Supplementary sex-specific transcriptome in human cortex
1
Supplementary Material
Transcriptome Analysis of Male-Female Differences in Prefrontal Cortical Development
Cynthia Shannon Weickert, Ph.D.1,2, Michael Elashoff, Ph.D.3, Allen Brent Richards, Ph.D.1, Duncan Sinclair, BSc.2, Sabine Bahn, Ph.D.4, Svante Paabo, Ph.D.5, Philipp Khaitovich, Ph.D.5,6, and Maree J. Webster, Ph.D.7
1MiNDS Unit,CBDB, NIMH, IRP, Bethesda, MD, 20894
2Schizophrenia Research Institute (SRI), University of New South Wales, Prince of Wales Medical Research Institute, Sydney, Australia
3Cardiodx, Palo Alto, CA94304
4Institute of Biotechnology, University of Cambridge, Cambridge, CB2 1QT, UK
5Department of Evolutionary Genetics, Max Plank Institute for Evolutionary Anthropology, Leipzig, Germany
6Institute for Computational Biology, Shanghai Institute for Biological Sciences, Chinese Academy of Sciences, 320 YueYang Rd., Shanghai, 200031 China
7Stanley Medical Research Institute, 9800 Medical Center Drive, Rockville,MD20850
Corresponding Author:
Maree J. Webster
Stanley Medical Research Institute
9800 Medical Center Drive, Suite C-050
Rockville, MD20850
Phone: 240 499 1171
Fax: 301 251 8602
Email:
Acknowledgements: We acknowledge the assistance of Dr. H. Ronald Zielke and Robert Vigorito of the University of Maryland Brain and Tissue Bank for Developmental Disorders.
Funding:Supported by the Stanley Medical Research Institute,the intramural program of NIMH, the Schizophrenia Research Institute, the University of New South Wales and the Prince of Wales Medical Research Institute.
Supplementary Introduction
Males and female humans differ in their cognitive, psychological and emotional development; however, it is not known if and when they differ in global gene expression patterns during brain development. Gender dimorphisms exist in spatial ability and aggressiveness (1) and in the propensity to develop psychiatric disorders such as autism, attention deficit hyperactivity disorder (ADHD) and depression (2). Sex differences exist in cortical complexity (3), in adult regional brain volumes (4) and in childhood gray matter volumes (5). Divergent development of the male and female brain is believed to result from the sex-specific development of gonads that indirectly influence the developing brain via gonadal hormones (6). However, this endocrine mechanism may work in concert with other more local acting sex specific developmental signals originating in the brain itself.
Neuronal nuclei of males and females contain a different complement of sex chromosomes that harbor distinct genes that may directly impact the developing brain. We hypothesized that gender dimorphisms in human brain development may include differential cortical expression of genes unique to the sex chromosomes (7, 8) as well as differential cortical expression of genes on the autosomes. In this study, we first give a broad overview of differential gene expression across human postnatal development, then we identify and functionally group 130 transcripts that are differentially expressed in the brain of developing males compared to females. While we identify thousands of transcripts that change with age, only a small subset of these change in a sex-specific manner during development. Therefore, sex does not appear to dictate ubiquitous transcriptional changes across development, but rather may impact human cortical development via selective control of specific genes. The sex-specific changes in gene expression that we find in the developing human frontal cortex represent part of the molecular biological substrate that distinguishes the male from the female brain.
Supplementary Material and Methods
Tissue Collection
Sixty cases ranging in age from 6 weeks to 49 years (Supplementary Table 1) were obtained from the University of Maryland Brain and Tissue Bank for Developmental Disorders (UMBB; NICHHD contract # NO1-HD8-3283; 37 males and 23 females, 33 African Americans and 27 Caucasians). Frozen tissue samples from 7-13 cases (defined as normal controls by the forensic pathologists at UMBB) were selected from each of seven developmental periods. Samples were included in the cohort if the pH was above 6.25 and if the RNA was of good quality [over 5.8, (9)] as determined by the high resolution Bioanalyzer electrophoresis system (Agilent Technologies, Palo Alto, CA, USA). The cases used in each experiment did not differ significantly within each age group according to brain pH or RNA Integrity Number (RIN) value.
Total RNA Isolation
Total RNA was extracted from 300mg gray matter of the middle frontal gyrus (Brodman’s area 46) using Trizol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions (10).To assess RNA quality, approximately 100-200ng RNA was applied to an RNA 6000 Nano LabChip, without heating prior to loading. The RIN was calculated by 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). cDNA was synthesised in three reactions per sample and pooled, from 3 µg of total RNA per 26.25 µl reaction using the Superscript First-Strand Synthesis Kit (Invitrogen) according to the manufacturer’s protocol.
Microarray Experimental Design
Total RNA from 45 cases was purified through a Qiagen RNA miniKit column (Qiagen Inc, Valencia CA USA) according to the manufacturers protocol. RNA was processed through the Affymetrix preparation protocol [ (11)] and hybridized to HG-U133 version 2.0+ (GeneChips, AffymetrixCA, USA). Hybridized arrays were subjected to rigorous quality control including analysis of 5’ 3’ ratios (included range 0.40-0.79), percent present (included range 37-47%), average pair-wise correlation analysis and principle component analysis (PCA), resulting in the exclusion of 3 individuals.
Affymetrix Microarray Suite (MAS 5.0) was used for image processing and data acquisition. The Bioconductor package was used to compute normalized expression values from the Affymetrix.cel files. Statistical analysis was performed using R and Bioconductor software. Probe sets that met the criteria of being 50% present in at least one of the age/gender subgroups were retained in the analysis (33 210 probes sets retained, 61% of total number). Differential gene expression across chronological age and between males and females were analyzed in a linear regression model including age (log scale), gender and their interaction as independent factors and gene expression (log scale) as the dependent variable.
Microarray Validation
We confirmed selected results from the microarray experiment using quantitative real-time PCR (RT-PCR) from RNA extracted from 58 individuals as more cases were added to the collection over time (9). Wetargeted expression levels of a selection of transcripts that showed large expression changes between genders. Pre-designed TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA, USA) with specific primer and probe combinations were chosen for each of the genes analyzed: SMCY (Hs01104401_g1), NLGN4Y (Hs01034378_s1), PCDH11X/Y (Hs00263173_m1), PCDH11X (Hs01673213_m1), HSPA1A (Hs00271229_s1), HSPH1 (Hs00971475_m1), DNAJB1 (Hs00428680_m1), HSPD1 (Hs01036746_g1), HSPB1 (Hs00356629_g1) and HSP90AA1 (Hs00743767_sH; Table S5). The location along the transcript that was targeted by the primer/probe set for each gene was chosen to match as closely as possible the location along the transcript targeted by the Affymetrix gene chip assay. The Y version of PCDH11 could not be targeted specifically by qPCR, so we measured both the "pan" transcript (PCDH11X/Y) and the X version (PCDH11X). The ‘pan’ probe (Hs00263173_m1) detected mRNA from both PCDH11Y (isoform c) and PCDH11X (isoforms c and d). The other probe (Hs01673213_m1) was specific to the X chromosome (PCDH11X isoforms c and d).
Each 10 µl qPCR reaction contained primers (final concentration 900 nmol/L), FAM-labeled probe (250 nmol/L), and 1.14 ng cDNA in 1x Taqman Universal Mastermix containing AmpliTaq Gold DNA polymerase, deoxynucleoside triphosphates (dNTPs), uracil-N-glycosylaseand passive reference. The PCR protocol used involved incubation at 50ºC for 2 min and 95ºC for 10 min, followed by 40 consecutive cycles of 95ºC for 15 s and 60ºC for 1 min. Serial dilutions of pooled cDNA, synthesized from seven samples, one from each developmental time point, were included on every qPCR plate and used by Sequence Detection Software (SDS; Applied Biosystems) to quantify sample expression by the relative standard curve method. Control wells containing no cDNA template displayed no amplification in any assay. Efficiencies of the qPCR reactions ranged from 77% to 104%, with R2 values of between 0.95 and 1.00. All reactions were performed in triplicate. Samples were excluded if standard deviation of the triplicates was greater than 30% of the mean. Expression levels were normalized to the geometric mean of three ’housekeeper’ genes that did not change expression with development: HMBS (Hs00609297_m1), GUSB (Hs99999908_m1), and PPIA (Hs99999904_m1; Table S5). A fourth housekeeper gene, UBC (Hs00824723_m1), was included in normalization of HSPA1A,HSPH1,DNAJB1,HSPD1,HSPB1 andHSP90AA1 as these qPCRs were run from a different RNA isolation performed later. Population outliers were excluded if the normalized expression value was greater than 2 standard deviations from the group mean. One-way and two-way ANOVAs were performed, with gender or gender and age group as independent variables and normalized expression levels for the gene of interest as the dependent variable. To maximize power to detect early gender differences in HSP mRNAs by qPCR, we combined the neonate and infant groups and contrasted this to the post-infancy time points.
Supplementary Results
At each range of alpha level we found many more transcripts to be significantly changed by subject age than what would be expected by chance (Table S2). We identified a total of 8061 probe sets (5,136 genes) that changed significantly in expression across postnatal development at a fairly conservative threshold (p<0.001, two-sided) Overall, a large number (24%) of the transcripts expressed in the human prefrontal cortex changed across development and this level of significance (p<0.001) corresponds to a false discovery rate of ~0.4% based on a permutation analysis. Therefore, the majority of the developmental changes detected are expected to be genuine. Transcripts that change in an age-dependent manner were distributed among many functional GO categories and are also evenly distributed across the human chromosomes (Table S2).
Of the 8,061 probe sets that changed significantly with age (p<0.001) the magnitude of the fold change (maximum value vs minimum) in gene expression spanned a considerable range. Most of the changes in transcript expression were moderate (under 2-fold; 6548 probe sets), many were substantial (over 2-fold but under 5-fold; 1397 probe sets), and some were quite large (over 5-fold; 116 probe sets). When grouping significant age-dependent changes (p<0.001) by functional groups or gene ontology (GO) categories we identified hundreds of functional groups with genes that changed significantly during the maturation of the frontal cortex. The overall percent of genes that were developmentally regulated in any given GO category that contained 10 or more members varied considerably (<1% to 78%, chi-square p<10-6). Thus, significant developmental changes in gene expression did not occur uniformly across all cellular pathways as defined by the GO classification. The GO categories were sorted according to the percentage of transcripts that changed significantly by regression analysis or by ANOVA. Twenty-one GO categories had at least 50% of their member transcripts changing significantly with age. They are listed in order of those with the highest percentage of changed transcripts to the lowest: 1) NADH dehydrogenase activity, 2) NADH dehydrogenase (ubiquinone) activity 3) calcium- and calmodulin-dependent protein kinase activity, 4) synapse, 5) cytosolic large ribosomal subunit (sensu Eukaryota), 6) neurotransmitter secretion, 7) MAP kinase activity, 8) androgen receptor signaling pathway, 9) nuclear pore, 10) actin cytoskeleton, 11) protein tyrosine/serine/threonine phosphatase activity, 12) synaptic vesicle, 13) glycolysis, 14) membrane fusion, 15) actin filament, 16) thyroid hormone receptor binding, 17) protein kinase binding, 18) axon guidance, 19) JNK cascade, 20) Ras protein signal transduction and 21) vesicle-mediated transport.
Sex specific changes in gene expression were not equally distributed among the chromosomes (Table S2). Sex differences were found among <1% of the transcripts located on the autosomes. Male versus female differences in expression of genes on the autosomes were found to remain significant by doing a permutation test excluding X and Y chromosomal genes.While expression of the Y-linked genes can be 35 times higher in males as compared to females (Table S3), the fold change is somewhat arbitrary because the level of expression of the Y-linked genes in the female, as reported by the array software, corresponds to the background level for those genes.
In order to rule out that gender effects observed in HSP gene expression as measured by qPCR were associated with a differential cause of death in infant females we determined that no significant differences in frequencies of specific causes of death between infant males [SIDS (2/9), asphyxia (4/9), other (3/9)] and infant females [SIDS (4/5), asphyxia (1/5), other (0/5)] existed by chi-squared analysis (df=2, chi-squared=4.71, less than p<0.05).Thus, overall we were able to replicate the gender difference in HSPs in 4/6 transcripts by qPCR and we determined the gender difference was due to an increase in expression in infant females which was not secondary to a gender difference in the cause of death.
Supplementary Discussion
Protocadherin (PCDH11Y) belongs to a family of cell adhesion molecules that direct the formation of specific neuronal circuits and synapses (12) in the developing brain through neuroanatomically restricted expression and gene regulation (13). PCDH11X differs from PCDH11Y in amino acid structure and function (14, 15). We found a threefold increase in levels of total PCDH11 in infant males as compared to females and our data demonstrate that the X-linked form of PCDH11 is expressed at similar levels in males and females, indicating that the X-linked homolog is not compensating for gender differences in expression levels by being up-regulated in females.The increased expression of PCDH11 in the frontal cortex of infant males suggests that it may play a unique role in the organization of the brain circuits involved in sexually dimorphic behaviors and perhaps in the propensity to manifest psychiatric illness in developing males. Protocadherin family members have been proposed as etiological factors in the development of schizophrenia; however conclusive genetic evidence linking protocadherin to schizophrenia is lacking (16-18). However, further investigation of the role of protocadherins in developmental brain disorders that particularly impact males and involve the prefrontal cortex seems warranted.
The increased mRNA for neuroligin (NLGN4Y) that we detect in males could also drive male-specific brain development in the infant. NLGN4Y is a member of the neuroligin family of cell adhesion molecules that can influence neuronal contacts by changing the balance of excitatory and inhibitory synapses during development (19). Genomic mutations in neuroligins have been linked to the human developmental brain disorder, autism-spectrum disorder (20). The absence of NLGN4Y found by microarray in females was confirmed by qPCR demonstrating the lack of NLGNY-like transcript expression in females. Our results suggest that some Y chromosome genes may be more active in early human life but the reasons for this are unknown. It is possible that the increased secretion of testosterone that occurs peri-natally (21)and then decreases to low levels in toddlers and school-age males, impacts the early expression of the protocadherin and neuroligin gene in the male brain. However, it is unlikely that testosterone alone controls the expression of these Y chromosome genes up-regulated early in life as they are not up-regulated again during or after puberty when blood testosterone increases again. Instead, it is possible that distinct transcription factors, DNA methylation events or mRNA transcript stability may change as males develop postnatally.
In contrast to the transcripts that increased in infant males, another set of transcripts was found to have increased in expression specifically in infant females. The microarray study revealed eight heat shock protein (HSP) genes that differed in expression levels according to gender. The qPCR analysis confirmed that four HSP genes were significantly increased in females. The gender difference in expression of these HSP mRNAs are driven by specific increases in infant females (3-12 months old) relative to older females and all males. HSPs are known to play a crucial role in the folding, maturation, translocation to the nucleus and transcription regulatory activity of hormone receptors in general (22-24), and thus may be acting in a coordinated fashion to mediate some aspects of gender-specific response to hormones during brain development. More specifically, since HSP40, HSP70 and HSP90 compose the majority of components that are critical for maturation of the glucocorticoid receptor into a high steroid affinity state (25, 26), the infant female brain may be better equipped to bind, process and modulate stress hormones thanthe male brain or than older female brain.
Alternately, HSPs are versatile molecules responsive to more generalized cellular stress and thus are not only capable of refolding denatured proteins and promoting cellular survival, butare also active in immune surveillance (27). Hence, the infant female brain may be more resilient to oxidative damage or immune challenge. Increased HSP70 expression has been shown to decrease neonatal hypoxic/ischemic brain injury in a mouse model by moderating the pro-apoptotic effects of apoptosis-inducing factor (AIF) (28) and cytochrome c (29). These data suggest that gender differences in HSP expression, acting to oppose apoptotic pathways, may contribute to the decreased risk of respiratory infant death in females (30). Female levels of HSP expression are twice as high in rat heart, muscle, and kidney and in human serum (31, 32) suggesting the female bias in HSP expression is not restricted to the brain. HSP gene transcription is regulated by estrogen in the brain (33), however infant females have much lower levels of estrogen as compared to adult females when HSPs are down-regulated suggesting that estrogen may not be critical for this early gender difference in HSP mRNA expression. In the adult brain, gender differences in HSP expression are not always apparent, however gender differences in cortical and hippocampal HSP70 and HSP90 expression in response to antidepressant administration(34) and in midbrain HSP70 after alcohol-feeding (35) have been found.Thus, gender differences in HSPs in brain may be restricted to infants unless provoked by environmental triggers.
Recent evidence has suggested a role for HSPs in the aetiology of a number of neurodevelopmental disorders, including schizophrenia and autism. Polymorphisms in the HSP70 gene have been associated with schizophrenia in a Korean cohort (36), and increased expression of HSPA1A, HSPA1B and HSPB1 has been observed in the prefrontal cortex of patients with schizophrenia (37). Increased prevalence of antibodies against HSP60 (38), HSP70 and HSP90 (39) have been reported in patients with schizophrenia, with particular increases in HSP60 in female patients (40). HSPA6, HSPB1, HSPA1A and DNAJB1 have been shown to be increased in expression in individuals with autism (41). If HSPs contribute to the aetiology of such developmental brain disorders, gender dimorphism in their expression during development could contribute to sex differences observed in their onset, prevalence and vulnerability to neurodevelopmental insults and merit further investigation. For example, the increased expression of HSP in infant females may serve to protect them from neonatal damage due to hypoxia or immune challenge, and in so doing reduce the risk of later psychiatric illness, particularly since foetal hypoxia and 2nd trimester infection are considered risk factors for schizophrenia (42, 43). Indeed the elevated levels of HSP mRNA found in the prefrontal cortex of adult patients with schizophrenia have been suggested to be a long lasting response to early infective immune challenge (37).