Similarities and Differences in Peripheral Blood Gene Expression Signatures of Individuals with Schizophrenia and their First-Degree Biological Relatives

Abstract

Several studies have evaluated the potential utility of blood-based whole-transcriptome signatures as a source of biomarkers for schizophrenia. This endeavor has been complicated by the fact that individuals with schizophrenia typically differ from appropriate comparison subjects on more than just the presence of the disorder; for example, individuals with schizophrenia typically receive antipsychotic medications, and have been dealing with the sequelae of this chronic illness for years. The inability to control such factors introduces a considerable degree of uncertainty in the results to date. To overcome this, we performed a blood-based gene-expression profiling study of schizophrenia patients (n=9) as well as their unmedicated, nonpsychotic, biological siblings (n=9) and unaffected comparison subjects (n=12). The unaffected biological siblings, who may harbor some of the genetic predisposition to schizophrenia, exhibited a host of gene-expression differences from unaffected comparison subjects, many of which were shared by their schizophrenic siblings, perhaps indicative of underlying risk factors for the disorder. Several genes that were dysregulated in both individuals with schizophrenia and their siblings related to nucleosome and histone structure and function, suggesting a potential epigenetic mechanism underlying the risk state for the disorder. Nonpsychotic siblings also displayed some differences from comparison subjects that were not found in their affected siblings, suggesting that the dysregulation of some genes in peripheral blood may be indicative of underlying protective factors. This study, while exploratory, illustrated the potential utility and increased informativeness of including unaffected first-degree relatives in research in pursuit of peripheral biomarkers for schizophrenia.

INTRODUCTION

The identification of genes that increase susceptibility to schizophrenia, and the biological and environmental mechanisms through which they act, remain among the most challenging issues in neuropsychiatric research. Progress in mapping the human genome increased the viability of candidate gene association studies which, while important, have focused largely on genes in biological signaling systems that are already widely implicated in schizophrenia, such as dopamine or glutamate transmission (Glatt and others 2007). A new era of genome-wide association studies has advanced the field further, in part by implicating genes that were not suspected of roles in the disorder previously, such as zinc-finger proteins and calcium channels, while also substantiating some prior candidate genes in the major histocompatibility region on chromosome 6p (O'Donovan and others 2008; Shi and others 2009; Stefansson and others 2009). Such studies may ultimately explain much of the heritable portion of the liability toward schizophrenia, which is estimated to range from 60–85% (Cardno and others 1999; Sullivan and others 2003); however, a considerable fraction of the variance in who actually becomes affected will remain unexplained until the effects of environmental factors and gene-environment interactions can be systematically integrated into genomic research.

Gene expression, as a general index of genomic functionality, may be useful in this regard as a final common pathway in which the effects of genetic and environmental risk factors converge. In an earlier study, we derived messenger RNA (mRNA) expression patterns in circulating peripheral blood samples from patients with schizophrenia or bipolar disorder and non-mentally ill control subjects (Tsuang and others 2005), which allowed us to distinguish a panel of relatively sensitive and specific biomarkers. Subsequently, we compared our blood-based biomarker set against a list of genes found to be dysregulated in schizophrenia in postmortem dorsolateral prefrontal cortex, a brain structure often implicated in the disorder, finding six putative risk genes that might also have biomarker potential (Glatt and others 2005b).

These findings are encouraging, but they raise another set of critical questions about the vulnerability to schizophrenia. Among these are the issues of the extent to which our initial findings reflected the true biological susceptibility toward then disorder versus the effects of treatment (e.g., antipsychotic medications) or other, less specific manifestations of mental illness (e.g., psychosis). One potentially effective way of disentangling these effects is to study non-psychotic, biological relatives of individuals with schizophrenia, together with their ill siblings, and with non-mentally ill community comparison subjects. This approach would allow us to determine, for example, whether “unaffected” relatives who are neither psychotic nor taking antipsychotic medication still differ from comparison subjects, or whether relatives and individuals with schizophrenia share dysregulated genes compared to comparison subjects. We utilized this strategy in the present pilot study with subjects who were recruited from the Harvard University/Beth Israel Deaconess Medical Center site of the Consortium on the Genetics of Schizophrenia (COGS) (Calkins and others 2007).

METHODS

Ascertainment and Clinical Characterization of Subjects

The COGS is a seven-site project, funded by the U.S. National Institutes of Health, which was designed to assess potential schizophrenia endophenotypes (e.g., social, psychophysiological, neurochemical or neuropsychological abnormalities) and perform genetic analyses on affected individuals (SZs), their unaffected biological siblings (SIBs) and other first-degree relatives, and community comparison subjects (CCSs) (Calkins and others 2007). Subjects in our study however, were ascertained only from one of the seven sites that participated in the COGS: the Harvard University site at the Massachusetts Mental Health Center (MMHC) Public Psychiatry Division of the Beth Israel Deaconess Medical Center (BIDMC). The institutional review boards of the MMHC and BIDMC approved the study, and all subjects signed informed consent.

The COGS methods will be summarized briefly as they have been described previously (Calkins and others 2007). Consortium-wide quality assurance procedures were exercised throughout the study. Each site followed an identical protocol to recruit, diagnose, assess endophenotypes, and collect blood samples for DNA analysis; in addition to these procedures, we introduced our previously validated protocol for mRNA analysis in peripheral blood. Medically healthy adults were recruited through flyers, print, and electronic media, and through community presentations. Individuals with schizophrenia were also referred by mental health providers. Eligible families had one of two pedigree structures. The first required availability of both of the proband’s parents, at least one of whom was not psychotic, and at least one non-psychotic sibling. The second included availability of one parent and at least two non-psychotic siblings. All subjects were administered a modified version of the Diagnostic Interview for Genetic Studies (Nurnberger and others 1994), the Family Interview for Genetic Studies (NIMH Genetics Initiative 1992), other clinical measures (see Calkins et al., 2007), and a medical record review. Premorbid IQ was estimated using the Wide Range Achievement Test, Third Edition (WRAT-3) Reading subtest (Jastak and Wilkinson 1993). All probands met DSM-IV diagnostic criteria for schizophrenia and were stable clinically (i.e., no psychiatric hospitalizations in the previous month).

Subjects who were eligible for this gene expression study were 18–65 years old and fluent in English. Other exclusion criteria included: 1) a history of electroconvulsive therapy in the past 6 months; 2) a positive drug or alcohol test result during study screening; 3) a diagnosis of a substance abuse disorder in the past 30 days or active substance dependence in the past six months; 4) an estimated premorbid IQ<70; 5) a history of head injury with loss of consciousness exceeding 15 minutes; 6) a seizure disorder; 7) any ocular, neurological, or systemic medical problem likely to cause neurocognitive or psychophysiological performance deficits; or 8) inability to provide informed consent. CCS subjects were also excluded if they had a history of any DSM-IV Cluster A personality disorder, psychosis, or a family history of psychosis among first- or second-degree relatives.

For the present study, we ascertained and completed assessments of 32 subjects, including 8 SZs, 12 unaffected SIBs, and 12 CCSs. Despite the fact that no subjects dropped out of the study, our final sample size was somewhat smaller (n=26) due to our own data-filtering which excluded six subjects. First, two siblings had no corresponding proband in the sample and so were eliminated. Second, to simplify the design of our analyses, we included only one unaffected SIB for each SZ; yet, in two of the included families, there were two eligible SIBs for each SZ. In one of these families, one of the “unaffected” SIBs was found to have been diagnosed previously with major depressive disorder while the other unaffected SIB in this family had no history of any mental illness; thus, we excluded the former SIB and included the latter SIB in our analyses. In the second family with multiple unaffected SIBs, we found no clinical basis to favor inclusion of one of the SIBs over the other, and thus we elected to include the SIB whose profile of gene expression quality metrics most closely resembled that of their SZ relative. Finally, in one additional family, the “unaffected” SIB of one SZ subject was found to have been diagnosed previously with bipolar disorder; thus, we removed this entire family (both the SIB and the related SZ) from our analyses, leaving seven sibling-pairs and 12 CCSs. All probands were treated with antipsychotic medication at the time of testing, primarily clozapine, olanzapine, or quietiapine, whereas none of the CCSs or SIBs currently or recently received antipsychotic or other psychotropic medications.

mRNA Sample Acquisition, Stabilization, Isolation, and Storage

Approximately 15ml of blood were collected from each participant after overnight fasting, using a Vacutainer™ tube (Becton Dickinson; Franklin Lakes, NJ; USA). The collected blood samples were immediately stored on ice until mRNA was extracted, which always occurred within six hours after the blood was drawn. Red blood cells were ruptured with hypotonic hemolysis buffer (1.6mM EDTA, 10mM KHCO₃, 153mM NH₄C1, pH 7.4), and peripheral blood mononuclear cells (PBMCs) were collected by centrifugation. PBMC total RNA was extracted with Trizol^® Reagent (Invitrogen, Carlsbad, CA, USA), according to the manufacturer's instructions.

mRNA Quantitation and Quality Assurance

Data from four subjects (including two SZs and two SIBs) did not meet mRNA quality-control standards and thus these samples were removed from further analysis. The removal of the two SZs due to poor mRNA quality caused two additional SIBs to be left with no corresponding SZ relative in the dataset, and as such these SIBs were also removed from consideration. This left a final sample for analysis of 12 CCSs, 7 SZs, and 7 SIBs.

Five micrograms of total RNA from each sample was used for hybridization on an Affymetrix GeneChip Human Genome U133 Plus 2.0 microarray following the manufacturer’s instructions. Gene expression intensities were imported into GeneSpring v7.3.1 software (Agilent Technologies, Palo Alto, CA, USA) for analysis. The quality of the hybridization of each transcript was assessed by using the cross-gene error model, and measurements with a base/proportional ratio lower than 9.6 in more than 50% of hybridizations were removed prior to subsequent analysis. Genes showing inconsistent annotation provided by Affymetrix (www.affymetrix.com/products/arrays/specific/hgu133plus.affx) and SOURCE (genome-www5.stanford.edu/cgi-bin/source/sourceBatchSearch) were also removed from subsequent analyses.

Microarray Data Import, Normalization, Transformation, Summarization, and Analyses

Partek Genomics Suite software, version 6.5 (Partek Incorporated; St. Louis, MO), was utilized for all analytic procedures performed on microarray scan data. First, interrogating probes on the microarray were imported. Next, corrections for background signal were applied using the robust multi-array average (RMA) method (Irizarry and others 2003), with further adjustments for the GC-content of probes. The set of GeneChips was standardized using quantile normalization, and expression levels of each probe underwent log-2 transformation to yield distributions of data that more closely approximated normality. As each transcript was typically measured by multiple probe sets, summarization of redundant probe sets was obtained by median polish. According to convention (Handran and others 2002), probe sets with a maximum signal:noise ratio of less than 3.0 were excluded from subsequent analyses.

Transformed, normalized, and summarized gene-expression intensity values from each subject were utilized in three orthogonal sets of comparisons of diagnostic groups, as follows: 1) SZ vs. CCS; 2) SIB vs. CCS; and 3) SZ vs. SIB. These comparisons were made using analyses of variance (ANOVAs), with diagnostic group and any distinguishing demographics as fixed factors; when comparing SZs and SIBs, family ID was also modeled as a fixed factor.

After all quality-control procedures were executed, 54,675 probes of full-length gene transcripts were included in the analyses. The type-I-error rate (α) in each initial two-group comparison was set at 0.05; however, due to the large number of statistical tests to be performed, the probability of committing type-I errors (i.e., finding false-positive results) in this study was greatly inflated. We addressed this threat in three ways. First, we utilized intersection-union tests (IUTs) on sets of nominally significant (p<0.05) results, an approach which has been shown to be relatively (if not overly) conservative for identifying shared effects across conditions (Deng and others 2008). Second, we reduced the data (and the corresponding number of tests) through secondary analyses of groups of genes: after performing the IUTs, the generated lists of significantly dysregulated genes were subjected to the DAVID algorithm (Dennis and others 2003) to determine if they were enriched for genes that disproportionately represented biological “terms”. Specifically, we evaluated if each list of genes was enriched for genes that aggregated in the same functional categories (defined by Clusters of Orthologous Groups [COG] (Tatusov and others 2000) ontologies, Protein Information Resource [PIR] (Wu and others 2003) keywords, and Universal Protein Resource [UniProt] (Apweiler and others 2004) features), represented similar ontologies (defined by the Gene Ontology Consortium [GOC] (Ashburner and others 2000)), participated in the same biological pathways (defined by BioCarta and the Kyoto Encyclopedia of Genes and Genomes [KEGG] (Kanehisa and Goto 2000)), or exhibited common protein domains (defined by the Integrative Protein Signature database [InterPro] (Hunter and others 2009), PIR, or the Simple Modular Architecture Research Tool [SMART] (Schultz and others 1998)). Third, we applied a Bonferroni correction to the p-values obtained in the enrichment analyses of these terms, only considering significant those tests that exceeded a threshold of α=0.05/the number of terms evaluated in a particular category.

RESULTS

Demographics

The three groups of subjects were comparable in age, with means (±standard deviations) of 43.0±11.4 years for CCS, 39.4±11.1 for SZs, and 40.0±13.3 for SIBs (p>0.500 for all comparisons). The groups were also uniform with regard to ancestry, as all subjects self-identified as Caucasian, except for two CCS who self-identified as African-American. Exclusion of the two African-American CCS did not substantially alter the results, so these subjects were retained in order to maximize inferential power with regard to diagnosis. There was a significant gender disparity between the groups, as both the CCS and SIB groups included both male and female subjects (CCS: 9 males and 3 females; SIB: 4 males and 3 females) while all seven SZs were male (χ²₍₂₎=10.1, p=0.006). As such, we controlled for sex (but not age or ancestry, in order to preserve degrees of freedom) in all subsequent statistical models.

SZ vs. CCS

Nominally significant (p<0.05) differences in levels of expression between SZs and CCS were observed for 2155 probes, of which 1246 were up-regulated and 909 down-regulated in SZs compared to CCS. Of these 2155 probes, 1593 “known probes” were complementary to 1473 recognized protein-coding mRNAs, KIAA or FLJ cDNAs, or open reading frames (collectively referred to as “known transcripts”), while 562 “unknown probes” complemented no presently recognized functional genomic element. Several known transcripts (k=110) were tagged by two or more dysregulated probes, further substantiating the evidence of their deviation between groups. Furthermore, ADAM28, CRKRS, HNRNPC, SMARCA5, and SPON1 each had three probes significantly dysregulated in SZs, while RPRD1A had four dysregulated probes and GM2A had five. In comparison to CCS, 819 known probes were up-regulated in SZ and 774 were down-regulated.

SIB vs. CCS

Nominally significant (p<0.05) differences in levels of expression between SIBs and CCS were observed for 2176 probes, of which 1418 were up-regulated and 758 down-regulated in SIBs compared to CCSs. Of these 2176 probes, 1636 known probes were complementary to 1493 known transcripts, while the remaining 540 probes were complementary to no known transcript. Several known transcripts (k=121) were tagged by two or more dysregulated probes. Furthermore, AGPAT3, AKT2, AP1S3, BNC2, CALU, HELQ, HNRNPC, KIAA0494, KLF6, LARP4, MBP, PDE4DIP, PIP5K1A, PPARA, PTGDS, TSPAN2, and ZNF81 each had three probes significantly dysregulated in SIBs, while CCDC50 had four dysregulated probes and ASPH had five. In comparison to CCSs, 1079 known probes were up-regulated in SIBs and 557 were down-regulated.

SZ vs. SIB

Nominally significant (p<0.05) differences in levels of expression between SZs and SIBs were observed for 1992 probes, of which 955 were up-regulated and 1037 down-regulated in SZs compared to SIBs. Of these 1992 probes, 1580 known probes were complementary to 1450 known transcripts, while the remaining 412 probes were complementary to no known transcript. Several known transcripts (k=115) were tagged by two or more dysregulated probes. Furthermore, C11ORF31, CD58, CNPY2, DCLK1, EPOR, KIAA1324, LAMP2, MUM1, RAB18, RAPGEF2, and STAM2 each had three probes significantly dysregulated in SZs, and CCPG1 and RUFY3 each had four. In comparison to SIBs, 733 known probes were up-regulated in SZs and 847 were down-regulated.

Intersection-Union Tests (IUTs)

Figure 1 shows a Venn diagram depicting the numbers of probes for known transcripts that were dysregulated in each of the three orthogonal comparisons of diagnostic groups.

FIGURE 1.

Venn Diagram of Genes Dysregulated between Groups. SZ: schizophrenia group; SIB: first-degree biological sibling of SZ subject group; CCS: unrelated non-mentally ill community comparison subject group.

[SZ vs. CCS] ∩ [SIB vs. CCS]

Compared to CCSs, the SZ and SIB groups showed significant evidence (p<0.05 in both comparisons) of common dysregulation of 172 probes for 168 known transcripts (Table 1). Two of these genes (MBP and PIGV) were each tagged by two probes that were dysregulated in both SZs and SIBs compared to CCSs, and one gene (HNRNPC) was tagged by three commonly dysregulated probes. All probes except one (for SPIRE1) were dysregulated in the same direction (up- or down-regulated) in both SZs and SIBs compared to CCSs. Ninety-one probes were up-regulated in both groups, 80 were down-regulated in both groups, and SPIRE1 was up-regulated in SIBs and down-regulated in SZs compared to CCSs.

Table 1.

Genes Significantly Dysregulated in both the SZ and SIB Groups Compared to the CCS Group¹

Probeset ID	Gene Symbol	Gene Product	SZ vs. CCS		SIB vs. CCS
			p	Fold-Change	p	Fold-Change
235931_at	FAM119A	family with sequence similarity 119, member A	2.49e−03	2.67	1.87e−02	1.75
212998_x_at	HLA-DQB1/LOC100294318	major histocompatibility complex, class II, DQ beta 1	4.07e−02	1.88	3.01e−02	1.43
244546_at	CYCS	cytochrome c, somatic	4.98e−02	1.61	8.83e−03	1.52
225155_at	SNHG5	small nucleolar RNA host gene 5 (non-protein coding)	3.82e−02	1.58	1.35e−02	1.46
235678_at	GM2A	GM2 ganglioside activator	8.36e−04	1.68	9.62e−04	1.25
222347_at	LOC644450	hypothetical protein LOC644450	1.97e−02	1.48	3.90e−02	1.29
203932_at	HLA-DMB	major histocompatibility complex, class II, DM beta	2.98e−02	1.40	1.50e−02	1.26
217478_s_at	HLA-DMA/HLA-DMB	major histocompatibility complex, class II, DM alpha/beta	2.20e−02	1.38	3.86e−02	1.18
219256_s_at	SH3TC1	SH3 domain and tetratricopeptide repeats 1	3.84e−02	1.30	3.38e−02	1.20
205306_x_at	KMO	kynurenine 3-monooxygenase (kynurenine 3-hydroxylase)	1.84e−02	1.28	4.55e−03	1.20
213244_at	SCAMP4	secretory carrier membrane protein 4	4.60e−02	1.27	3.71e−02	1.19
241446_at	ADAM28	ADAM metallopeptidase domain 28	8.42e−03	1.26	1.22e−02	1.16
223349_s_at	BOK	BCL2-related ovarian killer	4.33e−02	1.18	1.35e−03	1.23
244344_at	WNK4	WNK lysine deficient protein kinase 4	5.32e−03	1.25	3.19e−02	1.15
1555872_a_at	LOC728903	hypothetical LOC728903	4.96e−02	1.17	5.00e−04	1.22
219341_at	CLN8	ceroid-lipofuscinosis, neuronal 8 (epilepsy, progressive with mental retardation)	4.61e−02	1.21	1.11e−02	1.17
214693_x_at	NBPF10	neuroblastoma breakpoint family, member 10	2.80e−02	1.23	3.66e−02	1.12
221362_at	HTR5A	5-hydroxytryptamine (serotonin) receptor 5A	4.48e−04	1.23	3.49e−02	1.11
213987_s_at	CDC2L5	cell division cycle 2-like 5 (cholinesterase-related cell division controller)	2.05e−02	1.20	3.66e−02	1.13
1555390_at	C14orf21	chromosome 14 open reading frame 21	3.11e−02	1.16	5.04e−03	1.17
1558573_at	MCTS1	malignant T cell amplified sequence 1	3.11e−02	1.19	1.04e−02	1.13
1557991_at	METTL6	methyltransferase like 6	4.38e−02	1.18	1.47e−02	1.14
219518_s_at	ELL3/SERINC4	elongation factor RNA polymerase II-like 3/serine incorporator 4	3.91e−02	1.16	4.83e−03	1.15
232042_at	TTYH2	tweety homolog 2 (Drosophila)	1.11e−02	1.22	3.23e−02	1.09
208136_s_at	MGC3771	hypothetical LOC81854	3.46e−02	1.15	9.89e−04	1.16
238316_at	ZNF567	zinc finger protein 567	5.83e−03	1.19	2.74e−02	1.11
1552671_a_at	SLC9A7	solute carrier family 9 (sodium/hydrogen exchanger), member 7	2.24e−02	1.14	2.30e−02	1.16
227298_at	FLJ37798	hypothetical gene supported by AK095117	1.64e−02	1.19	9.48e−03	1.11
221412_at	VN1R1	vomeronasal 1 receptor 1	2.97e−02	1.14	1.41e−03	1.15
1554285_at	HAVCR2	hepatitis A virus cellular receptor 2	2.60e−02	1.16	4.31e−02	1.13
1563947_a_at	ERC1	ELKS/RAB6-interacting/CAST family member 1	4.96e−02	1.16	2.08e−02	1.12
226918_at	JPH4	junctophilin 4	2.63e−03	1.20	3.70e−02	1.08
216391_s_at	KLHL1	kelch-like 1 (Drosophila)	2.53e−02	1.14	8.23e−03	1.14
205009_at	TFF1	trefoil factor 1	1.40e−02	1.15	1.89e−02	1.12
207049_at	SCN8A	sodium channel, voltage gated, type VIII, alpha subunit	1.76e−02	1.15	7.92e−03	1.12
1569074_at	FLJ37078	hypothetical protein FLJ37078	4.47e−03	1.15	8.79e−03	1.11
223926_at	KIF2B	kinesin family member 2B	1.51e−02	1.16	3.49e−02	1.10
1562581_at	LOC254028	hypothetical LOC254028	3.38e−02	1.17	4.56e−02	1.09
1570432_at	LOC100133287	hypothetical protein LOC100133287	4.15e−02	1.14	1.64e−02	1.12
221132_at	CLDN18	claudin 18	5.48e−03	1.16	1.32e−02	1.09
244500_s_at	EVI5L	ecotropic viral integration site 5-like	1.71e−02	1.14	1.04e−02	1.11
220847_x_at	ZNF221	zinc finger protein 221	3.41e−02	1.12	1.57e−03	1.13
1559277_at	FLJ35700	hypothetical protein FLJ35700	2.41e−02	1.15	3.20e−02	1.10
1555108_at	SLC10A7	solute carrier family 10 (sodium/bile acid cotransporter family), member 7	3.09e−04	1.15	1.55e−02	1.09
208115_x_at	C10orf137	chromosome 10 open reading frame 137	1.78e−02	1.14	4.28e−02	1.10
208282_x_at	DAZ1/DAZ2/DAZ3/DAZ4	deleted in azoospermia 1/2/3/4	3.61e−02	1.13	2.09e−02	1.11
215585_at	KIAA0174	KIAA0174	2.14e−02	1.14	2.62e−02	1.10
1570128_at	DDX19A	DEAD (Asp-Glu-Ala-As) box polypeptide 19A	3.47e−02	1.13	1.71e−02	1.10
210261_at	KCNK2	potassium channel, subfamily K, member 2	3.06e−02	1.15	3.44e−02	1.09
212924_s_at	LSM4	LSM4 homolog, U6 small nuclear RNA associated (S. cerevisiae)	4.34e−02	1.13	2.11e−02	1.10
227512_at	MEX3A	mex-3 homolog A (C. elegans)	2.71e−02	1.13	2.63e−02	1.09
207024_at	CHRND	cholinergic receptor, nicotinic, delta	3.77e−02	1.12	1.74e−02	1.11
220120_s_at	EPB41L4A	erythrocyte membrane protein band 4.1 like 4A	1.50e−03	1.16	2.60e−02	1.07
1561082_at	NID1	nidogen 1	7.59e−03	1.14	4.66e−02	1.08
219839_x_at	TCL6	T-cell leukemia/lymphoma 6	3.66e−03	1.12	1.62e−02	1.10
206510_at	SIX2	SIX homeobox 2	1.77e−02	1.15	1.99e−02	1.07
214466_at	GJA5	gap junction protein, alpha 5, 40kDa	6.89e−03	1.14	9.31e−03	1.08
211624_s_at	DRD2	dopamine receptor D2	2.12e−02	1.13	3.10e−02	1.09
1569570_at	AGBL4	ATP/GTP binding protein-like 4	1.07e−02	1.12	2.00e−02	1.10
1554641_a_at	TET3	tet oncogene family member 3	3.33e−02	1.12	2.97e−02	1.09
1564367_at	CXorf25	chromosome X open reading frame 25	1.75e−02	1.14	8.67e−03	1.07
231338_at	C15orf55	chromosome 15 open reading frame 55	4.72e−02	1.12	4.01e−02	1.09
236205_at	ABCC6P1	ATP-binding cassette, sub-family C, member 6 pseudogene 1	3.23e−02	1.12	3.56e−02	1.09
1552885_a_at	NKX6-3	NK6 homeobox 3	1.67e−02	1.13	3.04e−02	1.08
220191_at	GKN1	gastrokine 1	4.84e−02	1.10	9.71e−03	1.10
216914_at	CDC25C	cell division cycle 25 homolog C (S. pombe)	4.29e−02	1.12	3.38e−02	1.09
204746_s_at	PICK1	protein interacting with PRKCA 1	4.48e−02	1.12	1.02e−02	1.09
1570366_x_at	ZNF564/ZNF709	zinc finger protein 564/zinc finger protein 709	2.64e−02	1.10	1.93e−02	1.11
210195_s_at	PSG1	pregnancy specific beta-1-glycoprotein 1	2.87e−02	1.10	2.48e−02	1.10
1555131_a_at	PER3	period homolog 3 (Drosophila)	3.80e−02	1.10	2.60e−03	1.10
1558960_a_at	MFGE8	Milk fat globule-EGF factor 8 protein	2.65e−02	1.11	2.53e−02	1.09
227800_at	FAM110B	family with sequence similarity 110, member B	2.25e−02	1.11	1.10e−02	1.09
1556095_at	UNC13C	unc-13 homolog C (C. elegans)	4.99e−02	1.09	1.19e−03	1.10
213621_s_at	GUK1	Guanylate kinase 1	4.78e−02	1.12	4.77e−02	1.07
40284_at	FOXA2	forkhead box A2	8.49e−03	1.12	3.90e−02	1.07
216720_at	CYP2U1	cytochrome P450, family 2, subfamily U, polypeptide 1	6.05e−03	1.10	2.46e−02	1.08
1565580_s_at	TATDN2	TatD DNase domain containing 2	4.45e−02	1.12	2.75e−02	1.07
237939_at	EPHA5	EPH receptor A5	9.87e−03	1.11	3.14e−02	1.08
1553930_at	TAAR1	trace amine associated receptor 1	2.94e−03	1.12	4.61e−02	1.06
1562261_at	AMZ1	archaelysin family metallopeptidase 1	2.77e−02	1.10	1.14e−02	1.08
207534_at	MAGEB1	melanoma antigen family B, 1	6.74e−03	1.10	2.17e−02	1.07
211093_at	PDE6C	phosphodiesterase 6C, cGMP-specific, cone, alpha prime	5.95e−03	1.10	3.33e−03	1.07
233767_at	HHLA1	HERV-H LTR-associating 1	4.37e−02	1.09	1.88e−02	1.08
242973_at	CACNA1C	calcium channel, voltage-dependent, L type, alpha 1C subunit	8.75e−03	1.09	5.48e−03	1.07
223673_at	RFX4	regulatory factor X, 4 (influences HLA class II expression)	3.58e−02	1.06	6.90e−03	1.10
1554329_x_at	STXBP4	syntaxin binding protein 4	3.20e−02	1.08	2.62e−02	1.06
221470_s_at	IL1F7	interleukin 1 family, member 7 (zeta)	4.53e−02	1.08	1.79e−02	1.06
216492_at	KIR3DX1	killer cell immunoglobulin-like receptor, three domains, X1	4.98e−02	1.08	3.28e−02	1.06
1560469_at	NR5A2	nuclear receptor subfamily 5, group A, member 2	3.32e−02	1.07	4.07e−02	1.05
1562223_at	LOC642426	hypothetical LOC642426	2.33e−02	1.05	1.17e−03	1.06
216966_at	ITGA2B	integrin, alpha 2b (platelet glycoprotein IIb of IIb/IIIa complex, antigen CD41)	4.23e−02	1.06	1.76e−02	1.04
224995_at	SPIRE1	spire homolog 1 (Drosophila)	3.82e−02	−1.19	4.18e−02	1.13
1553533_at	JPH1	junctophilin 1	4.52e−02	−1.06	2.57e−02	−1.05
212043_at	TGOLN2	trans-golgi network protein 2	3.65e−02	−1.06	1.40e−02	−1.06
242084_at	LOC339316	hypothetical protein LOC339316	3.30e−02	−1.07	3.80e−03	−1.06
218192_at	IP6K2	inositol hexakisphosphate kinase 2	3.08e−02	−1.10	1.56e−02	−1.06
213473_at	BRAP	BRCA1 associated protein	4.31e−02	−1.11	1.87e−02	−1.07
226357_at	USP19	ubiquitin specific peptidase 19	1.48e−02	−1.11	4.44e−02	−1.07
230374_at	LOC100294358	hypothetical protein LOC100294358	3.30e−02	−1.11	2.85e−02	−1.07
204236_at	FLI1	Friend leukemia virus integration 1	2.72e−02	−1.10	4.04e−02	−1.08
212626_x_at	HNRNPC	heterogeneous nuclear ribonucleoprotein C (C1/C2)	1.74e−03	−1.12	6.46e−03	−1.07
217673_x_at	GNAS	GNAS complex locus	4.19e−02	−1.12	1.75e−02	−1.08
203351_s_at	ORC4L	origin recognition complex, subunit 4-like (yeast)	2.25e−02	−1.13	1.87e−02	−1.07
1557639_at	NFIA	Nuclear factor I/A	1.21e−02	−1.13	4.33e−02	−1.08
208398_s_at	TBPL1	TBP-like 1	1.08e−02	−1.14	4.52e−02	−1.07
204665_at	SIKE1	suppressor of IKBKE 1	3.71e−02	−1.13	2.19e−02	−1.09
223459_s_at	C1orf56	chromosome 1 open reading frame 56	4.62e−02	−1.12	1.28e−02	−1.09
202293_at	STAG1	stromal antigen 1	8.82e−03	−1.16	2.65e−02	−1.07
200014_s_at	HNRNPC	heterogeneous nuclear ribonucleoprotein C (C1/C2)	7.40e−03	−1.16	4.32e−02	−1.07
214737_x_at	HNRNPC	heterogeneous nuclear ribonucleoprotein C (C1/C2)	2.97e−03	−1.15	1.10e−02	−1.08
222212_s_at	LASS2	LAG1 homolog, ceramide synthase 2	2.20e−02	−1.11	6.80e−04	−1.13
220663_at	IL1RAPL1	interleukin 1 receptor accessory protein-like 1	2.71e−03	−1.15	2.19e−02	−1.10
222984_at	PAIP2	poly(A) binding protein interacting protein 2	1.34e−02	−1.16	1.53e−02	−1.09
230123_at	NECAP2	NECAP endocytosis associated 2	2.72e−02	−1.16	1.56e−02	−1.09
208741_at	SAP18	Sin3A-associated protein, 18kDa	1.33e−02	−1.14	3.30e−02	−1.11
219238_at	PIGV	phosphatidylinositol glycan anchor biosynthesis, class V	1.55e−03	−1.14	3.01e−04	−1.12
205160_at	PEX11A	Peroxisomal biogenesis factor 11 alpha	6.61e−03	−1.16	1.51e−02	−1.10
222867_s_at	MED31	mediator complex subunit 31	3.37e−02	−1.16	1.96e−02	−1.10
51146_at	PIGV	phosphatidylinositol glycan anchor biosynthesis, class V	3.21e−02	−1.14	3.64e−02	−1.12
226083_at	TMEM70	transmembrane protein 70	4.24e−02	−1.10	2.72e−03	−1.16
223447_at	REG4	regenerating islet-derived family, member 4	9.35e−03	−1.16	3.83e−02	−1.11
223416_at	SF3B14	splicing factor 3B, 14 kDa subunit	3.50e−02	−1.12	9.53e−03	−1.15
39549_at	NPAS2	neuronal PAS domain protein 2	2.54e−02	−1.16	1.81e−02	−1.11
214052_x_at	BAT2D1	BAT2 domain containing 1	4.31e−02	−1.13	9.71e−03	−1.15
200071_at	SMNDC1	survival motor neuron domain containing 1	2.95e−03	−1.19	2.92e−02	−1.09
229594_at	SPTY2D1	SPT2, Suppressor of Ty, domain containing 1 (S. cerevisiae)	6.29e−04	−1.18	4.33e−02	−1.10
221118_at	PKD2L2	polycystic kidney disease 2-like 2	1.35e−03	−1.20	6.74e−03	−1.10
208731_at	RAB2A	RAB2A, member RAS oncogene family	1.96e−02	−1.20	4.51e−02	−1.11
227244_s_at	SSU72	SSU72 RNA polymerase II CTD phosphatase homolog (S. cerevisiae)	1.14e−02	−1.19	1.14e−02	−1.12
237052_x_at	GIGYF2	GRB10 interacting GYF protein 2	4.70e−03	−1.18	5.46e−03	−1.13
222414_at	MLL3	myeloid/lymphoid or mixed-lineage leukemia 3	6.43e−03	−1.20	4.51e−02	−1.12
212293_at	HIPK1	homeodomain interacting protein kinase 1	1.26e−02	−1.20	3.09e−03	−1.12
213579_s_at	EP300	E1A binding protein p300	2.76e−02	−1.18	2.16e−02	−1.14
229723_at	TAGAP	T-cell activation RhoGTPase activating protein	1.09e−02	−1.22	1.23e−02	−1.11
1568764_x_at	LOC728613/PDCD6	programmed cell death 6 pseudogene/programmed cell death 6	4.33e−02	−1.17	2.82e−02	−1.17
217745_s_at	NAT13	N-acetyltransferase 13 (GCN5-related)	2.21e−03	−1.24	3.60e−02	−1.11
1552426_a_at	TM2D3	TM2 domain containing 3	2.73e−02	−1.23	2.12e−02	−1.12
208583_x_at	HIST1H2AJ	histone cluster 1, H2aj	2.92e−02	−1.21	3.14e−02	−1.14
209001_s_at	ANAPC13	anaphase promoting complex subunit 13	4.36e−04	−1.26	1.76e−02	−1.10
224315_at	DDX20	DEAD (Asp-Glu-Ala-Asp) box polypeptide 20	3.86e−02	−1.21	4.19e−02	−1.16
226565_at	TMEM99	transmembrane protein 99	2.73e−02	−1.22	3.18e−02	−1.15
209606_at	CYTIP	cytohesin 1 interacting protein	1.61e−02	−1.25	3.12e−02	−1.12
212665_at	TIPARP	TCDD-inducible poly(ADP-ribose) polymerase	1.55e−02	−1.24	4.61e−02	−1.12
217814_at	CCDC47	coiled-coil domain containing 47	2.67e−03	−1.24	5.08e−03	−1.15
225133_at	KLF3	Kruppel-like factor 3 (basic)	1.72e−03	−1.24	1.36e−03	−1.16
209187_at	DR1	down-regulator of transcription 1, TBP-binding (negative cofactor 2)	1.65e−02	−1.24	1.28e−02	−1.16
223939_at	SUCNR1	succinate receptor 1	1.22e−02	−1.23	1.34e−03	−1.19
209433_s_at	PPAT	phosphoribosyl pyrophosphate amidotransferase	1.39e−02	−1.28	3.50e−02	−1.14
1569349_at	C11orf30	chromosome 11 open reading frame 30	4.89e−04	−1.33	3.18e−02	−1.10
214331_at	TSFM	Ts translation elongation factor, mitochondrial	2.86e−02	−1.25	3.21e−02	−1.22
227137_at	LOC100292024	hypothetical protein LOC100292024	5.39e−03	−1.30	3.21e−02	−1.17
228749_at	ZDBF2	zinc finger, DBF-type containing 2	6.61e−03	−1.29	2.83e−03	−1.19
226731_at	PELO	Pelota homolog (Drosophila)	1.89e−02	−1.30	3.62e−02	−1.17
223809_at	RGS18	regulator of G-protein signaling 18	3.41e−02	−1.31	2.79e−02	−1.19
205857_at	SLC18A2	solute carrier family 18 (vesicular monoamine), member 2	3.53e−03	−1.35	3.46e−02	−1.16
212930_at	ATP2B1	ATPase, Ca++ transporting, plasma membrane 1	1.13e−02	−1.34	1.42e−02	−1.17
222067_x_at	HIST1H2BD	histone cluster 1, H2bd	4.38e−02	−1.34	4.05e−02	−1.21
215071_s_at	HIST1H2AC	histone cluster 1, H2ac	8.26e−03	−1.36	3.14e−03	−1.19
222642_s_at	TMEM33	transmembrane protein 33	3.09e−02	−1.32	2.79e−02	−1.24
238462_at	UBASH3B	ubiquitin associated and SH3 domain containing, B	3.68e−03	−1.39	1.49e−02	−1.17
205072_s_at	XRCC4	X-ray repair complementing defective repair in Chinese hamster cells 4	4.17e−03	−1.37	1.71e−02	−1.20
218446_s_at	FAM18B	family with sequence similarity 18, member B	1.22e−02	−1.35	2.99e−03	−1.24
208523_x_at	HIST1H2BI	histone cluster 1, H2bi	2.82e−02	−1.37	2.78e−02	−1.23
208490_x_at	HIST1H2BF	histone cluster 1, H2bf	2.66e−02	−1.39	3.74e−02	−1.22
208546_x_at	HIST1H2BH	histone cluster 1, H2bh	3.38e−02	−1.38	4.47e−02	−1.24
1554544_a_at	MBP	myelin basic protein	2.67e−02	−1.40	2.79e−02	−1.29
1562321_at	PDK4	pyruvate dehydrogenase kinase, isozyme 4	2.79e−02	−1.44	2.51e−02	−1.34
210136_at	MBP	myelin basic protein	1.45e−02	−1.47	1.12e−02	−1.33
214469_at	HIST1H2AE	histone cluster 1, H2ae	2.46e−02	−1.53	6.60e−03	−1.41
202708_s_at	HIST2H2BE	histone cluster 2, H2be	2.18e−02	−1.58	1.93e−03	−1.45
221958_s_at	GPR177	G protein-coupled receptor 177	3.38e−02	−1.90	3.64e−02	−1.49
241403_at	CLK4	CDC-like kinase 4	3.04e−02	−1.96	4.12e−02	−1.50

¹Rows are sorted by average fold-change in the SZ and SIB groups compared to the CCS group, with largest positive fold-change at the top and largest negative fold-change at the bottom of the Table.

The list of 168 known transcripts (represented by the 172 known probes) dysregulated in both SZs and SIBs compared to CCSs was enriched with genes that represented various functional categories, ontologies, pathways, and protein domains. Those terms that surpassed a Bonferroni-corrected threshold for significant enrichment (α=0.05/number of terms evaluated in a particular category) are shown in Table 2. Notably, six of the seven significantly enriched terms represented histone- and nucleosome-related functions, ontologies, or protein domains.

Table 2.

Functional Categories, Ontologies, Pathways, and Protein Domains Significantly Over-Represented Among Genes Significantly Dysregulated in both the SZ and SIB Groups Compared to the CCS Group

Domain	Category	Term	Genes on List	n (%) of Genes on List	Fold- Enrichment	p	Bonferroni- corrected p
Functional Category	PIR Keywords	nucleosome core	HIST1H2AC HIST1H2AE HIST1H2BD HIST1H2BF/I HIST2H2BE HIST1H2BH	6 (3.7)	19.9	1.14e−05	2.89e−03
Ontology	Cellular Component	nucleosome	HIST1H2AC HIST1H2BD HIST2H2BE HIST1H2BF/I HIST1H2AE HIST1H2BH	6 (3.7)	14.0	6.15e−05	1.43e−02
Pathway	KEGG Pathway	systemic lupus erythematosus	HLA-DQB1 HIST1H2AC HIST1H2BD HIST2H2BE HIST1H2BF/I HIST1H2AE HIST1H2BH HLA-DMB HLA-DMA	9 (5.6)	11.5	6.25e−07	4.44e−05
Protein Domain	InterPro Domain	histone core	HIST1H2AC HIST1H2BD HIST2H2BE HIST1H2BF/I HIST1H2AE HIST1H2BH	6 (3.7)	20.1	1.05e−05	3.82e−03
	InterPro Domain	histone fold	HIST1H2BD HIST2H2BE HIST1H2BF/I DR1 HIST1H2AE HIST1H2BH	6 (3.7)	14.6	5.30e−05	1.91e−02
	PIR Superfamily	Histone H2B	HIST1H2BD HIST2H2BE HIST1H2BF/I HIST1H2BH	4 (2.5)	33.3	1.89e−04	1.31e−02
	SMART Domain	H2B	HIST1H2BD HIST2H2BE HIST1H2BF/I HIST1H2BH	4 (2.5)	33.7	1.89e−04	1.71e−02

[SZ vs. CCS] ∩ [SZ vs. SIB]

Compared to both the CCS and SIB groups, SZ subjects showed significant evidence (p<0.05 in both comparisons) of common dysregulation of 96 probes for 94 known transcripts (Table 3). Two of these genes (PPM1A and RAPGEF2) were each tagged by two probes that were dysregulated in SZs compared to both SIBs and CCSs. The majority of changes observed in SZs were in the same direction vis-à-vis both SIBs and CCSs. Fifty-one probes were down-regulated and 37 probes were up-regulated in SZs compared to both groups; six probes were up-regulated in SZs compared to SIBs but down-regulated in SZs compared to CCSs; and two probes were down-regulated in SZs compared to SIBs but up-regulated in SZs compared to CCSs.

Table 3.

Genes Significantly Dysregulated in the SZ Group Compared to Both SIB and CCS Groups¹

Probeset ID	Gene Symbol	Gene Product	SZ vs. CCS		SZ vs. SIB
			p	Fold-Change	p	Fold-Change
201050_at	PLD3	phospholipase D family, member 3	2.81e−02	1.29	4.85e−02	1.32
242663_at	LOC148189	Hypothetical LOC148189	1.78e−02	1.13	5.09e−03	1.29
205618_at	PRRG1	proline rich Gla (G-carboxyglutamic acid) 1	2.73e−02	1.17	4.04e−02	1.22
244597_at	LOC26010	Viral DNA polymerase-transactivated protein 6	1.62e−02	1.16	4.78e−02	1.22
214612_x_at	MAGEA6	melanoma antigen family A, 6	1.74e−02	1.15	7.10e−03	1.21
207325_x_at	MAGEA1	melanoma antigen family A, 1 (directs expression of antigen MZ2-E)	4.24e−02	1.15	2.24e−03	1.20
207773_x_at	CYP3A43	cytochrome P450, family 3, subfamily A, polypeptide 43	4.99e−02	1.13	1.08e−02	1.22
243570_at	SPCS2	signal peptidase complex subunit 2 homolog (S. cerevisiae)	1.42e−02	1.19	3.38e−02	1.15
1561378_at	C12orf42	chromosome 12 open reading frame 42	2.75e−02	1.13	7.26e−03	1.21
205447_s_at	MAP3K12	mitogen-activated protein kinase kinase kinase 12	1.20e−02	1.18	3.12e−02	1.15
1560432_at	CLRN1OS	clarin 1 opposite strand	3.43e−02	1.14	9.10e−03	1.18
209842_at	SOX10	SRY (sex determining region Y)-box 10	1.37e−02	1.16	3.44e−02	1.16
1570128_at	DDX19A	DEAD (Asp-Glu-Ala-As) box polypeptide 19A	3.47e−02	1.13	4.53e−02	1.17
236688_at	FRMPD3	FERM and PDZ domain containing 3	3.73e−02	1.14	2.61e−02	1.16
1553746_a_at	C12orf64	chromosome 12 open reading frame 64	8.03e−04	1.15	2.04e−03	1.13
219699_at	LGI2	leucine-rich repeat LGI family, member 2	3.88e−02	1.14	1.62e−02	1.14
1561332_at	ATP13A5	ATPase type 13A5	7.05e−03	1.15	1.84e−02	1.12
1559459_at	LOC613266	hypothetical LOC613266	3.48e−02	1.14	9.74e−03	1.12
231994_at	CHDH	choline dehydrogenase	4.21e−03	1.16	4.97e−04	1.10
242301_at	CBLN2	cerebellin 2 precursor	2.12e−03	1.15	2.41e−02	1.10
231155_at	DEFB119	defensin, beta 119	1.05e−02	1.13	4.60e−02	1.12
204796_at	EML1	echinoderm microtubule associated protein like 1	3.31e−02	1.11	2.68e−03	1.13
236730_at	GIPC3	GIPC PDZ domain containing family, member 3	2.37e−02	1.13	4.01e−02	1.11
213209_at	TAF6L	TAF6-like RNA polymerase II, p300/CBP-associated factor (PCAF)-associated factor	1.59e−02	1.11	1.16e−02	1.13
238262_at	SPDYA	speedy homolog A (Xenopus laevis)	2.92e−02	1.15	3.30e−02	1.09
1552458_at	MBD3L1	methyl-CpG binding domain protein 3-like 1	2.52e−02	1.13	4.65e−02	1.10
222328_x_at	MEG3	Maternally expressed 3 (non-protein coding)	4.38e−02	1.11	2.30e−02	1.12
1562271_x_at	ARHGEF7	Rho guanine nucleotide exchange factor (GEF) 7	4.24e−02	1.11	1.20e−02	1.12
218182_s_at	CLDN1	claudin 1	2.47e−02	1.10	3.75e−02	1.12
220771_at	LOC51152	melanoma antigen	7.99e−03	1.11	4.75e−03	1.11
230394_at	TCP10L	t-complex 10 (mouse)-like	7.80e−03	1.13	1.77e−02	1.08
204235_s_at	GULP1	GULP, engulfment adaptor PTB domain containing 1	2.18e−02	1.09	1.64e−02	1.10
232424_at	PRDM16	PR domain containing 16	4.27e−02	1.12	1.79e−02	1.07
221177_at	MIA2	melanoma inhibitory activity 2	3.56e−02	1.12	1.01e−02	1.07
206858_s_at	HOXC6	homeobox C6	4.14e−02	1.11	4.37e−02	1.08
1566734_at	LOC283454	hypothetical protein LOC283454	4.79e−02	1.10	1.80e−02	1.08
1557312_at	C12orf61	chromosome 12 open reading frame 61	9.72e−03	1.10	2.55e−02	1.07
244344_at	WNK4	WNK lysine deficient protein kinase 4	5.32e−03	1.25	4.23e−02	−1.16
1562048_at	LOC152225	hypothetical LOC152225	8.37e−03	1.12	3.74e−02	−1.08
1568764_x_at	LOC728613/PDCD6	programmed cell death 6 pseudogene/programmed cell death 6	4.33e−02	−1.17	2.93e−02	1.21
214331_at	TSFM	Ts translation elongation factor, mitochondrial	2.86e−02	−1.25	2.97e−02	1.28
238923_at	SPOP	speckle-type POZ protein	2.08e−02	−1.15	1.98e−02	1.16
201111_at	CSE1L	CSE1 chromosome segregation 1-like (yeast)	4.71e−03	−1.32	2.27e−02	1.27
236026_at	GPATCH2	G patch domain containing 2	1.46e−02	−1.20	8.24e−03	1.14
230954_at	C20orf112	chromosome 20 open reading frame 112	2.02e−03	−1.17	1.40e−02	1.12
203605_at	SRP54	signal recognition particle 54kDa	4.69e−02	−1.08	3.43e−02	−1.10
218241_at	GOLGA5	golgi autoantigen, golgin subfamily a, 5	2.75e−02	−1.11	4.79e−02	−1.08
217877_s_at	GPBP1L1	GC-rich promoter binding protein 1-like 1	3.18e−02	−1.11	1.89e−02	−1.13
201371_s_at	CUL3	cullin 3	3.58e−03	−1.15	4.07e−02	−1.12
202209_at	LSM3	LSM3 homolog, U6 small nuclear RNA associated (S. cerevisiae)	1.41e−02	−1.16	4.79e−02	−1.12
209326_at	SLC35A2	solute carrier family 35 (UDP-galactose transporter), member A2	4.99e−02	−1.13	1.89e−02	−1.15
200066_at	IK	IK cytokine, down-regulator of HLA II	8.39e−03	−1.14	1.70e−02	−1.15
220958_at	ULK4	unc-51-like kinase 4 (C. elegans)	2.51e−02	−1.12	2.70e−02	−1.17
225268_at	KPNA4	karyopherin alpha 4 (importin alpha 3)	2.36e−02	−1.17	2.85e−02	−1.13
218171_at	VPS4B	vacuolar protein sorting 4 homolog B (S. cerevisiae)	3.90e−02	−1.14	3.76e−02	−1.16
226648_at	HIF1AN	hypoxia inducible factor 1, alpha subunit inhibitor	2.02e−02	−1.19	1.88e−02	−1.11
209986_at	ASCL1	achaete-scute complex homolog 1 (Drosophila)	4.17e−02	−1.13	4.32e−02	−1.18
227357_at	MAP3K7IP3	mitogen-activated protein kinase kinase kinase 7 interacting protein 3	3.58e−02	−1.14	1.62e−02	−1.18
57703_at	SENP5	SUMO1/sentrin specific peptidase 5	3.65e−03	−1.17	3.98e−03	−1.15
209463_s_at	TAF12	TAF12 RNA polymerase II, TATA box binding protein (TBP)-associated factor, 20kDa	4.93e−02	−1.11	1.65e−02	−1.21
201099_at	USP9X	ubiquitin specific peptidase 9, X-linked	2.47e−03	−1.17	2.16e−02	−1.16
226952_at	EAF1	ELL associated factor 1	6.68e−03	−1.18	4.81e−02	−1.15
1554351_a_at	TIPRL	TIP41, TOR signaling pathway regulator-like (S. cerevisiae)	1.41e−02	−1.15	3.91e−02	−1.19
218761_at	RNF111	ring finger protein 111	4.01e−02	−1.18	5.00e−04	−1.16
202352_s_at	PSMD12	proteasome (prosome, macropain) 26S subunit, non-ATPase, 12	3.95e−02	−1.12	1.81e−02	−1.22
212665_at	TIPARP	TCDD-inducible poly(ADP-ribose) polymerase	1.55e−02	−1.24	3.29e−02	−1.11
200071_at	SMNDC1	survival motor neuron domain containing 1	2.95e−03	−1.19	4.83e−02	−1.17
224974_at	SUDS3	suppressor of defective silencing 3 homolog (S. cerevisiae)	2.08e−02	−1.24	3.64e−02	−1.12
217795_s_at	TMEM43	transmembrane protein 43	3.01e−02	−1.17	2.57e−02	−1.20
214865_at	DOT1L	DOT1-like, histone H3 methyltransferase (S. cerevisiae)	4.06e−02	−1.19	7.84e−03	−1.21
223288_at	USP38	ubiquitin specific peptidase 38	4.70e−02	−1.17	4.70e−02	−1.24
221873_at	ZNF143	zinc finger protein 143	2.90e−02	−1.16	1.19e−02	−1.26
221244_s_at	PDPK1	3-phosphoinositide dependent protein kinase-1	3.18e−02	−1.18	4.06e−02	−1.25
204507_s_at	PPP3R1	protein phosphatase 3 (formerly 2B), regulatory subunit B, alpha isoform	2.53e−02	−1.22	4.47e−02	−1.20
227413_at	UBLCP1	ubiquitin-like domain containing CTD phosphatase 1	4.25e−02	−1.15	7.16e−03	−1.27
231588_at	PRCP	Prolylcarboxypeptidase (angiotensinase C)	4.91e−04	−1.29	2.09e−02	−1.15
222432_s_at	CCDC47	coiled-coil domain containing 47	1.26e−03	−1.17	6.55e−03	−1.28
204513_s_at	ELMO1	engulfment and cell motility 1	2.05e−02	−1.18	4.41e−02	−1.27
218668_s_at	RAP2C	RAP2C, member of RAS oncogene family	2.05e−02	−1.24	1.65e−02	−1.22
231640_at	LYRM5	LYR motif containing 5	5.58e−03	−1.29	1.44e−02	−1.21
202539_s_at	HMGCR	3-hydroxy-3-methylglutaryl-Coenzyme A reductase	1.82e−02	−1.26	4.25e−02	−1.30
33494_at	ETFDH	electron-transferring-flavoprotein dehydrogenase	1.89e−02	−1.31	3.82e−02	−1.26
212585_at	OSBPL8	oxysterol binding protein-like 8	4.07e−02	−1.23	4.47e−02	−1.34
219532_at	ELOVL4	elongation of very long chain fatty acids (FEN1/Elo2, SUR4/Elo3, yeast)-like 4	1.55e−02	−1.37	2.24e−02	−1.22
223809_at	RGS18	regulator of G-protein signaling 18	3.41e−02	−1.31	7.82e−04	−1.31
1554588_a_at	TTC30B	tetratricopeptide repeat domain 30B	4.70e−02	−1.23	2.63e−03	−1.39
229027_at	PPM1A	protein phosphatase 1A (formerly 2C), magnesium-dependent, alpha isoform	1.75e−02	−1.32	6.14e−03	−1.30
202006_at	PTPN12	protein tyrosine phosphatase, non-receptor type 12	2.32e−02	−1.30	3.82e−02	−1.33
227728_at	PPM1A	protein phosphatase 1A (formerly 2C), magnesium-dependent, alpha isoform	3.14e−02	−1.24	2.86e−02	−1.40
225598_at	SLC45A4	solute carrier family 45, member 4	4.87e−02	−1.44	4.17e−02	−1.31
203097_s_at	RAPGEF2	Rap guanine nucleotide exchange factor (GEF) 2	7.51e−03	−1.46	4.33e−02	−1.31
215071_s_at	HIST1H2AC	histone cluster 1, H2ac	8.26e−03	−1.36	1.61e−02	−1.43
226106_at	RNF141	ring finger protein 141	2.29e−02	−1.30	1.85e−02	−1.53
217494_s_at	LOC100290144	hypothetical protein LOC100290144	1.31e−02	−1.39	5.66e−03	−1.57
240744_at	CPA5	carboxypeptidase A5	1.09e−02	−1.65	2.42e−02	−1.37
203096_s_at	RAPGEF2	Rap guanine nucleotide exchange factor (GEF) 2	4.80e−02	−1.54	1.47e−02	−1.75

¹Rows are sorted by average fold-change in the SZ group compared to SIB and CCS groups, with largest positive fold-change at the top and largest negative fold-change at the bottom of the Table.

The list of 94 known transcripts (represented by the 96 known probes) dysregulated in SZs compared to both SIBs and CCSs was enriched at a nominal level of significance with genes that represented various functional categories, pathways, ontologies, and protein domains; however, none of these terms surpassed a Bonferroni-corrected threshold for significant enrichment (α=0.05/number of terms evaluated in a particular category).

[SIB vs. SZ] ∩ [SIB vs. CCS]

Compared to both the CCS and SZ groups, SIBs showed significant evidence (p<0.05 in both comparisons) of common dysregulation of 82 probes for 81 known transcripts (Table 4). One of these genes (ZNF81) was tagged by two probes that were dysregulated in SIBs compared to both SZs and CCSs. The majority of changes observed in SIBs were in the same direction vis-à-vis both SZs and CCSs. Forty-nine probes were up-regulated and 22 probes were down-regulated in SIBs compared to both groups; seven probes were down-regulated in SIBs compared to CCSs but up-regulated in SIBs compared to SZs; and four probes were up-regulated in SIBs compared to CCSs but down-regulated in SIBs compared to SZs.

Table 4.

Genes Significantly Dysregulated in the SIB Group Compared to Both SZ and CCS Groups¹

Probeset ID	Gene Symbol	Gene Product	SIB vs. CCS		SIB vs. SZ
			p	Fold-Change	p	Fold-Change
210233_at	IL1RAP	interleukin 1 receptor accessory protein	1.66e−02	1.50	2.17e−05	2.06
232197_x_at	ARSB	arylsulfatase B	4.03e−02	1.24	4.94e−02	1.58
217222_at	IGHG1	Immunoglobulin heavy constant gamma 1 (G1m marker)	1.58e−02	1.17	4.75e−02	1.27
1569076_a_at	ZNF836	zinc finger protein 836	3.17e−02	1.17	3.49e−02	1.23
227777_at	C10orf18	chromosome 10 open reading frame 18	2.14e−02	1.15	1.10e−02	1.23
236246_x_at	LOC653160	Hypothetical protein LOC653160	2.89e−02	1.14	4.20e−02	1.23
220894_x_at	PRDM12	PR domain containing 12	6.27e−03	1.17	2.11e−02	1.20
205925_s_at	RAB3B	RAB3B, member RAS oncogene family	1.72e−02	1.13	5.90e−03	1.22
239356_at	LOC100129122	Hypothetical protein LOC100129122	4.22e−02	1.10	4.93e−02	1.24
211403_x_at	VCX2	variable charge, X-linked 2	1.26e−03	1.16	1.24e−02	1.17
223895_s_at	EPN3	epsin 3	1.62e−03	1.19	1.59e−02	1.14
237095_at	ASXL2	additional sex combs like 2 (Drosophila)	8.40e−04	1.17	3.46e−02	1.14
244344_at	WNK4	WNK lysine deficient protein kinase 4	3.19e−02	1.15	4.23e−02	1.16
230217_at	RLBP1L1	retinaldehyde binding protein 1-like 1	9.71e−03	1.15	4.34e−02	1.15
214197_s_at	SETDB1	SET domain, bifurcated 1	1.37e−02	1.11	1.96e−02	1.19
219465_at	APOA2	apolipoprotein A-II	4.15e−02	1.10	1.09e−02	1.19
236150_at	AGPHD1	aminoglycoside phosphotransferase domain containing 1	1.73e−02	1.13	3.59e−02	1.15
222092_at	PTPN21	Protein tyrosine phosphatase, non-receptor type 21	2.48e−02	1.10	3.18e−02	1.18
206586_at	CNR2	cannabinoid receptor 2 (macrophage)	3.62e−02	1.08	6.81e−03	1.19
232393_at	ZNF462	zinc finger protein 462	4.56e−02	1.08	9.47e−04	1.18
217675_at	ZBTB7C	zinc finger and BTB domain containing 7C	1.73e−02	1.10	3.59e−03	1.16
1561039_a_at	ZNF81	zinc finger protein 81	6.86e−03	1.10	1.52e−02	1.16
205150_s_at	TRIL	TLR4 interactor with leucine rich repeats	5.21e−03	1.10	2.40e−06	1.16
238280_at	CYB5RL	cytochrome b5 reductase-like	1.23e−02	1.12	2.62e−02	1.14
207465_at	PRO0628	uncharacterized protein PRO0628-like	2.51e−02	1.05	1.94e−02	1.20
218549_s_at	FAM82B	family with sequence similarity 82, member B	1.02e−02	1.12	4.63e−02	1.13
210820_x_at	COQ7	coenzyme Q7 homolog, ubiquinone (yeast)	3.26e−02	1.10	5.60e−03	1.14
230908_at	TACR1	tachykinin receptor 1	4.81e−02	1.06	3.50e−04	1.18
233368_s_at	DNAJC27	DnaJ (Hsp40) homolog, subfamily C, member 27	1.90e−03	1.11	1.36e−02	1.13
228651_at	VWA1	von Willebrand factor A domain containing 1	1.00e−02	1.11	3.12e−02	1.13
238925_at	SNTB2	syntrophin, beta 2 (dystrophin-associated protein A1, 59kDa, basic component 2)	3.97e−02	1.10	3.18e−02	1.13
229839_at	SCARA5	Scavenger receptor class A, member 5 (putative)	3.24e−02	1.09	3.35e−02	1.13
1559342_a_at	SNRPN	small nuclear ribonucleoprotein polypeptide N	2.76e−02	1.07	1.32e−02	1.15
1555774_at	ZAR1	zygote arrest 1	1.81e−02	1.09	1.88e−02	1.13
1560692_at	LOC285878	hypothetical protein LOC285878	3.54e−02	1.07	4.98e−02	1.14
1562675_at	LOC100128003	hypothetical protein LOC100128003	3.04e−02	1.07	2.50e−02	1.14
230819_at	FAM148C	family with sequence similarity 148, member C	2.30e−02	1.06	3.21e−03	1.13
231783_at	CHRM1	cholinergic receptor, muscarinic 1	2.61e−02	1.09	1.42e−02	1.10
215721_at	IGHG1/LOC90925	immunoglobulin heavy constant gamma 1 (G1m marker)/hypothetical protein LOC9	6.52e−03	1.09	3.65e−02	1.10
1569729_a_at	ASZ1	ankyrin repeat, SAM and basic leucine zipper domain containing 1	8.13e−04	1.10	2.39e−02	1.09
1562876_s_at	LOC541471	Hypothetical LOC541471	1.04e−02	1.06	1.12e−02	1.12
1563223_a_at	CENPI	centromere protein I	4.69e−02	1.08	3.99e−02	1.10
210035_s_at	RPL5/SNORA66	ribosomal protein L5/small nucleolar RNA, H/ACA box 66	4.67e−02	1.06	4.04e−02	1.12
208416_s_at	SPTB	spectrin, beta, erythrocytic	4.42e−02	1.07	3.39e−02	1.10
215655_at	GRIK2	Glutamate receptor, ionotropic, kainate 2	5.46e−03	1.06	1.94e−02	1.11
204189_at	RARG	retinoic acid receptor, gamma	2.38e−02	1.10	4.65e−02	1.07
240079_at	ZNF81	Zinc finger protein 81	2.46e−02	1.07	1.45e−02	1.08
236967_at	LOC645249	hypothetical protein LOC645249	2.18e−02	1.07	4.43e−02	1.08
1561137_s_at	GYPE	glycophorin E	4.65e−02	1.05	4.20e−05	1.08
215071_s_at	HIST1H2AC	histone cluster 1, H2ac	3.14e−03	−1.19	1.61e−02	1.43
229090_at	LOC220930	hypothetical LOC220930	4.27e−02	−1.22	1.08e−02	1.43
223809_at	RGS18	regulator of G-protein signaling 18	2.79e−02	−1.19	7.82e−04	1.31
235012_at	LRCH1	leucine-rich repeats and calponin homology (CH) domain containing 1	1.27e−02	1.20	4.44e−02	−1.09
200071_at	SMNDC1	survival motor neuron domain containing 1	2.92e−02	−1.09	4.83e−02	1.17
205307_s_at	KMO	kynurenine 3-monooxygenase (kynurenine 3-hydroxylase)	4.53e−02	1.18	5.78e−03	−1.14
200620_at	TMEM59	transmembrane protein 59	3.98e−02	−1.09	3.41e−02	1.13
201580_s_at	TMX4	thioredoxin-related transmembrane protein 4	4.14e−03	−1.26	3.13e−02	1.29
230784_at	PRAC	prostate cancer susceptibility candidate	3.99e−03	1.11	3.99e−02	−1.09
212665_at	TIPARP	TCDD-inducible poly(ADP-ribose) polymerase	4.61e−02	−1.12	3.29e−02	1.11
1570128_at	DDX19A	DEAD (Asp-Glu-Ala-As) box polypeptide 19A	1.71e−02	1.10	4.53e−02	−1.17
205669_at	NCAM2	neural cell adhesion molecule 2	2.85e−02	−1.09	2.71e−02	−1.09
236034_at	ANGPT2	angiopoietin 2	2.14e−02	−1.09	4.44e−02	−1.09
230057_at	LOC285178	hypothetical protein LOC285178	2.93e−02	−1.09	1.80e−02	−1.11
221594_at	C7orf64	chromosome 7 open reading frame 64	3.52e−02	−1.09	1.24e−02	−1.11
210906_x_at	AQP4	aquaporin 4	7.50e−03	−1.07	1.80e−02	−1.13
1554932_at	ZSWIM2	zinc finger, SWIM-type containing 2	2.49e−02	−1.07	3.95e−02	−1.14
230957_at	PCDHB19P	Protocadherin beta 19 pseudogene	3.49e−02	−1.08	8.23e−03	−1.14
244822_at	GART	Phosphoribosylglycinamide formyltransferase	1.60e−02	−1.06	1.46e−02	−1.15
241604_at	ATP11A	ATPase, class VI, type 11A	2.75e−02	−1.06	3.11e−02	−1.17
233141_s_at	ST7L	suppression of tumorigenicity 7 like	3.87e−02	−1.06	1.25e−02	−1.18
233082_at	ZNF630	zinc finger protein 630	1.49e−02	−1.10	2.12e−02	−1.15
215366_at	SNX13	sorting nexin 13	1.53e−02	−1.15	3.08e−02	−1.11
1568663_a_at	PWRN2	Prader-Willi region non-protein coding RNA 2	2.04e−02	−1.09	1.01e−02	−1.18
210315_at	SYN2	synapsin II	2.02e−02	−1.10	3.06e−02	−1.17
231130_at	FKBP7	FK506 binding protein 7	3.44e−02	−1.08	2.09e−02	−1.20
211722_s_at	HDAC6	histone deacetylase 6	3.11e−02	−1.09	1.50e−02	−1.20
221424_s_at	OR51E2	olfactory receptor, family 51, subfamily E, member 2	1.39e−02	−1.14	2.44e−02	−1.15
219617_at	C2orf34	chromosome 2 open reading frame 34	4.15e−02	−1.08	2.95e−02	−1.23
1564000_at	ANKRD31	ankyrin repeat domain 31	2.92e−03	−1.16	4.96e−02	−1.18
1568764_x_at	LOC728613/PDCD6	programmed cell death 6 pseudogene/programmed cell death 6	2.82e−02	−1.17	2.93e−02	−1.21
214331_at	TSFM	Ts translation elongation factor, mitochondrial	3.21e−02	−1.22	2.97e−02	−1.28
232281_at	LOC148189	Hypothetical LOC148189	6.52e−03	−1.12	1.48e−03	−1.39

¹Rows are sorted by average fold-change in the SIB group compared to SZ and CCS groups, with largest positive fold-change at the top and largest negative fold-change at the bottom of the Table.

The list of 81 known transcripts (represented by the 82 known probes) dysregulated in SIBs compared to both SZs and CCSs was nominally significantly enriched with genes that represented various functional categories, pathways, ontologies, and protein domains; however, none of these terms surpassed a Bonferroni-corrected threshold for significant enrichment (α=0.05/number of terms evaluated in a particular category).

[SZ vs. CCS] ∩ [SZ vs. SIB] ∩ [SIB vs. CCS]

Eight probes for eight known transcripts were dysregulated in all three orthogonal comparisons of diagnostic groups (Table 5). Four transcripts (SMNDC1, TIPARP, HIST1H2AC, and RGS18) were expressed at intermediate levels in SIBs with highest expression in CCSs and lowest expression in SZ. Conversely, one transcript (DDX19A) that was intermediately expressed in SIBs had the highest expression level in SZs and the lowest expression level in CCSs. Two transcripts (LOC728613/PDCD6 and TSFM) had the lowest expression levels in SIBs, the highest expression levels in CCSs, and intermediate expression in SZs. The remaining transcript (WNK4) had its highest expression level in SIBs and lowest expression level in CCSs, with SZs exhibiting intermediate expression levels. These eight known transcripts were not enriched with genes that represented any functional categories, pathways, ontologies, and protein domains.

Table 5.

Genes Significantly Dysregulated between SZ, SIB, and CCS Groups

Probeset ID	Gene Symbol	Gene Product	SZ vs. CCS		SZ vs. SIB		SIB vs. CCS
			p	Fold-Change	p	Fold-Change	p	Fold-Change
1570128_at	DDX19A	DEAD (Asp-Glu-Ala-As) box polypeptide 19A	3.47e−02	1.13	4.53e−02	1.17	1.71e−02	1.10
244344_at	WNK4	WNK lysine deficient protein kinase 4	5.32e−03	1.25	4.23e−02	−1.16	3.19e−02	1.15
1568764_x_at	LOC728613/PDCD6	programmed cell death 6 pseudogene/programmed cell death 6	4.33e−02	−1.17	2.93e−02	1.21	2.82e−02	−1.17
214331_at	TSFM	Ts translation elongation factor, mitochondrial	2.86e−02	−1.25	2.97e−02	1.28	3.21e−02	−1.22
212665_at	TIPARP	TCDD-inducible poly(ADP-ribose) polymerase	1.55e−02	−1.24	3.29e−02	−1.11	4.61e−02	1.12
200071_at	SMNDC1	survival motor neuron domain containing 1	2.95e−03	−1.19	4.83e−02	−1.17	2.92e−02	−1.09
223809_at	RGS18	regulator of G-protein signaling 18	3.41e−02	−1.31	7.82e−04	−1.31	2.79e−02	−1.19
215071_s_at	HIST1H2AC	histone cluster 1, H2ac	8.26e−03	−1.36	1.61e−02	−1.43	3.14e−03	−1.19

¹Rows are sorted by average fold-change across all three comparisons, with largest positive fold-change at the top and largest negative fold-change at the bottom of the Table.

DISCUSSION

In the past five years, much work has been done to establish the validity of blood-based gene-expression signatures as a means of detecting meaningful biomarkers for neuropsychiatric disorders. For example, we (Glatt and others 2005a) and others (Sullivan and others 2006) have found a reasonable level of correspondence in gene expression levels between peripheral blood and various brain structures, including some relevant to schizophrenia such as dorsolateral prefrontal cortex. Others have demonstrated the considerable heritability and temporal stability of gene expression levels in peripheral blood (Meaburn and others 2009). However, one problem that has consistently plagued this area of research has been the inability to determine if transcriptomic abnormalities reflect “trait” or “state” conditions due to the inherent confounds of comparing non-mentally ill individuals to individuals with schizophrenia who undergo psychopharmacotherapy and deal with a chronic debilitating disorder and its sequelae. A recent study by Takahashi et al. (Takahashi and others 2010) partially overcame this conundrum by studying antipsychotic-free schizophrenia patients; yet, while many of the subjects in this study were truly drug- or at least antipsychotic-naïve, others had previously been on antipsychotics as recently as eight weeks prior to testing, and some subjects were actively on other classes of drugs such as antidepressants, benzodiazepines, or mood stabilizers at the time of testing. The work of Takahashi et al. has markedly advanced the field by identifying profiles of as few as 14 probes that yielded 82.4% sensitivity and 93.8% specificity in classifying a separate set of schizophrenia patients and control subjects. However, because these patients had already been ill for some time, it remains unknown if the differences expressed in their peripheral blood transcriptomes were static markers of the underlying genetic susceptibility to the disorder or were consequences of their illness. In an attempt to further clarify the contributions of trait and state to peripheral blood transcriptomic abnormalities in schizophrenia, we have assessed biological siblings, who share both genetic and early environmental factors in common with their affected relatives. In addition to helping rule out various confounder effects as major drivers behind observed transcriptomic abnormalities in the patient population, the inclusion of relatives was also intended to shed light on potential protective factors operating to keep these genetically susceptible individuals from expressing the illness.

At the intersection of probes that were differentially expressed in the peripheral blood of both SZ and SIB groups relative to the CCS group, we found 168 commonly dysregulated genes which may reflect risk factors for SZ that are independent of illness-associated factors, such as chronic treatment with antipsychotic medication. One of these 168 commonly dysregulated genes, SPIRE1, was significantly dysregulated in both the SZ and SIB groups but in opposite directions, suggesting that up-regulation of this particular gene may be associated with some protective capacity among genetically susceptible individuals (SIBs). Collectively, these 168 commonly dysregulated genes contained an over-representation of genes related to histone and nucleosome function, ontology, and structure, suggesting that SZ may be associated with a global dysregulation of the histone system. Of note, in an early study, we showed that lysergic acid diethylamide, which can elicit psychotic symptoms, had effects on histone acetylation (Brown and Liew 1975). Histones are proteins around which nuclear double-stranded DNA is coiled, and the functionality of histones can be altered by post-translational modifications such as methylation, acetylation, phosphorylation, ubiquitination, SUMOylation, citrullination and ADP-ribosylation. Such epigenetic modifications can influence the accessibility of the surrounding DNA leading to up- or down-regulation of the mRNA transcripts of genes in the vicinity of that modification. It is conceivable, therefore, that a singular abnormality or a collection of abnormalities (such as increased intensity of gene expression) leading to a common endpoint (e.g., histone dysfunction) in SZ and its genetically influenced risk state could lead to the consequent dysregulation of a whole host of other genes relevant to the development or presentation of the disorder.

In addition to gene-set enrichment analyses in DAVID, we compared this list of jointly dysregulated known transcripts to the list of 45 “Top Results” from the Schizophrenia Gene (SZGene) Database (Allen and others 2008) as of March 15, 2011, which collates the evidence from genetic association studies of schizophrenia and identifies which genes have significant meta-analytic evidence as risk factors for the disorder. Only one of the 45 Top Results from the SZGene database was also significantly dysregulated in both the SZ and SIB groups compared to the CCS group: DRD2, which was commonly up-regulated in the SZ and SIB groups. This result is strikingly similar to that of Zvara et al. (2005), who found up-regulation of DRD2 in peripheral blood from drug-naïve schizophrenia patients as well. Of course, the D2 dopamine receptor, which is encoded by this gene, is the major antagonistic target of all effective antipsychotic medications, and has been the protein around which the dopamine hypothesis of schizophrenia was built and modified over the course of the last four decades (Baumeister and Francis 2002; Seeman and others 2005; Van Rossum 1967). We and others have also found consistent evidence for association of polymorphisms in this gene with susceptibility to the disorder (Glatt and others 2009; Glatt and Jonsson 2006). Collectively, these findings suggest that variation in DRD2, particularly those modifications which result in over-expression of its transcript, are a trait marker of the liability toward schizophrenia, not a state marker or merely a response to treatment with D2 receptor antagonists.

We also performed a literature search in PubMed for keyword pairs of “schizophrenia” and the official gene symbol for each jointly dysregulated transcript, and found that, in addition to DRD2, several other genes that were on our list of 168 genes (Table 1) had been previously associated with the disorder in at least one other study. These genes included: 1) CACNA1C, which encodes a calcium channel and has been identified as a genome-wide significant risk factor for bipolar disorder (Ferreira and others 2008) and subsequently identified as a risk factor for SZ as well (Green and others 2009); 2) GNAS, which encodes an adenylate cyclase-stimulating G alpha protein and has been associated with deficit SZ (Minoretti and others 2006); 3) HLA-DQB1, which encodes a HLA class II histocompatibility antigen and has shown mixed (but mostly negative) evidence for association with the disorder across a number of studies (Nimgaonkar and others 1993; Nimgaonkar and others 1997; Schwab and others 2002); 4) HNRNPC, which encodes a heterogeneous nuclear ribonucleoprotein that was found to be down-regulated in SZ (as was its mRNA transcript in the present study) in postmortem left posterior superior temporal gyrus (Wernicke's area) (Martins-de-Souza and others 2009); 5) MBP, which encodes a myelin basic protein and is one of the most commonly observed dysregulated transcripts in functional genomic studies of postmortem brain tissue from SZ subjects (Segal and others 2007); 6) NPAS2, which encodes a transcription factor and putative circadian clock protein which has been associated with both SZ and bipolar disorder (Mansour and others 2009); 7) PER3, which encodes another circadian clock protein associated with SZ (Mansour and others 2006); 8) PICK1, which encodes a protein-kinase-C-alpha-interacting protein that was found to be up-regulated in SZ (as it was in the present study) in postmortem dorsolateral prefrontal cortex (Sarras and others); and 9) SLC18A2, which encodes vesicular monoamine transporter 2 and has been associated with SZ (Talkowski and others 2008).

At the intersection of probes that were differentially expressed in the SZ group compared to both the SIB and CCS groups we found 94 jointly dysregulated genes whose up- or down-regulation may lead to SZ while typical levels of expression of these genes may spare the genetically susceptible SIBs from illness; however, these genes may also reflect the influence of environmental factors or non-genetic biological or developmental changes that are associated with the disorder. A third option is that such genes are induced by or responsive to pharmaco- or other therapies, or sequelae of the disorder, to which the SZ group was exclusively exposed.

We also observed a set of 82 transcripts that were dysregulated exclusively in unaffected SIBs and not altered in their affected biological relatives. In some instances, SIBs had expression levels that were intermediate to the two comparison groups (SZs and CCSs), suggestive of some illness-associated (possibly genetically influenced) risk genes whose level of expression is associated with the likelihood of expressing the disorder. In other words, these changes may indicate that the level or intensity of gene expression in SIBs is inadequate to bring about manifestations of the disorder; however, a minority of commonly dysregulated transcripts had this signature. Instead, most of the 82 transcripts were either up- or down-regulated in SIBs in both comparisons (with SZs and with CCSs). These changes in particular may indicate the presence of protective factors operating only within those who also have a genetic susceptibility to the disorder; i.e., only in SIBs and not in CCSs.

The results must be interpreted in the context of several limitations. First, this initial demonstrative study utilized a small sample. This small sample size imposed limits on inferential power and thus inhibited our ability to observe results for individual biomarkers that would withstand rigorous corrections for multiple testing (e.g., Bonferroni correction); nevertheless, we anticipate our findings will have substantial utility in highlighting candidate biomarkers and their associated effect sizes which can be exploited in the design of future studies designed prospectively to test these hypotheses with suitable levels of power. Second, despite our attempts to match the three subject groups on relevant demographics, a gender disparity was encountered which conceivably might alter the results. We statistically accounted for sex as a covariate in our analyses, but lacked inferential power to model sex-by-diagnosis interactions that might be operating. Thus, future work should focus on larger samples that are identical or nearly so with regard to sex and other relevant factors, such as ancestry, that have the potential to confound such analyses. Third, the results here were all derived by microarray which, while highly efficient, is not the most sensitive assay for measuring gene expression intensities. Although we lacked the ability in this small pilot study to verify microarray-derived results by a more sensitive method, such as quantitative real-time PCR, this is obviously a prerequisite for advancing any of the specific candidate biomarkers identified here to subsequent stages of experimentation, such as replication efforts in other samples, genetic association studies, or functional investigations into the biology of these genes. Nevertheless, our prior work and that of others has shown that a good proportion of microarray-derived results are ultimately verifiable by other methods; this, coupled with the fairly large numbers of candidate biomarkers identified in the various group comparisons, suggests that some true-positive results are likely contained within this set of putative biomarkers. Lastly, we would highlight that one strength of our study (our cell-isolation method) can not entirely overcome some of the weaknesses associated with examining peripheral blood as a source of biomarkers. Thus, while the technique we have used to isolate PBMCs produces a sample of cells for biomarker identification that is far more homogeneous than that obtained through whole-blood RNA extraction (a commonly used approach), PBMCs are themselves a functionally heterogeneous class of blood cells. For example, if different types of PBMCs (e.g., lymphocytes and macrophages) express different transcriptome signatures, and if our subject groups systematically differed in the proportion of different PBMC cell types in their blood samples, then some observed group differences in gene expression may actually reflect diagnostic-group differences in blood constitution rather than true group differences in gene expression within the same cells.

In summary, we have identified signatures of gene expression in peripheral blood that distinguish individuals with schizophrenia from non-mentally ill comparison subjects, as well as the differences and similarities that these affected individuals share with their unaffected biological siblings. We do not expect our work will necessarily shed light on the underlying neurobiological and molecular causes of schizophrenia or its genetically influenced risk state. Rather, this work is intended to identify potential biomarkers that will have utility for the purposes of clinical classification and risk prediction. In order to be useful in the clinical context, a potential biomarker is not also required to directly reflect or represent the core source of its abnormal expression, though it may do so in some instances (c.f., Sullivan and others 2006), and we expect that some (but certainly not all) of the putative biomarkers we have identified presently may illuminate neurobiological and molecular processes that may have gone awry in schizophrenia. Subsequent work in postmortem brain tissue will be required to reveal which of our putative biomarkers also have an etiologic role reflected in the brain and which solely have utility as peripheral classifiers. At this relatively early stage, much work remains to be done before we can declare any of the observed transcriptomic alterations true biomarkers for the disorder or its genetically influenced risk or protective states. Yet, this pilot study has accomplished our intention of demonstrating, in the classical tradition of genetic epidemiology, the potential utility of including first-degree biological relatives in biomarker studies of schizophrenia. The abnormalities of gene expression identified in the unaffected relatives were not entirely overlapping with nor entirely distinct from those seen in the schizophrenia patients themselves, but in many cases the changes were in a similar direction and had common biological threads suggesting that they reflect the inherent (but in some cases unrealized) risk for the disorder. In other instances, the changes seen in unaffected siblings were opposite to those seen in their affected relatives, perhaps shedding light on the protective biological architecture that these individuals (but not their affected siblings) have inherited or acquired. If even a portion of the results reported here can be verified and extended in subsequent studies, it may provide a basis for advanced diagnostics, targeted intervention strategies, and a better understanding of the biological susceptibility toward schizophrenia.