neurogenetics

, Volume 15, Issue 4, pp 267–287 | Cite as

A quantitative transcriptome reference map of the normal human brain

  • Maria Caracausi
  • Lorenza Vitale
  • Maria Chiara Pelleri
  • Allison Piovesan
  • Samantha Bruno
  • Pierluigi Strippoli
Original Article

Abstract

We performed an innovative systematic meta-analysis of 60 gene expression profiles of whole normal human brain, to provide a quantitative transcriptome reference map of it, i.e. a reference typical value of expression for each of the 39,250 known, mapped and 26,026 uncharacterized (unmapped) transcripts. To this aim, we used the software named Transcriptome Mapper (TRAM), which is able to generate transcriptome maps based on gene expression data from multiple sources. We also analyzed differential expression by comparing the brain transcriptome with those derived from human foetal brain gene expression, from a pool of human tissues (except the brain) and from the two normal human brain regions cerebellum and cerebral cortex, which are two of the main regions severely affected when cognitive impairment occurs, as happens in the case of trisomy 21. Data were downloaded from microarray databases, processed and analyzed using TRAM software and validated in vitro by assaying gene expression through several magnitude orders by ‘real-time’ reverse transcription polymerase chain reaction (RT-PCR). The excellent agreement between in silico and experimental data suggested that our transcriptome maps may be a useful quantitative reference benchmark for gene expression studies related to the human brain. Furthermore, our analysis yielded biological insights about those genes which have an intrinsic over-/under-expression in the brain, in addition offering a basis for the regional analysis of gene expression. This could be useful for the study of chromosomal alterations associated to cognitive impairment, such as trisomy 21, the most common genetic cause of intellectual disability.

Keywords

Human brain Gene expression profile Integrated transcriptome reference map Meta-analysis 

Introduction

The entire set of transcripts of a given cell (or cell population), on which its differentiation and functioning or development depends, has been defined as transcriptome [1]. The production of transcripts occurs through the molecular mechanism of gene expression, the process through which DNA molecules are converted into RNA molecules, including protein-coding messenger RNA (mRNA), ribosomal RNA, transfer RNA and non-coding RNA.

Whole-genome expression profiling enables the identification of differentially regulated genes under various conditions, distinguishing, for example, different types of cells or tissues and investigating differences between a pathological and a normal condition [2]. The first powerful technology used to study the global patterns of gene expression—hence used to measure the expression level for thousands of genes simultaneously and to explore the functional identities of different tissues or the same tissue under different biological conditions—was DNA microarray. This technique is the most frequently used method of studying either DNA genotypes or whole-genome expression, using single colour or two-colour fluorescent hybridization of target fragments with specific DNA probes attached to the solid surface of a biochip [3]. In spite of the recent advent of techniques of direct sequencing of the transcripts (RNA-seq), microarrays are considered the most accurate tool to measure the levels of gene expression [4]. With expression profiling assays of many kinds in routine use, transcriptome analysis has become a general phenotyping method [5].

The most complex organ for gene expression profile investigation is the brain [6]. The brain is the main organ of the central nervous system (CNS). It is an immensely complex organ composed of billions of precisely interconnected neurons which together allow it to carry out sensory, motor and cognitive functions. Impaired development of cognitive functions is the cause of intellectual disability (ID). The causes of ID may be genetic or non-genetic: they are of genetic origin if they are due to abnormalities in a single gene or to structural or numerical abnormalities of chromosomes. Trisomy 21 (Down syndrome) is the best known genetic cause of ID [7].

Homo sapiens is the mammal with the highest number of genes expressed in the brain [8, 9]. To date, our current knowledge of human brain gene expression is based on reports of studies that used microarray techniques for transcriptome profiling performed on post-mortem brain tissue [2]. There have been several microarray experiments performed on the human brain to analyze the global gene expression level of tissue, to compare gene expression patterns between its different regions [10, 11] or to compare pathophysiological states [12, 13, 14, 15].

The European Bioinformatics Institute (EBI) and National Center for Biotechnology Information (NCBI) provide ArrayExpress [16] and Gene Expression Omnibus (GEO) [17] databases respectively, which are repositories of high throughput sequencing studies and hybridization array data that can be searched for and downloaded.

Our goal in this study was to perform a systematic meta-analysis of the gene expression profile of the whole normal human brain in order to provide a quantitative transcriptome reference map of it, i.e. a reference typical value of expression for each of the 39,250 known, mapped and 26,026 uncharacterized (unmapped) transcripts assayed by any of the experimental platforms used to this regard. This task implied the possibility of performing cross-platform, inter-array and intra-array data normalization in order to incorporate any publicly available dataset in the calculation. To this aim, we used the software named Transcriptome Mapper (TRAM), which is able to generate transcriptome maps of this type from any source listing gene expression values for a given gene set, e.g. expression microarray [18]. In addition to providing reference gene expression values (in the form of percentages of the mean value), allowing quantitative comparisons among the expression of all investigated genes, it maps gene expression along the chromosome’s physical map, thus also allowing discovery of regional over- or under-expression within a biological condition or while comparing two different biological conditions.

The use of the whole (in toto) brain is justified by its recognition as a specific organ performing superior cognitive functions, and a wealth of gene expression data has been produced from whole brain RNA by different methods (among many others, for example Northern blot analysis, expressed sequence tag (EST) libraries, polymerase chain reaction (PCR) assays and microarray experiments). Due to the very recent demonstration of differences in adult brain regional gene expression according to sex [19], we extended our analysis of the whole brain to the comparison of male- versus female-derived samples.

In addition, we extended our analysis to a comparison with human foetal brain gene expression, with a pool of human tissues and to two normal human brain regions: cerebellum and cerebral cortex, two of the main brain regions severely affected when cognitive impairment occurs [20, 21, 22], as happens in the case of trisomy 21 [23]. This also yielded biological insights about those genes which have an intrinsic over-/under-expression in the brain or brain subregions, thereby offering a basis for the regional (chromosomal) analysis of gene expression. This could be useful for the study of chromosomal alterations associated to cognitive impairment, such as trisomy 21, the most common genetic cause of ID.

Finally, data obtained by our meta-analysis were validated by assaying gene expression through several magnitude orders by ‘real-time’ reverse transcription PCR (RT-PCR). We found an excellent agreement between in silico and experimental data, thus suggesting that our transcriptome map may be a useful tool as a quantitative reference for gene expression studies related to the human brain.

Materials and methods

Database search

In order to retrieve datasets derived from whole normal adult human brain, foetal brain, cerebellum and cerebral cortex, we made a systematic search in gene expression data repositories for any single sample available listing gene expression values for these tissues. GEO functional genomics repository was searched for: ‘Homo sapiens [ORGANISM] AND brain’ (or ‘cerebellum’, or ‘cortex’). ArrayExpress database of functional genomics experiments was searched at http://www.ebi.ac.uk/arrayexpress/ for the terms ‘brain’ (or ‘cerebellum’, or ‘cortex’), choosing ‘Homo sapiens’ as organism. This strategy was used to ensure high sensitivity in the search. Search results were then filtered using inclusion and exclusion criteria as explained below in the Dataset selection section.

In addition, the term ‘Tissue’ was used to retrieve datasets derived from collection of different human tissues analyzed in the same databases (in GEO: ‘Homo sapiens [ORGANISM] AND tissue*[TI] OR organ*[TI]’; in ArrayExpress: ‘Tissue’, choosing ‘Homo sapiens’ as organism). This led to generate a pool of samples including all the main human organs and tissues except the brain, serving as a comparison set to highlight brain-specific differential gene expression compared to all the other anatomical human structures.

The searches were made up to May 2013.

Dataset selection

The inclusion criteria of datasets in the analysis were experiments carried out on the whole organ or tissue, normal phenotype of individuals, adult or foetal (for the foetal brain transcriptome map only) age of the subject from whom the sample was obtained and availability of the raw or pre-processed data.

Exclusion criteria were exon arrays (hampering data elaboration by TRAM due to an exceedingly high number of data rows) or platforms using probes split into several distinct arrays for each sample (hampering intra-sample normalization); lack of identifiers corresponding to those found in the GEO sample records (GSM) or ArrayExpress sample records; platforms assaying an atypical number of genes (i.e. <5.000 or >60.000); and data derived from cell lines, pathological or treated tissue and children or foetal (except in foetal brain transcriptome map) tissues.

In order to obtain a quantitative transcriptome map, values from each dataset were linearized when provided as logarithms. In some cases, we used raw files (e.g. CEL file) to be converted into pre-processed data, using the software ‘AltAnalyze’ [24].

TRAM analysis

TRAM software [18] allows the import of gene expression data recorded in the NCBI GEO and EBI ArrayExpress databases in tab-delimited text format. It also allows the integration of all data by decoding probe set identifiers to gene symbols via UniGene data parsing [25], normalizing data from multiple platforms using intra-sample and inter-sample normalization (scaled quantile normalization) [26], creating graphical representation of gene expression profile through two ways, ‘Map’ and ‘Cluster’ mode, and determining the statistical significance of results.

We created a directory (folder) for each tissue, containing all the sample datasets related to the same source and selected for the study. To compare brain samples with a pool of human tissues (without brain samples), we collected the first in a folder named Pool 'A' and the second in a folder named Pool 'B'; to compare foetal brain with adult total brain, we collected the first in a folder named Pool 'C' and the second as above (Pool 'A'); to compare the brain regions with total brain, we collected the samples from cerebellum in a folder named Pool 'D' and the samples from cerebral cortex in a folder named Pool 'E' (Table 1). In addition, we also provided two datasets deriving from the adult brain samples for which the sex of the sample donor was available: male brain (Pool 'A.1') and female brain (Pool 'A.2'). We did not consider samples deriving from male/female mix of tissue. The comparisons allowed us the analysis of differential transcriptome maps, using the ratio of the mean expression values for each locus, in addition to the maps related to each single type of sample.
Table 1

Sample pools selected for the meta-analysis of gene expression profiles in whole adult brain (pool 'A'), male adult brain (pool 'A.1'), female adult brain (pool 'A.2'), pool of non-brain tissues (pool 'B'), foetal brain (pool 'C'), cerebellum (pool 'D') and cerebral cortex (pool 'E')

Pool

Sample type

Sample number

TRAM mapped loci

Pool 'A'

Whole adult brain

n = 60

39,250

Pool 'A.1'

Male adult brain

n = 13

27,437

Pool 'A.2'

Female adult brain

n = 5

25,954

Pool 'B'

Pool of non-brain tissues

n = 622

34,985

Pool 'C'

Foetal brain

n = 35

38,483

Pool 'D'

Cerebellum

n = 140

38,163

Pool 'E'

Cerebral cortex

n = 18

27,504

Information for each individual sample used is provided in Supplementary Tables 1 and 2

We ran the whole set of analyses permitted by TRAM (in both Map and Cluster mode, although we focused on the Map mode) using default parameters as described [18]. We used an updated version of TRAM including enhanced resolution of gene identifiers and updated UniGene and Entrez Gene databases (TRAM 1.1, June 2013), in comparison with the original 2011 version [18]. TRAM is freely available at http://apollo11.isto.unibo.it/software. Briefly, gene expression values were assigned to individual loci via UniGene, intra-sample normalized as percentage of the mean value and inter-sample normalized by scaled quantile. The value for each locus, in each biological condition, is the mean value of all available values for that locus. The whole-genome gene expression median value was used in order to determine percentiles of expression for each gene. Although TRAM is a map-centred transcriptome analysis tool, it can also summarize and allow the analysis of gene expression data of unmapped genes, exploiting its capability of parsing and normalization in order to highlight differential expression of single genes between two biological conditions even in the absence of data about genomic location of the gene [18].

Using the Map mode graphical representation, we searched for over-/under-expressed genome segments, which have a window size of 500,000 bp and a shift of 250,000 bp. The expression value for each genomic segment is the mean of the expression values of the loci included in that segment. A segment is defined over-/under-expressed if it has an expression value which is significantly different between two conditions analyzed and contains at least three individually over-/under-expressed genes, e.g. genes which have expression values within the highest and the lowest 2.5th percentile. Significance of the over-/under-expression for single genes was determined by running TRAM in Map mode with a segment window of 12,500 bp. This window size corresponds to about a quarter of the mean length of a gene, so the significant over-/under-expression of a segment almost always corresponds with that of a gene. When the segment window contains more than one gene, the significance is maintained if the expression value of the over-/under-expressed gene prevails over the others.

For the creation of the maps, TRAM software does not consider probes where the expression values are not available, assuming that an expression level has not been measured. Furthermore, it gives 95 % of the minimum positive value present in a sample to those expression values equal to or lower than ‘0’, in order to obtain meaningful numbers when we need to obtain a ratio between values in pool 'A' and pool 'B'. Assuming that in these cases, an expression level is too low to be detected under the experimental conditions used, and this transformation is useful to highlight differential gene expression.

Housekeeping gene search

We determined the predicted genes that behave like housekeeping genes, in that they are mainly involved in fundamental cellular function and are universally and constantly expressed in all tissues [27, 28, 29]. A search of housekeeping genes in the transcriptome maps was performed using the following parameters in combination: expression value > 100, in order to select genes expressed above the mean value and so at an appreciable level; data points number ≥ half the number of samples of the map, in order to select commonly expressed genes (≥311 for pool of tissues minus brain transcriptome map; ≥30 for brain transcriptome map; ≥17 for foetal brain transcriptome map); standard deviation (SD), expressed as a percentage of the mean value, ≤30 or ≤40, in order to identify genes with a low expression variation among different samples. We searched for SD ≤ 30 to identify the first gene with the lowest SD, instead SD ≤ 40 to identify multiple genes which behave like housekeeping genes.

Gene selection for map validation

In order to obtain a sample experimental confirmation of the meta-analysis derived map, for each tissue, we selected a group of genes with these features: range of expression values covering the whole range of the expression magnitude order as calculated by TRAM; regular spacing of the expected expression values, i.e. each gene is expected to have a fold increase detectable through real-time RT-PCR (at least 1 PCR cycle) in comparison with the subsequent gene with a lower expression in the group; the gene is known and characterized; and when possible, the chosen gene is known to have a specific function in the brain.

RT-PCR

Complementary DNA (cDNA) templates were obtained from RT of commercial human brain total RNA, human cerebellum total RNA and human cortex total RNA (Clontech, Mountain View, CA). They were derived from a normal whole brain (an 18-year-old Caucasian male), normal cerebellum (pooled from ten male/female Caucasians, age 22–68 years) and normal cerebral cortex (pooled from five male Asians, age 20–44 years), respectively.

RT conditions used were 4 μg of total RNA (1 μg/μL), SuperScript III First-strand Synthesis Supermix (Invitrogen by Life Technologies, Grand Island, NY, USA) containing RT enzyme mix (includes SuperScript III RT 200 U/μL) 8 μL and RT reaction mix (includes oligo dT-20 2.5 μM, random hexamers 2.5 ng/μL, MgCl2 10 mM and dNTPs 10 mM) 40 μL. The RT-PCR reaction was performed in a final volume of 80 μL to have the same template for all the subsequent reactions.

The reaction consisted of three steps: an incubation of 10 min at 25 °C, followed by an incubation at 50 °C for 30 min and a final step of 5 min at 85 °C. Escherichia coli RNase H 4 μL (8 U) was then added to the reaction for 20 min at 37 °C.

Primer design

Primers pairs were designed with ‘Amplify 3’ software [30] following standard criteria [31]. They are designed to specifically recognise expressed sequences (each primer being designed on a different exon) and to bind to regions common to all isoforms of the same gene because microarray probe sequences complement the known isoform sequences of the same gene. These constrains cause a variation in the amplicon lengths between 82 and 247 bp. Each primer is about 20–23 nt long, with an annealing temperature of 61 °C.

PCR

First, a qualitative analysis of RT products was performed using PCR and agarose gel electrophoresis. PCR experiments were performed in a 25 μL final volume, containing 2.5 μL of cDNA, 1 U Taq polymerase (TaKaRa, Shiga, Japan) with companion reagents (0.2 mM each dNTPs, 2 mM MgCl2, 10× PCR buffer) and 0.2 μM of each primer. An initial denaturation step of 2 min at 94 °C was followed by 25 cycles of 30 s at 94 °C, 30 s at annealing temperature (Ta 61 °C), 30 s at 72 °C and a final extension of 7 min at 72 °C. All RT-PCR products obtained were gel analyzed following a standard method [32].

Real-time PCR profile and melting curve

Real-time PCR assays were performed in triplicate, using the CFX96 instrument (Bio-Rad Laboratories, Hercules, CA). The reactions were performed in a total volume of 20 μL containing the following: 2.5 μL of cDNA (final concentration 0.78 ng/μL); 10 μL of SYBR Select Master Mix 2× for CFX (Applied Biosystem, by Life Technologies) containing AmpliTaq® DNA Polymerase, UP (high purified), SYBR® GreenER™ dye and Heat-labile uracil-DNA glycosylase (UDG); 0.8 μL (0.3 μM) of both forward and reverse primer (MWG, Life Technologies); and 5.9 μL of RNase-free water. Cycling parameters were 2 min at 50 °C (UDG activation), 2 min at 95 °C (AmpliTaq Fast DNA Polymerase UP activation), 40 cycles of 15 s at 95 °C (denature) and of 1 min at 60 °C (anneal and extend). A melting step needs to be performed to assay amplification specificity. This step consisted of an increase in temperature of 0.5 °C/s from 65 to 95 °C.

For each gene, we used the primer pair that gave between 90 and 110 % efficiency. We used the ∆Ct (delta cycle threshold) method, a variation of the Livak method [33], that uses the difference between reference and target gene Ct values for each sample to do a relative quantification normalized to a reference gene (Observed Ratio(reference/target) = 2Ct(reference) − Ct(target)). For each gene group, we set the gene with an intermediate expression value and a low SD (expressed as percentage of the mean value) in TRAM analysis as reference gene. The ratio among the transcriptome map expression values was calculated by dividing each expression value of the target gene by the expression value of the reference (expected ratio). Then, we compared this value with the observed ratio, and we examined the relationship between these two variables through bivariate statistical analysis using JMP 5.1 software (SA Institute, Campus Drive Cary, NC).

Results

Database search and database building

The performed search, followed by checking for exclusion and inclusion criteria as described in the Materials and methods section above, retrieved 60 samples from 15 microarray experiments on whole adult brain, 35 samples from 13 microarray experiments on whole foetal brain, 140 samples from 15 microarray experiments on whole cerebellum, 18 samples from 4 microarray experiments on whole cerebral cortex and 622 samples from 12 microarray experiments on non-brain tissue pool. Dataset search on different human tissues retrieved ten articles describing gene expression profile for 53 different tissues or organs.

Sample identifiers and main sample features are listed in Supplementary Table 1 (brain, male brain, female brain, foetal brain, cerebellum and cortex) and Supplementary Table 2 (human non-brain comparison pool) (available also at http://apollo11.isto.unibo.it/suppl).

Datasets were loaded into TRAM and analyzed obtaining 11 transcriptome maps: adult brain (pool 'A'), adult brain (pool 'A') versus pool of tissues minus brain (pool 'B'), male brain (pool 'A.1'), female brain (pool 'A.2'), male brain (pool 'A.1') versus female brain (pool 'A.2'), foetal brain (pool 'C'), foetal brain (pool 'C') versus adult brain (pool 'A'), cerebellum (pool 'D'), cerebellum (pool 'D') versus adult brain (pool 'A'), cerebral cortex (pool 'E'), and cerebral cortex (pool 'E') versus adult brain (pool 'A').

Transcriptome differential maps

Each map provides the total of data points analyzed for each tissue, i.e. gene expression values (expressed as percentage of the mean value) for all human mapped loci following intra- and inter-sample normalization [18], the number of loci for each tissue and for which the comparison between two conditions (different tissues) was possible, due to the presence of values for those loci in both sample pools considered, and the number (at least three over-/under-expressed genes) and the gene content of each genomic segment found to be statistically significantly over-/under-expressed in the comparison between the two tissues. Each genomic segment was identified among the 12,373 segments generated using the default window of 500,000 bp with a sliding window of 250,000 bp and following removal of overlapping segments with similar gene content.

A segment or a gene was considered to be statistically significantly over-/under-expressed for q < 0.05, where q is the p value obtained by the method of hypergeometric distribution [18] and corrected for multiple comparison. Detailed results for each map are provided below and are also available at: http://apollo11.isto.unibo.it/suppl.

Adult brain versus pool of tissues minus brain transcriptome map analysis

In the adult brain transcriptome map analysis, 1,803,680 data points corresponding to 39,250 mapped loci (Supplementary Table 3) were included. A total of 22,699 data points of the map correspond to 565 chr21 mapped loci. Results obtained by analysis included 28 significantly over-expressed segments.

The genome segment that has the highest statistically significant expression value is on chromosome 12 (12q13.12) (Table 2a), including the over-expressed known genes TUBA1A, TUBA1B and TUBA1C, encoding respectively for brain-specific tubulin alpha 1a, alpha 1b and alpha 1c. There are no statistically significant under-expressed segments.
Table 2

List of the three most over-expressed in (a) or under-expressed in (b) genomic segments and genes (all significantly, with q < 0.05)

Segment or gene

 

Tissue

  

Whole adult brain

Whole foetal brain

Cerebellum

Cerebral cortex

a

 Over-expressed segments

  I

12q13.12

566.00

12q13.12

981.68

11q13.1

492.10

12q13.12

1051.22

 

TUBA1B TUBA1A TUBA1C

TUBA1B TUBA1A TUBA1C

Hs.736281 Hs.593027 MALAT1 Hs.712678 Hs.732685

TUBA1B TUBA1A TUBA1C

  II

Xp11.23a

561.23

4q21.3a

702.21

6p21.3

490.85

12q12

785.70

BEX1 BEX4 TCEAL5 BEX2 TCEAL7 NGFRAP1

RPS29 RPL36AL KLHDC2

GRM4 Hs.592692 RPS10 PACSIN1

ARF3 DDN TUBA1B TUBA1A TUBA1C

  III

6p21.3

524.86

6p21.3

658.07

11q13.1a

487.38

11q13.1

738.20

Hs.743967 Hs.592692 Hs.597332 RPS10 PACSIN1

C6orf1 Hs.743967 RPS10

MALAT1 Hs.712678 Hs.732685 CFL1

Hs.736281 Hs.593027 MALAT1 Hs.712678 Hs.732685

 Over-expressed genes

  I

BCYRN1

7,875.54

TUBA1B

6,954.44

BCYRN1

12,372.42

Hs.732685

12,387.16

  II

TUBA1B

4,876.25

TUBA1A

6,810.31

Hs.732685

8,742.11

TUBA1B

6,343.86

  III

Hs.732685

4,680.24

EEF1A1

6,336.99

CALM2

4,704.56

TUBA1C

6,093.47

 Over-expressed chr21 genes

  I

SOD1

1,727.39

DNAJC28

2,817.20

DNAJC28

1,447.08

OLIG1

726.82

  II

DNAJC28

878.57

SOD1

1,657.36

SOD1

1,432.22

PCP4

700.08

  III

APP

791.87

ATP5O

1,074.36

TIAM1

1,229.18

SOD1

629.70

b

 Under-expressed segments

  I

  

13q21.31a

3.19

 
  

Hs.375745 Hs.735749 Hs.551057

 Under-expressed genes

  I

TRG

2.93

Hs.439634

0.17

Hs.707129

0.99

ZNF852

3.59

  II

Hs.737002

3.00

Hs.674562

0.73

Hs.680393

1.01

Hs.28723

3.71

  III

Hs.729885

3.73

Hs.599650

1.01

OR2T29

1.1

Hs.441636

3.74

 Under-expressed chr21 genes

  I

Hs.542623

5.08

Hs.729539

2.31

Hs.561029

1.82

Hs.50927

5.48

  II

Hs.561029

5.13

LOC339622

3.10

Hs.677645

1.89

LOC339622

6.29

  III

Hs.542565

5.18

Hs.666775

3.16

Hs.580903

1.93

Hs.290805

6.44

Analysis was performed using default parameters (see Materials and methods section). Under each segment, the expression value and the genes included in the segment are indicated; under each gene, the expression value is indicated. In order to simplify, some segments are not shown because they overlap with those highlighted in one of the listed regions. The complete results for these models are available in the online supplementary material. Segments and genes are sorted by decreasing expression value in (a) and by increasing expression value in (b). The overlapping results among the transcriptome maps are marked in bold. In the ‘Map’ mode, TRAM displays UniGene EST clusters (with the prefix ‘Hs.’ in the case of H. sapiens) only if they have an expression value. Data refer to the transcriptome maps of whole adult brain, whole foetal brain, and cerebellum and cerebral cortex

aCytoband was derived from the UCSC Genome Browser

At single gene level, BCYRN1 (chr2), encoding for a brain cytoplasmic RNA 1, has the highest expression value (7,875.54) (Table 2a), and TRG (chr7), encoding for T cell receptor gamma, has the lowest expression value (2.93) (Table 2b). Among the chr21 genes, SOD1, encoding for superoxide dismutase 1 soluble, has the highest expression value (1,727.39), followed by DNAJC28 (878.57), encoding for DnaJ (Hsp40) homolog, subfamily C, member 28, and APP (791.87), encoding for amyloid beta (A4) precursor protein (Table 2a).

In the analysis of the adult brain versus pool of tissues minus brain TRAM map, regional differential expression of pool 'A' (60 total brain samples) versus pool 'B' (622 pool of tissues minus brain samples and 34,985 mapped loci listed in the Supplementary Table 4) was investigated. Results included 23 significantly over-expressed (n = 7) or under-expressed (n = 16) segments.

The genome segment that has the highest statistically significant expression value is on chromosome 5 (5q34) (Table 3a), including the over-expressed known genes GABRA6, GABRA1 and GABRG2, encoding respectively for gamma-aminobutyric acid (GABA) A receptor, subunit alpha 6, GABA A receptor, subunit alpha 1 and GABA A receptor, gamma 2. The genome segment that has the lowest statistically significant expression value is on chromosome 2 (2p12) (Table 3b), including the under-expressed known genes REG1B, REG1A and REG3A, encoding respectively for regenerating islet-derived 1 beta, regenerating islet-derived 1 alpha and regenerating islet-derived 3 alpha.
Table 3

List of the three most over-expressed in (a) or under-expressed in (b) segments and genes (all significantly, with q < 0.05)

Segment or gene

 

Tissue

  

Whole adult brain versus pool of non-brain tissues

Whole foetal brain versus whole adult brain

Cerebellum versus whole adult brain

Cerebral cortex versus whole adult brain

a

 Over-expressed segments

  I

5q34

11.35

8q13.1a

7.51

15q22.2

5.23

1q23.1

17.55

GABRA6 GABRA1 GABRG2

LOC100130155 Hs.388788 BHLHE22

RORA Hs.660127 Hs.655820

OR6Y1 OR6K2 OR6N1 PYHIN1

  II

13q21.32

9.25

2q33

6.99

12q22a

4.05

11q12.1

7.8

PCDH9 Hs.656886 Hs.676018

SATB2 Hs.151184 SATB2-AS1

Hs.585087 BTG1 Hs.434392

OR5F1 OR5T2 OR8U1

  III

3q24

6.86

10p14a

4.06

1p31.3

3.87

11q11

7.33

ZIC4 Hs.720460 ZIC1

Hs.734120 Hs.676705 Hs.735675 Hs.655681

INADL Hs.737385 Hs.673484

OR5L2 OR5F1 OR5T2

 Over-expressed genes

  I

ANKRD30B

79.69

TMSB15A

63.15

CBLN3

81.34

ZNF790

255.06

  II

DNAJC28

65.06

DCX

42.79

Hs.665664

17.65

Hs.197693

110.78

  III

CDR1

57.38

Hs.712990

40.03

Hs.12316

16.97

Hs.594912

86.53

 Over-expressed chr21 genes

  I

DNAJC28

65.06

CXADR

7.40

Hs.657183

6.83

KRTAP13-2

32.77

  II

LINC00320

39.82

LRRC3DN

4.86

ADAMTS5

6.36

KRTAP15-1

7.10

  III

OLIG1

11.12

Hs.675532

4.24

TIAM1

5.29

Hs.657999

6.49

b

 Under-expressed segments

  I

2p12

0.16

2q31.1

0.38

  

REG1B REG1A REG3A

LRP2 BBS5 KLHL41 Hs.593163

  II

12q21.3-q22

0.17

5q31

0.41

  

LUM Hs.539252 DCN

Hs.658232 NR3C1 Hs.703520

  III

18q12.1

0.19

7q22.2

0.54

  

DSC3 DSC2 DSC1

Hs.718842 LOC100216546 Hs.656426 Hs.657627

 Under-expressed genes

  I

Hs.633942

0.01

TNNC1

0.01

NRGN

0.01

Hs.683165

0.01

  II

Hs.554169

0.01

Hs.80714

0.01

LCE5A

0.01

CKM

0.01

  III

CSTA

0.01

GGT6

0.01

CKM

0.01

DNAJC28

0.02

 Under-expressed chr21 genes

  I

COL6A2

0.09

LINC00320

0.03

LINC00320

0.04

DNAJC28

0.02

  II

Hs.663673

0.11

S100B

0.05

KRTAP10-10

0.11

LINC00320

0.10

  III

FAM3B

0.14

C21orf91

0.08

LINC00323

0.15

LINC00323

0.11

Analysis was performed using default parameters (see Materials and methods section). Under each segment, the expression value and the genes included in the segment are indicated; under each gene, the expression value is indicated. In order to simplify, some segments are not shown because they overlap with those highlighted in one of the listed regions. The complete results for these models are available as online supplementary material. Segments and genes are sorted by decreasing gene expression ratio in (a) and by increasing gene expression ratio in (b). The overlapping results among the transcriptome maps are marked in bold. In the ‘Map’ mode, TRAM displays UniGene EST clusters (with the prefix ‘Hs.’ in the case of H. sapiens) only if they have an expression value. Data refer to the differential transcriptome maps: whole adult brain versus pool of non-brain tissues, whole foetal brain versus whole adult brain, cerebellum versus whole adult brain and cerebral cortex versus whole adult brain

aCytoband was derived from the UCSC Genome Browser

At single gene level, an increase of more than ten times was observed in all the first 125 loci (Supplementary Table 5). In particular, a fold increase of 79.69 was observed for the known gene ANKRD30B (Table 3a), encoding for ankyrin repeat domain 30B. We also observed that in this range of expression ratio, five chr21 genes are included: DNAJC28, LINC00320, LINC00323, OLIG1 and OLIG2 (Supplementary Table 5). In particular, DNAJC28 had a 65.06-fold increase with respect to all tissues, followed by LINC00320, encoding for a long intergenic non-protein coding RNA 320, that had a 39.82-fold increase (Table 3a).

Among the genes with the lowest A/B expression ratio, a fold decrease of 100 was observed for the EST clusters Hs.633942, Hs.554169, Hs.727036 and Hs.720702 and for the known genes CSTA, KRT13 and MSMB, encoding for cystatin A (stefin A), keratin 13 and beta-microseminoprotein, respectively (Supplementary Table 5).

When the 24 spinal cord samples were removed from the non-brain sample pool, using the ‘Exclude’ sample function provided by the TRAM graphical interface in order to test if the removal of a central nervous system organ from the human tissue set could alter the transcriptome comparison of the brain with non-brain tissue pool, substantially analogous results were obtained (data not shown).

Male brain versus female brain transcriptome map analysis

The adult brain samples for which the sex of the sample donor was available were grouped into two additional datasets: pool 'A.1' including 13 male samples and pool 'A.2' including 5 female samples and the corresponding transcriptome maps were generated and compared.

The gene expression value for each of the 27,437 loci of the male brain transcriptome map (Supplementary Table 6) and for each of the 25,954 loci of the female brain transcriptome map (Supplementary Table 7) is available. The differential transcriptome map produced a gene expression ratio for each of the 25,954 shared loci (Supplementary Table 8). Results included six significantly over-expressed segments. The genome segment that has the highest statistically significant expression value is on chromosome Y, including the known genes TTTY15, USP9Y and DDX3Y encoding respectively for testis-specific transcript, Y-linked 15 (non-protein coding), ubiquitin-specific peptidase 9, Y-linked and DEAD (Asp-Glu-Ala-Asp) box helicase 3, Y-linked (supplementary full results for transcriptome maps are available at: http://apollo11.isto.unibo.it/suppl/).

At single gene level, a more than 10-fold increase was observed in all the first 84 loci (Supplementary Table 8). In particular, a more than 100-fold increase was observed for the uncharacterized locus LOC613126 (378.56) and the known genes LBX1 (366.49), DRD4 (332.51) and CDC42EP5 (115.26), encoding respectively for ladybird homeobox 1, D4 subtype of the dopamine receptor and for the CDC42 effector protein (Rho GTPase binding) 5 (Table 4).
Table 4

List of the ten most over- or under-expressed genes in male brain (pool 'A.1') versus female brain (pool 'A.2') transcriptome map

Gene name

Value 'A.1'

Value 'A.2'

Ratio 'A.1'/'A.2'

Location

Data points 'A.1'

Data points 'A.2'

SD as % of expression 'A.1'

SD as % of expression 'A.2'

Over-expressed genes

 LOC613126

2,019.28

5.33

378.56

chr7

6

3

110.66

71.16

LBX1

529.57

1.44

366.49

chr10

16

5

214.23

75.11

DRD4

1,268.15

3.81

332.51

chr11

16

5

213.44

125.06

CDC42EP5

1,433.81

12.44

115.26

chr19

6

3

109.49

88.48

 LOC100133315

578.64

7.97

72.59

chr11

9

3

148.30

106.70

EDARADD

126.93

1.89

67.15

chr1

6

3

103.76

35.84

HIST3H3

182.62

2.73

66.94

chr1

13

5

182.15

70.33

HES7

1,244.07

18.79

66.22

chr17

6

3

110.37

48.39

PPIAL4A

279.32

4.75

58.77

chr1

25

5

110.99

120.25

FTHL17

154.68

2.78

55.58

chrX

6

3

117.92

4.55

Under-expressed genes

CCL4

27.16

201.40

0.13

chr17

13

5

54.58

172.86

IL1B

27.82

208.20

0.13

chr2

50

10

92.63

112.94

 Hs.434622

0.90

6.78

0.13

chr1

3

3

50.62

63.74

 Hs.611927

2.23

19.15

0.12

chr6

3

3

70.55

85.33

 Hs.25345

0.38

3.42

0.11

chr6

3

3

19.05

44.69

NPAS4

8.37

80.07

0.10

chr11

9

6

144.65

158.20

 LOC400768

0.41

4.24

0.10

chr1

3

3

8.38

57.81

CCL3

19.43

201.78

0.10

chr17

13

5

136.51

140.44

 Hs.710548

0.75

7.94

0.09

chr17

3

3

27.09

63.80

XIST

7.97

203.97

0.04

chrX

44

31

78.75

131.58

‘Value’ is the mean gene expression value normalized across all the pool samples. ‘Data points’ are the number of spots associated to an expression value for the locus. ‘SD’ is the standard deviation for the expression value expressed as a percentage of the mean. All these gene results are statistically significant in the transcriptome map with a segment window of 12,500 bp. Full results are available in the supplementary material (see text)

Among the genes with the lowest A/B expression ratio, a 25-fold decrease was observed for the known gene XIST, encoding for the X-inactive-specific transcript (non-protein coding) (Table 4).

Foetal brain versus adult brain transcriptome map analysis

In the foetal brain transcriptome map analysis, 855,662 data points corresponding to 38,483 mapped loci were included (Supplementary Table 9). A total of 11,583 data points of the map correspond to 559 chr21 mapped loci. Results obtained by analysis included 38 significantly over-expressed segments.

The genome segment that has the highest statistically significant expression value is on chromosome 12 (Table 2a), including the over-expressed known genes TUBA1A, TUBA1B and TUBA1C. There are no statistically significant under-expressed segments.

At single gene level, TUBA1B (chr12) has the highest expression value (6,954.45) (Table 2a), and the EST cluster Hs.439634 (chr14) has the lowest expression value (0.17) (Table 2b). Among the genes of chromosome 21 (chr21), DNAJC28 has the highest expression value (2,817.20), followed by SOD1 (1,657.36) and ATP5O (1,074.36), encoding for ATP synthase, H+ transporting, mitochondrial F1 complex, O subunit (Table 2a).

In the analysis of the foetal brain versus adult brain TRAM map, regional differential expression of pool 'C' (35 foetal brain samples) versus pool 'A' (60 adult brain samples) was investigated. Results included 22 significantly over-expressed (n = 19) or under-expressed (n = 3) segments.

The genome segment that has the highest statistically significant expression value is on chromosome 8 (Table 3a), including the over-expressed known gene BHLHE22, encoding for basic helix-loop-helix family, member e22, the uncharacterized locus LOC100130155 and the EST cluster Hs.388788. The genome segment that has the lowest statistically significant expression value is on chromosome 2 (2q31.1) (Table 3b), including the under-expressed known genes LRP2, BBS5 and KLHL41, encoding for low density lipoprotein receptor-related protein 2, Bardet-Biedl syndrome 5, and kelch-like family member 41, respectively, and the EST cluster Hs.593163.

At single gene level, an increase of more than ten times was observed in all the first 44 loci (Supplementary Table 10), including TMSB15A, encoding for thymosin beta 15a, and DCX, encoding for doublecortin, which have a fold increase of 63.15 and 42.79 (Table 3a), respectively.

All the chr21 loci have an increase of less than ten times, and an increase between two and ten times was observed for 28 loci (Supplementary Table 10). In particular, CXADR, encoding for coxsackie virus and adenovirus receptor, has a 7.4-fold increase compared to adult brain (Table 3a). LINC00320 has the lowest C/A expression ratio (0.03) preceded by the known gene S100B (0.05), encoding for S100 calcium-binding protein B (Table 3b). Among the genes with the lowest C/A expression ratio, a fold decrease of 100 was observed for the EST clusters Hs.80714 and for the known genes LCE5A, ANKK1, GGT6 and TNNC1, encoding respectively for late cornified envelope 5A, ankyrin repeat and kinase domain containing 1, gamma-glutamyltransferase 6 and troponin C type 1 (slow) (Supplementary Table 10).

Cerebellum versus adult brain transcriptome map analysis

In the analysis of the cerebellum TRAM map, 3,862,253 data points corresponding to 38,163 mapped loci were included (Supplementary Table 11). A total of 49,004 data points of the map correspond to 554 chr21 mapped loci. Results obtained by analysis included 28 significantly over-expressed segments and 1 significantly under-expressed segment.

The genome segment that has the highest statistically significant expression value is on chromosome 11 (11q13.1) (Table 2a), including four over-expressed UniGene EST clusters (Hs.736281, Hs.593027, Hs.712678 and Hs.732685) and the known gene MALAT1, encoding for metastasis-associated lung adenocarcinoma transcript 1 (non-protein coding). The genome segment that has the lowest statistically significant expression value is on chromosome 13 (Table 2b), including the under-expressed EST clusters Hs.375745, Hs.735749 and Hs.551057.

At single gene level, BCYRN1 has the highest expression value (12,372.42) (Table 2a), and the UniGene EST cluster Hs.707129 has the lowest expression value (0.99) (Table 2b). Among the chr21 genes, DNAJC28 has the highest expression value (1,447.08), followed by SOD1 (1,432.22) (Table 2a).

In the analysis of the cerebellum versus adult brain TRAM map, regional differential expression of pool 'D' (140 cerebellum samples) versus pool 'A' (60 total brain samples) was investigated. Results included 42 significantly over-expressed segments.

The genome segment that has the highest statistically significant expression value is on chromosome 15 (15q22.2) (Table 3a), including the over-expressed known gene RORA, encoding for RAR-related orphan receptor A, and the two UniGene EST clusters Hs.660127 and Hs.655820. There are no statistically significant under-expressed segments.

At single gene level, an increase of more than ten times was observed in all the first 23 loci, including the gene FSTL5, encoding for follistatin-like 5 (Supplementary Table 12). In particular, a fold increase of 81.34 was observed for the known gene CBLN3, encoding for cerebellin 3 precursor (Table 3a). Among the genes with the lowest D/A expression ratio, a fold decrease of 100 was observed for the known gene NRGN, encoding for neurogranin (protein kinase C substrate, RC3) (Table 3b). All the chr21 loci have an increase of less than ten times, and an increase between two and ten times was observed for 40 loci in particular. The DNAJC28 and LINC00320 genes mentioned above have a 1.65-fold change and a 0.04-fold change respectively compared to total brain (Supplementary Table 12).

Cerebral cortex versus adult brain transcriptome map analysis

In the analysis of the cerebral cortex TRAM map, 780,661 data points corresponding to 27,504 mapped loci were included (Supplementary Table 13). A total of 9,846 data points of the map correspond to 336 chr21 loci. Results obtained by analysis included 21 significantly over-expressed segments.

The segment that has the highest statistically significant expression value is on chromosome 12 (12q13.12) (Table 2a), including the over-expressed known genes TUBA1A, TUBA1B and TUBA1C. There are no statistically significant under-expressed segments.

At single gene level, the EST cluster Hs.732685 has the highest expression value (12,387.16), followed by TUBA1B, which is the first known gene with the highest expression value (6,343.86) (Table 2a). The known gene ZNF852 has the lowest expression value (3.59) (Table 2b). Among the chromosome 21 genes, OLIG1, encoding for oligodendrocyte transcription factor 1, has the highest expression value (726.82), followed by PCP4, encoding for Purkinje cell protein 4 (700.08), and SOD1 (629.7) (Table 2a).

In the analysis of the cerebral cortex versus adult brain TRAM map, regional differential expression of pool 'E' (18 cortex samples) versus pool 'A' (60 total brain samples) was investigated. Results included ten significantly over-expressed segments.

The segment that has the highest statistically significant expression value is on chromosome 1 (1q23.1), including the over-expressed known genes OR6Y1, OR6K2 and OR6N1, encoding olfactory receptors, and PYHIN1, encoding pyrin and HIN domain family, member 1 (Table 3a). There are no statistically significant under-expressed segments.

At single gene level, an increase higher than ten times was observed in all the first 178 loci (Supplementary Table 14). In particular, an increase of 255.06 times was observed for the gene ZNF790, encoding for zinc finger protein 790 (Table 3a). Among the genes with the lowest E/A expression ratio, a fold decrease of 100 was observed for the EST cluster Hs.683165 and for the known gene CKM, encoding for creatine kinase, muscle (Table 3b). We also observed that only one chr21 gene, KRTAP13-2, encoding for keratin-associated protein 13–2, has an increase of 32.77 times, while the other chr21 loci have an increase of less than 7.1 times (Table 3a). The DNAJC28 gene has the lowest E/A expression ratio (0.02), preceded by LINC00320 (0.1), both previously mentioned (Table 3b). Furthermore, we noted that SOD1, despite being among the genes of chr21 with a high expression value, has a low expression ratio compared to total brain (0.36) (Supplementary Table 14).

Validation of TRAM map results through real-time RT-PCR

In order to confirm the results of our meta-analysis study, we performed real-time RT-PCR experiments which are useful for validating the transcriptome maps of the whole brain, cerebellum and cerebral cortex. The primer pairs used to validate results are listed in Table 5.
Table 5

Primer pairs used to validate TRAM Maps

Gene symbol

Gene name

RefSeq RNA GenBank accession No.

Primer pairs sequence (5′ → 3′)a

RT-PCR product size (base pair)

BCYRN1

Brain cytoplasmic RNA 1

NR_001568

tagcgagaccccgttctccag

gttgctttgagggaagttacgc

82

TUBA1B

Tubulin, alpha 1b

NM_006082

cctcgactcttagcttgtcgg

aggcagtagagctcccagcag

159

SOD1

Superoxide dismutase 1, soluble

NM_000454

tagcgagttatggcgacgaag

ggtacagcctgctgtattatctc

153

BEX5

Brain expressed, X-linked 5

NM_001012978

gtggtagaagctgacccctgag

gctcctggtattcaccacctc

225

RCAN1

Regulator of calcineurin 1

NM_004414

ctggagcttcattgactgcgag

gtgatgtccttgtcatacgtcc

186

GPCPD1

Glycerophos-phocholine phosphodiesterase GDE1 homolog (S. cerevisiae)

NM_019593

actcatggacctcagatctcg

ggatcattggtatcatcaccc

174

OPRD1

Opioid receptor, delta 1

NM_000911

ctgggcaacgtgcttgtcatg

catcaggtacttggcactctg

141

TBX18

T-box 18

NM_001080508

ctcgggggagacttggatgag

ctgattctgataggcagtgacg

247

PABPN1L

Poly(A) binding protein, nuclear 1-like (cytoplasmic)

NM_001080487

aggagaaggtggaggctgacc

gtccagagaacttgtcacacag

141

NPTX1

Neuronal pentraxin 1

NM_002522

gggcaaactttgcaatcgctc

tctcctcggtgtcgttcctg

193

NUAK1

NUAK family, SNF1-like kinase, 1

NM_014840

tcaatgggagaccttaccgag

tactctccgctgctgatttg

139

NTNG1

Netrin G1

NM_001113226

gaaagtgaaactcgatcctccg

ctcgcatcacactcattattgc

102

ANKRD55

Ankyrin repeat domain 55

NM_024669

ggcttgaaggctgtgtgagtc

gcagcatttgtgtgtgttgaggc

228

GAPDH

Glyceraldehyde-3-phosphate dehydrogenase

NM_002046

caacgaccactttgtcaagc

ctgtgaggaggggagattca

214

NCDN

Neurochondrin

NM_001014839

ctgccacatcttcctcaacctc

ccagggtggccacattagcag

155

SERPFIN1

Serpin peptidase inhibitor, clade F (alpha-2 antiplasmin, pigment epithelium derived factor), member 1

NM_002615

ggctgtctccaacttcggcta

cggtgaatgatggattctgttcg

150

INADL

InaD-like (Drosophila)

NM_176877

ctcacacttcagcagtccatc

ggaaccatctgtgaacactaac

117

aForward primer (top); Reverse primer (bottom) (for each gene)

Using criteria as described in the Materials and methods section above, we selected nine genes from the adult brain transcriptome map: BCYRN1, TUBA1B, SOD1, BEX5, RCAN1, GPCPD1, OPRD1, TBX18 and PABPN1L. BEX5 gene was chosen as reference gene. The in vitro observed gene expression ratios between each target gene and the reference gene are provided in Table 6. The correlation between the observed and expected gene expression ratios as calculated by bivariate analysis was statistically highly significant (Pearson correlation coefficient = 0.98 and p = 0) (Fig. 1a).
Table 6

Selected genes to validate in vitro through real-time RT-PCR adult brain, cerebellum and cerebral cortex transcriptome maps

Official gene symbol

Official gene name

EEV

SD

ER

Ct mean

OR

Adult brain

BCYRN1

Brain cytoplasmic RNA 1

7,875.54

13.02

12.80

18.67

14.22

TUBA1B

Tubulin, alpha 1b

4,876.25

55.86

8.00

19.63

7.31

SOD1

Superoxide dismutase 1, soluble

1,727.39

55.37

2.80

20.12

5.21

BEX5

Brain expressed, X-linked 5

615.23

81.22

1.00

22.05

1.00

RCAN1

Regulator of calcineurin 1

212.66

218.52

0.34

24.10

0.33

GPCPD1

Glycerophosphocholine phosphodiesterase GDE1 homolog (S. cerevisiae)

63.29

92.61

0.10

26.08

0.084

OPRD1

Opioid receptor, delta 1

27.07

217.68

0.044

29.69

0.0068

TBX18

T-box18

9.65

61.70

0.015

29.92

0.0058

PABPN1L

Poly(A) binding protein, nuclear 1-like

4.57

8.77

0.007

34.09

0.00032

Cerebellum

TUBA1B

Tubulin, alpha 1b

2,991.80

35.16

7.05

20.26

20.53

NPTX1

Neuronal pentraxin 1

1,081.87

86.51

2.55

22.53

4.26

NUAK1

NUAK family, SNF1-like kinase, 1

424.37

41.82

1.00

24.62

1.00

RCAN1

Regulator of calcineurin 1

142.60

166.56

0.34

23.41

2.31

GPCPD1

Glycerophosphocholine phosphodiesterase GDE1 homolog (S. cerevisiae)

167.03

95.59

0.39

23.32

2.46

NTNG1

Netrin G1

39.16

74.20

0.090

26.38

0.30

ANKRD55

Ankyrin repeat domain 55

17.19

52.41

0.041

31.34

0.010

TBX18

T-box 18

8.96

122.35

0.021

28.74

0.058

Cerebral cortex

TUBA1B

Tubulin, alpha 1b

6,343.86

43.54

7.99

18.50

55.72

GAPDH

Glyceraldehyde-3-phosphate dehydrogenase

2,947.49

103.08

3.71

17.31

127.12

NUAK1

NUAK family, SNF1-like kinase, 1

793.97

62.54

1.00

24.30

1.00

NCDN

Neurochondrin

333.60

125.52

0.42

22.35

3.86

SERPINF1

Serpin peptidase inhibitor, clade F (alpha-2 antiplasmin, pigment epithelium derived factor), member 1

131.02

59.12

0.17

24.29

1.0070

RCAN1

Regulator of calcineurin 1

81.21

90.00

0.10

24.00

1.23

OPRD1

Opioid receptor, delta 1

42.71

80.72

0.13

27.77

0.090

INADL

InaD-like (Drosophila)

22.34

71.29

0.07

29.32

0.031

In bold, the gene chosen as reference, underlined genes having a high standard deviation for each map. From left to right, official gene symbol of selected gene, official full gene name, expected expression value (EEV), i.e. expression value as provided by TRAM software; standard deviation (SD) as percentage of expression; expected ratio (ER) between reference and target gene expression value; threshold cycle (Ct) provided by Bio-Rad CFX Manager Software 2.1 manually positioning the threshold line; observed ratio (OR) determined between each target gene and the reference gene using the delta Ct (ΔCt) method, according to the formula 2ΔCt = 2Ctref − Cttarget

Fig. 1

Bivariate analysis between observed (real-time RT-PCR) and expected (TRAM) expression ratio in brain of selected genes (Table 6). a The fit line and a 95 % bivariate normal density ellipse are shown; Pearson correlation coefficient is 0.98 and p value 0. b The fit line and a 95 % bivariate normal density ellipse determined after the exclusion from the statistical analysis of those genes having a high standard deviation (SD as percentage of expression >95) of expected expression value (Table 6). This step did not affect the correlation between the two variables; indeed, Pearson correlation coefficient is 0.98 and p value is 0.0001

From the whole cerebellum transcriptome map, we selected the seven genes: TUBA1B, NPTX1, NUAK1, RCAN1, GPCPD1, NTNG1, ANKRD55 and TBX18. NUAK1 gene was chosen as reference gene. We determined the in vitro observed expression ratio between each target gene and the reference gene (Table 6) as discussed above, and we determined the correlation through bivariate analysis. The result was again statistically highly significant (Pearson correlation coefficient = 0.97 and p = 0) (Fig. 2a).
Fig. 2

Bivariate analysis between observed (real-time RT-PCR) and expected (TRAM) expression ratio in cerebellum of selected genes (Table 6). a The fit line and a 95 % bivariate normal density ellipse are shown; Pearson correlation coefficient is 0.97 and p value 0. b The fit line and a 95 % bivariate normal density ellipse determined after the exclusion from the statistical analysis of those genes having a high standard deviation (SD as percentage of expression >95) of expected expression value (Table 6). This step improved the correlation between the two variables; indeed, Pearson correlation coefficient is 0.98 and p value is 0.0019

From the whole cerebral cortex transcriptome map, we selected the eight genes: TUBA1B, GAPDH, NUAK1, NCDN, SERPINF1, RCAN1, OPRD1 and INADL. NUAK1 gene was chosen as reference gene. We determined the in vitro observed expression ratio between each target gene and the reference gene (Table 6) as discussed above, and we performed the bivariate analysis between the expected and the observed ratios. The correlation between the two variables was not statistically significant (Fig. 3a).
Fig. 3

Bivariate analysis between observed (real-time RT-PCR) and expected (TRAM) expression ratio in cerebral cortex of selected genes (Table 6). a The fit line and a 95 % bivariate normal density ellipse are shown. The correlation between the two variables was not statistically significant, the Pearson correlation coefficient is 0.65 and p value is 0.0762. b The fit line and a 95 % bivariate normal density ellipse determined after the exclusion from the statistical analysis of those genes having a high standard deviation (SD as percentage of expression >95) of expected expression value (Table 6). This step showed an increase of the correlation between the two variables; indeed, the Pearson correlation coefficient is 0.99 and p value is 0

When we attempted to exclude those genes that have an expected expression value with a high SD (as percentage of expression >95) from the statistical analysis, assuming that a high SD implies a greater variability in the observed data, the correlation between the two variables became statistically highly significant (Pearson correlation coefficient = 0.99 and p < 0.0001) (Fig. 3b).

Applying this criterion to the previous statistical analysis, we observed that for the adult brain data, there was no change (Fig. 1b), and for the whole cerebellum, data occurred an increase of the Pearson correlation coefficient to 0.98, while the p value became 0.0019 (Fig. 2b). Despite these variations, all correlations remained significant.

Housekeeping gene search

In the pool of normal tissues minus brain transcriptome map, using the lower SD as a percentage of the mean value (≤30), two EST clusters (Hs.714416 and Hs.728191) and only one known gene (POM121C, encoding for the transmembrane nucleoporin C) were retrieved, while using the higher SD as a percentage of the mean value at ≤ 40, 29 genes common to all pool tissues fulfilled the selected criteria. By applying functional enrichment analysis with the ‘ToppGene Suite’ Gene Ontology tool [34], we found that the statistically significant (p ≤ 0.05) enriched function is structural constituent of ribosome (GO:0003735), associated to the RPS17L, MRPL18 and RPS18 genes. Ten out of 29 genes resulted ‘not found’: they are the EST clusters.

In the adult brain, using the lower SD as a percentage of the mean value (≤30), one EST cluster (Hs.705664) and only one known gene (FUNDC1, encoding for an integral mitochondrial outer membrane protein) were retrieved, while using the higher SD as a percentage of the mean value at ≤40, 15 genes fulfilled the selected criteria. By applying functional enrichment analysis, we found that the statistically significant (p ≤ 0.05) enriched function is unfolded protein binding (GO:0051082), associated to the PDRG1 and TTC1 genes. Six out of 15 genes resulted ‘not found’: they are the EST clusters.

The same thing was done for the foetal brain transcriptome map. In this case, using the lower SD as a percentage of the mean value (≤30), 40 genes including some ribosomal proteins (MRPS24, MRPS35 and RPS18) were retrieved, while using the higher SD as percentage of the mean value at ≤40, 229 genes came out by the selection. The statistically significant (p ≤ 0.05) enriched function is a structural constituent of ribosome (GO:0003735), associated to the MRPL18, RPL23, RPL4, RPS17L, MRPL51, RPS13, MRPL14, MRPS24, RPL24, RPS24, RPS18 and MRPL32 genes.

Analyzing common results among the three maps: RPS17L, encoding for S17-like ribosomal protein, is the only gene common to all three maps; the EST cluster Hs.714416 is common to the pool of tissues and adult brain transcriptome maps; PDRG1, encoding for p53 and DNA-damage regulated 1, TCEAL8, encoding for transcription elongation factor A (SII)-like 8, and TIMM9, encoding for translocase of inner mitochondrial membrane 9 homolog (yeast), are common to the foetal brain and brain transcriptome maps; and ACTG1, encoding for actin, gamma 1, EMC4, encoding for ER membrane protein complex subunit 4, MRPL18, encoding for mitochondrial ribosomal protein L18, and YTHDF1, encoding for YTH domain family, member 1, are common to the pool of tissues and foetal brain transcriptome maps. These results are provided in the Supplementary Table 15.

Discussion

In this work, we proposed an integrated model of the human transcriptome of the whole brain (adult and foetal) and of two of the brain regions severely affected in Down syndrome (DS), such as the cerebellum and cerebral cortex [23].

TRAM software allowed us to integrate data from different sources, including data from different microarray experiments, also performed on different platforms through a method of intra-sample and inter-sample normalization (quantile scaled normalization). We used a set with a relevant number of samples to create a quantitative transcriptome reference map for a very complex tissue such as the brain. This map allowed us to identify critical genome regions which are over-/under-expressed in normal brain and are typical of this tissue compared to the pool of tissues minus brain. It also allowed us to assign a linear reference expression value to each locus.

The adult brain transcriptome map showed the highest expression value of a segment on chromosome 12, including the known genes TUBA1A, TUBA1B and TUBA1C [35, 36]. This finding appears to be an evident genetic correlate of the well-known main structural characteristic of the nervous system functional unit, the neuron. The cytoskeleton is most responsible for the neuron shape and function; it consists of microtubules and the expression of tubulins is typically used as a marker of neural differentiation. The alpha and beta tubulins represent the major components of microtubules. The analysis at single gene level showed that the first over-expressed gene is BCYRN1, while TUBA1B is the second. BCYRN1 encodes a small neural non-mRNA, the sequence contains a 5′ portion homologous to the Alu Lm (left-hand Alu monomer) and seems to have a regulative function [37, 38]. This RNA probably belongs to the subfamily of Alu-elements over-represented in the transcriptome of the brain tissue [39, 40]. As pointed out by Lejeune [41], some studies supported the hypothesis that deficiency in the neurotubule network could be associated with mental disorder [42]. The relationship between some mental disorders and increased levels of BCYRN1 was demonstrated. Its over-dosage seems to be associated with the increase of the cytoskeleton fibres causing a blockage in transport of the RNA within the cell due to the fibre overcrowding [38]. Among chr21 genes, SOD1, APP and DNAJC28 have the highest expression values. A preclinical research study conducted on mice used both SOD1 and APP as therapeutic targets in support of the hypothesis of the over-dosage effect of chr21 genes in the pathogenesis of Down syndrome [43]. The DNAJC28 gene has not yet been characterized and its function could be interesting to investigate.

To highlight the brain-specific gene expression profile, a comparison pool composed of 622 datasets derived from 53 different human tissues or organs has been accurately assembled. Spinal cord samples (n = 24) have been included in this pool, considering the distinctive anatomical and functional features of the brain and spinal cord, and the extensive overlapping of results when the spinal cord samples are included in or removed from the non-brain sample pool.

When we compare the adult brain gene expression with the pool of tissues minus brain gene expression, the highest statistically significant expression value is attributed to a segment on chromosome 5, including the known genes GABRA6, GABRA1 and GABRG2 encoding for subunits of the GABAergic receptor, one of the major inhibitory receptors in the mammalian brain. An altered expression of GABA receptors is typical in the brains of subjects with autism and fragile X syndrome, another common form of inherited mental retardation [44, 45]. The significantly under-expressed segment is on chromosome 2 and contains REG1B, REG1A and REG3A genes, encoding for proteins secreted by the pancreas [46, 47, 48]. These data allow the affordable identification of genes expressed specifically in the brain and not in other tissues.

At single gene level, ANKRD30B is over-expressed in the adult brain compared to the pool of non-brain tissues. It is also known as the NY-BR-1.1 gene, a homologous gene of NY-BR-1, with which it shares 54 % of amino acids, but unlike NY-BR-1 gene expressed in breast and testis, it is also expressed in the brain. The ankyrin proteins carry out a wide variety of biological activities; they have a repeat motif recognized in more than 400 proteins, including cyclin-dependent kinase inhibitors, transcriptional regulators, cytoskeletal organizers, developmental regulators and toxins [49].

Among chr21 genes, DNAJC28, LINC00320, LINC00323, OLIG1 and OLIG2 are over-expressed. OLIG1 and OLIG2 are known to have the function of transcription factors in oligodendrocytes, cells that perform the main function of myelination of neurons in the central nervous system. A recent study suggested a link between the over-dosage of OLIG2 in trisomy 21 and the reduced brain size in affected individuals [50]. Instead, LINC00320 and LINC00323 are non-coding sequences that may have a regulatory role and that seem to have a fundamental role in the adult brain because they maintain a low expression value in the foetal brain transcriptome map. Their function could be interesting to investigate.

The foetal brain transcriptome map, as the adult brain transcriptome map, showed the highest expression value of the segment on chromosome 12, including the known genes TUBA1A, TUBA1B and TUBA1C. It is interesting to note that at two very different developmental stages of the human brain, the same chromosomal segment (12q13.12) is over-expressed. At single gene level, TUBA1B has the highest expression value. Among the genes of chr21, DNAJC28, followed by SOD1 (1,657.36) and ATP5O, is over-expressed.

Comparison between the foetal brain and the adult brain transcriptome map showed the highest expression value of a segment on chromosome 8, including the known gene BHLHE22 and the uncharacterized loci, LOC100130155 and Hs.388788. BHLHE22 gene encodes for a transcription factor that regulates cell fate determination, proliferation and differentiation [51]. It seems to be involved in neuronal development rather than in the functions performed by the adult brain. LOC100130155 and Hs.388788 are loci which have not yet been characterized but are probably fundamental in brain development.

At single gene level, there is a high expression ratio for TMSB15A gene and for DCX gene, which have an increase of 63.15 times and of 42.79 times, respectively. The former is involved in cellular motility [52, 53]. Its expression profile was analyzed in normal and pathological tissues [54], and it was considered a neuroblastoma-specific gene. The derivation of neuroblastome from neuroblasts or nerve undifferentiated cells allows us to draw the further conclusion that neuroblastoma-specific genes might include genes that are especially expressed in the developing nervous system. The DCX gene is involved in neuronal migration during development by regulating the organization and stability of microtubules [55, 56]. It was demonstrated that DCX mutations are associated with disorders caused by an abnormal migration of neurons during development [57]. The low expression ratio of LINC00320 could be significant of the regulatory role played by this locus in the adult brain only. This information favours the hypothesis that the foetal brain transcriptome map can be used as a reference for the rapid identification of genes involved in the development of the central nervous system.

We noted that many differences exist between the gene expression patterns of the adult and foetal brain; indeed, many adult brain genes are not yet active at such an early phase of development as the foetal stage, while many foetal brain genes which are less active in the adult brain are possibly those involved in brain development.

The experimental screening was performed on the adult brain transcriptome map through a real-time RT-PCR experiment using nine genes selected from the brain TRAM map (see Materials and methods). One of these genes is chosen as reference, according to its medium-high expression value among the nine genes selected. The observed and expected expression ratios are listed in Table 6. It is interesting to note that the correlation between the values calculated by TRAM using 60 normal human brains and an experimental comparison using independent samples is maintained through several orders of magnitude and from the lowest to the highest values.

The significant correlation indicates that the alternative method of sample collection—a different tissue source of the sample and the different post-mortem interval of tissue retrievement—was neutralized by the high number of samples. In fact, with regard to the post-mortem interval (PMI), it was proven that undegraded mRNA may be isolated from most brain regions many hours post-mortem (up to 30 h), so the PMI is not actually predictive of mRNA integrity [58].

Although there are studies showing a sex-biased brain gene expression (e.g. [19]), it was also proven that gender does not have a significant influence on the observed gene expression changes, at least when a significant number of samples are available for study [59]. Following the observation by Trabzuni and colleagues [19] that a minority of genes (2.6 %) presents sex-biased expression level in specific brain subregions, it is expected that differential gene expression analysis between male and female whole brains affects a small number of genes. Table 4 shows the first ten over-/under-expressed genes identified by our analysis. For several of them, we have found literature evidence about their relationship with gender (e.g. DRD4 [60, 61], FTHL17 [62], XIST [63]). DRD4 gene, encoding for D4 subtype of the dopamine receptor, seems to be involved in the difference in the number of dompaminergic neuron between males and females [60]. DRD4 gene has been extensively studied for its possible association with increased vulnerability to antisocial and impulsive behaviour, including behaviour recurrent in males and not in females (e.g. [61]). FTHL17 gene, encoding for ferritin, heavy polypeptide-like 17, was described as a typical example of X-linked genes expressed only in male germ cells [62]. It could be interesting to investigate its linkage with brain tissue. XIST gene, encoding for X-inactive-specific transcript (non-protein coding) [63], has the lowest expression ratio between male and female brain in the differential transcriptome map, demonstrating that it is a gene typically expressed in females. These results are a confirmation of the reliability of the data produced by our map.

For the other genes shown in Table 4, it would be interesting to investigate their relationship with gender, in particular for the neuronal gene NPAS4, encoding for neuronal PAS protein 4 and seeming to be involved in the transcriptional regulation of some neural genes in the brain [64], and for CCL3, IL1B and CCL4 genes encoding respectively for chemokine (C-C motif) ligand 3, interleukin 1, beta and chemokine (C-C motif) ligand 4, all three of which are involved in immune system response [65, 66].

In summary, the high correlation (r = 0.94, data not shown) between the male- and female-derived gene expression values supports the utility of a human brain map as a general standard reference, while the creation of differential expression maps comparing male- and female-derived samples remains feasible using our approach and may highlight sex-biased differences for specific genes, whether they are located on the sex chromosomes or not.

The experimental validation of our map allows us to consider all the other intervening values among the points we selected as bona fide values, thus for the first time providing a complete and quantitative representation of the reference expression values for 39,250 mapped and 26,026 unmapped transcripts of the normal human brain.

The initial analysis of the biological meaning of these data clearly shows a relationship of our findings with what is already known about genes which have either a relevant function or are almost silenced in the brain. While the correlation between the values obtained by meta-analysis of multiple gene profiling datasets and independent samples assayed by a different method (real-time RT-PCR) is excellent, some remaining differences may easily be attributed to the physiological biological differences among the samples, as well as to experimental error. In some cases, specific differences in the method of analysis can explain the variability of the expression value. For instance, BCYRN1 contains a 5′ portion homologous to the Alu Lm, and the corresponding array probe (37 bp) recognizes this domain. So, it is possible that, in addition to BCYRN1, other members of this family of interspersed repetitive DNA [67] hybridized to the probe. This could explain a hybridization signal higher than that observed; indeed, the observed gene expression relative to the reference gene is 7.11 by RT-PCR and 12.8 by meta-analysis of cDNA hybridization microarrays.

The cerebellum transcriptome map showed the highest expression value of a segment on chromosome 11 containing an EST cluster and the long non-coding RNA (ncRNA), MALAT1 [68]. This ncRNA is very abundant in neurons where it controls the expression of a subset of genes significantly involved in nuclear and synapse function and regulates synaptogenesis [69]. At single gene level, BCYRN1 is over-expressed as in the whole brain, while among chr21 genes, DNAJC28 and SOD1 remain the genes with the highest statistically significant expression values.

When we compare the cerebellum gene expression map with the adult brain gene expression map, the highest value is attributed to CBLN3. It is a new member of the precerebellin family coding for a neuropeptide detected in the granule neurons of cerebellum that, together with CBLN1, regulate synapse integrity and plasticity in Purkinje cells [70]. The lowest expression value is attributed to NRGN. This gene has specific functions in the cerebral cortex [71], and it is a target of thyroid hormone and in hypothyroidism conditions suffers an altered control of expression which causes mental disorder during development [72]. The experimental validation of the expected data was made on eight genes of the whole cerebellum transcriptome map, again finding a highly significant statistical correlation. We noted that the observed expression ratios of RCAN1 and GPCPD1 indicate their over-expression in comparison with the reference gene instead of being under-expressed as the data anticipated. This result could be due to a SD as a percentage of expression calculated by TRAM >95 for RCAN1 and GPCPD1. A high SD could explain the significant distance between expected and in vitro observed results; indeed, the exclusion of RCAN1 and GPCPD1 from the statistical analysis determines the increase of the statistical correlation from 0.96 to 0.97.

In the cerebral cortex TRAM map, the significantly over-expressed segment is again on chromosome 12, including the over-expressed known genes TUBA1A, TUBA1B and TUBA1C. At single gene level, the EST cluster Hs.732685 is the first over-expressed locus, followed by TUBA1B. Among chr21 genes, OLIG1, PCP4 and SOD1 have the highest statistically significant expression values. PCP4 is known to be highly expressed in the brain, primarily in the Purkinje cells, a class of GABAergic neurons located in the cerebellar cortex (OMIM entry #601629).

When we compare the cerebral cortex gene expression with the adult brain gene expression, the over-expression of the KRTAP13-2 gene encoding for a protein associated to keratin is interesting. In contrast, DNAJC28 and LINC00320 are under-expressed: the only time this happens to DNAJC28 in our maps. The experimental validation of the expected data was made on eight genes of the whole cerebral cortex transcriptome map. The gene expression value as calculated by TRAM for GAPDH, NCDN and RCAN1 was not observed. The first two have a SD as a percentage of expression calculated by TRAM >95, while the third has a SD of 90 %. The statistical correlation between expected and observed ratios is not significant. To improve the correlation, we excluded those genes with a SD as percentage of expression value >95 from the statistical analysis and, in fact, after this selection, the correlation became highly significant. Applying the same criterion to the previous maps which were experimentally screened, the correlation remained the same in the whole brain transcriptome map (Fig. 1b), whereas it further increased in the whole cerebellum transcriptome map (Fig. 2b). That the significance was maintained confirms the theory that our maps could be used as a reference for expression studies performed on the whole brain.

Until now, no study had been carried out by relating such a large number of samples, integrating data from several experiments conducted on different platforms and providing information about 39,250 loci in the adult brain map, about 38,163 loci in the cerebellum map and about 27,504 loci in the cerebral cortex map. The new TRAM version issued includes updated gene tables of Entrez Gene and UniGene, improving gene localization data and parsing. It solves the problem of many aliases of each gene, allowing the correct assignment of each probe to the loci for known transcripts and EST clusters. The EST clusters and the uncharacterized loci mentioned in our work would be interesting to investigate. Their localization was derived from UCSC ‘ESTs’ track in the UCSC Genome Browser [73], which was also imported and processed during the TRAM set-up [18].

There are several databases and atlases of brain gene expression with specific features publicly available online, for example GeneCards [74], Allen Atlas [75] and BrainSpan Atlases [http://www.brainspan.org/]. GeneCards is an integrated database that provides genomic, transcriptomic, proteomic, genetic, clinical and functional information on human genes. However, it provides transcriptomic information that comes from a limited number of hybridization experiments performed only on the Affymetrix GeneChip HG-U95 set. Furthermore, this database does not give information about uncharacterized loci, such as EST clusters. Allen Brain Atlas is a transcriptional atlas of the adult human brain comprising microarray profiling performed on macrodissected and on laser microdissected brain regions. It provides expression data on brain-specific regions. It provides a very useful amount of data for the experimenter but remains incomplete when searching for data derived from whole tissue and from a large number of samples. Furthermore, the number of loci and the number of hybridizations per single locus are lower than those analyzed by TRAM software. This is confirmed by the data attributed to SOD1 expression in the brain. Although it is possible to see the expression of SOD1 for each macrodissected brain region, it is not possible to have a reference expression value for the entire tissue. The hybridization signal is only provided by two probes which are both from the same platform, whereas the one provided by our transcriptome map was obtained from the normalization of 96 data point expression values and 9 different platforms. The TRAM data for BCYRN1, the gene with the highest expression value in the brain, is not searchable in the Allen Atlas, probably due to the low number of genes about which it gives information, and the same goes for clusters of EST. The different types of data do not allow us to make a comparison between the expression values provided by TRAM and those provided by the Allen Atlas. BrainSpan Atlas is an atlas of spatio-temporal gene expression profiles obtained by RNA sequencing and exon-array performed on macrodissected tissues at different stages of human development. Neither of these atlases correspond with our goal of determining the whole brain gene expression profile using a wide range of data to provide a normalized reference expression value for each human transcript, with the chance of integrating them even if they come from multiple experiments carried out on different platforms [18].

The availability of the systematic and detailed expression maps presented here for several human tissues represented an excellent occasion to investigate, among many features of the transcriptome, the suitability of individual genes as the best ones for selection as reference genes in gene expression studies. This is because they fulfil criteria including a widely diffused, constant and high expression. The results consist of many known genes and uncharacterized loci, mostly EST clusters. The enhanced function is related to the known genes of each group. It is interesting to note that the non-brain tissue pool and the foetal brain share the same enriched function, i.e. structural constituent of ribosome (GO:0003735), a function associated with genes that result constitutively expressed in the cells although it is not associated with the same genes. Furthermore, the enriched function of the adult brain is unfolded protein binding (GO:0051082). We analyzed only these maps to verify if the housekeeping genes shared among the pool of 53 tissues are also present in the whole adult and foetal brain.

Finally, it would be interesting to investigate the role of certain ncRNAs emerging from our analysis when the corresponding probes were present on the experimental platform.

Our maps can provide a useful and ready reference benchmark to test hypotheses about localized gene expression levels of human transcripts in the brain and in two brain subregions such as the cerebellum and cortex, two of the main brain regions severely affected in ID, while further specific study will be done about gene expression in hippocampus. These data could also contribute to a better understanding on a regional (chromosomal) basis of the chr21 genes expression [76]. Its over-expression in trisomy 21 is associated with the most common form of constitutional ID.

In addition, the transcriptome maps can easily be extended to many other tissues and pathological conditions to obtain a quantitative dissection of regional gene expression levels within a certain tissue or of differential expression between two biological conditions.

Notes

Acknowledgments

We would like to give special thanks to the Fondazione Umano Progresso, Milano, Italy, for supporting the research on trisomy 21 conducted at the DIMES Dept. MC fellowship is funded by a donation from the company Illumia, Bologna, Italy, that we greatly thank for their interest in our research. We are grateful to Rotary Club, Cesena, Italy (President: Ing. Giuliano Arbizzani) for the generous donation of the thermal cycler ‘GenePro’ (Bioer). We thank all the other people that very kindly contributed by individual donations to support part of the work that we are conducting on the subject. In particular, we are profoundly grateful to Matteo and Elisa Mele, to the Costa, Dal Monte, Ghignone and Morini families as well as to the architects of the Jérôme Lejeune exhibition at the Rimini Meeting in 2012, to Rina Bini, to the ‘Gruppo Arzdore’ and the ‘Associazione Turistica Pro Loco di Dozza’ (Dozza, Bologna, Italy) for their generous support to our trisomy 21 research. We are grateful to Danielle Mitzman for her expert revision of the manuscript.

Supplementary material

10048_2014_419_MOESM1_ESM.pdf (150 kb)
Supplementary Table 1(PDF 150 kb)
10048_2014_419_MOESM2_ESM.pdf (462 kb)
Supplementary Table 2(PDF 461 kb)
10048_2014_419_MOESM3_ESM.xlsx (2.7 mb)
Supplementary Table 3(XLSX 2806 kb)
10048_2014_419_MOESM4_ESM.xlsx (2.5 mb)
Supplementary Table 4(XLSX 2529 kb)
10048_2014_419_MOESM5_ESM.xlsx (3.7 mb)
Supplementary Table 5(XLSX 3752 kb)
10048_2014_419_MOESM6_ESM.xlsx (1.9 mb)
Supplementary Table 6(XLSX 1966 kb)
10048_2014_419_MOESM7_ESM.xlsx (1.8 mb)
Supplementary Table 7(XLSX 1874 kb)
10048_2014_419_MOESM8_ESM.xlsx (3.2 mb)
Supplementary Table 8(XLSX 3265 kb)
10048_2014_419_MOESM9_ESM.xlsx (2.7 mb)
Supplementary Table 9(XLSX 2759 kb)
10048_2014_419_MOESM10_ESM.xlsx (4.7 mb)
Supplementary Table 10(XLSX 4781 kb)
10048_2014_419_MOESM11_ESM.xlsx (2.7 mb)
Supplementary Table 11(XLSX 2716 kb)
10048_2014_419_MOESM12_ESM.xlsx (4.7 mb)
Supplementary Table 12(XLSX 4812 kb)
10048_2014_419_MOESM13_ESM.xlsx (1.9 mb)
Supplementary Table 13(XLSX 1976 kb)
10048_2014_419_MOESM14_ESM.xlsx (3.4 mb)
Supplementary Table 14(XLSX 3480 kb)
10048_2014_419_MOESM15_ESM.pdf (205 kb)
Supplementary Table 15(PDF 204 kb)

References

  1. 1.
    Strachan T and Read AP (2011) High parallel analysis of gene expression. In: Garland Science/Taylor & Francis Group editors. Human Molecular Genetics, 4th edn. New York, pp 245–253Google Scholar
  2. 2.
    Naumova OY, Lee M, Rychkov SY, Vlasova NV, Grigorenko EL (2013) Gene expression in the human brain: the current state of the study of specificity and spatiotemporal dynamics. Child Dev 84:76–88. doi:10.1111/cdev.12014 PubMedCrossRefPubMedCentralGoogle Scholar
  3. 3.
    Myers AJ (2012) The age of the “ome”: genome, transcriptome and proteome data set collection and analysis. Brain Res Bull 88:294–301. doi:10.1016/j.brainresbull.2011.11.015 PubMedCrossRefGoogle Scholar
  4. 4.
    Malone JH, Oliver B (2011) Microarrays, deep sequencing and the true measure of the transcriptome. BMC Biol 9:34. doi:10.1186/1741-7007-9-34 PubMedCrossRefPubMedCentralGoogle Scholar
  5. 5.
    Mortazavi A, Williams BA, McCue K, Schaeffer L, Wold B (2008) Mapping and quantifying mammalian transcriptomes by RNA-Seq. Nat Methods 5:621–628. doi:10.1038/nmeth.1226 PubMedCrossRefGoogle Scholar
  6. 6.
    Oldham MC, Konopka G, Iwamoto K, Langfelder P, Kato T, Horvath S, Geschwind DH (2008) Functional organization of the transcriptome in human brain. Nat Neurosci 11:1271–1282. doi:10.1038/nn.2207 PubMedCrossRefPubMedCentralGoogle Scholar
  7. 7.
    Mégarbané A, Noguier F, Stora S, Manchon L, Mircher C, Bruno R, Dorison N, Pierrat F, Rethoré MO, Trentin B, Ravel A, Morent M, Lefranc G, Piquemal D (2013) The intellectual disability of trisomy 21: differences in gene expression in a case series of patients with lower and higher IQ. Eur J Hum Genet 21:1253–1259. doi:10.1038/ejhg.2013.24 PubMedCrossRefGoogle Scholar
  8. 8.
    Lockhart DJ, Barlow C (2001) Expressing what’s on your mind: DNA arrays and the brain. Nat Rev Neurosci 2:63–68. doi:10.1038/35049070 PubMedCrossRefGoogle Scholar
  9. 9.
    Enard W, Khaitovich P, Klose J, Zöllner S, Heissig F, Giavalisco P, Nieselt-Struwe K, Muchmore E, Varki A, Ravid R, Doxiadis GM, Bontrop RE, Pääbo S (2002) Intra- and interspecific variation in primate gene expression patterns. Science 296:340–343. doi:10.1126/science.1068996 PubMedCrossRefGoogle Scholar
  10. 10.
    Khaitovich P, Muetzel B, She X, Lachmann M, Hellmann I, Dietzsch J, Steigele S, Do HH, Weiss G, Enard W, Heissig F, Arendt T, Nieselt-Struwe K, Eichler EE, Pääbo S (2004) Regional patterns of gene expression in human and chimpanzee brains. Genome Res 14:1462–1473. doi:10.1101/gr.2538704 PubMedCrossRefPubMedCentralGoogle Scholar
  11. 11.
    Roth RB, Hevezi P, Lee J, Willhite D, Lechner SM, Foster AC, Zlotnik A (2006) Gene expression analyses reveal molecular relationships among 20 regions of the human CNS. Neurogenetics 7:67–80. doi:10.1007/s10048-006-0032-6 PubMedCrossRefGoogle Scholar
  12. 12.
    Iwamoto K, Kakiuchi C, Bundo M, Ikeda K, Kato T (2004) Molecular characterization of bipolar disorder by comparing gene expression profiles of postmortem brains of major mental disorders. Mol Psychiatry 9:406–416. doi:10.1038/sj.mp.4001437 PubMedCrossRefGoogle Scholar
  13. 13.
    Iwamoto K, Bundo M, Kato T (2005) Altered expression of mitochondria-related genes in postmortem brains of patients with bipolar disorder or schizophrenia, as revealed by large-scale DNA microarray analysis. Hum Mol Genet 14:241–253. doi:10.1093/hmg/ddi022 PubMedCrossRefGoogle Scholar
  14. 14.
    Hodges A, Strand AD, Aragaki AK, Kuhn A, Sengstag T, Hughes G, Elliston LA, Hartog C, Goldstein DR, Thu D, Hollingsworth ZR, Collin F, Synek B, Holmans PA, Young AB et al (2006) Regional and cellular gene expression changes in human Huntington’s disease brain. Hum Mol Genet 15:965–977. doi:10.1093/hmg/ddl013 PubMedCrossRefGoogle Scholar
  15. 15.
    Ryan MM, Lockstone HE, Huffaker SJ, Wayland MT, Webster MJ, Bahn S (2006) Gene expression analysis of bipolar disorder reveals downregulation of the ubiquitin cycle and alterations in synaptic genes. Mol Psychiatry 11:965–978. doi:10.1038/sj.mp.4001875 PubMedCrossRefGoogle Scholar
  16. 16.
    Brooksbank C, Bergman MT, Apweiler R, Birney E, Thornton J (2014) The European Bioinformatics Institute’s data resources 2014. Nucl Acids Res 42:D18–D25. doi:10.1093/nar/gkt1206 PubMedCrossRefPubMedCentralGoogle Scholar
  17. 17.
    Barrett T, Edgar R (2006) Gene expression omnibus: microarray data storage, submission, retrieval, and analysis. Methods Enzymol 411:352–369. doi:10.1016/S0076-6879(06)11019-8 PubMedCrossRefPubMedCentralGoogle Scholar
  18. 18.
    Lenzi L, Facchin F, Piva F, Giulietti M, Pelleri MC, Frabetti F, Vitale L, Casadei R, Canaider S, Bortoluzzi S, Coppe A, Danieli GA, Principato G, Ferrari S, Strippoli P (2011) TRAM (Transcriptome Mapper): database-driven creation and analysis of transcriptome maps from multiple sources. BMC Genomics 12:121. doi:10.1186/1471-2164-12-121 PubMedCrossRefPubMedCentralGoogle Scholar
  19. 19.
    Trabzuni D, Ramasamy A, Imran S, Walker R, Smith C, Weale ME, Hardy J, Ryten M, North American Brain Expression Consortium (2013) Widespread sex differences in gene expression and splicing in the adult human brain. Nat Commun 4:2771. doi:10.1038/ncomms3771 PubMedCrossRefPubMedCentralGoogle Scholar
  20. 20.
    Kesslak JP, Nagata SF, Lott I, Nalcioglu O (1994) Magnetic resonance imaging analysis of age-related changes in the brains of individuals with Down’s syndrome. Neurology 44:1039–1045PubMedCrossRefGoogle Scholar
  21. 21.
    Nadel L (2003) Down’s syndrome: a genetic disorder in biobehavioral perspective. Genes Brain Behav 2:156–166. doi:10.1034/j.1601-183X.2003.00026.x PubMedCrossRefGoogle Scholar
  22. 22.
    Pennington BF, Moon J, Edgin J, Stedron J, Nadel L (2003) The neuropsychology of Down syndrome: evidence for hippocampal dysfunction. Child Dev 74:75–93. doi:10.1111/1467-8624.00522 PubMedCrossRefGoogle Scholar
  23. 23.
    Haydar TF, Reeves RH (2012) Trisomy 21 and early brain development. Trends Neurosci 35:81–91. doi:10.1016/j.tins.2011.11.001 PubMedCrossRefPubMedCentralGoogle Scholar
  24. 24.
    Emig D, Salomonis N, Baumbach J, Lengauer T, Conklin BR, Albrecht M (2010) AltAnalyze and DomainGraph: analyzing and visualizing exon expression data. Nucleic Acids Res 38:W755–W762. doi:10.1093/nar/gkq405 PubMedCrossRefPubMedCentralGoogle Scholar
  25. 25.
    Lenzi L, Frabetti F, Facchin F, Casadei R, Vitale L, Canaider S, Carinci P, Zannotti M, Strippoli P (2006) UniGene Tabulator: a full parser for the UniGene format. Bioinformatics 22:2570–2571. doi:10.1093/bioinformatics/btl425 PubMedCrossRefGoogle Scholar
  26. 26.
    Piovesan A, Vitale L, Pelleri MC, Strippoli P (2013) Universal tight correlation of codon bias and pool of RNA codons (codonome): the genome is optimized to allow any distribution of gene expression values in the transcriptome from bacteria to humans. Genomics 101:282–289. doi:10.1016/j.ygeno.2013.02.009 PubMedCrossRefGoogle Scholar
  27. 27.
    Butte AJ, Dzau VJ, Glueck SB (2001) Further defining housekeeping, or “maintenance,” genes Focus on “A compendium of gene expression in normal human tissues”. Physiol Genomics 7:95–96PubMedGoogle Scholar
  28. 28.
    Tu Z, Wang L, Xu M, Zhou X, Chen T, Sun F (2006) Further understanding human disease genes by comparing with housekeeping genes and other genes. BMC Genomics 7:31. doi:10.1186/1471-2164-7-31 PubMedCrossRefPubMedCentralGoogle Scholar
  29. 29.
    Pilbrow AP, Ellmers LJ, Black MA, Moravec CS, Sweet WE, Troughton RW, Richards AM, Frampton CM, Cameron VA (2008) Genomic selection of reference genes for real-time PCR in human myocardium. BMC Med Genomics 1:64. doi:10.1186/1755-8794-1-64 PubMedCrossRefPubMedCentralGoogle Scholar
  30. 30.
    Engels WR (1993) Contributing software to the internet: the Amplify program. Trends Biochem Sci 18:448–450PubMedCrossRefGoogle Scholar
  31. 31.
    Sharrocks AD (1994) The design of primer for PCR. In: Griffin HG, Griffin AM (eds) PCR technology – current innovations. CRC Press, Boca Raton, pp 5–11Google Scholar
  32. 32.
    Davis LG, Kuehl WM, Battey JF (1994) Basic methods in molecular biology. Appleton & Lange, NorwalkGoogle Scholar
  33. 33.
    Livak KJ, Schmittgen TD (2001) Analysis of relative gene expression data using real-time quantitative PCR and the 2(−Delta Delta C(T)) Method. Methods 25:402–408. doi:10.1006/meth.2001.1262 PubMedCrossRefGoogle Scholar
  34. 34.
    Chen J, Bardes EE, Aronow BJ, Jegga AG (2009) ToppGene Suite for gene list enrichment analysis and candidate gene prioritization. Nucleic Acids Res 37:W305–W311. doi:10.1093/nar/gkp427 PubMedCrossRefPubMedCentralGoogle Scholar
  35. 35.
    Baumann MH, Wisniewski T, Levy E, Plant GT, Ghiso J (1996) C-terminal fragments of alpha- and beta-tubulin form amyloid fibrils in vitro and associate with amyloid deposits of familial cerebral amyloid angiopathy, British type. Biochem Biophys Res Commun 219:238–242. doi:10.1006/bbrc.1996.0211 PubMedCrossRefGoogle Scholar
  36. 36.
    Okumura A, Hayashi M, Tsurui H, Yamakawa Y, Abe S, Kudo T, Suzuki R, Shimizu T, Shimojima K, Yamamoto T (2013) Lissencephaly with marked ventricular dilation, agenesis of corpus callosum, and cerebellar hypoplasia caused by TUBA1A mutation. Brain Dev 35:274–279. doi:10.1016/j.braindev.2012.05.006 PubMedCrossRefGoogle Scholar
  37. 37.
    Wang H, Iacoangeli A, Popp S, Muslimov IA, Imataka H, Sonenberg N, Lomakin IB, Tiedge H (2002) Dendritic BC1 RNA: functional role in regulation of translation initiation. J Neurosci 22:10232–10241PubMedPubMedCentralGoogle Scholar
  38. 38.
    Mus E, Hof PR, Tiedge H (2007) Dendritic BC200 RNA in aging and in Alzheimer’s disease. Proc Natl Acad Sci U S A 104:10679–10684. doi:10.1073/pnas.0701532104 PubMedCrossRefPubMedCentralGoogle Scholar
  39. 39.
    Faulkner GJ, Kimura Y, Daub CO, Wani S, Plessy C, Irvine KM, Schroder K, Cloonan N, Steptoe AL, Lassmann T, Waki K, Hornig N, Arakawa T, Takahashi H, Kawai J et al (2009) The regulated retrotransposon transcriptome of mammalian cells. Nat Genet 41:563–571. doi:10.1038/ng.368 PubMedCrossRefGoogle Scholar
  40. 40.
    Xu AG, He L, Li Z, Xu Y, Li M, Fu X, Yan Z, Yuan Y, Menzel C, Li N, Somel M, Hu H, Chen W, Pääbo S, Khaitovich P (2010) Intergenic and repeat transcription in human, chimpanzee and macaque brains measured by RNA-Seq. PLoS Comput Biol 6:e1000843. doi:10.1371/journal.pcbi.1000843 PubMedCrossRefPubMedCentralGoogle Scholar
  41. 41.
    Lejeune J (1988) Research on pathogeny of mental retardation in trisomy 21. Working group on: “Aspects of the uses of genetic engineering”. Commentarii Vol. III N° 31. Pontificia Academia Scientiarum, Rome, ItalyGoogle Scholar
  42. 42.
    Nunez J (1985) Microtubules and brain development: the effects of thyroid hormones. Neurochem Int 7:959–968. doi:10.1016/0197-0186(85)90144-5 PubMedCrossRefGoogle Scholar
  43. 43.
    Costa AC, Scott-McKean JJ (2013) Prospects for improving brain function in individuals with Down syndrome. CNS Drugs 27:679–702. doi:10.1007/s40263-013-0089-3 PubMedCrossRefGoogle Scholar
  44. 44.
    D’Hulst C, De Geest N, Reeve SP, Van Dam D, De Deyn PP, Hassan BA, Kooy RF (2006) Decreased expression of the GABAA receptor in fragile X syndrome. Brain Res 1121:238–245. doi:10.1016/j.brainres.2006.08.115 PubMedCrossRefGoogle Scholar
  45. 45.
    Fatemi SH, Reutiman TJ, Folsom TD, Thuras PD (2009) GABA(A) receptor downregulation in brains of subjects with autism. J Autism Dev Disord 39:223–230. doi:10.1007/s10803-008-0646-7 PubMedCrossRefPubMedCentralGoogle Scholar
  46. 46.
    Sanchez D, Figarella C, Marchand-Pinatel S, Bruneau N, Guy-Crotte O (2001) Preferential expression of reg I beta gene in human adult pancreas. Biochem Biophys Res Commun 284:729–737. doi:10.1006/bbrc.2001.5033 PubMedCrossRefGoogle Scholar
  47. 47.
    Zhou L, Zhang R, Wang L, Shen S, Okamoto H, Sugawara A, Xia L, Wang X, Noguchi N, Yoshikawa T, Uruno A, Yao W, Yuan Y (2010) Upregulation of REG Ialpha accelerates tumor progression in pancreatic cancer with diabetes. Int J Cancer 127:1795–1803. doi:10.1002/ijc.25188 PubMedCrossRefGoogle Scholar
  48. 48.
    Christa L, Carnot F, Simon MT, Levavasseur F, Stinnakre MG, Lasserre C, Thepot D, Clement B, Devinoy E, Brechot C (1996) HIP/PAP is an adhesive protein expressed in hepatocarcinoma, normal Paneth, and pancreatic cells. Am J Physiol 271:G993–G1002PubMedGoogle Scholar
  49. 49.
    Jäger D, Stockert E, Güre AO, Scanlan MJ, Karbach J, Jäger E, Knuth A, Old LJ, Chen YT (2001) Identification of a tissue-specific putative transcription factor in breast tissue by serological screening of a breast cancer library. Cancer Res 61:2055–2061PubMedGoogle Scholar
  50. 50.
    Lu J, Lian G, Zhou H, Esposito G, Steardo L, Delli-Bovi LC, Hecht JL, Lu QR, Sheen V (2012) OLIG2 over-expression impairs proliferation of human Down syndrome neural progenitors. Hum Mol Genet 21:2330–2340. doi:10.1093/hmg/dds052 PubMedCrossRefPubMedCentralGoogle Scholar
  51. 51.
    Xu ZP, Dutra A, Stellrecht CM, Wu C, Piatigorsky J, Saunders GF (2002) Functional and structural characterization of the human gene BHLHB5, encoding a basic helix-loop-helix transcription factor. Genomics 80:311–318. doi:10.1006/geno.2002.6833 PubMedCrossRefGoogle Scholar
  52. 52.
    Bao L, Loda M, Janmey PA, Stewart R, Anand-Apte B, Zetter BR (1996) Thymosin beta 15: a novel regulator of tumor cell motility upregulated in metastatic prostate cancer. Nat Med 2:1322–1328. doi:10.1038/nm1296-1322 PubMedCrossRefGoogle Scholar
  53. 53.
    Gu YM, Li SY, Qiu XS, Wang EH (2008) Elevated thymosin beta15 expression is associated with progression and metastasis of non-small cell lung cancer. APMIS 116:484–490. doi:10.1111/j.1600-0463.2008.00918.x PubMedCrossRefGoogle Scholar
  54. 54.
    Yokoyama M, Nishi Y, Yoshii J, Okubo K, Matsubara K (1996) Identification and cloning of neuroblastoma-specific and nerve tissue-specific genes through compiled expression profiles. DNA Res 3:311–320. doi:10.1093/dnares/3.5.311 PubMedCrossRefGoogle Scholar
  55. 55.
    Fung SJ, Joshi D, Allen KM, Sivagnanasundaram S, Rothmond DA, Saunders R, Noble PL, Webster MJ, Weickert CS (2011) Developmental patterns of doublecortin expression and white matter neuron density in the postnatal primate prefrontal cortex and schizophrenia. PLoS One 6:e25194. doi:10.1371/journal.pone.0025194 PubMedCrossRefPubMedCentralGoogle Scholar
  56. 56.
    Bechstedt S, Brouhard GJ (2012) Doublecortin recognizes the 13-protofilament microtubule cooperatively and tracks microtubule ends. Dev Cell 23:181–192. doi:10.1016/j.devcel.2012.05.006 PubMedCrossRefPubMedCentralGoogle Scholar
  57. 57.
    Hehr U, Uyanik G, Aigner L, Couillard-Despres S, Winkler J (2007) DCX-related disorders. In: Pagon RA, Adam MP, Bird TD, Dolan CR, Fong CT, Smith RJH, Stephens K (eds) GeneReviews® [Internet]. University of Washington, Seattle, Seattle, 1993–2014Google Scholar
  58. 58.
    Ervin JF, Heinzen EL, Cronin KD, Goldstein D, Szymanski MH, Burke JR, Welsh-Bohmer KA, Hulette CM (2007) Postmortem delay has minimal effect on brain RNA integrity. J Neuropathol Exp Neurol 66:1093–1099. doi:10.1097/nen.0b013e31815c196a PubMedCrossRefGoogle Scholar
  59. 59.
    Lockstone HE, Harris LW, Swatton JE, Wayland MT, Holland AJ, Bahn S (2007) Gene expression profiling in the adult Down syndrome brain. Genomics 90:647–660. doi:10.1016/j.ygeno.2007.08.005 PubMedCrossRefGoogle Scholar
  60. 60.
    Vawter MP, Evans S, Choudary P, Tomita H, Meador-Woodruff J, Molnar M, Li J, Lopez JF, Myers R, Cox D, Watson SJ, Akil H, Jones EG, Bunney WE (2004) Gender-specific gene expression in post-mortem human brain: localization to sex chromosomes. Neuropsychopharmacology 29:373–384. doi:10.1038/sj.npp.1300337 PubMedCrossRefPubMedCentralGoogle Scholar
  61. 61.
    Dmitrieva J, Chen C, Greenberger E, Ogunseitan O, Ding YC (2011) Gender-specific expression of the DRD4 gene on adolescent delinquency, anger and thrill seeking. Soc Cogn Affect Neurosci 6:82–89. doi:10.1093/scan/nsq020 PubMedCrossRefPubMedCentralGoogle Scholar
  62. 62.
    Wang PJ, McCarrey JR, Yang F, Page DC (2001) An abundance of X-linked genes expressed in spermatogonia. Nat Genet 27:422–426. doi:10.1038/86927 PubMedCrossRefGoogle Scholar
  63. 63.
    Brown CJ, Hendrich BD, Rupert JL, Lafrenière RG, Xing Y, Lawrence J, Willard HF (1992) The human XIST gene: analysis of a 17 kb inactive X-specific RNA that contains conserved repeats and is highly localized within the nucleus. Cell 71:527–542. doi:10.1016/0092-8674(92)90520-M PubMedCrossRefGoogle Scholar
  64. 64.
    Ooe N, Saito K, Mikami N, Nakatuka I, Kaneko H (2004) Identification of a novel basic helix-loop-helix-PAS factor, NXF, reveals a Sim2 competitive, positive regulatory role in dendritic-cytoskeleton modulator drebrin gene expression. Mol Cell Biol 24:608–616. doi:10.1128/MCB.24.2.608-616.2004 PubMedCrossRefPubMedCentralGoogle Scholar
  65. 65.
    Guo CJ, Douglas SD, Lai JP, Pleasure DE, Li Y, Williams M, Bannerman P, Song L, Ho WZ (2003) Interleukin-1beta stimulates macrophage inflammatory protein-1alpha and -1beta expression in human neuronal cells (NT2-N). J Neurochem 84:997–1005. doi:10.1046/j.1471-4159.2003.01609.x PubMedCrossRefPubMedCentralGoogle Scholar
  66. 66.
    Zunszain PA, Anacker C, Cattaneo A, Choudhury S, Musaelyan K, Myint AM, Thuret S, Price J, Pariante CM (2012) Interleukin-1β: a new regulator of the kynurenine pathway affecting human hippocampal neurogenesis. Neuropsychopharmacology 37:939–949. doi:10.1038/npp.2011.277 PubMedCrossRefPubMedCentralGoogle Scholar
  67. 67.
    Martignetti JA, Brosius J (1993) BC200 RNA: a neural RNA polymerase III product encoded by a monomeric Alu element. Proc Natl Acad Sci U S A 90:11563–11567PubMedCrossRefPubMedCentralGoogle Scholar
  68. 68.
    Kryger R, Fan L, Wilce PA, Jaquet V (2012) MALAT-1, a non protein-coding RNA is upregulated in the cerebellum, hippocampus and brain stem of human alcoholics. Alcohol 46:629–634. doi:10.1016/j.alcohol.2012.04.002 PubMedCrossRefGoogle Scholar
  69. 69.
    Bernard D, Prasanth KV, Tripathi V, Colasse S, Nakamura T, Xuan Z, Zhang MQ, Sedel F, Jourdren L, Coulpier F, Triller A, Spector DL, Bessis A (2010) A long nuclear-retained non-coding RNA regulates synaptogenesis by modulating gene expression. EMBO J 29:3082–3093. doi:10.1038/emboj.2010.199 PubMedCrossRefPubMedCentralGoogle Scholar
  70. 70.
    Iijima T, Miura E, Matsuda K, Kamekawa Y, Watanabe M, Yuzaki M (2007) Characterization of a transneuronal cytokine family Cbln–regulation of secretion by heteromeric assembly. Eur J Neurosci 25:1049–1057. doi:10.1111/j.1460-9568.2007.05361.x PubMedCrossRefGoogle Scholar
  71. 71.
    Li HY, Li JF, Lu GW (2003) Neurogranin: a brain-specific protein. Sheng Li Ke Xue Jin Zhan 34:111–115PubMedGoogle Scholar
  72. 72.
    Shen YC, Tsai HM, Cheng MC, Hsu SH, Chen SF, Chen CH (2012) Genetic and functional analysis of the gene encoding neurogranin in schizophrenia. Schizophr Res 137:7–13. doi:10.1016/j.schres.2012.01.011 PubMedCrossRefGoogle Scholar
  73. 73.
    Kuhn RM, Karolchik D, Zweig AS, Wang T, Smith KE, Rosenbloom KR, Rhead B, Raney BJ, Pohl A, Pheasant M, Meyer L, Hsu F, Hinrichs AS, Harte RA, Giardine B, Fujita P, Diekhans M, Dreszer T, Clawson H, Barber GP, Haussler D, Kent WJ (2009) The UCSC Genome Browser Database: update 2009. Nucleic Acids Res 37:D755–D761. doi:10.1093/nar/gkn875 PubMedCrossRefPubMedCentralGoogle Scholar
  74. 74.
    Safran M, Dalah I, Alexander J, Rosen N, Iny Stein T, Shmoish M, Nativ N, Bahir I, Doniger T, Krug H, Sirota-Madi A, Olender T, Golan Y, Stelzer G, Harel A, Lancet D (2010) GeneCards Version 3: the human gene integrator. Database (Oxford) 2010:baq020. doi: 10.1093/database/baq020
  75. 75.
    Hawrylycz MJ, Lein ES, Guillozet-Bongaarts AL, Shen EH, Ng L, Miller JA, van de Lagemaat LN, Smith KA, Ebbert A, Riley ZL, Abajian C, Beckmann CF, Bernard A, Bertagnolli D, Boe AF, Cartagena PM et al (2012) An anatomically comprehensive atlas of the adult human brain transcriptome. Nature 489:391–399. doi:10.1038/nature11405 PubMedCrossRefGoogle Scholar
  76. 76.
    Strippoli P, Pelleri MC, Caracausi M, Vitale L, Piovesan A, Locatelli C, Mimmi MC, Berardi AC, Ricotta D, Radeghieri A, Barisani B, Basik M, Monaco MC, Ghezzo A, Seri M, Cocchi G (2013) An integrated route to identifying new pathogenesis-based therapeutic approaches for trisomy 21 (Down Syndrome) following the thought of Jérôme Lejeune. Sci Postprint 1:e00010. doi:10.14340/spp.2013.12R0005 CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Maria Caracausi
    • 1
  • Lorenza Vitale
    • 1
  • Maria Chiara Pelleri
    • 1
  • Allison Piovesan
    • 1
  • Samantha Bruno
    • 1
  • Pierluigi Strippoli
    • 1
  1. 1.Department of Experimental, Diagnostic and Specialty Medicine (DIMES), Unit of Histology, Embryology and Applied BiologyUniversity of BolognaBolognaItaly

Personalised recommendations