	7.8	L	c	Acute Ml	N	Bid Ale: 0.251 g per 100mL paracetamol <3mgL
47	CON	F	60	6.52	13	7.4	L	c	Ischaemic heart disease	N	Nil detected
48	CON	M	62	6.57	37.5	7.5	R	c	Acute Ml	N	TOX not performed
49	CON	M	50	6.61	19	6.9	L	c	Ischaemic heart disease	N	Nil detected
50	CON	M	58	6.56	12	7.6	L	c	Ischaemic heart disease	N	Nil detected
51	CON	M	73	6.57	48.0	8.1	R	c	Ischaemic heart disease.	N	TOX not performed
52	CON	M	46	6.65	25	7.3	R	c	Mitral valve prolapse	N	Nil detected
53	CON	F	49	6.64	15	8.3	R	c	Probable ARVD	N	Nil detected
54	CON	M	56	6.64	24	7.1	R	c	Coronary artery atheroma	N	TOX not performed
55	CON	M	34	6.51	20.5	7.1	R	c	Acute exacerbation of	N	TOX not performed
56	CON	M	44	6.72	50	7.2	L	c	Ischaemic heart disease	N	Nil detected
57	CON	M	50	6.68	29	8.1	R	c	Ischaemic Heart Disease	N	Nil detected
58	CON	M	53	6.70	27	7.1	R	C	Acute Ml	N	TOX not performed
59	CON	M	43	6.71	13	7.4	R	C	Thrombotic coronary artery occlusion	N	Nil detected
60	CON	M	38	6.73	13.5	7.7	L	c	ASCVD	N	Nil detected
61	CON	M	60	6.95	21.5	6.9	R	c	Ischaemic heart disease	N	TOX not performed
62	CON	M	54	6.80	29	6.0	R	A	Coronary artery atheroma	N	TOX not performed
63	CON	M	18	6.79	33	7.0	L	C	Probable hypertrophic cardiomyopathy	N	Paracetamol (24mg/L); lignocaine (1mg/L)
64	CON	M	37	6.80	21	7.6	L	C	Ischaemic heart disease	N	Nil detected
65	CON	F	21	6.83	39.5	8.1	R	C	Primary cardiac arrhythmia	N	Nil detected
66	CON	F	51	7.15	37.5	7.7	R	C	Acute Ml	N	TOX not performed
67	CON	F	33	6.91	24	8.0	L	C	Cardiac arrythmia; myocardial fibrosis	N	Nil detected
68	CON	M	59	6.96	20	7.4	R	C	Coronary thrombosis	N	Nil detected
69	CON	M	64	7.01	39.5	7.0	R	C	Coronary artery thrombosis	N	TOX not performed
70	CON	M	24	6.95	43	7.1	R	C	Probable idiopathic cardiac arrhythmia	N	Nil detected
71	CON	M	57	6.86	188	8.4	L	c	Ischaemic heart disease	N	less 20g
72	CON	M	37	6.94	24	7.5	R	c	Electrocution	A	Codeine <0.5mg/L; paracetamol 3mg/L; lignocaine< 0. 5mg/L
73	CON	M	56	6.98	37	6.2	R	c	Hypertension, cardiomegaly	N	TOX not performed
74	CON	M	55	7.19	20	7.3	L	c	Cardiac arrest due to emphzyma, pneumothorax	N	Nil detected

Dx = diagnosis; SCZ = Schizophrenia patient, SCZ-A = schizoaffective CON = Control case; C = Caucasian; A = Asian; Agonal States-1 = good, 2 = fair, 3 = poor; COPD = Chronic obstructive pulmonary disease; Ml = Myocardial infarction; ASCVD = Atherosclerotic cardiovascular disease; ARVD = arrhytmogenic right ventricular dysplasia; Manner of Death: A = accidental, N = natural, S = suicide, U = undetermined.

Table 2S

Clinical characteristics of schizophrenia cohort

Case	Subclass	Syndrome	Age of onset	Duration of illness	Last CPZ	Daily CPZ	Lifetime CPZ	Antidepressant history	TD	Hallucinations Auditory / Visual		Paranoia	Grandiose	Thought disorder	Suicide attempts
1	Residual	Negative	35	16	570	556.25	3248500	Yes	Yes	Yes	No	Yes	No	Yes	No
2	Disorganised	Negative	21	35	450	425.00	5429375	No	Yes	Yes	No	No	No	Yes	No
3	Undifferentiated	Positive	19	33	500	590.80	7116186	No	Yes	Yes	No	Yes	No	Yes	Yes
4	Undifferentiated	Positive	23	45	190	224.00	3679200	No	No	Yes	No	Yes	No	Yes	No
5	Paranoid	Positive	19	35	1560	325.00	4151875	No	Yes	Yes	No	Yes	No	Yes	Yes
6	Paranoid	Positive	19	39	450	350.00	4982250	No	No	Yes	No	Yes	No	Yes	No
7	Paranoid	Positive	19	47	N/A	1850.00	31736750	Yes	Yes	Yes	No	Yes	No	No	No
8	Undifferentiated	Positive	17	38	1300	2362.50	32767875	Yes	No	Yes	Yes	Yes	No	Yes	Yes
9	Paranoid	Positive	30	25	N/A	900.00	8212500	No	No	Yes	No	Yes	Yes	Yes	No
10	Paranoid	Positive	19	35	700	762.50	9740937.5	Yes	Yes	Yes	No	Yes	Yes	Yes	No
11	Bipolar type	Positive	30	27	415	1100.00	10840500	No	No	Yes	Yes	Yes	Yes	Yes	No
12	Bipolar type	N/A	31	30	100	1866.50	20438175	No	No	Yes	No	No	Yes	Yes	Yes
13	Disorganised	Negative	21	46	350	625.00	10493750	Yes	Yes	Yes	No	Yes	No	Yes	No
14	Paranoid	Negative	32	43	500	700.00	10986500	Yes	Yes	Yes	No	No	Yes	Yes	No
15	Paranoid	Positive	17	23	986	1012.50	8499937.5	Yes	No	Yes	No	Yes	Yes	Yes	No
16	Undifferentiated	Positive	27	17	500	575.00	3567875	No	No	Yes	No	Yes	No	No	Yes
17	Residual	Negative	27	24	N/A	400.00	3504000	No	No	Yes	No	Yes	No	Yes	No
18	Undifferentiated	Positive	18	9	300	649.50	2133607.5	No	No	Yes	No	Yes	No	Yes	Yes
19	Paranoid	Positive	21	30	150	800.00	8760000	No	Yes	Yes	No	Yes	Yes	Yes	No
20	Paranoid	Positive	21	31	680	430.00	4865450	Yes	No	Yes	No	Yes	Yes	Yes	No
21	Paranoid	Positive	40	17	618	600.00	3723000	Yes	No	No	No	Yes	No	Yes	No
22	Paranoid	Positive	22	11	250	222.00	891330	No	No	Yes	No	Yes	Yes	Yes	No
23	Paranoid	Positive	26	4	300	552.50	806650	Yes	No	Yes	No	Yes	No	Yes	Yes
24	Depressive type	N/A	21	52	380	650.00	12337000	Yes	Yes	Yes	No	Yes	No	Yes	No
25	Disorganised	Positive	17	39	760	1000.00	14235000	Yes	No	Yes	Yes	No	No	Yes	Yes
26	Undifferentiated	Negative	23	4	200	287.50	419750	Yes	No	Yes	Yes	Yes	No	Yes	No
27	Bipolar type	Positive	27	7	250	195.00	498225	Yes	No	Yes	No	Yes	Yes	Yes	Yes
28	Depressive type	N/A	14	19	95	162.50	1126937.5	Yes	No	Yes	No	Yes	No	No	Yes
29	Paranoid	Positive	30	27	700	250.00	2463750	Yes	No	Yes	No	Yes	No	Yes	Yes
30	Paranoid	Positive	19	8	N/A	315.00	919800	No	No	No	No	Yes	No	Yes	Yes
31	Disorganised	Negative	21	38	750	784.00	10874080	No	No	Yes	No	Yes	No	Yes	No
32	Depressive type	N/A	36	37	300	300.00	4051500	Yes	No	Yes	No	Yes	No	Yes	Yes
33	Depressive type	N/A	27	3	285	187.50	205312.5	Yes	No	No	No	Yes	No	Yes	Yes
34	Paranoid	Positive	19	42	1200	1150.00	17629500	No	No	Yes	No	Yes	No	Yes	Yes
35	Disorganised	Positive	19	13	190	780.00	3701100	Yes	No	Yes	No	No	No	Yes	Yes
36	Undifferentiated	Positive	26	41	1340	1300.00	19454500	No	No	Yes	Yes	Yes	No	Yes	No
37	Paranoid	Positive	24	32	580	350.00	4088000	No	No	Yes	No	Yes	No	Yes	No

CPZ=chlorpromazine equivalent neuroleptic dose; TD=Tardive dyskinesia.

RNA extraction and quality assessment

Grey matter tissue from dorsolateral prefrontal cortex (DLPFC) was pulverized over dry ice. Approximately 300 mg was weighed while frozen and stored at −80°C until extraction. Total RNA was extracted with TRIzol Reagent (Cat. No. 15596-018; Invitrogen Life Science, Melbourne, Vic., Australia), using a polytron as previously described [19]. Total RNA pellets were resuspended in DEPC-treated water (Cat. No. 95284-100ML; Sigma-Aldrich, Castle Hill, NSW, Australia). The yield of total RNA was then analysed using a spectrophotometer (Nanodrop ND-1000; Thermo Scientific, Wilmington, DE, USA). Total RNA was not further purified through a column. The quality of extracted total RNA was determined on high-resolution capillary electrophoresis using an Agilent Bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA, USA). Approximately 100–200 ng RNA was applied to an RNA 6000 Nano LabChip (Ajilent Technologies, Waldbronn, Germany), without heating prior to loading, and subsequent steps performed according to the manufacturer's protocol. Two internal controls (samples with established high-quality RNA integrity number (RIN) = 10) were loaded on each chip, and all samples analysed on the same day. The RIN was calculated using an algorithm incorporating information from the entire electrophoretic trace and used as an indicator of RNA quality, ranging from 1 (lowest quality) to 10 (highest quality) [20]. Three separate aliquots of total RNA were used to synthesize cDNA from 3 μg of total RNA in a 26.25 μL reaction using the SuperScript First-Strand Synthesis kit (Invitrogen, Carlsbad, CA, USA) and were pooled to reduce any potential variability in cDNA reactions prior to diluting for reverse transcription–polymerase chain reaction (RT-PCR).

Quantitative reverse transcription-polymerase chain reaction

Transcript levels for nine housekeeper genes – glucuronidase-β (GUSB), cyclophilin A (PPIA), ubiquitin C (UBC), porphobilinogen deaminase (PBGD), succinate dehydrogenase complex subunit A (SDHA), β-actin (ACTB), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), TATA box binding protein (TBP) and β-2-microglobulin (B2M) – were measured on quantitative real-time PCR (qPCR) in each cohort sample using an ABI Prism 7900HT Fast Real Time PCR system with a 384-well format (Applied Biosystems, Foster City, CA, USA) and TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA, USA; Table 2). Samples were run with a seven-point standard curve using serial dilutions of pooled cDNA derived from a representative sample of subjects (three controls and three patients). The no-template control did not produce a signal in any assay. All amplifications from each subject were performed in triplicate and relative quantities were determined from the standard curve. Outliers due to measurement errors were omitted if the percent variance of the triplicates was >30% of the quantity mean value as previously described [21–23] and the mean re-calculated based on two values (this occurred in <5% of the samples). Stability of mRNA expression was measured using the program geNorm VBA applet for Microsoft Excel version 3.5, developed by Vandesompele et al., in which more stable genes will have a lower M [24].

Table 2.

Housekeeper genes with ABI Taqman Gene Expression assay part numbers

Gene symbol	Gene name	Taqman assay
GUSB	Glucuronidase-β	Hs99999908_m1
PBDG	Hydroxymethylbilane synthase	Hs00609297_m1
PPIA	Peptidylprolyl isomerase A (Cyclophilin A)	Hs99999904_m1
UBC	Ubiquitin C	Hs00824723_m1
SDHA	Succinate dehydrogenase complex, subunit A, flavoprotein (Fp)	Hs00188166_m1
ACTB	β-Actin	Hs99999903_m1
TBP	TATA box binding protein	Hs00427620_m1
B2M	β-2-Microglobulin	Hs99999907_m1
GAPDH	Glyceraldehyde-3-phosphate dehydrogenase	Hs99999905_m1

Statistical analysis

Statistical analyses were conducted using Statistica version 7 (StatSoft, Tulsa, OK, USA). After measurement errors were removed, population outliers were determined and samples >2SD from the mean for each group were removed (on average 1–2 subjects per group). We also defined population outliers using the Grubb's test, which yielded similar results. Both paired and non-paired Student's t-tests were used to determine if the diagnostic groups differed on individual housekeepers or geomean. To maximize power, the analyses were carried out with the schizophrenia patients combined with the schizoaffective patients as a single diagnostic category, but it is recognized that other analyses focusing only on the schizophrenia group are also planned. Correlations were run on continuous variables to determine if any of the demographic factors (age, pH, PMI or RIN) related to housekeeper mRNA levels or their geometric mean. If this was the case, analysis of covariance (ANCOVA) was used. Independent t-tests were conducted to examine whether gender or hemisphere were related to housekeeper gene expression. Factorial ANOVA was used to determine the effect of gender or hemisphere on gene expression in the diagnostic groups. Statistical significance was set at p ≤ 0.05. Significance level is reported as p rounded to three significant figures.

Results

Matching of diagnostic groups

The patients with schizophrenia and schizoaffective disorder (n = 37) were on average 51 years old, with a brain pH of 6.61 and a PMI of 28.8 hours (Table 1). The normal control subjects selected to match these patients (n = 37) had an average age of 51 years, a brain pH of 6.66 and a PMI of 24.8 hours. The mean RINs were 7.3 for both groups and had a low coefficient of variation (between 7.8 and 8.0%). As anticipated age, pH, PMI, freezer storage time, and RIN did not differ significantly between the groups (all t < 1.3, all p > 0.210). Freezer storage time of brain tissue for patients averaged 81 months and did not differ statistically as compared to normal controls at 70 months. The number of men was higher than women in both groups, although the number of men was slightly lower in the schizophrenia group (n = 24) as compared to the normal control group (n = 30). Women with schizophrenia (n =13) outnumbered their control group counterparts (n = 7). The number of right and left hemispheres studied was comparable for both the control (L = 14, R = 23) and patient (L = 20, R = 17) groups. One patient with a comorbid diagnosis for alcohol dependency and another schizophrenia patient with comorbid drug abuse at the time of death were included in the cohort.

Relationship of demographic variables to each other and RIN

The subjects spanned a considerable age range (18–78 years) and we found that age at death negatively correlated with the overall brain-weight (r = −0.32, p = 0.005, n = 74), overall brain volume (r = −0.26, p = 0.026, n = 74), DLPFC pH (r = −0.27, p = 0.018, n = 74) and RIN (r = −0.27, p = 0.020, n = 74; Figure 1). As expected, women (n = 20, 1300 mL) had smaller brains as compared to men (n = 54, 1468 mL, t = 5.13, df = 72, p < 0.0001), and women (1288 g) had lighter brains as compared to men (1469 g; (t = 5.55, df = 72, p < 0.0001). Patients with schizophrenia (1394 g) had a lighter brain-weight compared to non-affected controls (1446 g) although this difference was not statistically significant (t = −1.53, df = 72, p = 0.131). People with schizophrenia tended to smoke (n = 23) rather than not smoke (n = 7) and were more likely to be confirmed smokers than controls (n = 9). Smokers tended to have lower brain pH in both frontal cortex (6.63 vs 6.66) and cerebellum (6.45 vs 6.57), and frontal cortex RIN was marginally lower in the smokers (7.2 vs 7.3). The pH and RINs in the smokers and non-smokers, however, did not differ statistically.

Figure 1.

Overall pH and RNA integrity number (RIN) similarly decreased with increasing age (both r = −0.27, p < 0.020, n = 74).

As expected, the RIN measurement for DLPFC total RNA positively correlated with brain pH taken from the same tissue (r = 0.39, p < 0.001) and also correlated with cerebellar pH (r = 0.31, p = 0.007; Figure 6a). Supplementary material to be found online, at http://www.informaworld.com/10.1080/00048670903393662. We found that the pH measurements from the cerebellum and middle frontal gyrus correlated significantly with each other (r = 0.48, p < 0.001, n = 74). PMI did not correlate with RIN (r = 0.02, p = 0.873; Figure 6b-Supp) or cerebellar pH (r = 0.17, p = 0.216). Additionally, we extracted the same amount of RNA per gram of frozen tissue in people with schizophrenia and unaffected controls (both, 0.66 μg mg⁻¹, t = −0.06, df = 72, p = 0.958).

Figure 6S.

(A) RIN values are positively correlated with pH (p=0.001) (B) and not with PMI (p=0.87).

Transcript stability and calculation of geometric mean

We determined that ACTB and GADPH were the two most stable genes generating the lowest M, and that B2M and GUSB were the least stable genes with the highest M (Figure 2A). Pairwise variation analysis suggests that using the two most stable housekeeping genes (ACTB and GADPH) is sufficient to determine the normalization factor because V2/3 is below the suggested threshold of 0.15 (Figure 2B) [24]. For the purposes of normalizing the data to housekeeper control genes, however, we chose a further two genes because it has also been recommended that genes used for normalization are not significantly altered in schizophrenia compared to controls, and represent transcripts of different levels of expression. We chose four genes, two (ACTB and GAPDH) with higher expression levels, one (UBC) being in the midrange of expression, and one (TBP) with lower expression levels [25]. When the geomean was recalculated using these four housekeeping genes (herein geomean(4)) and outliers removed (one case removed) we did not detect a significant difference between the normalizing factor geomean in patients with schizophrenia compared to controls (unpaired t-test, t = 0.82, df = 71, p = 0.415; paired t-test, t = −0.79, df = 35, p = 0.420; Figure 3A).

Figure 2.

Average expression stability and optimal number of housekeeper genes for normalization determined using the program geNorm. (A) Genes were ranked according to average expression stability. ACTB and GAPDH were the most stable genes (lowest M), while GUSB was the least stable gene (highest M) (B) Pairwise variation with the stepwise addition of more housekeeping genes shows that the use of the two most stable housekeeping genes will be sufficient for accurate normalization because this satisfies the cut-off (0.15 M) recommended by Vandesompele et al. [24].

Figure 3.

(A) The calculated geomean(4) was comparatively similar in both diagnostic groups whether paired (t = −0.79, df = 35, p = 0.420) or unpaired (t = 0.82 df = 71, p = 0.415); (B) geomean(4) decreased with an increase in age (r = −0.28, p = 0.018); (C) overall pH in dorsolateral prefrontal cortex (DLPFC) tissue was positively correlated with geomean(4) for controls (r = 0.41, p = 0.012) and patients (r = 0.52, p = 0.001); (D) agonal state 1 vs 3 showed a significant decrease in the geomean(4) (p = 0.031 by planned post-hoc LSD); (E) RNA integrity number (RIN) strongly correlated with the geomean(4) (r = 0.57, p < 0.0001); (F) postmortem interval (PMI) had no effect on the geomean(4) (r = 0.006, p = 0.96).

Correlation of demographic variables and housekeeper mRNAs

Age

When considering the whole cohort, the expression of six of nine housekeeping genes correlated negatively with age at death (SDHA: r = −0.26, p = 0.026; ACTB: r = −0.28, p = 0.018; GAPDH: r = −0.28, p = 0.016; PBGD: r = −0.26, p = 0.030; UBC: r = −0.26, p = 0.03; PPIA: r = −0.26, p = 0.025) as did the geomean(4) (r = −0.28, p = 0.018; Figure 3B). When we explored which diagnostic group contributed more to these overall correlations, we found that the normal group did not show any significant correlations with any housekeeper control genes or geomean(4) and age at death (all p > 0.08). In contrast, the expression of five of nine housekeeping genes and geomean(4) correlated negatively with age in schizophrenia patients only (SDHA: r = −0.45, p = 0.006; ACTB: r = −0.47, p = 0.004; GAPDH: r = −0.37, p = 0.024; UBC: r = −0.37, p = 0.028, PPIA: r = −0.40, p = 0.016; geomean(4): r = −0.45, p = 0.006).

pH

Most (seven of nine) reference genes showed a significant positive correlation with DLPFC tissue pH (range: r = 0.26–0.58, p = 0.02 to <0.001). The only two potential reference genes that did not show a significant relationship with tissue pH were those deemed least stable, B2M (r = 0.14, p = 0.237) and GUSB (r = 0.11, p = 0.364). In general, most reference mRNAs were positively correlated to tissue pH in both control and schizophrenia subjects. Additionally, geomean(4) showed a significant correlation to pH in both diagnostic groups (controls: r = 0.41, p = 0.012; patients: r = 0.52, p = 0.001; Figure 3C).

Agonal state

Agonal state had an influence on DLPFC pH (F = 3.07. df = 2,71, p = 0.051) but this did not reach statistical significance with cerebellar pH (F = 3.01, df = 2,71, p = 0.057). In both the prefrontal cortex and the cerebellum, the individuals with the highest agonal state rating had the lowest pH (6.44 and 6.32, respectively). Each agonal state rating had significantly different DLPFC pH on post-hoc least significant difference testing (LSD; all p ≤ 0.05). Although individuals with the highest agonal state scores had the lowest RNA quality as assessed on RIN, this did not reach statistical significance. Those individuals with the longest agonal state in the present study did show a 27% decrease (F = 2.68, df = 2,33, p = 0.086) in expression of the geomean(4) of the housekeeper normalizer genes (agonal state: 1 vs 3, p = 0.031 on planned post-hoc LSD; Figure 3D).

RNA integrity number

All nine potential reference genes had a significant positive correlation with DLPFC RNA quality as determined on RIN. Most of the genes (eight of nine) and the geomean(4) (Figure 3E) showed a strong and statistically significant correlation (range r = 0.43–0.63, all p < 0.001, except GUSB mRNA (r = 0.24, p = 0.042)). All nine housekeeper mRNAs in the control group displayed a correlation with RIN (all r > 0.46, all p < 0.006). Similarly, eight of nine housekeeping genes showed a correlation with RIN in the schizophrenia group (excluding GUSB; all r > 0.36, all p < 0.029).

Postmortem interval

Of the housekeeper control mRNAs, most of the housekeeper genes and the geomean(4) were not correlated with PMI (Figure 3F; all r < 0.067, all p > 0.572 except B2M r = −0.20, p = 0.088). The expression of GUSB in schizophrenia patients only (r = −0.35, p = 0.041) was significantly negatively correlated with PMI.

Diagnostic differences in housekeeper control mRNA

T-tests

In the present study we found approximately 30% coefficient of variation in housekeeper mRNA in both groups (Table 3), which is less variable than previously found [6]. Surprisingly, we found that some housekeeping control genes were reduced in the DLPFC of patients with schizophrenia by as much as 10–15% (Figure 4A). We found that SDHA (t = −2.21, df = 69, p = 0.030 unpaired; t = −2.55, df = 33, p = 0.016 paired), PPIA mRNA (t = −2.32, df = 71, p = 0.023 unpaired; t = −2.55, df = 35, p = 0.015 paired) and GUSB (t = 1.93, df = 68, p = 0.057 unpaired; t = 2.02, df = 32, p = 0.052 paired) were significantly lower in the schizophrenia and schizoaffective disorder patient group by 14.4%, 14.9%, and 12.4%, respectively. The five other housekeeper genes examined (UBC, ACTB, GAPDH, TBP, PBGD) and the geomean(4) did not significantly differ in patients compared to controls (all p > 0.343; Figure 4B). After ANCOVA analysis, the significant reduction in SDHA mRNA (F = 4.54, df = 69, p = 0.036) and PPIA mRNA (F = 5.78, df = 68, p = 0.018) was maintained. No significant differences were found in any of the other mRNAs or the geomean(4) on ANCOVA (p > 0.05).

Figure 4.

mRNA expression of housekeeping genes in control and schizophrenia dorsolateral prefrontal cortex (DLPFC). mRNA expression was measured by quantitative reverse transcription-polymerase chain reaction and all cases expressed as a percentage of the control mean for that gene. (A) A significant reduction in expression of SDHA, and cyclophilin (PPIA) and strong trend in GUSB mRNAs was found in schizophrenia patients compared to controls (14.4%, p = 0.016; 14.9%, p = 0.015; and 12.4%, p = 0.052, respectively; paired t-test). (B) mRNA expression levels for all remaining housekeeping genes that did not significantly differ between diagnostic groups, including the geomean(4) (all p > 0.34).

Table 3.

Housekeeper coefficients of variance

		Coefficient of variance
Gene	Expression in schizophrenia relative to controls (%)	Controls (%)	Schizophrenia subjects (%)
SDHA	85.6	30.7	27.3
ACTB	93.6	34.2	32.1
GAPDH	93.4	36.6	36.3
B2M	88.7	29.1	25.5
TBP	103.1	35.7	38.5
GUSB	87.6	29.6	26.7
PPIA	85.1	29.5	29.5
PBGD	93.2	28.1	34.3
UBC	96.1	34.9	34.4
Geomean(4)	93.5	36.2	33.2

Effects of hemisphere and gender on geomean(4)

We found that there was a main effect of hemisphere on the geomean(4) by two-way ANOVA (F = 6.26, df = 1, 69, p = 0.021), but no overall effect of diagnosis (F = 6.26, df = 1, 69, p = 0.228) and no interaction effect (F = 6.26, df = 1,69, p = 0.989). Both people with schizophrenia and controls had significantly more expression of the housekeeper geomean(4) on the left hemisphere as compared to the right. When we covaried for factors that correlated with the geomean(4) (age, pH and RIN), we found that the main effect of hemisphere was similar (F = 3.86, df = 1,66, p = 0.053) and the effect of diagnosis and the diagnosis × hemisphere interaction remained non-significant. We did not detect a main effect of gender or diagnosis or an interaction between diagnosis and gender on the levels of geomean(4) (all F < 1.3, all p > 0.260).

Consideration of clinical variables

The patients with schizophrenia in the present cohort had 4-45 years duration of illness, which was positively correlated with daily chlorpromazine equivalent neuroleptic dose (CPZ; r = 0.38, p = 0.020, n = 37) and the last recorded dose of CPZ (r = 0.38, p = 0.027, n = 37). Total brainweight of people with schizophrenia and/or schizoaffective disorders was negatively correlated with daily CPZ (r = −0.34, p = 0.038, n = 37; Figure 5A), duration of illness (r = −0.35, p = 0.031, n = 37; Figure 5B), and lifetime CPZ estimates (r = −0.39, p = 0.018, n = 37). The relationship of brain volume with duration of illness and lifetime CPZ estimates reached trend levels of significance (p = 0.060–0.084, n = 37). The only control mRNA that showed a statistically significant correlation with any medication measure was UBC with the last recorded CPZ (r = −0.36, p = 0.048, n = 31). Duration of illness did, however, negatively correlate with SDHA and ACTB (r > −0.37 p = 0.028, n = 35; r = −0.38, p = 0.022, n = 36, respectively) mRNA expression, and with the geomean(4) (r = −0.35, p = 0.034, n = 36; Figure 5C). Because these same dependent variables correlated with age in the people with schizophrenia only, we ran partial correlations with age and found that only SDHA mRNA retained a relationship with duration of illness at a trend level (r = −0.32, p = 0.066).

Figure 5.

The clinical variables that correlated with brainweight were (A) daily chlorpromazine equivalent neuroleptic dose (CPZ) (r = −0.34, p = 0.038, n = 37) and (B) duration of illness (r = −0.35, p = 0.031, n = 37). (C) Duration of illness negatively correlated with the geomean(4) (r = −0.35, p = 0.034, n = 36).

Power analysis

Power calculations, performed using means and standard deviations of housekeeping genes in controls and patients with schizophrenia, were used to determine the sample sizes required to reject a null hypothesis of no difference in housekeeping gene expression between diagnostic groups. Power analysis indicated that the present cohort contained a sufficient sample size to detect a 1.5-fold difference in the mRNA expression of all housekeeping genes between patients and controls, with 90% statistical power (α = 0.05), requiring between six and 12 samples per group to detect this fold change. Additionally, we had sufficient sample size to detect a 1.25-fold change in expression for all of the genes at 80% power and for six of nine genes at 90% power (Table 4). To detect a 1.1-fold difference with 90% statistical power would require 141–299 samples per group or, relaxing stringency to 70% statistical power, 83–176 samples per group (Table 4).

Table 4.

Calculations of sample sizes (α = 0.05)

	Statistical power 70%			Statistical power 80%			Statistical power 90%
	1.1-fold difference	1.25-fold difference	1.5-fold difference	1.1-fold difference	1.25-fold difference	1.5-fold difference	1.1-fold difference	1.25-fold difference	1.5-fold difference
SDHA	91	15	4	116	19	5	154	25	7
PPIA	92	15	4	118	19	5	157	26	7
B2M	83	14	4	106	17	5	141	23	6
GUSB	87	14	4	111	18	5	148	24	6
ACTB	127	21	6	163	26	7	217	35	9
GAPDH	154	25	7	196	32	8	262	42	11
TBP	176	29	8	225	36	9	299	48	12
PBGD	112	18	5	143	23	6	190	31	8
UBC	143	23	6	183	30	8	244	39	10

Discussion

Some consensus concerning the effects and relative importance of premortem and postmortem factors on human brain RNA quality is emerging. We find, as others do, that RIN is a powerful indicator of mRNA levels and appears to be the most critical factor for matching schizophrenia cases and controls [6,26,27]. But because RIN measures are available only after RNA isolation, we found that tissue pH can be used as a general guide to select potential controls to match to cases prior to RNA isolation [26,27]. Because cerebellar and cortical pH correlate, determining cerebellar pH may be a valuable first step in evaluating the suitability of cases for RNA analysis [26,28], but high tissue pH does not always indicate good RNA [26,27], and low pH does not necessarily yield low-quality RNA. The current cut-off for what is an acceptable RIN varies considerably, with some studies suggesting that a moderate amount of degradation and a RIN as low as 3.8–4.0 may be usable [6,8,29]. Also, it should be mentioned that RINs are typically higher after total RNA is purified by columns, but this purification step would exclude some small RNAs of interest such as microRNAs, and thus columns were not used in the present study. Ultimately, the cut-off for a usable RIN would depend on criteria set by the end user depending on the assay used and the molecule of interest, and thus, will vary. Earlier studies emphasized the importance of matching for tissue pH and RIN, and we have prioritized tissue pH, RIN, and age as variables to control for in assembling a cohort of schizophrenia and normal brains that are pair-selected within ranges, creating a closely matched and well-powered cohort.

Most brain banks attempt to collect cases with the shortest PMI possible, although researchers have challenged the importance of PMI in determining RNA quality [6,30–33]. A number of studies suggested that PMI has no correlation or only a weak correlation with RNA quality [10,26–28]. Studies that have examined human tissue collected during neurosurgery for epilepsy with minimal delay shows that the freshly acquired brain tissue pH and RIN are comparable or even lower to that collected after death [10,26]. In agreement with these published findings, we found no correlation between PMI and most of the RNA quality measures in the DLPFC [26,34], with only one slight negative correlation with SDHA as previously described [6]. A microarray study in mouse brain tissue identified >1000 transcripts that are degradation susceptible (labile) with increasing PMI, but they found that increased PMI does not affect GADPH mRNA levels [35]. Other studies in human brain suggest that another housekeeper control gene, B2M, is highly labile [36,37]. On one hand, B2M may serve to correct for degradation levels if used to normalize genes of interest; on the other hand, we found that B2M is a fairly unstable housekeeper control gene and thus, we would not recommend it as a good control for schizophrenia studies. Other studies in humans suggest that some housekeeper transcripts are slightly influenced by PMI, especially in the hippocampus [6], although this may not be the case for all hippocampal housekeeper RNAs [28]. Thus, although effects of PMI need to be evaluated for particular transcripts of interest through correlational analyses, PMI is unlikely to be a major determinant of RNA or transcript quality and may be helpful but not critical to consider when matching samples.

In contrast, it has been well demonstrated that tissue pH influences mRNA preservation in postmortem tissue [6,32,38,39]. In the present study we found that the majority of housekeeper RNAs assayed and the geomean(4) correlated strongly with tissue pH. Studies that have measured the same housekeeper controls have also found positive correlations with brain tissue pH [6,40]. Some previous reports suggest that a pH cut-off (pH <5.9–6.0) can be used as a screen to choose which brains from which to extract RNA and analyse further [4], or that tissue pH could be used as a critical lower threshold for transcriptome analysis (pH range 6.54–6.8) [28,41]. Because most postmortem samples used here and in other current cohorts span a large pH range [6,26–28], with as many as 39–56% of subjects from psychiatric groups having a pH <6.5 [8] and as many as 28% of the present current sample, it would not be reasonable to exclude this large number of cases. Thus it is of critical importance to either pair-match for, or group (stratum) match for tissue pH. Several research groups have noted that pH is not always a reliable predictor of RNA quality and we find that the individual with the lowest brain pH (5.7) had a RIN of 6.2. Hence, we agree with the recommendation that RNA isolation and RIN quality assessment should be completed on all samples before precious material is excluded from cohorts based on tissue pH alone [6,27].

The importance of agonal state in postmortem research has been increasingly recognized over the past several decades [36,41–44]. We found that the longer the agonal state (prolonged death) the lower the brain pH; but we did not find an overall effect of agonal state on RIN. This agrees with studies that show that as agonal state scores increase, the tissue pH decreases [27,28,45], but RIN may not differ significantly with agonal state [27]. We did find a lower value for geomean(4) with higher agonal state scores. Prolonged agonal state can downregulate some mRNAs (energy metabolism genes), while increasing others (stress response and inflammation control genes) [45], suggesting that agonal state is an important factor to consider when matching samples or groups of samples.

Studies of the effects of aging on housekeeper gene expression in the human brain have led to the conclusion that age at death influences mRNA expression, especially when individuals span a large age range [6,40]. In the present study, individuals spanned six decades of life with an average age of 51, which is comparable to several other cohorts used to study people with schizophrenia compared to controls [6,46]. One of the surprising findings of the present study was that five of nine housekeeper control genes and the geomean(4) negatively correlated with age in the schizophrenia patients only. We also found that women but not men with schizophrenia had significantly smaller brains with increasing age. Lighter brains in both sexes with schizophrenia has been reported previously [47], but the neurobiological underpinnings of this are not understood. We also found that overall brainweight and brain volume negatively correlated with the mean daily dose of neuroleptic medication and duration of illness, suggesting that the longer someone was ill or the higher the dose of daily neuroleptics, the lower the brainweight and the smaller the brain at death. Patients with a longer duration of illness tended to have higher average lifetime antipsychotic dose, higher antipsychotic dose at the time of death, and lower housekeeper mRNA expression. Because the clinical measure of duration of illness is linked to age at death, we attempted to control for the effect of age by using partial correlation and found that SDHA mRNA correlated negatively with duration of illness at a trend level. The fact that the correlations between duration of illness and housekeeper gene mRNAs are no longer statistically significant is difficult to interpret because age is not independent from duration of illness and thus some of the true variance coming from the duration of illness will be subtracted in the partial correlation for age. This is also true for earlier reports that partialled out the effect of age and failed to find an effect of duration of illness on brainweight using meta-analysis [47]. The fact that older people with schizophrenia with a longer duration of illness have smaller brains and have decreased control mRNAs suggests that the longer someone lives with schizophrenia the greater the chance that they have accumulated deleterious impacts on their brain. This observation, although made on cross-sectional data across the lifespan, fits well with volumetric studies showing decreased grey matter volume over time in people suffering from chronic schizophrenia [48,49]. The present data suggest that some of these deleterious effects may be linked to antipsychotic medication effects, or at the very least are not prevented by current treatment strategies. Recent reports suggest that chronic antipsychotic use (both typical and atypical) causes a 10% decrease in brainweight and does in fact decrease cortical brain volume in controlled stereological studies of non-human primates [50,51]. This volume reduction was linked to a significant reduction in astrocytes and oligodendrocytes, a finding that parallels the reduction in astrocytic and oligodendrocyte markers found in some [52–55] but not all [56,57] postmortem studies of people with schizophrenia.

In the present study we found that two housekeeper genes, β-actin and GAPDH mRNAs, do not differ between people with schizophrenia compared to control as previously found [58,59], and the present data suggest that both are particularly stable mRNAs and useful housekeeping control genes for postmortem human brain studies. It should be emphasized, however, that the utility of GAPDH as a control varies from brain collection to brain collection, because some laboratories find GAPDH to be stable and to not differ in people with schizophrenia as compared to controls, as we have done [59,60], while others suggest that it is one of the least stable and thus, least reliable control genes [22,61]. We found that three putative housekeepers, GUSB, PPIA, and SDHA, are lower in people with schizophrenia. PPIA, and SDHA mRNA differences reached statistical significance and both of these housekeeper mRNAs may be regulated by cellular stress. PPIA encodes a member of the peptidyl-prolyl cis-trans isomerase (PPIase) family, which is a cyclosporin binding protein and can accelerate the folding of proteins. PPIA mRNA is not typically reduced in patients with schizophrenia in other cohorts [60,62–64]. Thus, the reductions in PPIA mRNA that we have found may represent increased power to detect a difference (approx. 10% reduction) or may represent a finding specific to the present cohort. The other transcript we found to be decreased, SDHA, encodes a major catalytic subunit of succinate-ubiqui-none oxidoreductase, part of the mitochondrial respiratory chain complex II on the mitochondrial inner membrane. Mutations in SDHA have been associated with a form of mitochondrial respiratory chain deficiency known as Leigh syndrome and with late-onset optic atrophy [65,66]. Dysfunction of mitochondrial respiration can lead to cognitive decline in attention and mental flexibility because respiratory chain diseases most often affect muscle and the central nervous system (CNS) [67]. CNS symptoms can include epilepsy, demyelination, dementia and psychosis. Oxidative stress related to alterations in metabolism linked to mitochondrial distress has been observed in schizophrenia [43,68–70]. GUSB, or glucuronidase-β, mRNA encodes a protein that is required to degrade glycosaminoglycans (i.e. heparin sulphate), and genetic changes in GUSB can result in deafness, behavioural deficits and mental retardation [71]. Thus, although certain genes are chosen to represent the integrity of the transcript pool and are thus used as housekeeper control or reference genes, they can have a critical function in brain as evidenced when mutated. The question regarding whether slightly lowered reference gene transcript levels have any functional significance requires further investigation. The present study suggests that GUSB, PPIA and SDHA mRNAs are not optimal choices for reference genes in schizophrenia studies.

Taken together, the present data suggest that although only a small amount of variance in RNA quality or housekeeper mRNA level is explained by any one demographic (age) or tissue characteristic (pH), their effects are consistent across many laboratories and represent known confounders/covariates that can be controlled for when constructing postmortem cohorts. We suggest that matching diagnostic groups minimally for age, pH and RIN may help to resolve differences in gene expression that occur in schizophrenia as either a cause, a compensation or a consequence of the disease state [17].

Footnotes

Acknowledgements

Since its establishment, Schizophrenia Research Institute (SRI) has supported schizophrenia research across a broad range of disciplines and has also been a major supporter of the New South Wales Tissue Resource Centre, an infrastructure facility that collects, stores and distributes fixed and frozen postmortem human brain tissue for projects related to schizophrenia. This work was supported by Schizophrenia Research Institute, utilizing infrastructure funding from NSW Health, University of New South Wales School of Psychiatry, and Prince of Wales Medical Research Institute. We thank Dusan Hadzi-Pavlovic for statistical advice.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.

Methods and Materials

References

Kleinman

Hyde

Herman

. Methodological issues in the neuropathology of mental illness. Bloom

Kupfer

, Psychopharmacology: the fourth generation of progress. New York: Raven Press, 1995:859–864.

Lewis

. The human brain revisited: opportunities and challenges in postmortem studies of psychiatric disorders. Neuropsychopharmacology 2002; 26:143–154.

Ravid

Van Zwieten

Swaab

. Brain banking and the human hypothalamus: factors to match for, pitfalls and potentials. Prog Brain Res 1992; 93:83–95.

Bahn

Augood

Ryan

Standaert

Starkey

Emson

. Gene expression profiling in the post-mortem human brain: no cause for dismay. J Chem Neuroanat 2001; 22:79–94.

Bunney

Vawter

. Microarray technology: a review of new strategies to discover candidate vulnerability genes in psychiatric disorders. Am J Psychiatry 2003; 160:657–666.

Lipska

Deep-Soboslay

Weickert

. Critical factors in gene expression in postmortem human brain: focus on studies in schizophrenia. Biol Psychiatry 2006; 60:650–658.

Siegmund

Connor

Campan

. DNA methylation in the human cerebral cortex is dynamically regulated throughout the life span and involves differentiated neurons. PLoS ONE 2007; 2:e895.

Weis

Llenos

Dulay

Elashoff

Martinez-Murillo

Miller

. Quality control for microarray analysis of human brain samples: the impact of postmortem factors, RNA characteristics, and histopathology. J Neurosci Methods 2007; 165:198–209.

Monoranu

Apfelbacher

Grunblatt

. pH measurement as quality control on human postmortem brain tissue: a study of the BrainNet Europe Consortium. Neuropathol Appl Neurobiol 2009; 35:329–337.

10.

Chevyreva

Faull

Green

Nicholson

. Assessing RNA quality in postmortem human brain tissue. Exp Mol Pathol 2008; 84:71–77.

11.

Johnston

Cervenak

Shore

Torrey

Yolken

. Multivariate analysis of RNA levels from postmortem human brains as measured by three different methods of RT-PCR. Stanley Neuropathology Consortium. J Neurosci Methods 1997; 77:83–92.

12.

Harrison

Weinberger

. Schizophrenia genes, gene expression, and neuropathology: on the matter of their convergence. Mol Psychiatry 2005; 10:40–68.

13.

Mirnics

Middleton

Marquez

Lewis

Levitt

. Molecular characterization of schizophrenia viewed by microarray analysis of gene expression in prefrontal cortex. Neuron 2000; 28:53–67.

14.

Perlman

Weickert

Akil

Kleinman

. Postmortem investigations of the pathophysiology of schizophrenia: the role of susceptibility genes. J Psychiatry Neurosci 2004; 29:287–293.

15.

Pongrac

Middleton

Lewis

Levitt

Mirnics

. Gene expression profiling with DNA microarrays: advancing our understanding of psychiatric disorders. Neurochem Res 2002; 27:1049–1063.

16.

Tamminga

Holcomb

. Phenotype of schizophrenia: a review and formulation. Mol Psychiatry 2005; 10:27–39.

17.

Lewis

Gonzalez-Burgos

. Neuroplasticity of neocortical circuits in schizophrenia. Neuropsychopharmacology 2008; 33:141–165.

18.

Draganic

Catts

Carr

. Neuroscience Institute of Schizophrenia and Allied Disorders (NISAD): 10 years of Australia's first virtual research institute. Aust N Z J Psychiatry 2007; 41:78–88.

19.

Kozlovsky

Shanon-Weickert

Tomaskovic-Crook

Kleinman

Belmaker

Agam

. Reduced GSK-3beta mRNA levels in postmortem dorsolateral prefrontal cortex of schizophrenic patients. J Neural Transm 2004; 111:1583–1592.

20.

Schroeder

Mueller

Stocker

. The RIN: an RNA integrity number for assigning integrity values to RNA measurements. BMC Mol Biol 2006;7:3.

21.

Law

Kleinman

Weinberger

Weickert

. Disease-associated intronic variants in the ErbB4 gene are related to altered ErbB4 splice-variant expression in the brain in schizophrenia. Hum Mol Genet 2007; 16:129–141.

22.

Silberberg

Baruch

Navon

. Detection of stable reference genes for real-time PCR analysis in schizophrenia and bipolar disorder. Anal Biochem 2009; 391:91–97.

23.

Wong

Weickert

. Transcriptional interaction of an estrogen receptor splice variant and ErbB4 suggests convergence in gene susceptibility pathways in schizophrenia. J Biol Chem 2009; 284:18 824–18 832.

24.

Vandesompele

De Preter

Pattyn

. Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol 2002; 3 (7): RESEARCH0034.

25.

Bustin

Benes

Nolan

Pfaffl

. Quantitative real-time RT-PCR: a perspective. J Mol Endocrinol 2005; 34:597–601.

26.

Stan

Ghose

Gao

. Human postmortem tissue: what quality markers matter? Brain Res 2006; 1123:1–11.

27.

Webster

. Tissue preparation and banking. Prog Brain Res 2006; 158:3–14.

28.

Mexal

Berger

Adams

Ross

Freedman

Leonard

. Brain pH has a significant impact on human postmortem hippocam-pal gene expression profiles. Brain Res 2006; 1106:1–11.

29.

Schoor

Weinschenk

Hennenlotter

. Moderate degradation does not preclude microarray analysis of small amounts of RNA. Biotechniques 2003; 35:1192–1196, 1198–1201.

30.

Barton

Pearson

Najlerahim

Harrison

. Pre- and postmortem influences on brain RNA. J Neurochem 1993; 61: 1–11.

31.

Ervin

Heinzen

Cronin

. Postmortem delay has minimal effect on brain RNA integrity. J Neuropathol Exp Neurol 2007; 66:1093–1099.

32.

Kingsbury

Foster

Nisbet

. Tissue pH as an indicator of mRNA preservation in human post-mortem brain. Brain Res Mol Brain Res 1995; 28:311–318.

33.

Schramm

Falkai

Tepest

. Stability of RNA transcripts in post-mortem psychiatric brains. J Neural Transm 1999; 106:329–335.

34.

Trotter

Brill

II Bennett

. Stability of gene expression in postmortem brain revealed by cDNA gene array analysis. Brain Res 2002; 942:120–123.

35.

Catts

Fernandez

Taylor

Coulson

Lutze-Mann

. A microarray study of post-mortem mRNA degradation in mouse brain tissue. Brain Res Mol Brain Res 2005; 138:164–177.

36.

Atz

Walsh

Cartagena

. Methodological considerations for gene expression profiling of human brain. J Neurosci Methods 2007; 163:295–309.

37.

Auer

Lyianarachchi

Newsom

Klisovic

Marcucci

Kornacker

. Chipping away at the chip bias: RNA degradation in microarray analysis. Nat Genet 2003; 35:292–293.

38.

Harrison

Heath

Eastwood

Burnet

McDonald

Pearson

. The relative importance of premortem acidosis and postmortem interval for human brain gene expression studies: selective mRNA vulnerability and comparison with their encoded proteins. Neurosci Lett 1995; 200:151–154.

39.

Miller

Diglisic

Leister

Webster

Yolken

. Evaluating RNA status for RT-PCR in extracts of postmortem human brain tissue. Biotechniques 2004; 36:628–633.

40.

Preece

Cairns

. Quantifying mRNA in postmortem human brain: influence of gender, age at death, postmortem interval, brain pH, agonal state and inter-lobe mRNA variance. Brain Res Mol Brain Res 2003; 118:60–71.

41.

Meng

Tsavaler

. Sample matching by inferred agonal stress in gene expression analyses of the brain. BMC Genomics 2007;8:336.

42.

Tomita

Vawter

Walsh

. Effect of agonal and postmortem factors on gene expression profile: quality control in microarray analyses of postmortem human brain. Biol Psychiatry 2004; 55:346–352.

43.

Vawter

Tomita

Meng

. Mitochondrial-related gene expression changes are sensitive to agonal-pH state: implications for brain disorders. Mol Psychiatry 2006; 11:615, 663–679.

44.

Hardy

Wester

Winblad

Gezelius

Bring

Eriksson

. The patients dying after long terminal phase have acidotic brains; implications for biochemical measurements on autopsy tissue. J Neural Transm 1985; 61:253–264.

45.

Vawter

Walsh

. Systematic changes in gene expression in postmortem human brains associated with tissue pH and terminal medical conditions. Hum Mol Genet 2004; 13:609–616.

46.

Hashimoto

Arion

Unger

. Alterations in GABA-related transcriptome in the dorsolateral prefrontal cortex of subjects with schizophrenia. Mol Psychiatry 2008; 13:147–161.

47.

Harrison

Freemantle

Geddes

. Meta-analysis of brain weight in schizophrenia. Schizophr Res 2003; 64:25–34.

48.

DeLisi

Sakuma

Maurizio

Relja

Hoff

. Cerebral ventricular change over the first 10 years after the onset of schizophrenia. Psychiatry Res 2004; 130:57–70.

49.

DeLisi

Sakuma

Tew

Kushner

Hoff

Grimson

. Schizophrenia as a chronic active brain process: a study of progressive brain structural change subsequent to the onset of schizophrenia. Psychiatry Res 1997; 74:129–140.

50.

Dorph-Petersen

Pierri

Perel

Sun

Sampson

Lewis

. The influence of chronic exposure to antipsychotic medications on brain size before and after tissue fixation: a comparison of haloperidol and olanzapine in macaque monkeys. Neuropsychopharmacology 2005; 30:1649–1661.

51.

Konopaske

Dorph-Petersen

Pierri

Sampson

Lewis

. Effect of chronic exposure to antipsychotic medication on cell numbers in the parietal cortex of macaque monkeys. Neuropsychopharmacology 2007; 32:1216–1223.

52.

Johnston-Wilson

Sims

Hofmann

. Disease-specific alterations in frontal cortex brain proteins in schizophrenia, bipolar disorder, and major depressive disorder. The Stanley Neuropathology Consortium. Mol Psychiatry 2000; 5:142–149.

53.

Steffek

McCullumsmith

Haroutunian

Meador-Woodruff

. Cortical expression of glial fibrillary acidic protein and glutamine synthetase is decreased in schizophrenia. Schizophr Res 2008; 103:71–82.

54.

Toro

Hallak

Dunham

Deakin

. Glial fibrillary acidic protein and glutamine synthetase in subregions of prefrontal cortex in schizophrenia and mood disorder. Neurosci Lett 2006; 404: 276–281.

55.

Webster

O'Grady

Kleinman

Weickert

. Glial fibrillary acidic protein mRNA levels in the cingulate cortex of individuals with depression, bipolar disorder and schizophrenia. Neuroscience 2005; 133:453–461.

56.

Damadzic

Bigelow

Krimer

. A quantitative immunohistochemical study of astrocytes in the entorhinal cortex in schizophrenia, bipolar disorder and major depression: absence of significant astrocytosis. Brain Res Bull 2001; 55:611–618.

57.

Dean

Gray

Scarr

. Regionally specific changes in levels of cortical S100beta in bipolar 1 disorder but not schizophrenia. Aust N Z J Psychiatry 2006; 40:217–224.

58.

Dean

Keriakous

Scarr

Thomas

. Gene expression profiling in Brodmann's area 46 from subjects with schizophrenia. Aust N Z J Psychiatry 2007; 41:308–320.

59.

Lauriat

Dracheva

Chin

Schmeidler

McInnes

Haroutunian

. Quantitative analysis of glutamate transporter mRNA expression in prefrontal and primary visual cortex in normal and schizophrenic brain. Neuroscience 2006; 137:843–851.

60.

O'Connor

Muly

Arnold

Hemby

. AMPA receptor subunit and splice variant expression in the DLPFC of schizophrenic subjects and rhesus monkeys chronically administered antipsychotic drugs. Schizophr Res 2007; 90:28–40.

61.

Johansson

Fuchs

Okvist

. Validation of endogenous controls for quantitative gene expression analysis: application on brain cortices of human chronic alcoholics. Brain Res 2007; 1132:20–28.

62.

Hashimoto

Straub

Weickert

Hyde

Kleinman

Weinberger

. Expression analysis of neuregulin-1 in the dorso-lateral prefrontal cortex in schizophrenia. Mol Psychiatry 2004; 9:299–307.

63.

Weickert

Straub

McClintock

. Human dysbindin (DTNBP1) gene expression in normal brain and in schizophrenic prefrontal cortex and midbrain. Arch Gen Psychiatry 2004; 61:544–555.

64.

Weickert

Webster

Hyde

. Reduced GAP-43 mRNA in dorsolateral prefrontal cortex of patients with schizophrenia. Cereb Cortex 2001; 11:136–147.

65.

Bayley

Devilee

Taschner

. The SDH mutation database: an online resource for succinate dehydrogenase sequence variants involved in pheochromocytoma, paraganglioma and mitochondrial complex II deficiency. BMC Med Genet 2005;6:39.

66.

Horvath

Abicht

Holinski-Feder

. Leigh syndrome caused by mutations in the flavoprotein (Fp) subunit of succinate dehydrogenase (SDHA). J Neurol Neurosurg Psychiatry 2006; 77:74–76.

67.

Finsterer

. Cognitive decline as a manifestation of mitochondrial disorders (mitochondrial dementia). J Neurol Sci 2008; 272:20–33.

68.

Holmes

Tsang

Huang

. Metabolic profiling of CSF: evidence that early intervention may impact on disease progression and outcome in schizophrenia. PLoS Med 2006; 3:e327.

69.

Middleton

Mirnics

Pierri

Lewis

Levitt

. Gene expression profiling reveals alterations of specific metabolic pathways in schizophrenia. J Neurosci 2002; 22:2718–2729.

70.

Mimmack

Brooking

Bahn

. Quantitative polymerase chain reaction: validation of microarray results from postmortem brain studies. Biol Psychiatry 2004; 55:337–345.

71.

Tomatsu

Montano

Dung

Grubb

Sly

. Mutations and polymorphisms in GUSB gene in mucopolysaccharidosis VII (Sly Syndrome). Hum Mutat 2009; 30:511–519.

72.

Deep-Soboslay

Akil

Martin

Bigelow

Herman

Hyde

Kleinman

. Reliability of psychiatric diagnosis in postmortem research. Biol Psychiatry 2005; 57:96–101.

73.

Garrick

Howell

Terwee

Redenbach

Blake

Harper

. Brain donation for research: who donates and why? J Clin Neurosci 2006; 13:524–8.

74.

Hill

Keks

Roberts

Opeskin

Dean

MacKinnon

Copolov

. Problem of diagnosis in postmortem brain studies of schizophrenia. Am J Psychiatry 1996; 153:533–7.

75.

Sundqvist

Garrick

Bishop

Harper

. Reliability of postmortem psychiatric diagnosis for neuroscience research. Aust N Z J Psychiatry 2008; 42:221–7.

76.

Sards

Garrick

Sheedy

Harper

. Banking for the future: an Australian experience in brain banking. Pathology 2002; 34:225–9.

77.

Sheedy

Garrick

Dedova

Hunt

Miller

Sundqvist

Harper

. An Australian Brain Bank: a critical investment with a high return! Cell Tissue Bank 2008; 9:205–16.

78.

Hardy

Wester

Winblad

Gezelius

Bring

Eriksson

. The patients dying after long terminal phase have acidotic brains; implications for biochemical measurements on autopsy tissue. J Neural Transm 1985; 61:253–64.

79.

Tomita

Vawter

Walsh

Evans

Choudary

Overman

Atz

Myers

Jones

Watson

Akil

Bunney

Jr . Effect of agonal and postmortem factors on gene expression profile: quality control in microarray analyses of postmortem human brain. Biol Psychiatry 2004; 55:346–52.

80.

Harper

Kril

Daly

. Brain shrinkage in alcoholics is not caused by changes in hydration: a pathological study. J Neurol Neurosurg Psychiatry 1988; 51:124—7.

81.

Harrison

Heath

Eastwood

Burnet

McDonald

Pearson

82.

Kingsbury

Foster

Nisbet

Cairns

Bray

Eve

Lees

Marsden

. Tissue pH as an indicator of mRNA preservation in human post-mortem brain. Brain Res Mol Brain Res 1995;28:311–8.

83.

Stan

Ghose

Gao

Roberts

Lewis-Amezcua

Hatanpaa

Tamminga

. Human postmortem tissue: what quality markers matter? Brain Res 2006; 1123:1–11.

Selection of Reference Gene Expression in a Schizophrenia Brain Cohort

Abstract

Keywords

Methods

Subjects

RNA extraction and quality assessment

Quantitative reverse transcription-polymerase chain reaction

Statistical analysis

Results

Matching of diagnostic groups

Relationship of demographic variables to each other and RIN

Transcript stability and calculation of geometric mean

Correlation of demographic variables and housekeeper mRNAs

Age

pH

Agonal state

RNA integrity number

Postmortem interval

Diagnostic differences in housekeeper control mRNA

T-tests

Effects of hemisphere and gender on geomean(4)

Consideration of clinical variables

Power analysis

Discussion

Footnotes

Acknowledgements

Methods and Materials

References