Maturation of the Human Dorsolateral Prefrontal Cortex Coincides With a Dynamic Shift in MicroRNA Expression

Abstract

MicroRNA are small RNAs that provide specificity for the RNA induced silencing complex, which forms the basis of an exquisite combinatorial system for posttranscriptional regulation. This system, essential for complex metazoans, is exemplified in the development of the cerebral cortex. To explore the complexity of human cortical miRNA expression in detail, we analyzed RNA from postmortem prefrontal cortex from 97 subjects aged 2 months to 78 years using miRNA microarray. Global miRNA expression was highest in the early years before declining significantly after adolescence (n = 140 decreased, n = 32 increased). Late adolescence was also marked by an inflection point between miRNA on an upward trajectory vs the majority going down. Functional annotation of target genes displaying inverse mRNA expression patterns in the same tissue were overrepresented in neurodevelopmentally significant pathways including neurological disease (most significantly schizophrenia), nervous system development, and cell-to-cell signaling. As mature miRNA expression is largely posttranscriptionally regulated, miRNA biogenesis gene expression was also examined. Dicer and Exportin-5 displayed significant associations with age; however, neither correlated with global miRNA expression across the lifespan. This investigation of cortical miRNA expression provides a framework for understanding the complex posttranscriptional regulatory environment during development and aging that may form a substrate for changes observed in neurodevelopmental disorders.

Introduction

Development of the human prefrontal cortex and its vast array of projections and circuits is a complex dynamic process that involves an elaborate integration of genes through numerous regulatory systems. Efforts to understand the underlying mechanisms driving developmental gene expression have focused predominantly on genetic and epigenetic influences on transcription,¹ mediated by alterations in signal transduction pathways, their transcription factors, or gene promoter elements and associated chromatin structure. Emerging research, however, also highlights the potential for extensive posttranscriptional regulation of gene expression during neural development.² Small noncoding microRNA (miRNA) play a major role in these systems by functioning as a guide for the RNA induced silencing complex (RISC), which appears to be particularly important during the differentiation and development of the brain.^3–9 Indeed, a large proportion of human miRNA display a brain-enriched pattern of expression.⁷ Considering that each miRNA can potentially influence the expression of hundreds of target genes, the clinical implications of a disturbance to steady-state miRNA levels are substantial, particularly if such abnormality occurs during critical developmental periods. The human prefrontal cortex (PFC) is one of the last cortical regions to mature structurally and functionally and continues to develop into young adulthood.¹⁰ The PFC is one of the most functionally advanced regions of the brain, mediating working memory, attention, decision-making, and executive function.^11,12 It is suggested that any disruption in the development of the PFC may result in abnormalities in function that could increase vulnerability to psychiatric illness.^13–15 Abnormalities in miRNA expression have already been identified for numerous neurological and mental disorders including schizophrenia,^16–24 bipolar disorder,¹⁸ autism,²⁵ and fragile X mental retardation,²⁶ as well as Rett,²⁷ Tourette’s,²⁸ and DiGeorge²⁹ syndromes. In this study, we sought to identify miRNAs that were significantly associated with aging or displayed significant expression changes during neurodevelopmentally sensitive stages in an area of the PFC highlighted to be altered in several psychiatric illnesses, the dorsolateral PFC (DLPFC). As mature miRNA expression is largely posttranscriptionally regulated, we also examined the expression of genes involved in the miRNA biogenesis pathway. This was accomplished using whole-genome miRNA microarrays and quantitative real-time RT-PCR (qPCR) on postmortem tissue from nonpsychiatric individuals ranging from 2 months to 78 years of age. This revealed significant age-associated changes with a prominent decrease in miRNA expression appearing from teenage years onward, perhaps coinciding with cortical maturation.

Materials and Methods

Postmortem Brain Tissue

Fresh frozen postmortem prefrontal cortex tissue (Brodmann’s area 46) of 60 individuals was obtained from the National Institute of Child Health and Human Development Brain and Tissue Bank for Developmental Disorders (UMBB; NICHHD contract# NO1-HD8–3283) and of 37 individuals from the NSW Tissue Resource Centre (NSW TRC, University of Sydney; see online supplementary table 1). Use of these tissue cohorts was approved by the Institutional Review Board (University of Maryland, Baltimore, United States) and the Human Research Ethics Committee (University of Newcastle, Australia). Consent was obtained from the next of kin. All individuals were free of neurological and psychiatric symptoms at the time of death, and toxicological analyses showed them to be free of drug use. The overall distribution of ages between the 2 sample cohorts was different with the NICHHD covering the younger age group better (mean 12.21 years) than the NSW TRC (mean 51.14 years, P = 1.1^–20 ANOVA). The mean pH (P = .677), the freezer storage time (P = .287), and the control genes GUSB (P = .764) and HMBS (P = .818) did not differ between cohorts when using age as a covariate in ANCOVA. However, PMI was shorter the NICHHD cohort (16.52h) compared with the NSW TRC (24.79h, P = 3.0^–6), but this did not correspond with a reduction in RNA quality indicated by the respective RIN values, which were slightly lower in the NICHHD cohort (RIN = 6.63) compared with the NSW TRC (RIN = 7.30, P = .019), while still indicative of intact RNA.³⁰ For analysis, samples were subdivided into developmentally relevant age groups as done previously^31–34: neonate, infant, toddler, school age, teenage, young adult, adult, and older adults. Variables such as brain pH (F = 1.106, df = 7, P = .37), postmortem interval (PMI) (F = 0.983, df = 7, P = .45), and RNA integrity number (RIN; F = 2.071, df = 7, P = .14) do not differ between developmental groups (by ANOVA) nor do any of them significantly correlate with age.

RNA Extraction

Cortical grey matter tissue was carefully dissected from postmortem brain slices of the crown of the middle frontal gyrus, anterior to the genu of the corpus callosum.³⁵ Excess white matter was removed from each block of tissue with a scalpel or razor blade prior to extraction. Tissue was immediately homogenized and total RNA extracted using TRIzol reagent (Invitrogen, Life Technologies), according to the manufacturer’s instructions. RNA concentration and integrity was determined using the Bioanalyzer 2100 electrophoresis system (Agilent Technologies), where the mean RIN was 6.9. To eliminate any quality bias issues, all dissection, extraction, and quality control protocols were standardized and constant for both sets of tissue.

miRNA Microarray

Profiling of miRNA expression was achieved using a high throughput commercial bead-based miRNA microarray platform (Illumina). Each Sentrix array matrix contains oligonucleotide sequence probes for 470 annotated miRNA sequences corresponding to miRBase version 9.1. Total RNA (1 µg) was amplified according to the manufacturer’s instructions and labeled within a 96-well plate format for hybridization to the miRNA beadarray (Illumina Inc.). Microarray data were compiled and background subtracted within the BeadStudio software (Illumina, version 3.0) and normalized with respect to the geometric mean of U66 and U49 snoRNA expression. Of the 3 small RNA probe sets present on the microarray, geNorm³⁶ identified U66 and U49 to be the most stable across the cohort.

Linear Regression Analysis of Age

A linear model was used to analyze individual variables such as age, brain pH, PMI, and RIN. Age was the only variable to be significantly affecting miRNA expression. Simple regression analysis was then performed to identify the specific miRNA associated with age. In a series of linear regression models (1 model for each miRNA), age was included as the independent variable and miRNA expression as the dependent variable. To correct for multiple testing of genes, the P values were adjusted using Bonferroni step-down (Holm) correction. Because other demographic variables contributed significantly less than age (race 4%, RIN 3%, PMI 2%, gender <1% and pH <1%), they were not considered as covariates.

Hierarchical Cluster Analysis

Hierarchical clustering was performed on miRNA displaying significant association with age. Expression values were log₂-transformed and median-centered, then uncentered correlation by genes (average linkage) was accomplished using Gene Cluster v3.0 (Stanford University). Unsupervised clustering was also performed on sample Spearman correlation coefficients to identify groups of samples that displayed similar miRNA expression patterns. All clustering was displayed by heatmap and constructed using Java Treeview v1.1.6.³⁷

Quantitative Real-Time Reverse Transcription PCR

Microarray validation was performed by qPCR as described previously.¹⁶ Briefly, RNA was treated with DNase I (Invitrogen, Life Technologies, United States) and multiplex reverse transcription performed with Superscript II reverse transcriptase (Invitrogen, Life Technologies, United States) and a mix of miRNA sequence-specific primers along with primers for U6, U44, and U49 (for sequences, see online supplemantary table 2). Triplicate reactions were set up in a 96-well format using the epMotion 5070 automated pipetting system (Eppendorf, Germany) and carried out using the Applied Biosystems 7500 real-time PCR machine. Cycling conditions were 1 cycle for 2min at 50^oC, 1 cycle for 10min at 95^oC, 40 cycles of 15 s at 95^oC, and 1min at 50^oC (followed by standard dissociation curve analysis). qPCR was analyzed using the relative quantitation method with efficiency correction. cDNA was produced from human prefrontal cortex RNA samples, and serial dilutions were used as standards. Relative miRNA expression was calculated as the ratio of the miRNA and the geometric mean of controls U6 and U49 (the most stable of 3 controls as determined by geNorm). By ANOVA, the expression of U6 and U49 did not differ between age groups (P = .613 and P = .467 respectively, see online supplementary figure 1). Expression data were Pearson correlated with that obtained by microarray to determine the validity of the qPCR (PASW Statistics 20, SPSS Inc., IBM). To examine relative mRNA expression of miRNA biogenesis genes, cDNA synthesis was carried out as described above using random primers. Cycling conditions were 1 cycle for 2min at 50^oC, 1 cycle for 10min at 95^oC, 40 cycles of 15 s at 95^oC, and 1min at 60^oC (followed by standard dissociation curve analysis). Relative mRNA expression was calculated as the ratio of the gene and the geometric mean of controls hydroxymethylbilane synthase (HMBS) and beta-glucuronidase (GUSB). The expression of these genes did not differ (by ANOVA) between age groups (P = .155 and P = .679 respectively, see online supplementary figure 1). For sequences, see online supplementary table 2.

Target Gene and Pathway Analysis

The biological implications of age-associated changes in miRNA expression were explored using Ingenuity pathway analysis (IPA, Ingenuity Systems Inc.). The 172 altered miRNA mapped to 129 miRNA with unique seed-regions (those with common seed regions were grouped together). Putative and experimentally observed target genes were identified using the in-built functions within IPA: TargetScan, TarBase, miRecords, and the Ingenuity Knowledge Base. Gene expression analysis was performed in the same tissue previously,³⁸ so this mRNA data set was included to enable expression pairing between the miRNA and mRNA data sets. Only expression pairs displaying an inverse relationship were included in the analysis (miRNA up, target gene down and vice versa). This identified a total 1211 mRNA targets, and core analysis was performed on this gene list to identify relevant networks, pathways, and functions regulated by the age-associated miRNA/mRNA.

Results

miRNA Associated With Age

The relative expression of 470 miRNA (miRBase version 9.1) was analyzed in the DLPFC (BA46) of 97 postmortem samples (table 1). Of the 470 miRNA probes present on the microarray, 360 were expressed in this brain tissue. Global miRNA expression displayed a distinct profile, gradually decreasing from the neonatal period to adulthood, negatively associating with age (r = −.431, P = 1 × 10⁻⁵). Statistically significant decreases were observed from teenage years onward (13 years) compared with the neonate group (1–3 months) (ANOVA, F = 3.215, df = 7, P = .004, LSD post hoc P < .022), and adult miRNA expression is approximately 40% that of the neonate group (figure 1).

Table 1.

Demographic Information for Postmortem Prefrontal Cortex Tissue

Age Group	n	Age Range	Gender (M:F)	pH	Race (C:AA)	PMI	RIN
Neonate	7	0.1–0.2	5:2	6.5 (0.24)	5:2	21 (5.6)	6.0 (2.0)
Infant	13	0.3–0.9	8:5	6.6 (0.20)	2:11	17 (6.5)	6.9 (1.2)
Toddler	8	1.6–4.9	5:3	6.7 (0.26)	4:4	20 (4.7)	6.4 (1.2)
School age	9	5.4–13.0	5:4	6.6 (0.27)	7:2	15 (4.7)	6.7 (1.1)
Teenage	9	15.0–18.0	7:2	6.8 (0.08)	7:2	19 (6.6)	6.4 (0.9)
Young adult	10	20.0–24.9	7:3	6.7 (0.23)	7:3	19 (14.0)	6.9 (0.8)
Adult	30	33.0–49.2	24:6	6.7 (0.27)	26:4	21 (9.3)	7.2 (0.6)
Older adults	11	50.0–78.0	9:2	6.5 (0.29)	11:0	24 (13.4)	7.2 (0.6)

Note: C, Caucasian; AA, African American; PMI, postmortem interval (hours); RIN, RNA integrity number. Values listed for brain pH, PMI, and RIN are mean (standard deviation). All subjects returned negative toxicology results with the exception of one young adult subject (aged 20.97 years, cause of death: acute narcotic intoxication). Groups with fewer subjects (eg, neonate, toddler) may increase the risk of type II errors.

Fig. 1.

miRNA expression changes in the developing DLPFC. (A) The average expression of 360 miRNA was significantly decreased from the teenage group (13 years) onward. (B) Regardless of increasing/decreasing trajectory over the lifespan, age-associated miRNA were highly correlated until teenage years then proceeded in opposite directions. Bars are shown as the average expression for each developmental group plus standard error.

Individual variable correlations revealed that age was the most influential factor affecting miRNA expression. Using simple linear regression, 172 miRNA (48%) were identified to be significantly associated with age. The majority of miRNA decreased with age (140 miRNA) and only 32 increased with age (table 2, online supplementary table 3). Remarkably, irrespective of trajectory, the increasing (n = 138) and decreasing (n = 42) miRNA are highly correlated until teenage years (r = .722, P < .001) before proceeding in opposite directions (r = −.402, P = .001) (figure 1b). Quantitative real-time PCR array validation was performed on a selection of brain-associated miRNA including miR-16, miR-17-5p, miR-107, miR-181b, miR-195, and miR-219. These also represented most of the development-associated miRNA families. Expression data obtained by microarray and qPCR for each of these miRNA were significantly correlated over the lifespan (P < .05), with the exception of miR-16 displaying a strong trend (P = .063; see online supplementary figure 2).

Table 2.

Age-Associated miRNA Have Been Previously Implicated in Schizophrenia

Age-associated miRNA
Decrease with age (n = 140)
let-7b, let-7c, let-7d, let-7e, let-7f, let-7g, let-7i, 1, 101, 103, 105, 106b, 107, 124a, 126, 126*, 130a, 130b, 132, 135a, 136, 137, 138, 139, 142-3p, 142-5p, 146a, 148a, 148b, 153, 155, 15a, 15b, 16, 17-3p, 17-5p, 181b, 181c, 181d, 182*, 186, 187, 188, 190, 191, 193a, 199b, 19a, 19b, 20a, 212, 218, 22, 23a, 23b, 26b, 27b, 29a, 29c, 30a3p, 30a-5p, 30b, 30c, 30e-5p, 32, 324-5p, 33, 331, 335, 345, 361, 362, 365, 368, 369-3p, 374, 376a, 376a*, 376b, 377, 379, 381, 382, 383, 384, 409-3p, 409-5p, 410, 411, 421, 422b, 424, 432, 432*, 448, 450, 4543p, 455, 487a, 487b, 488, 490, 493-5p, 494, 496, 498, 503, 505, 506, 517b, 520d*, 525*, 539, 542-3p, 542-5p, 544, 545, 548b, 551b, 559, 577, 578, 582, 590, 600, 625, 631, 645, 647, 655, 656, 657, 660, 7, 769-5p, 802, 9, 95, 99a, 9*
Increase with age (n = 32)
133a, 189, 202*, 302b, 302b*, 31, 346, 363*, 412, 425, 507, 512-3p, 512-5p, 518a, 518b, 519b, 525, 527, 548a, 551a, 554, 567, 571, 583, 602, 610, 619, 622, 632, 641, 668, 801
Schizophrenia-associated miRNA
Decrease with age
DLPFC (BA46)	105, 107, 148b, 17-5p, 187, 199b, 382, 409-3p, 548b, 590
DLPFC (BA9)	let-7d, 101, 105, 107, 126*, 15a, 15b, 16, 181b, 181d, 190, 20a, 29c, 409-3p
STG (BA22)	let-7d, let-7e, 107, 130a, 138, 153, 155, 15a, 15b, 16, 17-3p, 17-5p, 191, 193a, 19a, 19b, 20a, 23a, 23b, 26b, 27b, 33, 331, 335, 379, 381, 409-5p, 432*, 448, 450, 455, 498, 517b, 9*, 99a
Increase with age
DLPFC (BA9)	31, 302b*, 512-3p, 519b
STG (BA22)	518b

Note: DLPFC, dorsolateral prefrontal cortex; BA, Brodmann’s Area; STG, superior temporal gyrus. A number of miRNA with altered expression in the postmortem SZ brain (BA9, 22, 46) appear to display changes in expression across the development of the normal brain (BA46). Of the 59 SZ-associated miRNA in the STG, 46% display age-related changes. Similarly in the BA46, 42% of the 33 SZ miRNA are age-related. The DLPFC BA9 however displays a bias whereby 65% of the 26 SZ miRNA display age-related changes in this study. Overall from an aging perspective, 20% of age-related miRNA are dysregulated in SZ, compared with 15% of non-age-related miRNAs.

A number of miRNA families were significantly associated with age including members of the let-7 family, miR-181 family, miR-15 family, and miR-17 family. Notably, these were shown to decrease in expression across the lifespan and have been previously implicated in schizophrenia (table 2).^16,17,39 Also displaying significant decreases with age were numerous brain-associated miRNA including miR-7, miR-9, miR-9*, miR-124a, miR-132, miR-137, and miR-138. miRNA belonging to the chromosome 19 miRNA cluster (C19MC) were overrepresented in the group of developmentally upregulated miRNA and included miR-512-3p, miR-512-5p, miR-518a, miR-518b, miR-525, and miR-527.

Hierarchical Clustering of miRNA Expression

Unsupervised cluster analysis was performed on the expression data for miRNA displaying significant associations with age. With red and blue indicative of high and low expression levels, respectively, the 140 miRNA decreasing with age and the 32 miRNA increasing with age are displayed clearly. In both groups, the distinct change in expression is apparent by the inflection point at approximately 17 years of age (figure 2).

Fig. 2.

Hierarchical clustering of 180 miRNA displaying age-related expression changes in the DLPFC. Hierarchical clustering was conducted on the 172 altered miRNA. Red = high expression and blue = low expression. Samples are arranged left to right in order of increasing age (years). The distinct change in miRNA levels to either increased or decreased expression occurred at approximately 17 years.

In order to identify groups of samples displaying similar miRNA expression patterns (in contrast to the predefined developmental age groups), each sample was correlated against all other samples utilizing all 360 expressed miRNA. Unsupervised clustering separated the cohort into 5 distinct groups based on similar patterns of expression (figure 3a). Age was the variable that significantly separated these groups (ANOVA on age by cluster, F = 28.56, df = 4, P < .0001). The median age for clusters 1–5 were 0.6 years, 15.7 years, 50.0 years, 48.0 years, and 51.0 years, respectively). Cluster 1 was comprised of predominantly infant samples; cluster 2 was comprised of children/adolescents; and clusters 3, 4, and 5 represented 3 subgroups of adults (not significantly varying in age, LSD post hocs P > .471). Variables such as brain pH, PMI, gender, and race were unable to separate clusters 3–5. Based on clusters 1–5 membership, average miRNA expression was shown to be significantly lower in clusters 3, 4, and 5 (adults) compared with the younger clusters 1 and 2 (figure 4b).

Fig. 3.

The expression of miRNA biogenesis genes across the lifespan. (A) DGCR8 mRNA displayed a 50% reduction from the neonate to infant groups but normalized by adulthood. (B) DROSHA mRNA did not display any age-related changes over the lifespan. (C) DICER mRNA increases over the lifespan, significant from young adulthood onward. (D) Exportin-5 mRNA decreases over the lifespan, significant from teenage years onward. Neonates 0.1–0.2 years, infants 0.3–0.9 years, toddlers 1.6–4.9 years, school age 5.4–12 years, teenage 13–18 years, young adults 20–24.9 years, adults 33–49.2 years, older adults 50–78 years. Bars are average miRNA expression for each developmental group. Error bars are standard error.

Fig. 4.

(A) Clustering of Spearman correlated samples reveals 5 distinct groups of expression. (B) Expression is grouped into infant (cluster 1), young adult (cluster 2), and adult groups (clusters 3–5). The adult clusters are significantly lower in expression that clusters 1 and 2. Bars are average miRNA expression. Error bars are standard error.

Expression of miRNA Biogenesis Genes Across the Lifespan

The overall pattern of global miRNA expression was a decrease with age and as such it was important to determine whether these changes were in response to changes in genes involved in miRNA biogenesis such as DGCR8, Drosha, Exportin-5, and Dicer. Components of the microprocessor complex, DGCR8 and Drosha, are responsible for cleaving pre-miRNA from the longer pri-miRNA transcripts in the nucleus. Neither of these displayed any significant associations with age (figure 3). Exportin-5 is a nuclear export protein responsible for transport of pre-miRNA from the nucleus into the cytoplasm. Exportin-5 mRNA expression was consistent with global miRNA expression, negatively correlating with age (r = −0.491, P = 8 × 10⁻⁵) and displaying significant decreases from the teenage group onward (ANOVA, F = 3.149, df = 7, P = .010, LSD post hoc P < .015). Over the lifespan, Exportin-5 mRNA decreased by 40%. Despite this significant decrease with age, Exportin-5 mRNA did not correlate with global miRNA expression on a sample-by-sample basis (r = .029, P = .828; figure 3). In the cytoplasm, pre-miRNA hairpins are cleaved by Dicer, resulting in double-stranded RNA and forming mature miRNA species. Dicer mRNA was shown to increase with age (r = .262, P = .045), and it displayed significant increases from young adulthood (19 years) onward (ANOVA, F = 2.793, df = 7, P = .020, LSD post hoc P < .029). Over the lifespan, Dicer mRNA increased 50%.

Target Gene and Pathway Analysis

To gain an appreciation of the biological implications of changes in miRNA expression with maturation and aging, the predicted targets of 129 age-associated miRNA (with unique seed regions) were examined using IPA. The predicted and experimentally verified target genes of the age-associated miRNA were filtered to include only those genes (mRNA) shown to be quantitatively associated with age, in this same tissue cohort.⁴⁰ Due to the predominately inverse relationship between miRNA and their target mRNAs, this gene list was further filtered to only include genes displaying inverse expression changes. Functional annotation performed on this list of 1211 genes revealed the most significantly associated biological functions to be neurological disease (251 genes), nervous system development/function (251 genes) and cell-to-cell signaling and interaction (123 genes; see online supplementary table 4). Schizophrenia- and psychosis-related genes were highly enriched (82 genes), and the most significant signaling pathways were ephrin receptor signaling, IL-3 signaling, integrin signaling, and axonal guidance.

Discussion

Genome-wide expression studies in postmortem brain have shown that many genes display age-related changes throughout development and ageing.^{31,32,34,40–43} Efforts to understand the underlying mechanisms driving changes in gene expression have focused predominantly on genetic influences on transcription¹; however, recent studies have emerged suggesting that there are also substantial posttranscriptional influences on gene expression. Posttranscriptional regulation mediated by miRNA, in particular, has been shown to play a major role in coordinating gene expression during the differentiation and development of the brain.^6,8,44,45 More specifically, there has been recent evidence to suggest that the unique developmental gene expression patterns of the human prefrontal cortex (compared to other primates) are driven by trans-acting regulators such as miRNA.⁴⁶

In this study, we examine genome-wide miRNA expression throughout the postnatal development of the dorsolateral prefrontal cortex. This investigation of 97 samples spanning 2 months to 78 years of age is the most comprehensive analysis of human neurodevelopmental miRNA expression to date. Nearly half of the miRNA expressed in this tissue displayed a significant association with age, and remarkably, a majority of these molecules were observed to decrease with age (figure 1; table 2, online supplementary table 3). This included most of the miRNA known to be functionally significant in the brain including miR-9, miR-17, miR-124a, miR-128ab, and miR-132 and miRNA previously dysregulated in schizophrenia including the miR-15 family, miR-107, miR-181b, miR-219, miR-7, miR-328, miR-382,^16,17,39 and miR-137, which was recently shown to be genetically associated with schizophrenia.⁴⁷ This significant age-dependent change in miRNA expression is consistent with a study by Somel et al. in the human superior frontal gyrus (SFG), where 31% of expressed miRNA were shown to be associated with age.⁴⁸ This age-dependent decrease is also consistent with the changes observed in the mouse cerebral cortex,⁴⁹ exemplified by decreases in miR-124a, miR-130a, miR-130b, and miR-9 across the postnatal lifespan. miR-128^48,50 has also been reported to display reduced expression during the development of the mouse and human cortices. This study also shows miR-23a,b and miR-29a,c strongly decreasing with age, in contrast to previous reports that show miR-23 and miR-29 expression increasing during postnatal development.^48,50 This disparity may be explained by discrete regional differences between human SFG and DLPFC miRNA expression. Similar temporospatial diversity has also been observed in the murine central nervous system, where substantial differences in miRNA expression is observed between brain regions.⁵¹ More recently, Somel et al.⁴⁶ identified decreasing trajectories of miR-320b, miR-454, and miR-92a in the prefrontal cortex across the lifespan (in 12 individuals from 0–98 years), which was consistent with our observations.

Global miRNA expression was observed to display significant reduction from the teenage group onward, and the cluster analysis of age-associated miRNA expression shows a prominent inflection point at approximately 17 years, marking the switch in both the low-to-high and high-to-low expressing miRNA molecules. A striking feature of the significantly altered miRNA was that despite their overall increasing or decreasing trajectory, these groups were highly correlated up until adolescence. This is a time important for cortical maturation and also involves a significant reorganization of the prevailing posttranscriptional regulatory environment. Unsupervised clustering of the entire cohort of samples separated the samples into 5 distinct clusters. Clusters 1 through 5 represented distinct groups of infants, children, and 3 groups of adults, which suggests miRNA expression could be divided into 3 separate stages throughout development and aging. The 3 adult subgroups were unable to be separated by variables such as brain pH, PMI, gender, or ethnicity; however, it is possible that the differences between these adult clusters is due to other environmental influences such as alcohol intake or nicotine use. Unfortunately, no additional patient information of this kind is available to further investigate this hypothesis. In accordance with our observed decrease in miRNA expression with age, each of the 3 adult clusters displayed significantly lower miRNA expression than the infant/childhood groups. These infant, childhood, and adult clusters of miRNA expression are broadly consistent with a recent gene expression study of the prefrontal cortex over the lifespan, where distinct transcriptional programs throughout development in the fetal stage, infancy, childhood, and adulthood/aging were observed.¹

miRNA have been shown to be implication in brain maturation and plasticity, so disruption of this program can be expected to have significant implications for neurodevelopmental disorders that are known to emerge at this stage of development. Dendritic spine structure is known to be indicative of brain maturation and their function forms a crucial part of brain plasticity. Cortical layer- and brain region-specific alterations in dendritic spine density have been reported in disorders such as autism and schizophrenia.⁵² Therefore, it is significant that a number of laboratories, including our own, have recently identified schizophrenia-associated alteration of cortical miRNA expression in postmortem samples.^16–20 Our investigations of miRNA expression in the DLPFC and STG from subjects with schizophrenia were characterized by an increase in numerous miRNA that display an age-associated expression pattern in normal development; the majority were also shown to decrease with age in the normal brain in contrast to their upregulation in schizophrenia^16,17,39 (table 2). It is possible that in schizophrenia, these neurodevelopment-associated miRNA are expressed at higher levels before adulthood, and instead of undergoing the transition to a lower expression during normal brain maturation, in schizophrenia, they remain elevated and enforce inappropriate levels of gene silencing. This elevation of miRNA expression observed in schizophrenia may have important implications for cognitive function in the mature brain, particularly in forebrain neurons. This is supported in a conditional Dicer knockout model, which was recently shown to be associated with increased cognitive performance.⁵³ Less is known about miRNA dysregulation in autism; however, studies utilising postmortem and peripheral tissue samples are emerging.⁵² A number of neurodevelopment-associated miRNA in this study have links to autism such as miR-23a and miR-381. Some are implicated in autism and schizophrenia (miR-181b, miR-195, and miR-212) including mIR-132, shown to be heavily involved in Fragile-X mental retardation.²⁶

miRNA expression is posttranscriptionally regulated to a large extent by proteins in their biogenesis pathway, enabling substantial changes to be coordinated by a few key molecules.⁵⁴ To investigate the developmental influence of this pathway on miRNA expression, we also examined the expression of these components to see if they could have any relationship to the prevailing miRNA expression at different ages. In contrast to expectation, no changes were observed in the microprocessor component Drosha and DGCR8 over the postnatal lifespan as a whole although DGCR8 did decline early before returning to postnatal levels in adults. Exportin-5, involved in the nuclear export of pre-miRNA, displayed high expression during infancy before halving by adulthood and was the most consistent with the global trend in miRNA expression. Dicer, involved in miRNA maturation, displayed increased expression from the infant stage through adulthood in contrast to the pattern of global miRNA expression, which declined with age.

We previously observed an increase in the expression of miRNA biogenesis-associated transcripts in the superior temporal gyrus¹⁶ and DLPFC^16,39 in schizophrenia that was consistent with broad changes in miRNA expression. The microprocessor components (DGCR8 and Drosha) are thought to be rate-limiting in the miRNA biogenesis pathway,⁵⁴ so their up-regulation in schizophrenia formed a possible explanation for increased miRNA expression in these samples. In this study, mRNA expression of genes in the biogenesis pathway overall did not display significant correlations with global miRNA expression or the significantly increasing/decreasing groups. This suggests that other factors or perhaps a combination are responsible for regulating developmental miRNA, as many influences on miRNA biogenesis beyond those tested have been identified.⁵⁵ It is also possible that there is a disconnect between the mRNA and protein expression trajectories and that functional protein levels correlate better with mature miRNA expression. Conversely, some of the miRNA biogenesis genes are themselves susceptible to regulation by miRNA and may be important targets of those altered during development and aging. According to miRNA target prediction analysis (microrna.org/miRanda), Dicer’s 1.5kb 3’UTR contains well-conserved binding sites for 65 different miRNA, of which nearly half were represented among the group of miRNA significantly decreasing with age, including some miRNAs with multiple binding sites such as miR-103/107 with 6 separate predicted sites. This contrasted with only 1 binding site for a miRNA (miR-31) displaying increase in cortical expression during development. This suggests that Dicer has significant potential for age-related regulation by miRNA, and hence, its increased expression could be attributed to a relaxation in miRNA-induced gene silencing. Approximately 25% of the binding sites are also for miRNA that we have identified to be dysregulated (upregulated) in schizophrenia. It is possible that Dicer may be serving a role in miRNA processing, as well as being susceptible to miRNA regulation due to its large 3’UTR containing many miRNA binding sites that is in contrast to the 3’UTRs of Exportin-5, Drosha, and DGCR8.

As mature miRNA can regulate the expression of hundreds of target genes, the substantial age-associated alteration is likely to have significant functional implications. To gain some insight into the developmental biology of these large shifts in miRNA expression, we cross-referenced putative miRNA targets with published genome-wide gene expression data from the same cohort.⁴⁰ Target genes reciprocally regulated by miRNA displaying age-related changes were identified and subjected to pathway analysis using IPA. The most significantly associated biological functions were cell cycle, neurological disease, nervous system development/function, and cell-to-cell signaling and interaction (see online supplementary table 4). The most significant signaling pathways were ephrin receptor signaling, IL-3 signaling, integrin signaling, and axonal guidance. These overrepresented pathways are likely indicative of the well-characterized decrease in cell proliferation and increase in neuronal differentiation throughout development. Schizophrenia-associated genes (compiled from the IPA database) were highly enriched among the miRNA targets, highlighting the importance of these miRNA in neurodevelopment and how disruptions to the miRNA system could alter gene expression patterns and contribute to disorders such as schizophrenia.

The brain-specific miR-9 was one of the most significantly altered miRNA over the lifespan, displaying a trajectory that decreased with age (r = −.47, adjusted P = .007), and miR-137 displayed the most robust changes over the lifespan, almost a 50% reduction (r = −.53, adjusted P = .00011). Recently, a large study of neurodevelopmental gene expression in the human DLPFC reported that following fetal development, there are dramatic changes in gene expression, with almost three-quarters of altered genes displaying a reversal in their trajectories between their fetal and postnatal stages.¹ A striking feature of genes displaying increased fetal expression followed by a postnatal decline was the significant enrichment of certain miRNA target genes. The miR-9 target genes were the most overrepresented in this cluster followed by those for miR-137 and remarkably the majority (139 of the 172, 81%) of the neurodevelopmentally associated miRNA. Functional analysis of these target genes suggested roles in axonal function, or more specifically the pruning of exuberant axons, which in turn is further evidence for the potential roles these miRNA likely play in brain development.

There are numerous molecular and cellular changes that occur throughout the normal development of the prefrontal cortex,^40,41 and there is increasing evidence implicating miRNA and miRNA functions in brain development, plasticity, and neurodevelopmental and psychiatric disorders. Disease-associated changes in miRNA expression may also have potential utility as biomarker candidates. We recently investigated miRNA expression in peripheral blood mononuclear cells and identified significant changes associated with schizophrenia.²⁴ Interestingly 80% of the neurodevelopmentally associated miRNA in this study are also expressed in peripheral cells and could potentially predict susceptibility to schizophrenia later on in life. Since the initiation of this study, there have been significant revisions and additions to the known human complement of miRBase annotated miRNA, which are not covered in this study. Therefore, it is likely that new technology will capture more miRNA species using higher resolution techniques such as RNA sequencing. Future studies will potentially extend the profiles captured in this work or perhaps identify new phenomena not observed with this approach. This study has provided support for the role of miRNA in the regulation of gene expression throughout the lifespan and also highlighted that disturbances to this regulatory system at critical time points in development could be an additional insult conferring increased vulnerability to psychiatric illness later in life.

Supplementary Material

Supplementary material is available at http:// schizophreniabulletin.oxfordjournals.org.

Funding

Schizophrenia Research Institute utilizing infrastructure funding from New South Wales Ministry of Health and the Macquarie Group Foundation; National Health and Medical Research Council of Australia, Schizophrenia Research Institute, and National Institute of Alcohol Abuse and Alcoholism (R24AA012725 to New South Wales Tissue Resource Centre at the University of Sydney); National Health and Medical Research Council Project Grant (631057 to MC); National Alliance for Research on Schizophrenia and DepressionYoung Investigator Award, the Hunter Medical Research Institute and an M.C. Ainsworth Research Fellowship in Epigenetics (to MC); National Health and Medical Research Council of Australia (1021970), Schizophrenia Research Institute, the University of New South Wales, and Neuroscience Research Australia (to CSW).