Cerebral Cortex, March 2016;26: 1059–1071
doi: 10.1093/cercor/bhu276
Advance Access Publication Date: 1 December 2014
Original Article
ORIGINAL ARTICLE
Cortical Tubers: Windows into Dysregulation of Epilepsy
Risk and Synaptic Signaling Genes by MicroRNAs
Alan A. Dombkowski1, Carlos E. Batista1, Daniela Cukovic1,
Nicholas J. Carruthers3, Ramya Ranganathan1, Upasana Shukla1,
Paul M. Stemmer3, Harry T. Chugani1,2, and Diane C. Chugani1
1
Carman and Ann Adams Department of Pediatrics, 2Department of Neurology, Wayne State University School of
Medicine, Children’s Hospital of Michigan, Detroit Medical Center, Detroit, MI, USA, and 3Institute of
Environmental Health Sciences, Wayne State University, Detroit, MI, USA
Address correspondence to Dr Alan A. Dombkowski, Division of Clinical Pharmacology and Toxicology, Children’s Hospital of Michigan, 3901 Beaubien Blvd,
Detroit, MI 48201, USA. Email: [email protected]
Abstract
Tuberous sclerosis complex (TSC) is a multisystem genetic disorder caused by mutations in the TSC1 and TSC2 genes. Over 80%
of TSC patients are affected by epilepsy, but the molecular events contributing to seizures in TSC are not well understood. Recent
reports have demonstrated that the brain is enriched with microRNA activity, and they are critical in neural development and
function. However, little is known about the role of microRNAs in TSC. Here, we report the characterization of aberrant microRNA
activity in cortical tubers resected from 5 TSC patients surgically treated for medically intractable epilepsy. By comparing
epileptogenic tubers with adjacent nontuber tissue, we identified a set of 4 coordinately overexpressed microRNAs (miRs 23a, 34a,
34b*, 532-5p). We used quantitative liquid chromatography-tandem mass spectrometry (LC-MS/MS) proteomic profiling to
investigate the combined effect of the 4 microRNAs on target proteins. The proportion of repressed proteins among the predicted
targets was significantly greater than in the overall proteome and was highly enriched for proteins involved in synaptic signal
transmission. Among the combinatorial targets were TSC1, coding for the protein hamartin, and several epilepsy risk genes. We
found decreased levels of hamartin in epileptogenic tubers and confirmed targeting of the TSC1 3′ UTR by miRs-23a and 34a.
Key words: expression profiling, hamartin, proteomics, synapse, tuberous sclerosis complex
Introduction
Tuberous sclerosis complex (TSC) is an autosomal dominant genetic disorder resulting from mutations in the TSC genes, leading to activation of mammalian/mechanistic target of rapamycin (mTOR),
and is characterized by hamartomatous lesions in many organs, including brain lesions termed tubers (Roach and Sparagana 2004).
Tubers are characterized by abnormal cortical lamination and the
presence of dysmorphic neurons and balloon cells. Epilepsy afflicts
over 80% of TSC patients, and many suffer seizures that do not respond to pharmacological therapy and require surgical resection of
epileptogenic tubers (Thiele 2004; Weiner et al. 2004). The molecular
events that contribute to seizures in TSC are not well characterized.
MicroRNAs are endogenous noncoding RNAs that provide
post-transcriptional regulation of many genes and most cell processes. Approximately one-third of all protein-coding genes are
believed to be regulated by microRNAs, and aberrant microRNA
expression has been implicated in a wide range of diseases (Filipowicz et al. 2008). Mature microRNAs are ∼22 nucleotides in
length and anneal to complementary sites in the 3′ untranslated
region (UTR) of target transcripts as part of an RNA-induced silencing complex, resulting in either transcript degradation or inhibition of protein translation. There are currently over 2000
known mature microRNAs in the human genome, and each microRNA may regulate dozens to hundreds of target transcripts.
© The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected]
1059
1060
| Cerebral Cortex, 2016, Vol. 26, No. 3
The 3′ UTR in a single-messenger RNA may contain binding sites
for numerous microRNAs, and a transcript can be concurrently
repressed by multiple microRNA species. Compared with other
organs, the brain is highly enriched for microRNA activity, likely
due to the remarkable level of structural and functional complexity in the tissue (Fineberg et al. 2009). Numerous microRNAs are
differentially expressed during corticogenesis and neuronal differentiation (Krichevsky et al. 2003; Nielsen et al. 2009). Recent reports have also implicated microRNAs in epileptogenesis
(Jimenez-Mateos et al. 2012; You et al. 2012). However, little is
known about the involvement of microRNAs in TSC, and there
are currently no published reports on their role in cortical tubers.
A potential role for microRNAs in neurological manifestations of
TSC is suggested by the prominent role of microRNAs in neurodevelopment and evidence indicating microRNA regulation downstream of mTOR. MicroRNAs have been reported to be regulated
by p53 in different cell types, and p53 levels and activity are regulated by mTOR (Lee et al. 2007; Shin et al. 2012). In addition, inflammation, which is present in epilepsy and cortical tubers
(Maldonado et al. 2003; Aronica and Crino 2011), has been
shown to induce microRNA expression through inflammatory
cytokines (Junker et al. 2009). In this work, we characterized microRNA expression patterns in cortical tubers resected to treat intractable epilepsy in TSC patients and compared expression
levels with patient-matched adjacent nontuber tissue that was
concurrently resected due to epileptiform electroencephalogram
(EEG) activity. Furthermore, we used quantitative proteomics to
assess the effects of altered microRNA levels on protein expression, as well as investigate the potential roles for p53 and inflammation as mechanisms involved in regulating microRNA levels in
TSC.
Materials and Methods
TSC Brain Tissue
Microarray analysis was performed on brain tissue samples obtained from 5 patients who underwent epilepsy surgery for medically refractory epilepsy at Children’s Hospital of Michigan,
Detroit. Table 1 shows patient demographics and mutation
types. Inclusion criteria were: 1) diagnosis of TSC defined by clinical criteria developed in a consensus conference (Roach et al.
1999); and 2) resective surgery for treatment of medically refractory epilepsy; 3) both epileptogenic tuber and adjacent nontuber
tissue available; and 4) written informed consent of parent or
guardian. Exclusion criteria were: history of previous brain surgery that may have affected the original epileptic focus (including
epilepsy surgery, surgery for subependymal giant cell astrocytoma, shunt placement for hydrocephalus). Preoperative assessment included clinical evaluation, neuroimaging using
magnetic resonance imaging (MRI) and positron emission tomography (PET) with alpha[C-11]methyl--tryptophan (AMT) (Chugani et al. 2013), and ictal/interictal EEG. Two tissue types were
sampled from the surgical resection determined by clinical
parameters for each patient: 1) tuber characterized by independent epileptiform activity (seizure onset); and 2) nontuber tissue
adjacent to the epileptogenic tuber (nononset). Portions of each
sample from frozen blocks were immediately placed in extraction buffer for RNA and protein isolation. All protocols were approved by the Human Investigation Committee at Wayne State
University.
RNA Isolation and Quality
Total RNA, including small RNAs, was isolated from brain tuber
and nontuber samples using 25 mg of each tissue sample and
the “All-in-One Purification” kit (Norgen Biotek, Thorold, ON,
Canada). All steps were performed according to the vendor’s
protocol. RNase-free DNaseI (Qiagen, Valencia, CA, USA) was
used to remove potential gDNA contamination from the samples.
RNA integrity was tested using Agilent 2100 Bioanalyzer with the
RNA 6000 Nano Assay kit or RNA 6000 Pico Assay kit (Agilent
Technologies, Palo Alto, CA, USA). The Bioanalyzer was used
to determine if the 18S and 28S ribosomal bands were defined
and to ensure minimal degradation of RNA. Optimal concentration of total RNA with the Nano assay was 100–200 ng/μL, and
0.5–1 ng/μL for testing with Pico assay. All steps were done according to the vendor’s protocol.
Sample Labeling and microRNA Microarray Hybridization
For each sample, 100 ng of total RNA along with Agilent microRNA spike-in controls were treated with calf intestine alkaline
phosphatase (GE Healthcare Bio-Sciences Corp., Piscataway, NJ,
USA), and incubated at 37 °C for 30 min. Then the samples were
denatured and labeled with Cyanine 3 per the vendor’s protocol
(Agilent Technologies). The labeled microRNAs were cleaned
using micro Bio-Spin 6 columns (Bio-Rad Laboratories, Hercules,
CA, USA), and dried in a speed-vac. Dried samples were resuspended in 17 μL of nuclease-free water. Also, 1 μL of Hyb
Spike-in, 4.5 μL of 10× GE Blocking, and 22.5 μL of 2× Hi-RPM
hybridization buffer (Agilent) were added, and the hybridization
mixture was incubated at 100 °C for 5 min and then put on ice for
5 min. Samples were immediately added to the microarrays in an
Agilent SureHyb hybridization chamber and rotated at 20 rpm in
a hybridization oven for 20 h at 55 °C. Agilent Human miRNA V3
microarrays were used. Microarrays were processed using the
miRNA Microarray System protocol v.2.2 (Agilent Technologies).
Slides were scanned using an Agilent dual laser scanner. Tiff
images were analyzed using Agilent’s feature extraction software
(version 10.7.1.1).
Microarray Data Analysis
Microarray data were imported into GeneSpring version 12 for
normalization and analysis (Agilent Technologies). The data on
Table 1 Patient demographics
Identifier
Gender
Age
Gene
Location of resected tissue
Mutation
A1409
E3007
H2407
81603
90602
F
F
M
M
M
2 years 6 months
10 years
8 years 9 months
8 years 11 months
7 years 10 months
TSC1
TSC1
TSC2
TSC2
TSC2
Right temporal
Left frontal
Left temporal
Right frontal
Left occipital/parietal
Frameshift, 8-bp insertion, nt 1609
Frameshift, 5-bp deletion, nt 1020–1024
Frameshift, 1-bp deletion nt 5444
Transition nt 2410 t->c, codon 804 Cys->Arg
Transition nt 5227 c->t, codon 1743 Arg -> Trp
Aberrant MicroRNA Expression in Cortical Tubers
each array were quantile normalized, an interarray normalization procedure that ensures equivalent signal distributions for
all arrays. Quantile normalization has been shown to be robust
and preferable for analysis of microRNA expression in tissue
samples (Rao et al. 2008). We then performed a series of stringent
filtering and statistical analyses, as outlined below. Filtering
microarray data using detection calls and variance metrics are
useful methods to improve statistical power while controlling
the false discovery rate (Hackstadt and Hess 2009). Following
this approach, we selected probes flagged as detected and also
having an expression level above the 25th percentile in at least
75% of the samples for at least one condition being compared
(tuber or nontuber). Additionally, we filtered out all probes having
a coefficient of variation ≤5.0%, calculated from all samples, as
these microRNAs are most likely to not be differentially expressed. Statistical analysis was performed using a moderated
t-test with the Westfall and Young family-wise error rate
(FWER) correction. The FWER indicates the probability that one
or more false positives exist among the identified set.
Quantitative RT-PCR of microRNA Expression
cDNAs of all microRNAs were prepared using a single reverse
transcription reaction for each sample. About 100 ng total RNA
and Universal cDNA Synthesis Kit (Exiqon, Inc., Woburn, MA,
USA) were used in a total volume of 20 μL. Incubation at 42 °C
for 60 min was followed by heat inactivation at 95 °C for 5 min
and immediate cooling to 4 °C. Undiluted cDNA was stored at
−20 °C. Immediately before use, cDNA of each sample was ×40 diluted and 8 μL used for the real-time PCR reaction with Exiqon
SYBR Green master mix. ROX passive reference dye (Affymetrix,
Inc., Santa Clara, CA, USA) was diluted 1:10 in nuclease-free water
and 0.4 μL of the ROX dilution was added per 20 μL qPCR reaction.
LNA primer sets for miRs-23a, 34a, and 532-5p were obtained
from Exiqon. Standard cycling condition for SybrGreen with
melting curve analysis was selected as well as manual baseline
and threshold setup. The ΔΔCt method was used to quantitate
differential expression. We used hsa-miR-26a as an endogenous
control in the microRNA qRT-PCR measurements. While miR-26a
has been reported as differentially expressed in some malignancies (Gao and Liu 2011), it was utilized as a control in our experiment because it had the lowest coefficient of variation (1.4%)
among all microRNAs as calculated from the microarray data of
all samples analyzed. We found that mir-26a has considerably
less variation in these tissues than U6 snRNA, a frequently
used control (data not shown). Additionally, miR-26a has a relatively high level of expression and has been used previously as an
endogenous control (Viprey et al. 2012). A two-tailed t-test with
equal variance was used to determine statistical significance.
Computational Prediction of microRNA Target Genes
and Ontology Analysis
We utilized miRWalk to perform a consensus analysis with 10 algorithms to predict target transcripts (Dweep et al. 2011). For the
combinatorial analysis of miRs 34a, 23a, 532-5p, and 34b*, we tabulated, for each target gene, the number of consensus agreements among the 10 algorithms for each of the 4 microRNAs. If
multiple sites were predicted for a given microRNA/transcript
interaction then the highest scoring site was used. The sum of
all 4 microRNAs was used as the total combinatorial score for
each transcript. Functional annotation analysis of target gene
sets was performed with DAVID (http://david.abcc.ncifcrf.gov)
using default statistical settings. To identify putative target
Dombkowski et al.
| 1061
genes associated with epilepsy, we utilized 2 public databases:
CarpeDB (http://www.carpedb.ua.edu) and epiGAD (http://www.
epigad.org). The interaction between miRs-34a, 23a, and 532-5p
and the 3′ UTR of TSC1 were modeled using RNAhybrid software
(Rehmsmeier et al. 2004).
Luciferase 3′ UTR Reporter Assay
The Cos-7 cell line was used in a dual-luciferase reporter assay
(Expresso single shot, Genlantis, San Diego, CA, USA). The fulllength TSC1 3′ UTR-luciferase reporter construct (SC222100), a
hsa-miR-34a precursor expression plasmid (SC400356), and control expression vector ( pCMV-MIR) were obtained from Origene
(Rockville, MD, USA). The miR-23a precursor expression plasmid
(HmiR0298-MR04) and scrambled expression control vector
(CmiR0001-MR04) were purchased from Genecopoeia (Rockville,
MD, USA). A pGL4.73 (SV40) Renilla-luc expression vector was
used to control for transfection efficiency (#E6911, Promega,
Madison, WI, USA). Cells were plated in DMEM (high glucose,
Gibco, Life Technologies, Grand Island, NY, USA) with 10%
FBS and 1% nonessential amino acids. The cells were plated in
a 96-well flat bottom white plate on the day of transfection.
After 3-h cells were transfected with Renilla-luc reporter,
3′ UTR-luc reporter, and microRNA expression vector or microRNA expression control vector (empty or scrambled) in the concentration ratio of 0.3:2.5:4.0 (Origene vector) or 0.25:2.5:4.25
(Genecopoeia vector) respectively, with Fugene 6 (Promega) as
per manufacturer’s instructions. Transfections were performed
using at least 3 replicates for each condition. After 48 h of transfection, the Renilla and Firefly luciferase activities were quantified by the Dual Glo luciferase assay system (Promega)
measured on a FlexStation 3 universal plate reader (Molecular
Devices, Madison, WI, USA). Data analysis was performed using
Softmax Pro. The activities of 3′UTR reporter gene were normalized by those of control Renilla-luc and expressed as a fold
change by comparing cells transfected with the miR-34a or 23a
expression vector to those transfected with empty or scrambled
expression vector. A two-tailed t-test with equal variance was
used to determine statistical significance.
Protein Extraction
Fresh 1× cell lysis buffer was prepared from 10× cell lysis buffer
( p/n 9803, Cell Signaling, Beverly, MA, USA) and protease inhibitor cocktail (DMSO solution) (Sigma p/n P8340) was added in 1:100
dilution. Frozen brain tissue (25–50 mg) was homogenized with a
mortar and a pestle for each sample. Liquid nitrogen was added
into the mortar during homogenization to keep the tissue frozen.
Cold 1× cell lysis buffer (500–800 μL) with protease inhibitor cocktail was added to the frozen tissue powder, mixed and scooped
into a 2-mL eppendorf tube. Then the protein extract was centrifuged for 10 min at 14 000 × g in a cold centrifuge. Supernatant
was transferred into a fresh 1.5-mL eppendorf tube and smaller
aliquots of each protein extract were stored at −80 °C.
Quantitative LC-MS/MS Proteomics Analysis
Each protein sample (95 µg) was reduced with dithiothreitol and
alkylated and iodoacetamide (Sigma, St Louis, MO, USA). Samples
were then diluted so that the Triton-X concentration was 0.05%
and digested with trypsin. Five microgram was analyzed for complete digestion by SDS-PAGE, and the remaining material was desalted using an Oasis 30-mg HLB cartridge (Waters, Milford, MA,
USA). A micro BCA assay (ThermoFisher Scientific, Waltham, MA,
1062
| Cerebral Cortex, 2016, Vol. 26, No. 3
USA) was used to confirm protein concentrations and then 43 µg
of each sample was taken for TMT-10plex labeling (ThermoFisher
Scientific). In addition, an equal aliquot of each sample was
pooled for a reference sample that was also 43 µg. Eight samples
and the reference pool were labeled according to the manufacturer’s protocol; one channel was left empty. The labeled samples
were combined and 200 µg was taken for analysis. Detergent was
removed by serial solid phase extraction using reversed phase,
strong cation exchange, and reversed phase materials. An Oasis
HLB cartridge (Waters), 4 strong cation exchange spin columns
(Nest Group, Southborough, MA) and a Pierce 100 µL C18 tip (ThermoFisher Scientific) were used, respectively. The sample was then
dried and resuspended and fractionated at pH 10 using an Agilent
PLRP-S 0.5 × 150 mm column (Agilent, Santa Clara, CA, USA) with
elution by a gradient of acetonitrile into 33 fractions. The fractions
were then dried and resuspended and submitted for LC-MS3 analysis on an Orbitrap Fusion mass spectrometer equipped with an
EASY-nLC UHPLC (ThermoFischer Scientific). Peptides from each
fraction were eluted into the mass spectrometer over a period of
110 min. Peptide fragmentation and reporter ion detection were
performed using simultaneous precursor selection, where the
top 10 most abundant MS2 fragments were selected for higher energy collisional dissociation fragmentation (McAlister et al. 2012).
MS2 fragmentation was at 30% collision energy and MS3 fragmentation was at 65% collision energy.
Proteome Discoverer 1.4.1.14 was used to extract and search
mass spectra against the Uniprot human canonical database
(February 2014; 20 264 entries) using Sequest HT with the percolator algorithm to assign confidence to identifications. Because of
the large number of spectra, each fraction was searched and
quantified separately, and the search results were then combined
in Proteome Discoverer. Peptides that were identified with high
confidence (1% false discovery rate) were used for subsequent
analysis. Proteins that were identified on the basis of those peptides were grouped to find the smallest list of proteins that explains all identified peptides. Reporter ions were quantified
relative to the all-sample-pool reporter intensity and the ratios
for each fraction were normalized to a median of 1. The resultant
proteomics data were imported into GeneSpring v12.6.1 for quantitative analysis. Ratios reflecting relative protein abundance to
the all-sample pool were quantile normalized for each of the
samples. Statistical significance of alterations in protein level between tuber and nontuber samples was determined using a twosided student’s t-test and the Storey bootstrap false discovery
rate method (GeneSpring). Fisher’s exact test was used to calculate significance of enrichment for repressed target proteins
(JMP 11).
For analysis of p53 pathway activation, we used a published
list of genes with multiple levels of evidence demonstrating direct induction by p53 transcriptional regulation (Riley et al. 2008).
Of the 136 genes annotated with p53 activator sites in Supplementary Table 2 of Riley et al., we found 28 with proteomics spectra for all samples in our dataset. These were subsequently used
to assess p53 activity. Likewise, we used a set of direct target
genes to measure NF-κB activity. We used the curated list of
NF-κB target genes from http://bioinfo.lifl.fr/NF-KB/ and downloaded all genes identified as “human gene with checked binding
sites.” Of the 104 genes meeting this criterion, 24 genes had proteomics spectra for all of our TSC samples. Significance of p53
and NF-κB pathway activation was determined by analysis of
log2(T/N) values for target proteins, where T represents protein
expression in tuber and N is expression in nontuber. The onesample t-test against zero was used to determine statistical significance (one-sided, JMP 11). Empirical cumulative distributions
of the log2(T/N) values from the proteome (5748 proteins), p53-activated proteins (28), and NF-κB target proteins (24) were derived
with JMP 11. Statistical significance of the difference between the
proteome distribution and each of the p53 and NF-κB distributions was determined using the two-sample Kolmogorov–Smirnov test (two-sided, JMP 11).
Results
Identification of Aberrant microRNA Expression in
Cortical Tubers
We performed microRNA expression analysis on matched pairs
of epileptogenic tubers (seizure onset) and adjacent nontuber tissue (nononset) for 5 subjects: 2 with TSC1 mutations and 3 with
TSC2 mutations (Table 1). Agilent microRNA microarrays having
probes for more than 900 human microRNAs were used to quantify microRNA expression for each of the 10 tissue samples. Stringent filtering and statistical analyses (Materials and Methods)
were used to arrive at a high-confidence list of differentially
expressed microRNAs. A FWER correction was used to ensure
that the probability of one or more false positives being among
the selected microRNAs was only 5%. This approach identified
5 statistically significant microRNAs having a minimum 2-fold
difference between tuber and nontuber tissues (Table 2). Four microRNAs exhibited higher expression in tuber compared with
nontuber tissue (miRs-34a, 23a, 34b*, 532-5p), while one microRNA had lower expression (miR-381).
Quantitative RT-PCR Confirms microRNA Overexpression
in Cortical Tubers
Computational analysis (below) suggested that miR-34b* targets
few transcripts, while miRs 23a, 34a, and 532-5p were predicted
to have a combinatorial effect on many genes involved in neurological processes. Therefore, we performed qRT-PCR analysis on
the latter group of microRNAs to confirm the microarray results.
Expression levels for each microRNA were compared between patient-matched tuber and nontuber tissue and the mean fold
change calculated from the 5 patients (Table 2). All 3 microRNAs
were confirmed to have higher expression in tubers than the adjacent nontuber tissue, consistent with the microarray results.
Table 2 Statistically significant microRNAs differentially expressed in TSC tubers
microRNA
P
P (corrected)
Microarray FC
qRT-PCR FC
Chr
miRbase accession
hsa-miR-34a
hsa-miR-23a
hsa-miR-532-5p
hsa-miR-34b*
hsa-miR-381
0.0012
0.0048
0.0101
0.0007
0.0098
0.0000
0.0000
0.0202
0.0000
0.0101
3.04
2.65
2.89
3.58
−2.18
6.79
2.80
1.86
1
19
X
11
14
MIMAT0000255
MIMAT0000078
MIMAT0002888
MIMAT0000685
MIMAT0000736
Aberrant MicroRNA Expression in Cortical Tubers
Comparison of microRNA Expression in Tubers and
Nontuber Tissue to Normal Cortex
Tissue was designated as tuber or nontuber by a neuropathologist, although the nontuber tissue may not be entirely normal.
The tissue was adjacent to the seizure-onset zone (tuber) and
also exhibited interictal spiking activity on the intracranial EEG.
Additionally, both tuber and nontuberal tissues are haploinsufficient for TSC1 or TSC2. We found differential expression of several
microRNAs in the matched tissue comparison; however, this approach did not indicate how the expression levels relate to normal cortical tissue from non-TSC subjects. To address that
question, we utilized microRNA data from a postmortem (PM)
study of primate brain tissue that included samples from
human, chimpanzee, and macaque (Somel et al. 2011). The
study investigated microRNA expression in prefrontal cortex
and cerebellar cortex from subjects representing a wide range
of ages. We downloaded the dataset from the NCBI Gene Expression Omnibus repository (GEO ID GSE29356) and utilized the data
for prefrontal cortex specimens from 8 human subjects ranging
in age from 2 days to 88 years. These data were generated using
the same microRNA microarray platform and version that we
used in our TSC study, thus facilitating direct comparison to
our results. To further ensure compatibility between the 2 datasets, we co-normalized the primate PM data with our TSC data
using quantile normalization. This procedure results in matched
quantiles of expression level across all microarrays. To determine
if the normalization approach was effective, we compared expression levels between the 2 datasets for 51 microRNAs that
were invariant in our TSC tuber versus nontuber analysis.
These microRNAs had expression coefficients of variations <5%,
indicating that they are not differentially expressed in TSC tubers
compared with nontuber tissue, and we would therefore expect
similar expression levels in the PM controls. Using normalized
expression values for the 51 microRNAs, a scatter plot comparison of the PM values versus the TSC data provided an excellent
linear correlation, with a Pearson’s correlation coefficient = 0.93
and a linear fit of y = 0.997x (data not shown). These metrics indicate that the normalized expression values from the 2 datasets
are similar, thus enabling the below comparison of microRNA expression in TSC tissues to their levels in normal PM cortex.
Analysis of the expression level of miR-34a in the PM control
samples revealed a pronounced postnatal increase in expression
that plateaus after the teenage years. Age-dependent elevation of
miR-34a expression in brain and cardiac tissue has previously
been reported (Somel et al. 2010; Li et al. 2011; Boon et al. 2013).
Since our pediatric subjects are within the timeframe of rapid increase in miR-34a expression, it is imperative that we consider
subject age when investigating microRNA expression events in
this population. We did not observe age-dependent expression
of miRs-23a, 34b*, 532-5p, and 381 in the PM dataset. Figure 1A
presents miR-34a expression as a function of age for each of the
PM subjects and the TSC samples in our study. Expression levels
for PM controls are represented by gray points and are fit with a
logarithmic curve (blue). Expression in nontuber tissue from
our TSC subjects is indicated by yellow inverted triangles, and
these points are generally situated on or near the curve fit to
the PM controls, indicating that miR-34a expression in the TSC
nontuber tissue is near normal. However, the expression of
miR-34a in epileptogenic tubers is clearly elevated above the expected age-associated level for each subject, as indicated by the
red diamonds. Expression of miRs-23a, 34b*, and 532-5p in nontuber tissue was similar to levels in the PM controls, and the corresponding expression level in epileptogenic tubers was clearly
Dombkowski et al.
| 1063
elevated for each microRNA (Fig. 1B). A measurable difference between PM controls and nontuber tissue was observed for miR381, with expression of tuber < nontuber < PM (data not shown).
Quantitative LC-MS/MS Proteomic Profiling
Demonstrates Significant Repression Among
Combinatorial Targets and Enrichment for Synaptic
Signal Transmission Proteins
We used liquid chromatography-tandem mass spectrometry
(LC-MS/MS) proteomics with tandem mass tags (TMT) to quantify
protein levels in tuber and nontuber tissues. This approach allowed us to: 1) perform quantitative analysis that is superior to
traditional Western blots (Aebersold et al. 2013); 2) perform global
analysis for a large portion of the proteome; and 3) perform concurrent analysis of all samples (TMT) thus providing an important internal control for the comparison of tuber and nontuber
protein levels. One limitation of LC-MS/MS is the requirement
for an ample amount of protein. We had sufficient protein from
tuber and nontuber from 3 of the 5 patients (E3007, 81603, H2407).
Patient 90602 had sufficient protein from nontuber tissue. We were
also able to extract protein from an additional tuber from subject
81603. Thus, the proteomics analysis was performed on 8 samples
in total: 4 tuber and 4 nontuber tissues. Overall 614 115 MS2 spectra
were submitted and 136 208 were identified (22%). About 42 060
peptides and 6309 proteins were identified on the basis of these
spectra. We obtained quantifiable mass spectra for 5749 proteins
in all 8 samples. The set of 5749 proteins quantified represents
somewhat less than half of all the proteins expected to be present
in these tissues (Michalski et al. 2011; Nagaraj et al. 2011). Very low
abundance proteins and proteins that do not generate tryptic peptides that produce good spectra are unlikely to be identified. We
first identified proteins with a significant change in abundance between tuber and nontuber samples. Statistical selection using a
10% false discovery rate and minimum 1.5-fold change resulted
in 1745 significant proteins. Of these, 842 proteins were elevated
and 903 had lower levels in tubers.
To investigate protein expression altered by the overexpressed microRNAs, we first identified high-confidence target
transcripts using a prediction consensus approach. It is widely
recognized that computational methods for microRNA target
prediction have a high false-positive rate (Yue et al. 2009;
Reyes-Herrera and Ficarra 2012). The number of false positives
can be reduced by applying a consensus of multiple prediction
methods (Kuhn et al. 2008). We used a consensus tool (miRWalk)
to identify putative targets for each of the 4 microRNAs overexpressed in tubers (Dweep et al. 2011). A consensus score was defined for each microRNA/transcript interaction as the number of
algorithms (of 10) that predicted that the transcript is targeted by
the given microRNA. We then focused on the set of transcripts
that are targets of more than one of the 4 microRNAs. Since a single microRNA may target many transcripts, and a given transcript 3′ UTR may have target sites for multiple microRNA
species, it is important to consider potential combinatorial targeting among a set of differentially expressed microRNAs (Dombkowski et al. 2011). To account for such combined effects a
combinatorial consensus score (CCS) was calculated for each
transcript by summing the individual consensus score from all
4 of the microRNAs. The target transcripts were ranked using
the CCS.
Our proteome profiling allowed us to empirically determine
the minimal CCS associated with significant target repression.
Since the 4 microRNAs were overexpressed in tubers their activity
1064
| Cerebral Cortex, 2016, Vol. 26, No. 3
Figure 1. Expression of 4 microRNAs is significantly elevated in epileptogenic tubers compared with adjacent nontuber tissue and postmortem controls. (A) Expression
data from a previously published postmortem study reveals age-dependent expression of miR-34a in human prefrontal cortex (gray points and blue curve). Considering
subject age, the expression of miR-34a in TSC nontuber tissue is near normal (yellow triangles), while expression in tubers is elevated (red diamonds). (B) MicroRNA
expression levels in TSC nontuber tissue (N, yellow boxes) are similar to postmortem controls (P, black boxes) for miRs-23a, 532-5p, and 34b*. Each microRNA is
overexpressed in epileptogenic tubers (T, red boxes) compared with controls.
would be evident by a reduction in target protein levels. Using the
set of 1745 differentially expressed proteins, we binned the proteins by CCS and then determined the fraction of proteins in
each bin (score) that were repressed in tubers. Considerable enrichment for repressed proteins was observed with targets having a CCS of 13 or greater. We found that 68% (85/125) of
proteins having a CCS ≥13 were repressed as compared with
50.5% (818/1620) of proteins with a CCS ≤12 (P = 0.00018) (Fig. 2).
The results demonstrate significant repression among proteins
having a high-confidence combinatorial target score, evidence
of targeting by the 4 overexpressed microRNAs.
We performed a functional analysis of the 85 repressed combinatorial target proteins (CCS ≥13) using DAVID (Database for Annotation, Visualization, and Integrated Discovery) to identify gene
ontologies and functional categories associated with the repressed targets (Huang da et al. 2009). Using the gene ontology
(GO) biological process, the most significant ontology was “transmission of nerve impulse. ” By definition of the GO Consortium,
this ontology is assigned to genes/proteins involved in synaptic
signaling and consequent electrochemical polarization/
depolarization in neurons. The fraction of repressed target proteins with this ontology (14/85, Table 3) is highly significant (P =
1.6e−8) and represents a 7.8-fold enrichment. The enrichment
value indicates that the number of repressed target proteins involved in synaptic signaling is nearly 8 times what is expected
by chance, given the number of genes with this ontology found
throughout the entire genome. The proportion of proteins in this
functional category (16.5%) is markedly higher in the repressed
combinatorial targets than in the group of repressed nontarget
proteins (5.9%). The results demonstrate that the set of transcripts
targeted by the collective group of 4 microRNAs is highly enriched
for genes/proteins involved in synaptic signaling. Additionally, 9
genes known to confer risk of epilepsy are among the 85 repressed
combinatorial target genes (Table 3), including the TSC1 protein
hamartin, synapsin II (SYN2), γ-aminobutyrate (GABA) A receptor
beta3 (GABRB3), doublecortin (DCX), and neurofibromin (NF1).
The consensus score is shown for each predicted microRNA/target
interaction. The scores for miR-34b* are consistently low, suggesting that this microRNA is not significantly involved in the regulation of these transcripts. Conversely, the scores for miR-34a are
Aberrant MicroRNA Expression in Cortical Tubers
consistently high indicating that this microRNA has a prominent
role in the post-transcriptional regulation of these genes. Other
high-ranking and significantly enriched ontologies among the
Dombkowski et al.
| 1065
repressed target proteins included “neurological system process”
and “neuron development.”
Hamartin is Repressed in Tubers
A notable repressed protein among the combinatorial targets is
hamartin, the protein product of the TSC1 gene. The quantitative
proteomics data show that, on average, hamartin levels were 2.2fold lower in epileptogenic tubers compared with nontuber tissue
(Fig. 3). Interestingly, hamartin levels were decreased in tubers of
the TSC1 patient and the TSC2 patients. TSC1 is predicted to be a
target of miRs-23a, 34a, and 532-5p based on a consensus of 5, 4,
and 5 prediction algorithms, respectively (Table 3).
Luciferase Reporter Assays Confirm Predicted
microRNA/Target Interactions of TSC1
Figure 2. The proportion of repressed proteins among predicted combinatorial
targets is significantly greater than in the overall proteome. Using quantitative
liquid chromatography-tandem mass spectrometry proteomics, a total of 1745
proteins were identified with significant differences in protein abundance
comparing epileptogenic tubers with adjacent nontuber tissue. The proteins
were segregated into 2 primary sets: 1) predicted combinatorial targets of the 4
overexpressed microRNAs; and 2) the remaining proteins ( proteome). The
fraction of repressed proteins in the predicted combinatorial target set is
significantly greater than in the overall proteome (P = 0.00018), reflecting the
repressive effect of the 4 microRNAs.
We utilized a luciferase reporter assay to validate targeting of the
3′ UTR of TSC1 by miRs-23a and 34a. A vector construct created
with the firefly luciferase reporter gene and the 3′UTR of TSC1
was transfected into COS-7 cells. This cell line has been widely
used for microRNA target validation, including studies of neural
development (Peng et al. 2012; Zhao et al. 2012). Importantly, we
used a construct that included the entire TSC1 3′ UTR since truncated UTRs may alter microRNA/UTR interactions (Kuhn et al.
2008). Co-transfection with a vector expressing pre-miR-34a
resulted in a 71% decrease in expression of the luciferase reporter
compared with control expression vector (P = 0.0005) (Fig. 4).
A 40% decrease in reporter expression was observed when
co-transfecting with pre-miR-23a compared with control expression vector transfection.
Table 3 Combinatorial target proteins significantly repressed in tubers and associated with synaptic signaling or epilepsy risk
Gene
GABRA3
CACNA1E
SCN3B
GABBR2
GABRB3
SLC6A1
SYN2
DCX
SLC30A3
APBA1
NPTX1
TSC1
KCNMA1
PVRL1
SNPH
NF1
KIF1B
Description
Gamma-aminobutyric acid receptor
subunit alpha-3
Voltage-dependent R-type calcium
channel subunit alpha-1E
Sodium channel subunit beta-3
Gamma-aminobutyric acid type B
receptor subunit 2
Gamma-aminobutyric acid receptor
subunit beta-3
Sodium- and chloride-dependent
GABA transporter 1
Synapsin-2
Neuronal migration protein
doublecortin
Zinc transporter 3
Amyloid beta A4 precursor proteinbinding family A member 1
Neuronal pentraxin-1
Hamartin
Calcium-activated potassium
channel subunit alpha-1
Poliovirus receptor-related protein 1
Syntaphilin
Neurofibromin
Kinesin-like protein KIF1B
UniProt
Accession
Protein FC
(T/N)
Trans. nerve
impulse
P34903
−6.7
♦
Q15878
−5.0
♦
Q9NY72
O75899
−4.4
−3.0
P28472
Epilepsy
risk
Prediction score
miR34a
miR23a
miR532-5p
miR34b*
5
3
3
2
▪
5
5
2
2
♦
▪
▪
5
6
5
0
4
5
2
3
−2.9
♦
▪
5
6
2
2
P30531
−2.7
♦
6
5
5
0
Q92777
O43602
−2.7
−2.6
♦
▪
▪
4
7
3
2
5
6
3
2
Q99726
Q02410
−2.4
−2.3
▪
♦
7
7
4
4
0
1
2
2
Q15818
Q92574
Q12791
−2.2
−2.2
−1.9
♦
♦
♦
6
4
5
1
5
5
4
5
5
2
2
3
Q15223
O15079
P21359
O60333
−1.9
−1.7
−1.6
−1.6
♦
♦
♦
♦
6
5
3
5
3
2
5
5
4
5
5
5
2
2
1
3
Note: Shading corresponds to the consensus prediction score.
▪
▪
1066
| Cerebral Cortex, 2016, Vol. 26, No. 3
Figure 3. Hamartin is significantly reduced in epileptogenic tubers compared with
perituberal tissue. Among the differentially expressed proteins identified in our
quantitative LC-MS/MS proteomics experiment, hamartin was repressed 2.2fold in tubers (T) compared with nontuber tissue (N). Reduced levels were
observed in subjects having mutations in either TSC1 or TSC2 genes.
To assess p53 pathway activation, we used 28 proteins known
to be directly induced by p53 through transcriptional regulation
(Materials and Methods). For each of these proteins, we calculated log2(T/N) using the average expression value for tuber (T)
and nontuber (N). This parameter is zero when no change in expression is measured, negative when repressed in tubers, and
positive when overexpressed in tubers compared with nontuber.
This metric was also calculated for all 5749 proteins having proteomics spectra for all samples. We refer to this set as the “proteome.” Expression levels for the group of 28 p53-activated proteins
were significantly elevated (P = 0.015), with mean log2(T/N) = 0.35
(1.27-fold). Mean log2(T/N) for the proteome was 0.0029, very close
to the expected value of zero.
We performed the same analysis for a group of 24 proteins established as direct targets of NF-κB, a key mediator of inflammation (Materials and Methods). We found that expression levels of
the overall set of NF-κB target proteins to be significantly elevated
in tubers (P < 0.0001), with mean log2(T/N) = 0.96 (1.94-fold). The
distributions of log2(T/N) values for the p53- and NF-κB-regulated
proteins were significantly different than the overall proteome
and reflect overall induction of the proteins. These results are evident in the empirical cumulative distributions shown in Figure 5.
Cumulative distributions provide the probability (y-axis) of observing a specified value or less (x-axis). The cumulative distribution of log2(T/N) values for the proteome is shown by the red, solid
curve. Distributions for p53- and NF-κB-regulated proteins are
clearly shifted to the right, reflecting increased expression in tubers for proteins regulated by the 2 transcription factors. Comparison of each distribution to the overall proteome reveals
statistical significance, P = 0.0112 and 0.0005, respectively, for
p53 and NF-κB target sets. Tables with proteomics data for p53
and NF-κB target proteins are found in the supplementary files.
Discussion
Figure 4. Luciferase reporter assays confirm that miRs-34a and 23a target the
3′ UTR of TSC1. A luciferase reporter construct with the full-length 3′ UTR of
TSC1 was co-transfected into COS-7 cells with pre-miR expression vector or
control vector. Co-transfection with pre-miR-34a resulted in a 71% reduction in
reporter expression, and miR-23a resulted in a 40% reduction.
Quantitative LC-MS/MS Proteomic Profiling
Demonstrates Activation of p53 and Inflammatory
Signaling in Epileptogenic Tubers
To investigate the potential role of p53 and proinflammatory
cytokine signaling in dysregulation of microRNAs in tubers, we
characterized expression profiles for proteins that are direct targets of p53 and NF-κB. While spectra for p53 and NF-κB themselves were not available in the proteomics experiment, their
abundance alone would not be an accurate determinant of activity since both proteins are also regulated through phosphorylation. Therefore, our analysis of direct transcriptional targets
provides a way to view pathway activity.
While microRNAs are enriched in the brain and implicated in
many diseases, the role of microRNAs in brain from TSC patients
has previously not been investigated. We report aberrant microRNA expression in tubers resected from TSC1 and TSC2 patients
to treat medically intractable seizures. Comparing expression
profiles in subject-matched pairs of tubers and adjacent nontuber tissue, we identified 5 differentially expressed microRNAs, 4
of which were overexpressed. Our integration of a previously
published microRNA expression dataset from PM brain tissue
provided control measurements. We found that expression levels
of the 4 microRNAs in nontuber cortical tissue were near normal,
and expression in epileptogenic tubers was higher than in the PM
controls. The analysis also revealed a pronounced age-dependent
expression pattern for miR-34a that highlights the need for ageor patient-matched controls.
While miR-34a has been reported as a tumor suppressor with
proapoptotic activity, a growing body of literature has shown
miR-34a to be involved in neurodevelopment and cortical neurogenesis (Aranha et al. 2010, 2011; Fineberg et al. 2012). Overexpression of miR-34a has been shown to target synaptic genes
and reduce the number of inhibitory synapses (Agostini et al.
2011). A recent study of neural stem cell differentiation demonstrated that overexpression of miR-34a resulted in significantly
impaired neuronal differentiation and synapse function (Morgado et al. 2014). The authors stated that downregulation of miR34a seems to be “crucial for neurogenesis progression.” That conclusion is consistent with low levels of miR-34a during the postnatal period, as we observed in the PM dataset. Overexpression of
Aberrant MicroRNA Expression in Cortical Tubers
Figure 5. Proteins regulated by p53 and NF-κB are overexpressed in tubers
compared with nontuber tissue. Empirical cumulative distributions of log2(T/N)
are shown for three groups of proteins, where T and N are average protein
expression in tuber and nontuber tissue, respectively. The distribution for the
overall proteome (5748 proteins) is shown with the red, solid line, the
distribution for proteins regulated by p53 (28) is shown by the dashed gray line,
and the distribution for proteins regulated by NF-κB (24) is shown by the dashed
blue line. The y-axis provides the probability that the specified value, or less, of
log2(T/N) will be observed. The proteome distribution is centered on zero
reflecting equal probability of induced or repressed proteins. The p53 and NF-κB
target sets are shifted to the right reflecting protein overexpression in tubers.
miR-34a during this critical period, as well as in utero, would be
expected to disrupt normal neurodevelopment.
Several groups have reported increased levels of miR-34a in
animal models of epilepsy (Hu et al. 2011, 2012; Sano et al.
2012; Risbud and Porter 2013). MiRs 23a and 381 have also been reported as differentially expressed in epilepsy, and we found these
microRNAs differentially expressed in epileptogenic tubers as
well (Song et al. 2011; Hu et al. 2012; Risbud and Porter 2013). Dysregulation of miR-34a has been associated with a range of neurological disorders. An analysis of genomic copy-number variants
(CNVs) associated with autism identified miR-34a as one of 10
“hub” microRNAs (microRNAs targeting multiple autism risk
genes) resident in the autism CNV regions (Vaishnavi et al. 2013).
MiR-34a has also been identified as a potential blood-based biomarker of schizophrenia (Lai et al. 2011). In a case/control comparison, miR-34a was found to be 2.5-fold higher in mononuclear
leukocytes of schizophrenia patients. This observation is consistent with a PM study that found miR-34a overexpression in the
prefrontal cortex of patients with schizophrenia (Kim et al. 2010).
Overexpression of miR-34a was also reported in the cerebral cortex
of a mouse model of Alzheimer’s disease (Wang et al. 2009).
The endpoint of microRNA activity is best observed at the protein level since many microRNA/transcript interactions impede
translation and may not alter transcript levels (Selbach et al.
2008; Li et al. 2012; Liu et al. 2013). We used quantitative LC-MS/
MS proteomic profiling to investigate alterations in protein abundance for targets of the 4 overexpressed microRNAs. We demonstrated significant repression in tubers for proteins of transcripts
with a high-confidence combinatorial target score. The set of repressed target transcripts was highly enriched for genes involved
in synaptic signaling. Additionally, the proteins for a number of
genes known to confer risk of epilepsy were among the combinatorial targets repressed in tubers. Notably, among the repressed
target proteins associated with epilepsy risk and synaptic signal
transmission was hamartin, the product of the TSC1 gene. We
used a luciferase reporter assay to demonstrate targeting of the
Dombkowski et al.
| 1067
TSC1 3′ UTR by miRs-34a and 23a. This finding is consistent
with a recent publication, and our own computational analysis,
that predict targeting of TSC1 by miR-23a (Romaker et al. 2014).
We also found decreased hamartin in tubers from the TSC1 patient, as well as those with TSC2 mutations.
Investigations from a number of laboratories suggest the TSC
genes may be post-transcriptionally regulated. A number of studies have reported repressed hamartin and/or tuberin levels in
cortical tubers, even where loss of heterozygosity (LOH) or other
second-hit mutations of the genes could not be identified (Kerfoot et al. 1996; Mizuguchi et al. 1996, 1997, 2000; Mizuguchi
and Takashima 2001; Vinaitheerthan et al. 2004; Boer et al.
2008). While reduced expression of one of the proteins could be
expected in the presence of inactivating mutations in the corresponding gene (e.g., truncations), many of the reports noted simultaneous repression of hamartin and tuberin with no clear
causative explanation. It has been speculated that mutation of
one of the TSC genes may reduce expression of the other gene
product, but no mechanism has been uncovered (Jozwiak et al.
2004). Niida et al. (2001) noted that one potential explanation is
that “expression of the wild-type protein being ‘turned off’ or reduced as a result of epigenetic events,” yet they found no evidence of methylation-based gene silencing. Collectively, these
observations describe hallmarks of microRNA activity. We have
found aberrant microRNA expression in cortical tubers and
their targeting of TSC1. Subsequent mechanistic studies are warranted to determine if the aberrant microRNAs represent a type
of “second hit,” contributing to tuber formation in the absence
of LOH (Crino et al. 2010) or if their dysregulation is an outcome
of tuber pathology.
In addition to TSC1, our integrated microRNA/proteomic profiling revealed a number of target transcripts which are known to
cause epilepsy when deficient or deleted (Table 3). Knockout of
SYN2 was shown to cause seizures in a mouse model (Rosahl
et al. 1995). Mice with inactivated GABRB3 have a seizure phenotype (Homanics et al. 1997). Mutations in DCX result in type I lissencephaly, with clinical features of disorganized cortical
layering, epilepsy, and mental retardation (Gleeson et al. 1998).
Mutations in NF1 lead to neurofibromatosis type I, a neurocutaneous syndrome characterized by benign brain lesions, cognitive
disorders, and seizures in ∼10% of patients (Ostendorf et al. 2013).
Each of these genes is a high-confidence combinatorial target of
the 4 overexpressed microRNAs, and their protein levels were significantly decreased in tubers.
A model for the mechanism of microRNA dysregulation and
their role in TSC tuber pathology emerges from our findings
and the current literature (Fig. 6). Considerable evidence in the literature has shown that: 1) loss of TSC1 or TSC2 results in activation of p53; and 2) activated p53 induces miRs-34a and 23a. Using
cell line knockouts of TSC1 and TSC2, a study confirmed that “loss
of TSC1 or TSC2 results in a dramatic elevation of p53 protein” due
to increased translation stimulated by mTOR activation (Lee et al.
2007). The same study reported elevated p53 levels in human angiomyolipomas due to TSC mutations. The authors presented an
intriguing hypothesis that the proapoptotic effect of p53 might be
the reason why TSC lesions are generally benign. A recent study
found that, in addition to increased synthesis of p53, mTOR activates p53 through phosphorylation of Ser46 (Krzesniak et al.
2014). Evidence of elevated p53 activity in TSC tubers is evident
in our proteomics dataset. We analyzed expression levels for
the proteins produced from a set of known p53-activated genes.
The overall set of proteins was significantly increased in tubers
compared with nontuber tissue, consistent with p53 activation.
A number of reports have demonstrated the direct
1068
| Cerebral Cortex, 2016, Vol. 26, No. 3
Figure 6. Mechanistic model depicting roles of microRNAs and signaling pathways in TSC tuber pathogenesis and epileptogenesis. Previous reports have shown that
mammalian/mechanistic target of rapamycin activation due to TSC mutations results in induction of miRs-23a and 34a mediated through p53. We have shown that these
microRNAs are overexpressed in epileptogenic tubers, along with miR-532-5p, and they target transcripts involved in synaptic signaling, neuron development, and a
number of genes known to confer risk of epilepsy. The repression of target proteins may contribute to tuber pathogenesis as denoted by MRI showing the locations of
multiple tubers. Events leading to inflammation in a subset of tubers may further exacerbate overexpression of the microRNAs mediated by proinflammatory cytokines.
The combined effects of these mechanisms may lead to epileptogenesis in a subset of tubers, as highlighted by the MRI and AMT-PET imaging shown in the figure. The
red arrow indicates an epileptogenic tuber with high AMT uptake, a marker of inflammation. The remaining arrows note the locations of nonepileptogenic tubers.
transactivation of miR-34a by p53, and this regulatory axis has
gained substantial interest (Chang et al. 2007; Raver-Shapira
et al. 2007; Rokavec et al. 2014). Induction of both miR-34a and
p53 due to loss of TSC1 was reported in a study detailing the
role of TSC1 in the survival of mast cells (Shin et al. 2012). MiR34b* has also been shown to be activated by p53 (He et al. 2007),
and a comparison of p53 null and wild-type colon cancer cell
lines identified miR-23a as a probable target of p53 (Chang et al.
2007). Recently, 2 studies have confirmed that miR-23a is induced
by p53 in hepatocellular carcinoma (Wang et al. 2013, 2014). Collectively, these reports and our results indicate that p53 activation is a likely mechanism involved in aberrant expression of
microRNAs with loss of TSC1 or TSC2 in cortical tubers, and
downstream microRNA targeting may contribute to tuber
Aberrant MicroRNA Expression in Cortical Tubers
pathogenesis through repression of genes involved in neuronal
development and synaptic signaling, as well as epilepsy risk
genes.
Inflammation and proinflammatory cytokines may also be
mechanistically involved in microRNA regulation in epileptogenic tubers. Focal epilepsy produces an inflammatory response
(Aronica and Crino 2011), and the role of microRNAs in neuroinflammation has recently drawn much attention (Thounaojam
et al. 2013). Numerous microRNAs have been reported as dysregulated in a variety of neurodegenerative disorders associated
with inflammation, including Alzheimer’s disease, Huntington’s
disease, multiple sclerosis, and Parkinson disease. Similar to our
finding in epileptogenic TSC tubers, expression levels of microRNAs 34a and 23a are both increased in active lesions of multiple
sclerosis patients (Junker et al. 2009). The study also reported that
the 2 miRs are induced in response to proinflammatory cytokines
interferon gamma, interleukin 1, and tumor necrosis factor (TNFα)
in cultured astrocytes. Among these cytokines, TNFα produced the
greatest increase in expression. Proinflammatory cytokine signaling along with elevated TNFα and NF-κB expression has been reported in TSC tubers (Maldonado et al. 2003). Additionally, a
microarray gene expression analysis of surgically resected tubers
found induction of many genes involved in inflammation (Boer
et al. 2010). Consistent with these reports, our proteomics data
provides evidence of proinflammatory signaling in epileptogenic
tubers. We analyzed the abundance of a group of proteins produced from genes transcriptionally regulated by NF-κB, a transcription factor that is a key mediator of the inflammatory
response (Tak and Firestein 2001). The overall set of proteins was
significantly increased in tubers compared with nontuber tissue,
showing considerable NF-κB activation. Notable among this set
is ICAM1 (Supplementary Table 2), previously identified as a marker of inflammation in TSC tubers (Maldonado et al. 2003). ICAM1
protein was 3.7-fold higher in tubers than in nontuber tissue.
The overall set of NF-κB target genes were induced nearly 2-fold,
on average. Our results, along with evidence from earlier reports,
suggest that proinflammatory cytokines in tubers may contribute
to the increased expression of the aberrant microRNAs.
In summary, epileptogenic tubers from patients with either
TSC1 or TCS2 mutations show coordinate upregulation of several
microRNAs. Quantitative LC-MS/MS proteomic profiling demonstrated significant repression of predicted combinatorial targets
of the overexpressed microRNAs. The repressed targets are enriched for transcripts involved in synaptic signaling and include
a number of genes known to confer risk of epilepsy, including
TSC1. Luciferase reporter assays demonstrated that miRs-23a
and 34a target the 3′ UTR of TSC1. Based on these results and
previous reports, we propose a model of the role of the aberrant
microRNAs in tuber pathology. The evidence indicate that
the microRNAs are induced downstream of mTOR activation,
mediated through p53. An open question is whether similar microRNA expression patterns are found in nonepileptogenic tubers. It is a challenging question to address since such tissue is
infrequently resected during epilepsy surgery and would require
PM analysis. However, given the changes observed in pathology
of tubers, it seems likely that microRNAs are involved in
alterations of synaptic density and loss of neuronal elements
compared with normal cortex. Additionally, inflammatory cytokines present in epileptogenic tubers likely contribute to the elevated microRNA expression and may be more highly expressed in
epileptogenic tubers. Aberrant microRNA expression due to TSC1
or TSC2 mutations and proinflammatory events may contribute
to tuber pathogenesis and recurrent seizures, and additional
mechanistic studies are warranted. Why the observed changes
Dombkowski et al.
| 1069
in microRNA expression occur in epileptogenic tubers and not
in adjacent histologically normal cortex in the presence of haploinsufficiency for TSC1 or TSC2 remains to be explained.
Supplementary Material
Supplementary material can be found at: http://www.cercor.
oxfordjournals.org/.
Funding
This work was supported by the U.S. National Institutes of Health
[1R01NS079429-01A1 (A.D.), 5R01NS064989-03 (H.C.), S10OD
010700 (P.S.), Center grants P30ES06639 and P30CA022453], and
the Epilepsy Foundation (grant #160313 to C.B.).
Notes
We thank S. Sundaram for helpful advice during the course of
this research and E. Asano for assistance with EEG designation
of epileptogenic tubers. Conflict of Interest: None declared .
References
Aebersold R, Burlingame AL, Bradshaw RA. 2013. Western blots
versus selected reaction monitoring assays: time to turn the
tables? Mol Cell Proteomics. 12:2381–2382.
Agostini M, Tucci P, Steinert JR, Shalom-Feuerstein R, Rouleau M,
Aberdam D, Forsythe ID, Young KW, Ventura A,
Concepcion CP, et al. 2011. microRNA-34a regulates neurite
outgrowth, spinal morphology, and function. Proc Natl Acad
Sci USA. 108:21099–21104.
Aranha MM, Santos DM, Sola S, Steer CJ, Rodrigues CM. 2011. miR34a regulates mouse neural stem cell differentiation. PLoS
One. 6:e21396.
Aranha MM, Santos DM, Xavier JM, Low WC, Steer CJ, Sola S,
Rodrigues CM. 2010. Apoptosis-associated microRNAs are
modulated in mouse, rat and human neural differentiation.
BMC Genomics. 11:514.
Aronica E, Crino PB. 2011. Inflammation in epilepsy: clinical observations. Epilepsia. 52(Suppl 3):26–32.
Boer K, Crino PB, Gorter JA, Nellist M, Jansen FE, Spliet WG, van
Rijen PC, Wittink FR, Breit TM, Troost D, et al. 2010. Gene expression analysis of tuberous sclerosis complex cortical tubers reveals increased expression of adhesion and
inflammatory factors. Brain Pathol. 20:704–719.
Boer K, Troost D, Jansen F, Nellist M, van den Ouweland AM,
Geurts JJ, Spliet WG, Crino P, Aronica E. 2008. Clinicopathological and immunohistochemical findings in an autopsy case
of tuberous sclerosis complex. Neuropathology. 28:577–590.
Boon RA, Iekushi K, Lechner S, Seeger T, Fischer A, Heydt S,
Kaluza D, Treguer K, Carmona G, Bonauer A, et al. 2013. MicroRNA-34a regulates cardiac ageing and function. Nature. 495:
107–110.
Chang TC, Wentzel EA, Kent OA, Ramachandran K, Mullendore M,
Lee KH, Feldmann G, Yamakuchi M, Ferlito M, Lowenstein CJ,
et al. 2007. Transactivation of miR-34a by p53 broadly influences gene expression and promotes apoptosis. Mol Cell.
26:745–752.
Chugani HT, Luat AF, Kumar A, Govindan R, Pawlik K, Asano E.
2013. alpha-[11C]-Methyl-L-tryptophan—PET in 191 patients
with tuberous sclerosis complex. Neurology. 81:674–680.
1070
| Cerebral Cortex, 2016, Vol. 26, No. 3
Crino PB, Aronica E, Baltuch G, Nathanson KL. 2010. Biallelic TSC
gene inactivation in tuberous sclerosis complex. Neurology.
74:1716–1723.
Dombkowski AA, Sultana Z, Craig DB, Jamil H. 2011. In silico analysis of combinatorial microRNA activity reveals target genes
and pathways associated with breast cancer metastasis.
Cancer Inform. 10:13–29.
Dweep H, Sticht C, Pandey P, Gretz N. 2011. miRWalk—database:
prediction of possible miRNA binding sites by "walking" the
genes of three genomes. J Biomed Inform. 44:839–847.
Filipowicz W, Bhattacharyya SN, Sonenberg N. 2008. Mechanisms
of post-transcriptional regulation by microRNAs: are the answers in sight? Nat Rev Genet. 9:102–114.
Fineberg SK, Datta P, Stein CS, Davidson BL. 2012. MiR-34a represses Numbl in murine neural progenitor cells and antagonizes neuronal differentiation. PLoS One. 7:e38562.
Fineberg SK, Kosik KS, Davidson BL. 2009. MicroRNAs potentiate
neural development. Neuron. 64:303–309.
Gao J, Liu QG. 2011. The role of miR-26 in tumors and normal tissues (Review). Oncol Lett. 2:1019–1023.
Gleeson JG, Allen KM, Fox JW, Lamperti ED, Berkovic S, Scheffer I,
Cooper EC, Dobyns WB, Minnerath SR, Ross ME, et al. 1998.
Doublecortin, a brain-specific gene mutated in human Xlinked lissencephaly and double cortex syndrome, encodes a
putative signaling protein. Cell. 92:63–72.
Hackstadt AJ, Hess AM. 2009. Filtering for increased power for
microarray data analysis. BMC Bioinformatics. 10:11.
He L, He X, Lim LP, de Stanchina E, Xuan Z, Liang Y, Xue W,
Zender L, Magnus J, Ridzon D, et al. 2007. A microRNA component of the p53 tumour suppressor network. Nature. 447:
1130–1134.
Homanics GE, DeLorey TM, Firestone LL, Quinlan JJ, Handforth A,
Harrison NL, Krasowski MD, Rick CE, Korpi ER, Makela R, et al.
1997. Mice devoid of gamma-aminobutyrate type A receptor
beta3 subunit have epilepsy, cleft palate, and hypersensitive
behavior. Proc Natl Acad Sci USA. 94:4143–4148.
Hu K, Xie YY, Zhang C, Ouyang DS, Long HY, Sun DN, Long LL,
Feng L, Li Y, Xiao B. 2012. MicroRNA expression profile of the
hippocampus in a rat model of temporal lobe epilepsy and
miR-34a-targeted neuroprotection against hippocampal
neurone cell apoptosis post-status epilepticus. BMC Neurosci.
13:115.
Hu K, Zhang C, Long L, Long X, Feng L, Li Y, Xiao B. 2011. Expression profile of microRNAs in rat hippocampus following lithium-pilocarpine-induced status epilepticus. Neurosci Lett.
488:252–257.
Huang da W, Sherman BT, Lempicki RA. 2009. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 4:44–57.
Jimenez-Mateos EM, Engel T, Merino-Serrais P, McKiernan RC,
Tanaka K, Mouri G, Sano T, O’Tuathaigh C, Waddington JL,
Prenter S, et al. 2012. Silencing microRNA-134 produces neuroprotective and prolonged seizure-suppressive effects. Nat
Med. 18:1087–1094.
Jozwiak S, Kwiatkowski D, Kotulska K, Larysz-Brysz M, LewinKowalik J, Grajkowska W, Roszkowski M. 2004. Tuberin and
hamartin expression is reduced in the majority of subependymal giant cell astrocytomas in tuberous sclerosis complex
consistent with a two-hit model of pathogenesis. J Child
Neurol. 19:102–106.
Junker A, Krumbholz M, Eisele S, Mohan H, Augstein F, Bittner R,
Lassmann H, Wekerle H, Hohlfeld R, Meinl E. 2009. MicroRNA
profiling of multiple sclerosis lesions identifies modulators of
the regulatory protein CD47. Brain. 132:3342–3352.
Kerfoot C, Wienecke R, Menchine M, Emelin J, Maize JC Jr,
Welsh CT, Norman MG, DeClue JE, Vinters HV. 1996. Localization of tuberous sclerosis 2 mRNA and its protein product tuberin in normal human brain and in cerebral lesions of
patients with tuberous sclerosis. Brain Pathol. 6:367–375.
Kim AH, Reimers M, Maher B, Williamson V, McMichael O,
McClay JL, van den Oord EJ, Riley BP, Kendler KS,
Vladimirov VI. 2010. MicroRNA expression profiling in the prefrontal cortex of individuals affected with schizophrenia and
bipolar disorders. Schizophrenia Res. 124:183–191.
Krichevsky AM, King KS, Donahue CP, Khrapko K, Kosik KS. 2003.
A microRNA array reveals extensive regulation of microRNAs
during brain development. RNA. 9:1274–1281.
Krzesniak M, Zajkowicz A, Matuszczyk I, Rusin M. 2014. Rapamycin prevents strong phosphorylation of p53 on serine 46
and attenuates activation of the p53 pathway in A549 lung
cancer cells exposed to actinomycin D. Mech Ageing Dev.
139:11–21.
Kuhn DE, Martin MM, Feldman DS, Terry AV Jr, Nuovo GJ, Elton TS.
2008. Experimental validation of miRNA targets. Methods.
44:47–54.
Lai CY, Yu SL, Hsieh MH, Chen CH, Chen HY, Wen CC, Huang YH,
Hsiao PC, Hsiao CK, Liu CM, et al. 2011. MicroRNA expression
aberration as potential peripheral blood biomarkers for
schizophrenia. PLoS One. 6:e21635.
Lee CH, Inoki K, Karbowniczek M, Petroulakis E, Sonenberg N,
Henske EP, Guan KL. 2007. Constitutive mTOR activation in
TSC mutants sensitizes cells to energy starvation and genomic damage via p53. EMBO J. 26:4812–4823.
Li C, Xiong Q, Zhang J, Ge F, Bi LJ. 2012. Quantitative proteomic
strategies for the identification of microRNA targets. Expert
Rev Proteomics. 9:549–559.
Li X, Khanna A, Li N, Wang E. 2011. Circulatory miR34a as an
RNA based, noninvasive biomarker for brain aging. Aging.
3:985–1002.
Liu Q, Halvey PJ, Shyr Y, Slebos RJ, Liebler DC, Zhang B. 2013. Integrative omics analysis reveals the importance and scope of
translational repression in microRNA-mediated regulation.
Mol Cell Proteomics. 12:1900–1911.
Maldonado M, Baybis M, Newman D, Kolson DL, Chen W,
McKhann G 2nd, Gutmann DH, Crino PB. 2003. Expression of
ICAM-1, TNF-alpha, NF kappa B, and MAP kinase in tubers
of the tuberous sclerosis complex. Neurobiol Dis. 14:279–290.
McAlister GC, Huttlin EL, Haas W, Ting L, Jedrychowski MP,
Rogers JC, Kuhn K, Pike I, Grothe RA, Blethrow JD, et al. 2012.
Increasing the multiplexing capacity of TMTs using reporter
ion isotopologues with isobaric masses. Anal Chem. 84:
7469–7478.
Michalski A, Cox J, Mann M. 2011. More than 100,000 detectable
peptide species elute in single shotgun proteomics runs but
the majority is inaccessible to data-dependent LC-MS/MS.
J Proteome Res. 10:1785–1793.
Mizuguchi M, Ikeda K, Takashima S. 2000. Simultaneous loss of
hamartin and tuberin from the cerebrum, kidney and heart
with tuberous sclerosis. Acta Neuropathol. 99:503–510.
Mizuguchi M, Kato M, Yamanouchi H, Ikeda K, Takashima S. 1996.
Loss of tuberin from cerebral tissues with tuberous sclerosis
and astrocytoma. Ann Neurol. 40:941–944.
Mizuguchi M, Kato M, Yamanouchi H, Ikeda K, Takashima S. 1997.
Tuberin immunohistochemistry in brain, kidneys and heart
with or without tuberous sclerosis. Acta Neuropathol.
94:525–531.
Mizuguchi M, Takashima S. 2001. Neuropathology of tuberous
sclerosis. Brain Dev. 23:508–515.
Aberrant MicroRNA Expression in Cortical Tubers
Morgado AL, Xavier JM, Dionisio PA, Ribeiro MF, Dias RB,
Sebastiao AM, Sola S, Rodrigues CM. 2014. MicroRNA-34a
modulates neural stem cell differentiation by regulating
expression of synaptic and autophagic proteins. Mol
Neurobiol.
Nagaraj N, Wisniewski JR, Geiger T, Cox J, Kircher M, Kelso J,
Paabo S, Mann M. 2011. Deep proteome and transcriptome
mapping of a human cancer cell line. Mol Syst Biol. 7:548.
Nielsen JA, Lau P, Maric D, Barker JL, Hudson LD. 2009. Integrating
microRNA and mRNA expression profiles of neuronal progenitors to identify regulatory networks underlying the
onset of cortical neurogenesis. BMC Neurosci. 10:98.
Niida Y, Stemmer-Rachamimov AO, Logrip M, Tapon D, Perez R,
Kwiatkowski DJ, Sims K, MacCollin M, Louis DN, Ramesh V.
2001. Survey of somatic mutations in tuberous sclerosis complex (TSC) hamartomas suggests different genetic mechanisms for pathogenesis of TSC lesions. Am J Hum Genet.
69:493–503.
Ostendorf AP, Gutmann DH, Weisenberg JL. 2013. Epilepsy in individuals with neurofibromatosis type 1. Epilepsia. 54:1810–1814.
Peng C, Li N, Ng YK, Zhang J, Meier F, Theis FJ, Merkenschlager M,
Chen W, Wurst W, Prakash N. 2012. A unilateral negative feedback loop between miR-200 microRNAs and Sox2/E2F3 controls neural progenitor cell-cycle exit and differentiation. J
Neurosci. 32:13292–13308.
Rao Y, Lee Y, Jarjoura D, Ruppert AS, Liu CG, Hsu JC, Hagan JP.
2008. A comparison of normalization techniques for microRNA microarray data. Stat Appl Genet Mol Biol. 7:Article22.
Raver-Shapira N, Marciano E, Meiri E, Spector Y, Rosenfeld N,
Moskovits N, Bentwich Z, Oren M. 2007. Transcriptional activation of miR-34a contributes to p53-mediated apoptosis.
Mol Cell. 26:731–743.
Rehmsmeier M, Steffen P, Hochsmann M, Giegerich R. 2004. Fast
and effective prediction of microRNA/target duplexes. RNA.
10:1507–1517.
Reyes-Herrera PH, Ficarra E. 2012. One decade of development
and evolution of microRNA target prediction algorithms.
Genomics Proteomics Bioinformatics. 10:254–263.
Riley T, Sontag E, Chen P, Levine A. 2008. Transcriptional control
of human p53-regulated genes. Nat Rev Mol Cell Biol.
9:402–412.
Risbud RM, Porter BE. 2013. Changes in microRNA expression in
the whole hippocampus and hippocampal synaptoneurosome fraction following pilocarpine induced status epilepticus. PLoS One. 8:e53464.
Roach ES, DiMario FJ, Kandt RS, Northrup H. 1999. Tuberous Sclerosis Consensus Conference: recommendations for diagnostic
evaluation. National Tuberous Sclerosis Association. J Child
Neurol. 14:401–407.
Roach ES, Sparagana SP. 2004. Diagnosis of tuberous sclerosis
complex. J Child Neurol. 19:643–649.
Rokavec M, Li H, Jiang L, Hermeking H. 2014. The p53/miR-34 axis
in development and disease. J Mol Cell Biol. 6:214–230.
Romaker D, Kumar V, Cerqueira DM, Cox RM, Wessely O. 2014. MicroRNAs are critical regulators of tuberous sclerosis complex
and mTORC1 activity in the size control of the Xenopus kidney. Proc Natl Acad Sci USA. 111:6335–6340.
Rosahl TW, Spillane D, Missler M, Herz J, Selig DK, Wolff JR,
Hammer RE, Malenka RC, Sudhof TC. 1995. Essential functions of synapsins I and II in synaptic vesicle regulation.
Nature. 375:488–493.
Sano T, Reynolds JP, Jimenez-Mateos EM, Matsushima S,
Taki W, Henshall DC. 2012. MicroRNA-34a upregulation
Dombkowski et al.
| 1071
during seizure-induced neuronal death. Cell Death Dis. 3:
e287.
Selbach M, Schwanhausser B, Thierfelder N, Fang Z, Khanin R,
Rajewsky N. 2008. Widespread changes in protein synthesis
induced by microRNAs. Nature. 455:58–63.
Shin J, Pan H, Zhong XP. 2012. Regulation of mast cell survival
and function by tuberous sclerosis complex 1. Blood. 119:
3306–3314.
Somel M, Guo S, Fu N, Yan Z, Hu HY, Xu Y, Yuan Y, Ning Z, Hu Y,
Menzel C, et al. 2010. MicroRNA, mRNA, and protein expression link development and aging in human and macaque
brain. Genome Res. 20:1207–1218.
Somel M, Liu X, Tang L, Yan Z, Hu H, Guo S, Jiang X, Zhang X, Xu G,
Xie G, et al. 2011. MicroRNA-driven developmental remodeling in the brain distinguishes humans from other primates.
PLoS Biol. 9:e1001214.
Song YJ, Tian XB, Zhang S, Zhang YX, Li X, Li D, Cheng Y,
Zhang JN, Kang CS, Zhao W. 2011. Temporal lobe epilepsy induces differential expression of hippocampal miRNAs including let-7e and miR-23a/b. Brain Res. 1387:134–140.
Tak PP, Firestein GS. 2001. NF-kappaB: a key role in inflammatory
diseases. J Clin Invest. 107:7–11.
Thiele EA. 2004. Managing epilepsy in tuberous sclerosis complex. J Child Neurol. 19:680–686.
Thounaojam MC, Kaushik DK, Basu A. 2013. MicroRNAs in
the brain: it’s regulatory role in neuroinflammation. Mol
Neurobiol. 47:1034–1044.
Vaishnavi V, Manikandan M, Tiwary BK, Munirajan AK. 2013.
Insights on the functional impact of microRNAs present in
autism-associated copy number variants. PLoS One. 8:
e56781.
Vinaitheerthan M, Wei J, Mizuguchi M, Greco A, Barness EG. 2004.
Tuberous sclerosis: immunohistochemistry expression of tuberin and hamartin in a 31-week gestational fetus. Fetal
Pediatr Pathol. 23:241–249.
Viprey VF, Corrias MV, Burchill SA. 2012. Identification of reference microRNAs and suitability of archived hemopoietic samples for robust microRNA expression profiling. Anal Biochem.
421:566–572.
Wang N, Zhu M, Tsao SW, Man K, Zhang Z, Feng Y. 2013. MiR-23amediated inhibition of topoisomerase 1 expression potentiates cell response to etoposide in human hepatocellular
carcinoma. Mol Cancer. 12:119.
Wang N, Zhu M, Wang X, Tan HY, Tsao SW, Feng Y. 2014.
Berberine-induced tumor suppressor p53 up-regulation gets
involved in the regulatory network of MIR-23a in hepatocellular carcinoma. Biochim Biophys Acta. 1839:849–857.
Wang X, Liu P, Zhu H, Xu Y, Ma C, Dai X, Huang L, Liu Y, Zhang L,
Qin C. 2009. miR-34a, a microRNA up-regulated in a double
transgenic mouse model of Alzheimer’s disease, inhibits
bcl2 translation. Brain Res Bull. 80:268–273.
Weiner HL, Ferraris N, LaJoie J, Miles D, Devinsky O. 2004. Epilepsy
surgery for children with tuberous sclerosis complex. J Child
Neurol. 19:687–689.
You G, Yan W, Zhang W, Wang Y, Bao Z, Li S, Li G, Song Y, Kang C,
Jiang T. 2012. Significance of miR-196b in tumor-related epilepsy of patients with gliomas. PLoS One. 7:e46218.
Yue D, Liu H, Huang Y. 2009. Survey of computational algorithms
for microRNA target prediction. Curr Genomics. 10:478–492.
Zhao X, Wu J, Zheng M, Gao F, Ju G. 2012. Specification and maintenance of oligodendrocyte precursor cells from neural progenitor cells: involvement of microRNA-7a. Mol Biol Cell.
23:2867–2878.