tdp-43 high throughput screening analyses in neurodegeneration: advantages and pitfalls

10
TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls Emanuele Buratti a , Maurizio Romano b , Francisco E. Baralle a, a International Centre for Genetic Engineering and Biotechnology (ICGEB) 34012 Trieste, Italy b Department of Life Sciences, University of Trieste, Via A. Valerio 28, 34127, Trieste, Italy abstract article info Article history: Received 5 December 2012 Revised 22 February 2013 Accepted 1 March 2013 Available online 14 March 2013 Dysfunctions in RNA processing and in particular the aberrant regulation of RNA binding proteins (RBPs) have recently been shown to play a fundamental role in the pathogenesis of neurodegenerative diseases. Understanding the pathogenic mechanisms involved will require the elucidation of the role(s) played by these RBPs in the general cell metabolism and neuronal survival in particular. In the past, the preferred approach has been to determine rst of all the functional properties of the factor(s) of interest and then use this knowledge to determine targets in biologically relevant events. More recently, novel experimental approaches such as microarrays, RNA-seq and CLIP-seq have also become very popular to study RBPs. The advantage of these approaches, collectively known as high throughput screening (HTS), is their ability to determine gene expression changes or RNA/protein targets at a global cellular level. In theory, HTS strategies should be ideal for uncovering novel functional roles/targets of any RBP inside the cell. In practice, however, there are still difculties in getting a coherent picture from all the huge amount of data they generate, frequently not validated experimentally and thus of unknown value. They may even act unfavorably towards a specic increase of knowledge of RBP functions, as the incomplete results are taken as solid data. In this work we will illustrate as an example the use of the HTS methodologies to characterize the interactions of a specic RBP: TDP-43. The multiple functions of this protein in RNA processing and its involvement in the pathogenesis of several forms of amyotrophic lateral sclerosis, frontotemporal lobar degeneration and other neurodegenerative diseases make it an excellent substrate for our analysis of the various advantages and limitations of different HTS experimental approaches. © 2013 Elsevier Inc. All rights reserved. Introduction In recent years, a huge amount of studies have highlighted the key role played by RNA metabolism in normal brain development and functioning, especially at the pre-mRNA splicing and microRNA ex- pression levels (Grabowski, 2011; Mattick, 2011; Norris and Calarco, 2012). Most importantly, the discovery that alterations at the level of several RNA binding proteins (RBPs) play a major and direct role in neurodegenerative processes has sparked extensive research in this eld that is still far from over (Hanson et al., 2012; Ugras and Shorter, 2012). Altogether, these ndings have shown that one of the major causes of neurodegeneration observed in diseases such as amyotrophic lateral sclerosis (ALS), Frontotemporal Lobar Degenera- tion (FTLD), Parkinson disease (PD), and Alzheimer disease (AD) may be ascribed to defects in RNA metabolism (Gagliardi et al., 2012; Mills and Janitz, 2012; Renoux and Todd, 2012; Youmans and Wolozin, 2012). As a result, considerable effort has been spent to try and identify the functions/target sequences of the major RBPs involved in these alterations (i.e. TDP-43 and FUS/TLS). This is not an easy task, as these proteins all belong to the class of heterogeneous ribonucleopro- teins (hnRNPs) that have been generally shown to affect all levels of an mRNA life cycle from transcription, processing, transport, to even- tually translation (Budini et al., 2011; Dormann and Haass, 2011; King et al., 2012; Lee et al., 2012). However, only a comprehensive character- ization of their biological properties will allow a better understanding of disease mechanisms. As several reviews have been recently published on this subject, the aim of this work will not concern that directly and the reader is further referred to them (Buratti and Baralle, 2012; Halliday et al., 2012; King et al., 2012; Rademakers et al., 2012; Sephton et al., 2012; Xu, 2012). Rather, the aim of this review will be to list the various methodologies employed to study the biological functions of these RBPs and critically examine their potential advantages and pitfalls. In particular, we will highlight some of the major limitations that have recently come to the fore when comparing similar screening procedures applied in different laboratories in various cells/model systems. Particular attention will be devoted to TDP-43, because for this protein there is a relatively high wealth of information regarding its major biological prop- erties (Buratti and Baralle, 2010) and because it has also been the subject of several high throughput screening (HTS) studies. At present, therefore, there is already a considerable amount of data available on this subject. Molecular and Cellular Neuroscience 56 (2013) 465474 Corresponding author at: Padriciano 99, 34012 Trieste, Italy. Fax: +39 040 3757361. E-mail address: [email protected] (F.E. Baralle). 1044-7431/$ see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.mcn.2013.03.001 Contents lists available at ScienceDirect Molecular and Cellular Neuroscience journal homepage: www.elsevier.com/locate/ymcne

Upload: francisco-e

Post on 15-Dec-2016

221 views

Category:

Documents


5 download

TRANSCRIPT

Page 1: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

Molecular and Cellular Neuroscience 56 (2013) 465–474

Contents lists available at ScienceDirect

Molecular and Cellular Neuroscience

j ourna l homepage: www.e lsev ie r .com/ locate /ymcne

TDP-43 high throughput screening analyses in neurodegeneration:Advantages and pitfalls

Emanuele Buratti a, Maurizio Romano b, Francisco E. Baralle a,⁎a International Centre for Genetic Engineering and Biotechnology (ICGEB) 34012 Trieste, Italyb Department of Life Sciences, University of Trieste, Via A. Valerio 28, 34127, Trieste, Italy

⁎ Corresponding author at: Padriciano 99, 34012 TriestE-mail address: [email protected] (F.E. Baralle).

1044-7431/$ – see front matter © 2013 Elsevier Inc. Allhttp://dx.doi.org/10.1016/j.mcn.2013.03.001

a b s t r a c t

a r t i c l e i n f o

Article history:Received 5 December 2012Revised 22 February 2013Accepted 1 March 2013Available online 14 March 2013

Dysfunctions in RNA processing and in particular the aberrant regulation of RNA binding proteins (RBPs)have recently been shown to play a fundamental role in the pathogenesis of neurodegenerative diseases.Understanding the pathogenic mechanisms involved will require the elucidation of the role(s) played bythese RBPs in the general cell metabolism and neuronal survival in particular. In the past, the preferredapproach has been to determine first of all the functional properties of the factor(s) of interest and then use thisknowledge to determine targets in biologically relevant events. More recently, novel experimental approachessuch as microarrays, RNA-seq and CLIP-seq have also become very popular to study RBPs. The advantage of theseapproaches, collectively known as high throughput screening (HTS), is their ability to determine gene expressionchanges or RNA/protein targets at a global cellular level. In theory, HTS strategies should be ideal for uncoveringnovel functional roles/targets of any RBP inside the cell. In practice, however, there are still difficulties in getting acoherent picture from all the huge amount of data they generate, frequently not validated experimentally andthus of unknown value. They may even act unfavorably towards a specific increase of knowledge of RBP functions,as the incomplete results are taken as solid data. In this work we will illustrate as an example the use of the HTSmethodologies to characterize the interactions of a specific RBP: TDP-43. The multiple functions of this protein inRNA processing and its involvement in the pathogenesis of several forms of amyotrophic lateral sclerosis,frontotemporal lobar degeneration and other neurodegenerative diseases make it an excellent substrate for ouranalysis of the various advantages and limitations of different HTS experimental approaches.

© 2013 Elsevier Inc. All rights reserved.

Introduction

In recent years, a huge amount of studies have highlighted the keyrole played by RNA metabolism in normal brain development andfunctioning, especially at the pre-mRNA splicing and microRNA ex-pression levels (Grabowski, 2011; Mattick, 2011; Norris and Calarco,2012). Most importantly, the discovery that alterations at the levelof several RNA binding proteins (RBPs) play a major and direct rolein neurodegenerative processes has sparked extensive research inthis field that is still far from over (Hanson et al., 2012; Ugras andShorter, 2012). Altogether, these findings have shown that one ofthe major causes of neurodegeneration observed in diseases such asamyotrophic lateral sclerosis (ALS), Frontotemporal Lobar Degenera-tion (FTLD), Parkinson disease (PD), and Alzheimer disease (AD) maybe ascribed to defects in RNA metabolism (Gagliardi et al., 2012; Millsand Janitz, 2012; Renoux and Todd, 2012; Youmans and Wolozin,2012).

As a result, considerable effort has been spent to try and identifythe functions/target sequences of the major RBPs involved in these

e, Italy. Fax: +39 040 3757361.

rights reserved.

alterations (i.e. TDP-43 and FUS/TLS). This is not an easy task, asthese proteins all belong to the class of heterogeneous ribonucleopro-teins (hnRNPs) that have been generally shown to affect all levels ofan mRNA life cycle from transcription, processing, transport, to even-tually translation (Budini et al., 2011; Dormann and Haass, 2011; Kinget al., 2012; Lee et al., 2012). However, only a comprehensive character-ization of their biological properties will allow a better understanding ofdisease mechanisms. As several reviews have been recently publishedon this subject, the aim of this work will not concern that directly andthe reader is further referred to them (Buratti and Baralle, 2012;Halliday et al., 2012; King et al., 2012; Rademakers et al., 2012; Sephtonet al., 2012; Xu, 2012). Rather, the aim of this review will be to list thevarious methodologies employed to study the biological functions ofthese RBPs and critically examine their potential advantages and pitfalls.In particular, we will highlight some of the major limitations that haverecently come to the fore when comparing similar screening proceduresapplied in different laboratories in various cells/model systems. Particularattention will be devoted to TDP-43, because for this protein there is arelatively high wealth of information regarding its major biological prop-erties (Buratti and Baralle, 2010) and because it has also been the subjectof several high throughput screening (HTS) studies. At present, therefore,there is already a considerable amount of data available on this subject.

Page 2: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

466 E. Buratti et al. / Molecular and Cellular Neuroscience 56 (2013) 465–474

Before examining this specific issue, however, it is best to briefly describethe major conceptual and technical approaches that have been used toinvestigate TDP-43 biological relevance until this moment.

Hypothesis-driven approaches in TDP-43 research

Traditionally, the search and characterization of novel regulatoryfactors in the field of RNA metabolism have been performed usingprinciples that were all basically hypothesis driven. Just to takeTDP-43 as an example, research in this area started after severalinitial reports from studies that highlighted an extremely reproduc-ible correlation between the number of UG(m)/U(n) repeats near the3′ splice site of CFTR exon 9 and its final inclusion in the mature CFTRmRNA (Cuppens et al., 1998; Larriba et al., 1998). From a pathologicalpoint of view, these observations represented one of the best genotype/phenotype correlations able to distinguish between normal individualsand patients affected by various monosymptomatic or full-blown casesof cystic fibrosis (Groman et al., 2004). Therefore, starting from this initialcorrelation, the effect on splicing recognition of this UG(m)/U(n) repeatswas first duplicated in a laboratory setting using minigene technology(Niksic et al., 1999). Subsequently, the development of this experimentalmodel coupled with classical biochemical purification techniques lead tothe identification of TDP-43 as themajor RBP functionally responsible forthis effect (Buratti et al., 2001) and adetailed dissection of its RNAbindingproperties (Buratti and Baralle, 2001). The results were then used tosearch for additional examples of other systems in which this proteincould play a role. This allowed to find several other examples in whichthis protein could act as a splicing regulator by binding to similar se-quences (Mercado et al., 2005; Passoni et al., 2012) and prompted severalstudies on its splicing properties (Ayala et al., 2005). The knowledge ofthis binding site has also allowed to pinpoint efficiently the binding siteof TDP-43 in various 3′UTR sequences of genes potentially important toneurodegeneration, such as hNFL (Volkening et al., 2009). Most impor-tantly, in all these steps the previous observation and/or results servedas the basis for constructing new sets of hypotheses that could then betested further. In this respect, TDP-43 is certainly not an exception andmost of the representative RBPs known to this date, such as the hnRNPA/Bs (Dreyfuss et al., 1993) and H/F protein family (Caputi and Zahler,2001), have been investigated using similar criteria and experimentalapproaches. It should be noted that this hypothesis-driven approach hasalso been applied to characterize TDP-43 involvement in microRNAbiogenesis. In fact, starting from the initial observations that it is presentin both the human and mouse microprocessor complexes (Fukudaet al., 2007; Gregory et al., 2004) and localized in peri-chromatin fibers(Casafont et al., 2009), two independent studies have confirmed thatthe expression of particular microRNAs sets potentially importantfor neuronal survival/development is regulated by this protein (Burattiet al., 2010; Kawahara and Mieda-Sato, 2012).

In general, the advantage of such a traditional approach is that it isusually very reproducible and, most importantly, always allows thedesign of clear sets of experiments leading either to the acceptanceor the rejection of the new hypotheses. As a result, at every stage ofthe study there is always a constant degree of certainty/confidencewith regards to the significance of the various candidates that fallunder investigation.

However, this in-depth characterization of any biological event isnecessarily a lengthy and laborious process. In recent times, therefore,the major perceived disadvantage of these classical approaches hasbeen the idea that focusing on single events or factors does not allowanalysis of “global” questions. In addition, the main drawback of notbeing able to analyzemany events in parallelmeans that there is alwaysthe possibility of missing important functions that do not fit in theoriginal hypothesis or that may not be readily apparent by straight in-ductive reasoning. This is, of course, an important drawback, as theuse of unbiased approaches to investigate well-characterized splicingfactors has led to several unforeseen discoveries in the RNA processing

field, for example, the role of U1snRNP in preventing premature cleav-age and polyadenylation in the general pre-mRNA population (Kaidaet al., 2010), or the role played by the classical splicing factor hnRNPA1 in selected microRNA biogenesis (Guil and Caceres, 2007). In bothcases, these specific functions could not have been easily predictedsimply based on our previous and rather extensive knowledge of theirbiochemical characteristics.

Non-hypothesis-driven approaches in TDP-43 research

In the past 10 years or so, therefore, several technical advances indata collection and management have allowed the appearance ofnovel experimental approaches that are non-hypothesis driven andallow obtaining information on a global scale. Basically, the biggestadvantage that all these techniques provide to the research commu-nity is the possibility to look at biological events at the level of thefull “transcriptome” or “proteome” of any protein of interest, withoutany subjective bias (outside that provided from the technical limita-tions of each technique).

As expected, these novel techniques have been found extremelyuseful when trying to acquire information on complex processes suchas changes in gene expression during different developmental stages,tissue-specific differentiation and, most importantly, during disease pro-gression. The challenge of these approaches, however, is represented bythe ability to obtain “results” out of the huge amount of “data” providedright from the very beginning. In fact, the most common output from allthese approaches consists in several hundred to several thousand genesexpression changes or RNA binding targets whose biological significanceis mostly unknown. As a result, several approaches have been attemptedto try andaddress this question: for example, the development of “systembiology” approaches, with all its pros and cons (Potthast, 2009). Theseanalyses attempt to codify from data generated by global approachesthemost effective procedures to identify the specific cell signaling ormet-abolic networks that are most affected. Then, following the identificationof these networks, it is presumed that researchers should be allowed toprogress back along the ladder of specific factors/events and eventuallydefine the biological significance of the factor under study. In somecases, this approach has yielded important results. For example, in theneurodegeneration field significant progress has been made using globaltechniques to identify cell signaling or metabolic networks involved indisease origin and progression, especially for ALS, AD and PD, as recentlyreviewed elsewhere (Cooper-Knock et al., 2012). One of themajor draw-backs of these data mining approaches is that they often maintain a highdegree of uncertainty with regards to the various candidates/hypothesesunder investigation until the validation stage.

Because of their potentialities, however, it is not surprising thatsince the discovery of its involvement in neurodegeneration (Araiet al., 2006; Neumann et al., 2006), TDP-43 has also become a majortarget of HTS studies that include microarray analysis, RNA sequenc-ing, and CLIP-seq approaches (schematically summarized in Fig. 1 andwill be described later in detail). Table 1 summarizes all HTS studiesperformed on TDP-43 and other ALS/FTLD studies discussed in thiswork with indication of the used samples and platforms.

Microarray technologies applied to TDP-43

To this date, general gene expression variation studies using mi-croarray analysis following the knock down of TDP-43 expressionusing siRNA approaches have been performed in culture cell lines suchas HeLa (Ayala et al., 2008), HEK293E (Fiesel et al., 2010, 2012), andmouse neuroblastoma Neuro2a cell line (Bose et al., 2011). In addition,a microarray study has also been recently performed in the context of atransgenic mouse model overexpressing wild type and mutant versionsof human TDP-43 (Igaz et al., 2011) and in iPSC cells derived from ALSpatients carrying disease-associated TDP-43 mutations (Egawa et al.,2012). In a more selective manner, the use of microarrays following

Page 3: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

.….… …

Sample 1 Sample 2

Fragmentation Random hexamer primed cDNA synthesis or addition of

sequencing adaptors and HTS-sequencing

exon/intron reads

junction reads

5’-linker ligationIsolation of RNA-

RBP complexProteinase K digestion

3’-linker ligation reverse transcription

PCR

3’UTR/polyA reads

Annotated genomic structure

.….….….….….….….…

products spotted on solid-support

.….….….….….….….…

Sample 1 Sample 2

Hybridization of differentially labeled

targets

Fluorescentscanning

Imageanalysis

Total RNA extraction

Fragmentation Random hexamer primed cDNA synthesis or addition of

sequencing adaptors and HTS-sequencing

exon/intron reads

junction reads

A. Microarray

B. RNA-seq

C. CLIP-seq

UV-crosslink + IP against RBP of

choice

Fragmentation and 5’-linker ligation

Isolation of RNA-RBP complex

Proteinase K digestion 3’-linker ligation

reverse transcription

PCR

HTS-sequencing and mapping to gene

3’UTR/polyA reads

Normalizationand data

processing

Oligos or PCR amplified

Fig. 1. Flowchart of Microarray, RNA-seq and Clip-seq experiments. (A) The scheme outlines the typical steps of a microarray screening carried out to determine differential mRNAexpression levels. Basically, after RNA extraction from selected cells or tissues the fluorescent samples are prepared by RNA-labeling directly or after an intermediate step ofRNA-amplification. These samples are then hybridized to the microarray of choice together with suitable and differentially labeled control samples and will compete for bindingto the probes. The differential amount of binding to each probe sequence on the microarray by either sample will determine the intensity of fluorescence at that location. Subse-quent data analysis will then allow quantification of the RNA transcripts in the sample and provide a value for this difference. An alternative possibility to this strategy that allows toestimate absolute gene expression levels is to use Affymetrix Gene Chip arrays in which only one sample is hybridized each time on the chip. (B) The scheme recapitulates the typ-ical steps of Next Generation Sequencing approaches, applied to determine the expression levels and sequence of a total RNA population. The experimental procedure is quite sim-ple in that it involves total RNA isolation, its fragmentation, and subsequent transformation in a library of cDNA fragments that can then be sequenced using high throughputsequencing technology. The reads can then be aligned on the reference transcriptome or genome (if available) and classified according to whether the fragments come from in-trons/exons, junctions read, or 3′UTR/polyA reads. (C) The scheme describes the classical workflow for a CLIP-seq experiment to determine, on a global scale, the binding positionsof any RBP of interest in the entire transcriptome. Briefly, this technique starts from the in vivo cross-linking of RBPs to the RNAs, followed by isolation through immunoprecipi-tation of the RNA–protein binding complexes, its purification, and addition of 5′ and 3′ linker sequences that will eventually allow the use of high throughput screeningmethodologies.

467E. Buratti et al. / Molecular and Cellular Neuroscience 56 (2013) 465–474

immunoprecipitation analysis has also been performed to find TDP-43mRNA targets in normal mouse brain (Narayanan et al., in press) and inthe cytoplasm of NSC-34 cells (Colombrita et al., 2012). Beside theirintrinsic worth in terms of gene expression analysis, one of the mostinteresting aspects of gathering several data sets is represented by thepossibility to compare them with each other, in order to determine ifthere are common themes/genes. Most of all, it would also be interestingto compare these results with similar microarray-based studies that havebeen performed on the brain of individuals affected by ALS or FTLD (Baciuet al., 2012; Chen-Plotkin et al., 2008; Kocerha et al., 2011; Mishra et al.,2007; Rabin et al., 2010). Before we discuss these comparisons, however,it is important to provide some background on the microarray techniqueitself.

Microarrays consist of a series of DNA sequences that are spotted atan extremely high density on a solid surface that initially consisted ofglass microscope slides but has now been mostly replaced by quartzwafers (Dalma-Weiszhausz et al., 2006). Although initially described in1995 (Schena et al., 1995), this is still a vigorously expanding field andnew generations of better microarray supports are underway, such ashigh-density bead arrays (Fan et al., 2006) or three-dimensional bead ar-rays (Dunbar, 2006). In general, these solid supports can be probed withtarget molecules (in the case of studies dealing with RNA metabolism,

usually fluorescently labeled, retrotranscribed cellular RNAs) to produceeither quantitative (i.e. gene expression) or qualitative (i.e. exon/intronusage) data (Dalma-Weiszhausz et al., 2006). The steps required in a typ-ical microarray experiment are outlined in Fig. 1A and described in thefigure legend. More recently, investigation in RNA splicing using thistechnology has beenmade easier by the availability of arrays that containsequences which span across predicted splicing junctions, thus greatlyaiding the identificationof changes in their usage. Regardingneurodegen-eration, for example, the use of this kind of microarrays has allowed toseparate splicing changes due to normal aging or to the presence of dis-ease (FTLD andAD) (Tollervey et al., 2011b) and has also allowed to iden-tify although not all validated (see subsequent discussions), almost 800splicing events that are altered in the brain of mice where TDP-43 wasknocked out using antisense oligonucleotides, with 83 of these eventscommonly regulated with FUS/TLS (Lagier-Tourenne et al., 2012).

There are however some limitations to the microarray approachthat should be considered. First of all, there is the prerequisite for se-quence information to design/synthesize the probes that need to bespotted on the solid surface. Nowadays, for many organisms, this isnot normally a problem, but it also means that researchers are some-what limited to the amount of knowledge present at any given time.For example, microarray data cannot be re-analyzed as more and more

Page 4: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

Table 1High-throughput analyses used for investigating TDP-43 gene targets and ALS/FTLD differential gene expression.

Aim of the study Main results Reference Sample Analysis Platform

Gene expression profiling of FTLD-U post-mortembrains

Identification of specific gene-linked pathways potentiallyinfluenced by neurodegenerative disease processes

(Mishra et al.,2007)

Humanbrain

Microarray Affymetrix GeneChip humangenome U133+ 2.0 arrays

Gene expression profiling of TDP-43-silencedHeLa cells

Loss of TDP-43 upregulated cyclin-dependent kinase 6(Cdk6) at the mRNA and protein level.

(Ayala et al.,2008)

HeLa Microarray Affymetrix GeneChip humangenome U133A 2.0 arrays

Gene expression profiling of FTLD-U brains with andwithout progranulin gene alterations

Identification of genes differentially expressed in FTLD-U cases compared to controls (Chen-Plotkinet al., 2008)

Humanbrain

Microarray Affymetrix GeneChip humangenome U133A 2.0 arrays

Gene expression profiling of TDP-43-silencedHEK293 cells

Loss of TDP-43 downregulated histone deacetylase 6 (HDAC6) at the mRNA and protein level. (Fiesel et al.,2010)

HEK293E Microarray Affymetrix GeneChip humangenome U133+ 2.0 arrays

Gene expression and exon splicing profiling ofsporadic ALS post-mortem lumbar spinal cords

Sporadic ALS displays compartment-specific aberrant exon splicing and alteredcell-matrix adhesion biology.

(Rabin et al.,2010)

Humanbrain

Microarray Affymetrix GeneChip human exon1.0 ST arrays

Gene expression profiling of TDP-43-silencedNeuro2A cells

Loss of TDP-43 downregulates ATG7 at the mRNA and protein level. (Bose et al.,2011)

Neuro2A Microarray Affymetrix mouse genome 430 2.0array

Gene expression profiling of transgenic miceoverexpressing hTDP-43-WT and hTDP-43-ΔNLSproteins

Dysregulation of the ALS-associated gene TDP-43 leads to neuronal death anddegeneration in mice.

(Igaz et al.,2011)

Mousebrain

Microarray Affymetrix mouse genome 430A2.0 array

Exon splicing profiling in mouse adult brains treatedwith TDP-43 antisense oligos

Splicing-sensitive junction arrays revealed alteration in 779 alternatively spliced events. (Polymenidouet al., 2011)

Mousebrain

Exon array Custom Affymetrix mouse exonjunction array

Exon splicing profiling of TDP-43 silenced HEK293and SH-SY5Y cells

Detection of splicing alteration of the SKAR/POLDIP3 gene in both neuronal andnon-neuronal cell lines

(Fiesel et al.,2012)

HEK293E/SH-SY5Y

Exon array Human exon 1.0-ST gene chip

High-resolution splice-junctionmicroarrays to evaluatesplicing changes in TDP-43 knockdown SH-SY5Y cells

Identification of 158 splicing changes in alternative cassette exons and 71 othertypes of splicing changes out of 30,154 alternative exons

(Tollerveyet al., 2011a)

SH-SY5Y Exon array Affymetrix AltSplicesplice-junction microarray

Alternative splicing profiling associated with agingand neurodegeneration in the human brain

Identification of age-related splicing changes in cognitively normal individuals andobservation that these were present also in 95% of individuals with FTLD or AD,independent of their age

(Tollerveyet al., 2011b)

Humanbrain

Microarray Affymetrix GeneChip human HJAYarray 17K

Gene expression profiling of ALS iPSC-derivedmotor neurons

Global gene expression analysis in motor neurons derived from ALS human iPSCs (Egawa et al.,2012)

HumaniPSCs

Microarray Whole human genome microarray4 × 44K

microRNA expression profiling of TDP-43 silencedHep3B cells

TDP-43 seems to affect the levels of at least two miRNAs, let-7b (positively) andmiR-663 (negatively).

(Buratti et al.,2010)

Hep-3B Microarray Exiqon miRCURY LNA microRNAarray

microRNA expression profiling of FTLD-TDP brainswith/without progranulin mutations

Identification of 5 candidate miRNAs (miR-548b-5p, miR-548c- 5p, miR-571, miR-922,and miR-516a-3p) and 18 genes potentially dysregulated in FTLD patients

(Kocerhaet al., 2011)

SH-SY5Y Microarray Applied Biosystems TaqManhuman MicroRNA low densityarrays 2.0

High-throughput DNA sequencing analysis onthe transcriptome of ES cells derived fromTDP-43 ko mice

Identification of Tbc1d1, a gene known to mediate leanness and linked to obesity,which becomes downregulated following TDP-43 silencing

(Chiang et al.,2010)

Mousebrain

RNA-Seq Illumina Genome Analyzer II

RNA-seq to identify TDP-43 RNA targets in mousebrain treated withTDP-43 antisense oligos

Massively parallel sequencing revealed 601 mRNAs level changes in mouse adult brainstreated with TDP-43 antisense oligos. High representation of both neuron-specific(such as doublecortin, the neuron-specific beta-tubulin and choline O-acetyl transferase (Chat))and glia-specific (such as glial fibrillary acidic protein, myelin binding protein, Glt1 and Mag) RNAs

(Polymenidouet al., 2011)

Mousebrain

RNA-Seq Illumina Genome Analyzer II

Co-immunoprecipitation of TDP-43/RNA complexesfrom the cytoplasm of NSC-34 cells and geneexpression profiling

Identification of neuronal RNA targets of TDP-43-containing ribonucleoprotein complexes (Sephton et al.,2011)

Rat brain RNA-Seq Illumina Genome Analyzer IIx

Co-immunoprecipitation of TDP-43/RNA complexesfrom rat cortical neurons

Identification of cytoplasmic RNA targets of TDP-43 in the mouse motoneuronalNSC-34 cell line.

(Colombritaet al., 2012)

NSC-34 RIP-chip Illumina MouseRef-8 v2.0BeadChips

Co-immunoprecipitation of TDP-43/RNA followed bymicroarray analysis (RIP-chip) from mouse brain

Isolation and identification of RNAs bound to TDP-43 protein in mouse brain (Narayananet al., in press)

Mousebrain

RIP-chip Affymetrix GeneChip mouse gene1.0 ST arrays

CLIP-seq to identify in vivo RNA targets of TDP-43 inadult mouse brain

Found strong representation of neuronal and glial mRNA targets, such as glutamatetransporter 1 (Glt1), myelin-associated glycoprotein (Mag) and myelin oligodendrocyteglycoprotein (Mog)

(Polymenidouet al., 2011)

Mousebrain

CLIP

iCLIP-seq of TDP-43–RNA targets in SH-SY5Y cell line,FTLD brains, and human embryonic stem cells

Increases in binding were found for the MALAT1 and NEAT1 noncoding RNAs. Binding of TDP-43 topre-mRNAs influenced alternative splicing in aposition-dependent manner.

(Tollerveyet al., 2011a)

SH-SY5Y CLIP

CLIP-seq of TDP-43–RNA targets in SH-SY5Y cells Identification of 127 TDP-43 specific RNA targets (Xiao et al.,2011)

SH-SY5Y CLIP

468E.Burattiet

al./Molecular

andCellular

Neuroscience

56(2013)

465–474

Page 5: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

469E. Buratti et al. / Molecular and Cellular Neuroscience 56 (2013) 465–474

microRNAs and other non-coding RNA molecules are being uncovered(Mattick, 2011).Moreover, microarray results are extremely dependenton such factors as experimental designs, in terms of equipment andimage scanning facilities, normalization methods used during data pro-cessing, and summarization of results to give an expression value foreach single transcript (Schlitt and Kemmeren, 2004). As a result, allthese potential variables in the analysis often result in data sets thatare difficult to compare both at quantitative but even at the qualitativelevel between different studies.

From TDP-43 microarray data to results: comparing data sets

In order to determine the overlap between results from HTS studiesthat focus on a single protein as opposed to disease, we compared thefull data set of microarray studies in which TDP-43 had been deletedfrom HeLa cells (Ayala et al., 2008), HEK293E cells (Fiesel et al., 2010),and a gene expressionmicroarray analysis on post-mortem brain tis-sue from patients affected by frontotemporal lobar degenerationwith ubiquitin-positive inclusions (FTLD-U), frontotemporal lobardegeneration with motor neuron disease (FTLD-MND), and controls(Mishra et al., 2007) (Fig. 2). Direct comparison of these studies wasfacilitated by the fact that they all basically employed the same solidsupport for the evaluation for changes in gene expression levels(Affymetrix Human Genome U1333 arrays). As shown in this figure,however, there is no common transcript that can be foundmisregulatedin all samples. Strikingly, out of the several hundred transcripts altered

H

684FTLD-Upatients

580

HEK293

39722

26

32

List of common genes (32)YWHAZHPCAL1KIF3AUGCGNLKGULP1MKRN1IDSDNAJB6PLEKHB2SLITRK5

PPP3R1SLC1A1ATP2B1DGUOKPEG10MBNL1PLCL2MRPL4CAP2ENC1

QKIHIPK2RIN2ZBTB20TNPO1PCSK1MEF2CHPRT1NDRG3CITED2LZTFL1

List of common genes (22)ERBB3CHGBGPRASP1DKFZP564O0823VSNL1SNAP25NMNAT2ACTL6BSYN2CDC42PTPRO

AP3B2SYT13B4GALT6SEZ6L2RAB18DZIP3GNASELL2FBXO3GABRDL1CAM

0

Fig. 2. Overlapping comparison of three microarray analyses for common genes altered in vadata sets analyzed include knockdown experiments of TDP-43 in HeLa cells (Ayala et al., 20patients (Mishra et al., 2007). All experimental evaluations were performed using basicallytered in each experiment is reported. The number of overlapping transcripts is reported in reHeLa and HEK293E cells due to the fact that both are knockdown studies.

in each experiment, only a 5%of overlap can be found between each twodata sets (Fig. 2). This is a very low figure even when considering thatone is comparing HeLa versus 293 versus patient samples. Because ofthe central role played by TDP-43 in cell survival and embryo develop-ment it would be expected that, notwithstanding eventual cell-specificdifferences, there should be a common core of genes/pathways regulat-ed by this protein in all cell lines/organs. Most importantly, none of theoverlaps between the patient and culture cell line results contain someof the rather ubiquitous, experimentally verified TDP-43 endogenoustargets that have been published until now, such as CDK6 (Ayala et al.,2008), HDAC6 (Fiesel et al., 2010), or ATG7 (Bose et al., 2011). More-over, this similarity goes down even lower if one analyzes in detailthese comparisons (see Table 2) in terms of expression-trend (eitherup or down) and presence of at least some RNP (ribonuclear particle)immunoprecipitation combined with deep sequencing (RIP-seq) signa-tures (see below) for TDP-43 (Sephton et al., 2011). In fact, Table 2shows that many of these common hits are actually regulated in oppo-site directions in amanner that should eventually be reconciled throughfurther experimentation.

Of course, it would have been very surprising to find a large overlapbetween studies that focus on a single protein, such as TDP-43, com-pared to studies that focus on a particular disease state. The reasonsare twofold. First, the pathological samples analyzed were mostlyobtained post-mortem, which puts a question mark to the reproduc-ibility of the process. Second, because the disease state is likelythe result of a complex interactions between environment and

eLa

E

List of common genes (26)ANXA1EIF4G3DIRAS3HSPB8GABPB2PTHLHPNPOPOLR3GSNX6DENRSMC1A

GPR137BZNF84CBLBGFPT1SKP2USP13MPP5TARDBP*FAM105AVAMP3TMEM33

MDFICC16orf53HAS2PFKP

rious high throughput screening microarray experiments shown as a Venn diagram. The08), HEK293E (Fiesel et al., 2010), and patient post-mortem tissue samples from FTLDthe same set of Affymetrix GeneChips. The total number of transcripts found to be al-d and their list is provided in the respective boxes. *TARDBP appears as a common hit in

Page 6: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

Table 2List of genes regulated by TDP-43 common to different microarrays and matching withthe top 25% enriched TDP-43 RNA targets in Sephton et al., 2011.

Gene Mishra Fiesel Ayala

ATP2B1 Down – DownB4GALT6 Down Down –

CDC42 Down Up –

CHGB Down Up –

DNAJB6 Down – –

L1CAM Down Up –

PPP3R1 Down – DownPTPRO Down Up –

RAB18 Down Down –

SNAP25 Down Up –

SYN2 Down Down –

UGCG Down – UpYWHAZ Down – –

GFPT1 – Down UpPFKP – Down DownTARDBPa – Down DownTMEM33 – Down DownVAMP3 – Down Up

For each array, the direction of gene expression is indicated. Mishra et al. (2007), genesdifferentially expressed in post-mortem brain tissue from FTLD-U versus controls, hy-bridized to HG-U133 Plus 2.0 human genome chips (Affymetrix). Fiesel et al. (2010),list of altered transcripts (>2-fold) upon TDP-43 knockdown in HEK293E cells mea-sured with the GeneChip human genome U133+ 2.0 array (Affymetrix). Ayala et al.(2008), genes differentially expressed in HeLa from TDP-43 and control siRNA treatedcells hybridized to Affymetrix GeneChip Human Genome U133A 2.0 arrays(Affymetrix). Sephton et al. (2011), RefSeq IDs for top 25% enriched TDP-43–RNAtargets from rat cortical neurons derived from TDP-43–RNA immunoprecipitationsequenced on the Illumina GA IIx genome analyzer.

a TARDBP is downregulated due to the fact that both Ayala and Fiesel are knockdownstudies. – indicates that the gene was not detected in the analyses.

470 E. Buratti et al. / Molecular and Cellular Neuroscience 56 (2013) 465–474

inheritance factors of multiple risk variants in multiple genes, suchas those recently highlighted in several studies (Hart and Gitler,2012; Luty et al., 2010; van Blitterswijk et al., 2012) and reviewedin recent articles on ALS and FTLD genetics (Al-Chalabi et al., 2012;Sieben et al., 2012). Nonetheless, although technical differences arekept to a minimum, the comparison of these studies does not painta satisfactory picture even when the two cell lines (HeLa vs.HEK293E) are compared. In fact, considering that TDP-43 is an essen-tial protein to ensure cellular survival and early embryo develop-ment in a variety of organisms (Feiguin et al., 2009; Kabashi et al.,2010; Kraemer et al., 2010; Sephton et al., 2010; Wu et al., 2010),at least for these two samples one would have expected to find arather large overlap of genes which might represent the core of bio-logical events controlled by this protein, and this is not the case. Inthis respect, it should be noted that a similar picture has also beenobserved when comparing the results obtained from the recentRIP-Chip analysis of TDP-43 in normal mouse brain (Narayanan etal., in press) with several other miscellaneous types of HTS studies(Colombrita et al., 2012; Sephton et al., 2011; Xiao et al., 2011)where the overlap of all pairwise comparisons was rather low andonly one gene, ARF1, was common to all these studies.

In conclusion, all these comparisons emphasize the need to choosecarefully the samples to be analyzed in terms of: which particular as-pect of TDP-43 biology is under investigation, which cell or tissue isused to obtain the samples, and whether this is performed either innormal or disease conditions. The chances of finding greater consis-tency between HTS studies will be considerably improved when allthese issues are successfully addressed by researchers.

Furthermore, this tendency to observe rather small overlaps is notjust confined to microarray-based studies that look at global gene ex-pression changes, but it also applies to RIP-Chip approaches wheremicroarray screening is performed following immunoprecipitationof the RNA–protein complex and reverse transcription of the boundRNAs. For example, a rather small degree of overlap has also beenthe case for the study by Colombrita et al. (Colombrita et al., 2012),

where RNAs bound by TDP-43 in the cytoplasm of NSC-34 cellswere identified by RIP-Chip analysis. In this case part of the reasonfor the small overlap could certainly be due to some physiologicalreasons (that is, cytoplasmic mRNAs bound by TDP-43 represent asubset of all mRNAs bound by these proteins). However, such a lowoverlap may have also depended on possible technical bias originat-ing from the different efficiency of the commercial antibodies usedto recognize and immunoprecipitate these two proteins when theyare complexed into RNP particles, a fact also observed in CLIP exper-iments (Polymenidou et al., 2011; Xiao et al., 2011).

Hence, there is considerable improvement to be made before wecan confidently think of correlating different microarray experiments,even when they are carried out with similar technical specifications.

RNA sequencing and TDP-43

More recently, advances in Next Generation Sequencing approacheshave allowed the direct sequencing of the entire transcriptome(Hallegger et al., 2010; Sanchez-Pla et al., 2012). With respect tomicroarrays, RNA-seq approaches (schematically described inFig. 1B) present several advantages. First of all, they do not requirethe previous knowledge of a transcriptome or genomic sequence(and therefore can be reanalyzed as our knowledge expands inthis direction). In addition, RNA-seq approaches are also capableof detecting RNA editing events, can quantize splicing isoforms,and can identify allele specific expression. Moreover, RNA-seqdoes not involve several of the data processing steps that can heavi-ly affect the quality of microarray outputs, such as background sub-traction (there is very little in each experiment), and possesses asomewhat easier normalization process (longer and most abun-dant transcripts will tend to yield the highest amount of reads).However, also RNA-seq suffers from some limitations that havenot been fully solved. First of all, they are still considerably moreexpensive than microarray technology and although they generallysuffer from less variability the need to use replicates for greaterdata management accuracy very rapidly escalates the price ofeach experiment to much higher levels than comparable microar-ray approaches (up to 10×). Second, in order to make sure thatmid- to low-expressed transcripts are sufficiently covered by frag-ment sequencing there is the need to obtain a considerable se-quence depth. In addition, there are still several limitations withregards to the best bioinformatic procedures for data handling inorder to avoid the still rather high rate of mis-annotations, mainlydue to their high dependence on sequencing depth (Tarazona etal., 2011). Such errors, in fact, can range from 45% to as high as84% depending on the type of mapper-assembler used for the anal-ysis (Palmieri et al., 2012). For a more comprehensive discussion ofthese comparative issues between RNA-seq and microarrays thereader is referred to recent reviews on the subject (Malone andOliver, 2011).

With regards to the TDP-43 example, the effect of knocking out andoverexpressing this protein has been analyzed through RNA sequencingin various knockdown models such as mouse (Chiang et al., 2010) andDrosophila (Hazelett et al., 2012). Moreover, RNA sequencing has beenused to identify TDP-43 mRNA targets following immunoprecipitationof TDP-43–RNA complexes in rat cortical neurons (Sephton et al., 2011).

Even when raw data are available it is not easy to make a formalcomparison between all these data sets in order to verify their degreeof overlap due to differences in dynamic range (Agarwal et al., 2010).In any case, comparison using available results and microarray studiescarried out previously on TDP-43/patients also shows a rather limitedoverlap (Fig. 3). In fact, when the full data sets from in HeLa cells(Ayala et al., 2008), HEK293E (Fiesel et al., 2010), and post-mortem tis-sue samples from FTLD patients (Mishra et al., 2007) are compared sepa-rately with themore than 4000 genes which can be immunoprecipitatedwith TDP-43 from rat neurons (Sephton et al., 2011) the number of

Page 7: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

471E. Buratti et al. / Molecular and Cellular Neuroscience 56 (2013) 465–474

overlapping hits between all these different assays does not substantiallygo beyond 5% (Fig. 3).

CLIP-seq approaches for TDP-43

Another consequence of all these recent advances in DNA sequencingtechnology has also provided researchers with the ability to explore thetarget sites of any RBP of interest. This has been initially achieved usingcross-linking, immunoprecipitation and HTS or CLIP-seq (Licatalosiet al., 2008; Ule et al., 2003) (Fig. 1C). These data are then used to identifythe binding site of the RBPs on the various RNAs (both coding and non-)and can ideally provide information regarding how the positioning of thebinding sites can influence the splicing regulation of a particular tran-script (Witten and Ule, 2011).

In recent times, CLIP has allowed the mapping in vivo of the bindingsites of several splicing factors, such as NOVA (Ule et al., 2005), SRSF1(ASF/SF2) (Sanford et al., 2008), hnRNP A1 (Guil and Caceres, 2007)andTra2β (Grellscheid et al., 2011). Recently, it has beenused in conjunc-tionwith another global approach such asmicroarray or exon array anal-ysis to gain more insight on eventual positional–functional relationships(Huelga et al., 2012). As with every technique, however, also theCLIP methodology has shown some limitations that are currentlyunder study (Zhang and Darnell, 2011). In addition, factors suchas abundance of the transcript itself, low cross-linking efficiencies,reverse-transcription stalling at the cross-linked site, differential ampli-fication efficiencies, and various technological biases can influencethe end picture. Finally, the main drawback is that a high CLIP readaround a specific RNA region is not necessarily more significant from

168

H

FTLD-Upatients

580

cn

ACLY, ACTR10, ARHGEF9, ASCL1, ATAD1, ATP1A1, ATP1A3, ATP2A2, ATP2B1, ATP5B, ATP6V0D1, ATP6V1B2, ATP6V1G2, B3GALT3, B4GALT6, BASP1, BCAT1, BNIP3, BSG, CABP1, CADPS, CALM1, CALM3, CAMK2B, CAMK2D, CAMKK2, CCNDBP1, CD200, CDC42, CDH13, CDK5, CHGB, CHN1, CHRM1, CMAS, CNR1, CUGBP2, CX3CL1, DCTN1, DLGAP1, DNAJA2, DNAJB6, DNM1, EEF1A2, ELAVL3, ENO2, ENPP5, FAM3C, FGFR1OP2, FHL1, GABRB3, GAD2, GLRB, GLS, GPI, GRIA4, GRIN2A, GSTA4, HNRPK, ILF3, INPP4A, KCND2, KCNJ3, KCNQ2, KIF1B, KIF5B, KLHDC2, L1CAM, LARS, LKAP, LMO4, LRPPRC, MAP2K1, MAPK9, MAPT, MATR3, MDH1, MRPL44, MYT1L, NAP1L2, NAP1L3, NAPA, NAPG, NDEL1, NDFIP1, NDRG4, NDUFS1, NEGR1, NLN, NME1, NRXN1, NRXN3, NSF, NTRK2, OAT, PACSIN1, PAFAH1B1, PCMT1, PCSK2, PCYOX1, PDE1A, PDE4DIP, PDGFRA, PIK4CA, PJA2, PLRG1, PPP2R1A, PPP2R2B, PPP3CA, PPP3R1, PRDM2, PRKAR1A, PRKAR2B, PRKCB1, PRNP, PSMD10, PSMD12, PTPRO, RAB14, RAB18, RAB31, RASA1, RGS2, RGS7, RIMS3, RIPX, RTCD1, RTN3, RTN4, SCAMP1, SCG2, SCN2B, SCOC, SDFR1, SFRS7, SLC12A5, SLC17A7, SMARCA2, SNAP25, SNAP91, SNCA, SNCB, SNRPN, SNX10, SSX2IP, STMN2, STXBP1, SYN2, SYT1, TAF9, TCP1, TMEFF1, TMEM30A, TUSC3, UBE2N, UBE2V2, UBQLN1, UGCG, USP11, UXS1, VDAC1, VPS35, YES1, YPEL4, YWHAB, YWHAH, YWHAZ, ZFHX1B

List of common genes (168)

List of common gAGPAT4, AK3L1, CDC42, CHGA, CHDYNC1LI1, EIF2SINSR, KLHL7, KPNNDUFC2, PFKP, PSCN3B, SLC12A2TARDBP*, TMEM1

Fig. 3. Paired comparison of microarray analyses for genes regulated by TDP-43. ComparisHEK293E (Fiesel et al., 2010) cells with the top 25% enriched TDP-43 RNA targets in rat cois the comparison between this data set and transcripts found to be altered in FTLD post-min case of the RIP-seq analysis, or altered transcripts in each of the other three microarray exlist is provided in the respective boxes. *TARDBP appears as a common hit due to the fact th

the functional point of view, than around regions that are CLIP-poor.As a result, the functional significance of the individual CLIPs cannotbe easily predicted on the basis of their abundance or position alone,and almost invariably need to be verified on a case-by-case basis.

Notwithstanding these limitations, CLIP-seq analysis for the TDP_43protein has been performed on mouse and human brains and variouscell lines including SHSY5Y (Polymenidou et al., 2011; Tollervey et al.,2011a; Xiao et al., 2011). In general, all these approaches have providedinteresting insights regardingwhat could be the roles played by TDP-43in the cells. For example, the data point out to the potential involvementof TDP-43 in maintaining steady state levels of messenger moleculeswith particularly long introns. There are also indications for a role ofTDP-43 in regulating long non-coding RNAs, or shared targets withother RBPproteins involved in neurodegeneration, as recently reviewedelsewhere (Gitler, 2012; Polymenidou et al., 2012; Sephton et al., 2012).

At the moment, no formal comparison has been made between thevarious CLIP studies to check for differences and similarities. However,preliminary indications suggest that just like microarray and RNA-seqstudies, a rather low degree of overlap may exist. As an example,none of the 45 RNAs that are downregulated following either FUS/TLS or TDP-43 depletion in adult mouse brain using AON technology(Lagier-Tourenne et al., 2012) were present in the CLIP targets ofFUS/TLS identified in an HEK293 immortalized cell line (Hoell et al.,2011). This total lack of overlap may of course be accounted by theseveral technical limitations that still exist for this technique. However,it cannot also be ruled out (although unlikely to a great extent) thatTDP-43 functions and target sites may vary considerably betweencell lines, thus partially explaining this disparity. In keeping with this

4095

EK293E

397

81 HeLa684

Ratorticaleurons

44

List of commongenes (81)

enes (44)

ACSL4, ACVR1B, ADD3, AKT2, ANK3, ARAF, ARNTL, ATP1B1, ATP2B1, ATP6V0E1, BIN1, BZW1, CD2AP, CDC23, CHKA, CIRBP, CLK1, CSNK1E, CSTB, CTCF, CXCR4, DNAJB6, DUSP6, DYRK1A, ENAH, ETF1, GFPT1, HIPK3, HNRPH1, IPPK, KATNB1, KLHDC3, LAMP2, LASS2, MAP1B, MAP2K6, MAX, NAP1L1, NASP, NET1, NF1, NFIB, PAK2, PDE3A, PEA15, PFKP, PIK3R3, POLD4, POM121,PPAP2A, PPM1B, PPP1R12A, PPP3R1, PRIM1, QDPR, RAC1, RHOQ, RNF38, RRAS2, SHMT2, SLC11A2,SLC16A1,SLC1A4, SLC39A9,SLC6A8, SNX24, STAM2, SUB1, TARDBP*, TGFB2, TMBIM1, TMCO1, TMEM33, TOP1, TRIO, TSN, UGCG, VAMP3, YWHAZ, ZBTB10, ZFP37

B4GALT6, CAPZA2, CCND2, CD47, GB, CROT, CTNNB1, DDX25, DEF8,

1, GABBR2, GAP43, GFPT1, HSPA5, A3, L1CAM, LEPR, LNPEP, NAB1, KN1, PTPRO, RAB18, RAB21, SCG3,

, SNAP25, SPINT2, SSU72, SYN2, 06B, TMEM33, UNC84A, VAMP3

on of gene expression changes after TDP-43 silencing in HeLa (Ayala et al., 2008) andrtical neurons as determined by RIP-seq analysis (Sephton et al., 2011). Also reportedortem samples from FTLD-U patients (Mishra et al., 2007). The total number of targets,periments is shown. The number of overlapping transcripts is indicated in red and theirat both Ayala (Ayala et al., 2008) and Fiesel (Fiesel et al., 2010) are knockdown studies.

Page 8: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

472 E. Buratti et al. / Molecular and Cellular Neuroscience 56 (2013) 465–474

view, cell specific differences may certainly play an important role inthese limited overlaps, especiallywhenwe consider the high specializa-tion of neuronal cell lines.

Another important issue that should also be kept well in mindwhen analyzing CLIP data is that having a TDP-43 binding site doesnot necessarily translate into functional consequences. Recently, forexample, it has been reported that TDP-43 can bind both humanand mouse SORT1 mRNA and regulate its expression at the mRNAsplicing level by affecting the inclusion of a cassette exon (17b) thatin humans carries a premature stop codon. Further dissection of thissplicing event has shown that a clear differential regulation of 17bexon in these two species is mostly due to an evolutionary conservedsplicing enhancer element within the sequence of exon 17b. This se-quence is consistently absent in primates but is present in mouseand other species that lack the premature stop codon (Prudencio etal., 2012).

Concluding remarks and future directions

The use of HTS technologies to investigate the role of RBPs inneurodegeneration is certainly appealing because of the great possibil-ities that this type of techniques have opened up in our understandingof complex disease mechanisms. Moreover, the use of these techniqueshas become more and more available to researchers due to the drop ininstrumental andHTS sequences prices togetherwith a reduction of runtimes and the introduction of standardized experimental procedures.There are however some gray areas that sorely need to be addressedin the future if we want to make the most out of these approaches.First of all, there is the need of greater accuracy in terms of our under-standing of the experimental bias associated with these approachesand of ways to reduce them. Most importantly, there is also the greatneed for developing simplified data analysis tools (both at the qualita-tive and quantitative level) in order tominimize the requirement of so-phisticated bioinformatics for data output and interpretation. Finally, asmost of these programs heavily rely on biological databases to providetheir output, there is also the independent need to always improvethe quality of the database annotations themselves, that are still full ofunintentional errors, fortunately this situation has been steadily im-proving in recent years.

In conclusion, the data collected up to now by HTS techniques andillustrated in this review by the TDP-43 example indicate how difficultand how limited is our ability to compare different HTS experimentsin order to gain further reliable insight in the biological properties ofRBPs. For these reasons, the routine, wide and successful use of HTSstudies needs further methodological and technical refining through asustained multidisciplinary collaboration between scientists with dif-ferent backgrounds in all the aspects of data collection, data processingand technical improvements of the HTS systems themselves.

Acknowledgements

This work was supported by AriSLA (TARMA) and the Universityof Trieste-Finanziamento per Ricercatori di Ateneo.

References

Agarwal, A., Koppstein, D., Rozowsky, J., Sboner, A., Habegger, L., Hillier, L.W., Sasidharan,R., Reinke, V., Waterston, R.H., Gerstein, M., 2010. Comparison and calibration oftranscriptome data from RNA-seq and tiling arrays. BMC Genomics 11, 383.

Al-Chalabi, A., Jones, A., Troakes, C., King, A., Al-Sarraj, S., van den Berg, L.H., 2012. Thegenetics and neuropathology of amyotrophic lateral sclerosis. Acta Neuropathol.124, 339–352.

Arai, T., Hasegawa, M., Akiyama, H., Ikeda, K., Nonaka, T., Mori, H., Mann, D., Tsuchiya, K.,Yoshida, M., Hashizume, Y., Oda, T., 2006. TDP-43 is a component of ubiquitin-positivetau-negative inclusions in frontotemporal lobar degeneration and amyotrophic lateralsclerosis. Biochem. Biophys. Res. Commun. 351, 602–611.

Ayala, Y.M., Pantano, S., D'Ambrogio, A., Buratti, E., Brindisi, A., Marchetti, C., Romano,M., Baralle, F.E., 2005. Human, Drosophila, and C. elegans TDP43: nucleic acid bind-ing properties and splicing regulatory function. J. Mol. Biol. 348, 575–588.

Ayala, Y.M., Misteli, T., Baralle, F.E., 2008. TDP-43 regulates retinoblastoma proteinphosphorylation through the repression of cyclin-dependent kinase 6 expression.Proc. Natl. Acad. Sci. U. S. A. 105, 3785–3789.

Baciu, C., Thompson, K.J., Mougeot, J.L., Brooks, B.R., Weller, J.W., 2012. The LO-BaFLmethod and ALS microarray expression analysis. BMC Bioinforma. 13, 244.

Bose, J.K., Huang, C.C., Shen, C.K., 2011. Regulation of autophagy by neuropathologicalprotein TDP-43. J. Biol. Chem. 286, 44441–44448.

Budini, M., Baralle, F.E., Buratti, E., 2011. Regulation of gene expression by TDP-43 andFUS/TLS in frontotemporal lobar degeneration. Curr. Alzheimer Res. 8, 237–245.

Buratti, E., Baralle, F.E., 2001. Characterization and functional implications of the RNAbinding properties of nuclear factor TDP-43, a novel splicing regulator of CFTRexon 9. J. Biol. Chem. 276, 36337–36343.

Buratti, E., Baralle, F.E., 2010. The multiple roles of TDP-43 in pre-mRNA processing andgene expression regulation. RNA Biol. 7, 420–429.

Buratti, E., Baralle, F.E., 2012. TDP-43: gumming up neurons through protein–proteinand protein–RNA interactions. Trends Biochem. Sci. 37, 237–247.

Buratti, E., Dork, T., Zuccato, E., Pagani, F., Romano, M., Baralle, F.E., 2001. Nuclear factorTDP-43 and SR proteins promote in vitro and in vivo CFTR exon 9 skipping. EMBO J.20, 1774–1784.

Buratti, E., De Conti, L., Stuani, C., Romano, M., Baralle, M., Baralle, F., 2010. Nuclear factorTDP-43 can affect selected microRNA levels. FEBS J. 277, 2268–2281.

Caputi, M., Zahler, A.M., 2001. Determination of the RNA binding specificity of theheterogeneous nuclear ribonucleoprotein (hnRNP) H/H'/F/2H9 family. J. Biol.Chem. 276, 43850–43859.

Casafont, I., Bengoechea, R., Tapia, O., Berciano, M.T., Lafarga, M., 2009. TDP-43 localizesin mRNA transcription and processing sites in mammalian neurons. J. Struct. Biol.167, 235–241.

Chen-Plotkin, A.S., Geser, F., Plotkin, J.B., Clark, C.M., Kwong, L.K., Yuan, W., Grossman,M., Van Deerlin, V.M., Trojanowski, J.Q., Lee, V.M., 2008. Variations in theprogranulin gene affect global gene expression in frontotemporal lobar degenera-tion. Hum. Mol. Genet. 17, 1349–1362.

Chiang, P.M., Ling, J., Jeong, Y.H., Price, D.L., Aja, S.M., Wong, P.C., 2010. Deletion of TDP-43 down-regulates Tbc1d1, a gene linked to obesity, and alters body fat metabo-lism. Proc. Natl. Acad. Sci. U. S. A. 107, 16320–16324.

Colombrita, C., Onesto, E., Megiorni, F., Pizzuti, A., Baralle, F.E., Buratti, E., Silani, V., Ratti,A., 2012. TDP-43 and FUS RNA-binding proteins bind distinct sets of cytoplasmicmessenger RNAs and differently regulate their post-transcriptional fate inmotoneuron-like cells. J. Biol. Chem. 287, 15635–15647.

Cooper-Knock, J., Kirby, J., Ferraiuolo, L., Heath, P.R., Rattray, M., Shaw, P.J., 2012. Geneexpression profiling in human neurodegenerative disease. Nat. Rev. Neurol. 8, 518–530.

Cuppens, H., Lin, W., Jaspers, M., Costes, B., Teng, H., Vankeerberghen, A., Jorissen, M.,Droogmans, G., Reynaert, I., Goossens, M., Nilius, B., Cassiman, J.J., 1998. Polyvariantmutant cystic fibrosis transmembrane conductance regulator genes. The polymor-phic (Tg)m locus explains the partial penetrance of the T5 polymorphism as adisease mutation. J. Clin. Invest. 101, 487–496.

Dalma-Weiszhausz, D.D., Warrington, J., Tanimoto, E.Y., Miyada, C.G., 2006. TheAffymetrix GeneChip platform: an overview. Methods Enzymol. 410, 3–28.

Dormann, D., Haass, C., 2011. TDP-43 and FUS: a nuclear affair. Trends Neurosci.Dreyfuss, G., Matunis, M.J., Pinol-Roma, S., Burd, C.G., 1993. hnRNP proteins and the

biogenesis of mRNA. Annu. Rev. Biochem. 62, 289–321.Dunbar, S.A., 2006. Applications of Luminex xMAP technology for rapid, high-throughput

multiplexed nucleic acid detection. Clin. Chim. Acta 363, 71–82.Egawa, N., Kitaoka, S., Tsukita, K., Naitoh, M., Takahashi, K., Yamamoto, T., Adachi, F.,

Kondo, T., Okita, K., Asaka, I., Aoi, T., Watanabe, A., Yamada, Y., Morizane, A.,Takahashi, J., Ayaki, T., Ito, H., Yoshikawa, K., Yamawaki, S., Suzuki, S., Watanabe,D., Hioki, H., Kaneko, T., Makioka, K., Okamoto, K., Takuma, H., Tamaoka, A.,Hasegawa, K., Nonaka, T., Hasegawa, M., Kawata, A., Yoshida, M., Nakahata, T.,Takahashi, R., Marchetto, M.C., Gage, F.H., Yamanaka, S., Inoue, H., 2012. Drugscreening for ALS using patient-specific induced pluripotent stem cells. Sci. Transl.Med. 4, 145ra104.

Fan, J.B., Gunderson, K.L., Bibikova, M., Yeakley, J.M., Chen, J., Wickham Garcia, E.,Lebruska, L.L., Laurent, M., Shen, R., Barker, D., 2006. Illumina universal bead arrays.Methods Enzymol. 410, 57–73.

Feiguin, F., Godena, V.K., Romano, G., D'Ambrogio, A., Klima, R., Baralle, F.E., 2009.Depletion ofTDP-43 affects Drosophila motoneurons terminal synapsis and locomotive behavior.FEBS Lett. 583, 1586–1592.

Fiesel, F.C., Voigt, A., Weber, S.S., Van den Haute, C., Waldenmaier, A., Gorner, K., Walter,M., Anderson,M.L., Kern, J.V., Rasse, T.M., Schmidt, T., Springer,W., Kirchner, R., Bonin,M., Neumann, M., Baekelandt, V., Alunni-Fabbroni, M., Schulz, J.B., Kahle, P.J., 2010.Knockdown of transactive response DNA-binding protein (TDP-43) downregulateshistone deacetylase 6. EMBO J. 29, 209–221.

Fiesel, F.C., Weber, S.S., Supper, J., Zell, A., Kahle, P.J., 2012. TDP-43 regulates globaltranslational yield by splicing of exon junction complex component SKAR. NucleicAcids Res. 40, 2668–2682.

Fukuda, T., Yamagata, K., Fujiyama, S., Matsumoto, T., Koshida, I., Yoshimura, K., Mihara,M., Naitou, M., Endoh, H., Nakamura, T., Akimoto, C., Yamamoto, Y., Katagiri, T.,Foulds, C., Takezawa, S., Kitagawa, H., Takeyama, K., O'Malley, B.W., Kato, S.,2007. DEAD-box RNA helicase subunits of the Drosha complex are required forprocessing of rRNA and a subset of microRNAs. Nat. Cell Biol. 9, 604–611.

Gagliardi, S., Milani, P., Sardone, V., Pansarasa, O., Cereda, C., 2012. From transcriptometo noncoding rnas: implications in ALS mechanism. Neurol. Res. Int. 2012, 278725.

Gitler, A.D., 2012. TDP-43 and FUS/TLS yield a target-rich haul in ALS. Nat. Neurosci. 15,1467–1469.

Page 9: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

473E. Buratti et al. / Molecular and Cellular Neuroscience 56 (2013) 465–474

Grabowski, P., 2011. Alternative splicing takes shape during neuronal development.Curr. Opin. Genet. Dev. 21, 388–394.

Gregory, R.I., Yan, K.P., Amuthan, G., Chendrimada, T., Doratotaj, B., Cooch, N., Shiekhattar,R., 2004. The microprocessor complex mediates the genesis of microRNAs. Nature432, 235–240.

Grellscheid, S., Dalgliesh, C., Storbeck, M., Best, A., Liu, Y., Jakubik, M., Mende, Y.,Ehrmann, I., Curk, T., Rossbach, K., Bourgeois, C.F., Stevenin, J., Grellscheid, D.,Jackson, M.S., Wirth, B., Elliott, D.J., 2011. Identification of evolutionarily conservedexons as regulated targets for the splicing activator tra2beta in development. PLoSGenet. 7, e1002390.

Groman, J.D., Hefferon, T.W., Casals, T., Bassas, L., Estivill, X., Des Georges, M., Guittard,C., Koudova, M., Fallin, M.D., Nemeth, K., Fekete, G., Kadasi, L., Friedman, K.,Schwarz, M., Bombieri, C., Pignatti, P.F., Kanavakis, E., Tzetis, M., Schwartz, M.,Novelli, G., D'Apice, M.R., Sobczynska-Tomaszewska, A., Bal, J., Stuhrmann, M.,Macek Jr., M., Claustres, M., Cutting, G.R., 2004. Variation in a repeat sequence deter-mines whether a common variant of the cystic fibrosis transmembrane conductanceregulator gene is pathogenic or benign. Am. J. Hum. Genet. 74, 176–179.

Guil, S., Caceres, J.F., 2007. Themultifunctional RNA-binding protein hnRNP A1 is requiredfor processing of miR-18a. Nat. Struct. Mol. Biol. 14, 591–596.

Hallegger, M., Llorian, M., Smith, C.W., 2010. Alternative splicing: global insights. FEBSJ. 277, 856–866.

Halliday, G., Bigio, E.H., Cairns, N.J., Neumann, M., Mackenzie, I.R., Mann, D.M., 2012.Mechanisms of disease in frontotemporal lobar degeneration: gain of function versusloss of function effects. Acta Neuropathol. 124, 373–382.

Hanson, K.A., Kim, S.H., Tibbetts, R.S., 2012. RNA-binding proteins in neurodegenerativedisease: TDP-43 and beyond. Wiley Interdiscip. Rev. RNA 3, 265–285.

Hart, M.P., Gitler, A.D., 2012. ALS-associated ataxin 2 polyQ expansions enhance stress-induced caspase 3 activation and increaseTDP-43pathologicalmodifications. J. Neurosci.32, 9133–9142.

Hazelett, D.J., Chang, J.C., Lakeland, D.L., Morton, D.B., 2012. Comparison of parallel high-throughput RNA sequencing between knockout of TDP-43 and its overexpressionreveals primarily nonreciprocal and nonoverlapping gene expression changes in thecentral nervous system of Drosophila. G3 (Bethesda) 2, 789–802.

Hoell, J.I., Larsson, E., Runge, S., Nusbaum, J.D., Duggimpudi, S., Farazi, T.A., Hafner, M.,Borkhardt, A., Sander, C., Tuschl, T., 2011. RNA targets of wild-type and mutantFET family proteins. Nat. Struct. Mol. Biol. 18, 1428–1431.

Huelga, S.C., Vu, A.Q., Arnold, J.D., Liang, T.Y., Liu, P.P., Yan, B.Y., Donohue, J.P., Shiue, L.,Hoon, S., Brenner, S., Ares Jr., M., Yeo, G.W., 2012. Integrative genome-wide analy-sis reveals cooperative regulation of alternative splicing by hnRNP proteins. CellRep. 1, 167–178.

Igaz, L.M., Kwong, L.K., Lee, E.B., Chen-Plotkin, A., Swanson, E., Unger, T., Malunda, J., Xu, Y.,Winton, M.J., Trojanowski, J.Q., Lee, V.M., 2011. Dysregulation of the ALS-associatedgene TDP-43 leads to neuronal death and degeneration in mice. J. Clin. Invest. 121,726–738.

Kabashi, E., Lin, L., Tradewell, M.L., Dion, P.A., Bercier, V., Bourgouin, P., Rochefort, D.,Bel Hadj, S., Durham, H.D., Velde, C.V., Rouleau, G.A., Drapeau, P., 2010. Gain andloss of function of ALS-related mutations of TARDBP (TDP-43) cause motor deficitsin vivo. Hum. Mol. Genet. 19, 671–683.

Kaida, D., Berg, M.G., Younis, I., Kasim, M., Singh, L.N., Wan, L., Dreyfuss, G., 2010. U1snRNP protects pre-mRNAs from premature cleavage and polyadenylation. Nature468, 664–668.

Kawahara, Y., Mieda-Sato, A., 2012. TDP-43 promotes microRNA biogenesis as a com-ponent of the Drosha and Dicer complexes. Proc. Natl. Acad. Sci. U. S. A. 109,3347–3352.

King, O.D., Gitler, A.D., Shorter, J., 2012. The tip of the iceberg: RNA-binding proteinswith prion-like domains in neurodegenerative disease. Brain Res. 1462, 61–80.

Kocerha, J., Kouri, N., Baker,M., Finch, N., DeJesus-Hernandez, M., Gonzalez, J., Chidamparam,K., Josephs, K.A., Boeve, B.F., Graff-Radford, N.R., Crook, J., Dickson, D.W., Rademakers, R.,2011. Altered microRNA expression in frontotemporal lobar degeneration with TDP-43pathology caused by progranulin mutations. BMC Genomics 12, 527.

Kraemer, B.C., Schuck, T., Wheeler, J.M., Robinson, L.C., Trojanowski, J.Q., Lee, V.M.,Schellenberg, G.D., 2010. Loss of murine TDP-43 disrupts motor function andplays an essential role in embryogenesis. Acta Neuropathol. 119, 409–419.

Lagier-Tourenne, C., Polymenidou, M., Hutt, K.R., Vu, A.Q., Baughn, M., Huelga, S.C.,Clutario, K.M., Ling, S.C., Liang, T.Y., Mazur, C., Wancewicz, E., Kim, A.S., Watt, A.,Freier, S., Hicks, G.G., Donohue, J.P., Shiue, L., Bennett, C.F., Ravits, J., Cleveland,D.W., Yeo, G.W., 2012. Divergent roles of ALS-linked proteins FUS/TLS and TDP-43 intersect in processing long pre-mRNAs. Nat. Neurosci. 15, 1488–1497.

Larriba, S., Bassas, L., Gimenez, J., Ramos, M.D., Segura, A., Nunes, V., Estivill, X., Casals,T., 1998. Testicular CFTR splice variants in patients with congenital absence of thevas deferens. Hum. Mol. Genet. 7, 1739–1743.

Lee, E.B., Lee, V.M., Trojanowski, J.Q., 2012. Gains or losses: molecular mechanisms ofTDP43-mediated neurodegeneration. Nat. Rev. Neurosci. 13, 38–50.

Licatalosi, D.D., Mele, A., Fak, J.J., Ule, J., Kayikci, M., Chi, S.W., Clark, T.A.,Schweitzer, A.C., Blume, J.E., Wang, X., Darnell, J.C., Darnell, R.B., 2008. HITS-CLIP yields genome-wide insights into brain alternative RNA processing. Nature456, 464–469.

Luty, A.A., Kwok, J.B., Dobson-Stone, C., Loy, C.T., Coupland, K.G., Karlstrom, H., Sobow,T., Tchorzewska, J., Maruszak, A., Barcikowska, M., Panegyres, P.K., Zekanowski, C.,Brooks, W.S., Williams, K.L., Blair, I.P., Mather, K.A., Sachdev, P.S., Halliday, G.M.,Schofield, P.R., 2010. Sigma nonopioid intracellular receptor 1 mutations causefrontotemporal lobar degeneration–motor neuron disease. Ann. Neurol. 68, 639–649.

Malone, J.H., Oliver, B., 2011. Microarrays, deep sequencing and the true measure of thetranscriptome. BMC Biol. 9, 34.

Mattick, J.S., 2011. The central role of RNA in human development and cognition. FEBSLett. 585, 1600–1616.

Mercado, P.A., Ayala, Y.M., Romano, M., Buratti, E., Baralle, F.E., 2005. Depletion of TDP43 overrides the need for exonic and intronic splicing enhancers in the humanapoA-II gene. Nucleic Acids Res. 33, 6000–6010.

Mills, J.D., Janitz, M., 2012. Alternative splicing of mRNA in the molecular pathology ofneurodegenerative diseases. Neurobiol. Aging 33 (1012), e1011–e1024.

Mishra, M., Paunesku, T., Woloschak, G.E., Siddique, T., Zhu, L.J., Lin, S., Greco, K., Bigio,E.H., 2007. Gene expression analysis of frontotemporal lobar degeneration of themotor neuron disease type with ubiquitinated inclusions. Acta Neuropathol. 114,81–94.

Narayanan, R.K., Mangelsdorf, M., Panwar, A., Butler, T.J., Noakes, P.G., Wallace, R.H., inpress. Identification of RNA bound to the TDP-43 ribonucleoprotein complex in theadult mouse brain. Amyotroph. Lateral Scler.

Neumann, M., Sampathu, D.M., Kwong, L.K., Truax, A.C., Micsenyi, M.C., Chou, T.T., Bruce, J.,Schuck, T., Grossman, M., Clark, C.M., McCluskey, L.F., Miller, B.L., Masliah, E.,Mackenzie, I.R., Feldman, H., Feiden, W., Kretzschmar, H.A., Trojanowski, J.Q., Lee, V.M.,2006. Ubiquitinated TDP-43 in frontotemporal lobar degeneration and amyotrophiclateral sclerosis. Science 314, 130–133.

Niksic, M., Romano, M., Buratti, E., Pagani, F., Baralle, F.E., 1999. Functional analysis ofcis-acting elements regulating the alternative splicing of human CFTR exon 9.Hum. Mol. Genet. 8, 2339–2349.

Norris, A.D., Calarco, J.A., 2012. Emerging roles of alternative pre-mRNA splicing regulationin neuronal development and function. Front. Neurosci. 6, 122.

Palmieri, N., Nolte, V., Suvorov, A., Kosiol, C., Schlotterer, C., 2012. Evaluation of differentreference based annotation strategies using RNA-seq—a case study in Drososphilapseudoobscura. PLoS One 7, e46415.

Passoni, M., De Conti, L., Baralle, M., Buratti, E., 2012. UG repeats/TDP-43 interactionsnear 5′ splice sites exert unpredictable effects on splicing modulation. J. Mol.Biol. 415, 46–60.

Polymenidou, M., Lagier-Tourenne, C., Hutt, K.R., Huelga, S.C., Moran, J., Liang, T.Y., Ling,S.C., Sun, E., Wancewicz, E., Mazur, C., Kordasiewicz, H., Sedaghat, Y., Donohue, J.P.,Shiue, L., Bennett, C.F., Yeo, G.W., Cleveland, D.W., 2011. Long pre-mRNA depletionand RNA missplicing contribute to neuronal vulnerability from loss of TDP-43. Nat.Neurosci. 14, 459–468.

Polymenidou, M., Lagier-Tourenne, C., Hutt, K.R., Bennett, C.F., Cleveland, D.W., Yeo,G.W., 2012. Misregulated RNA processing in amyotrophic lateral sclerosis. BrainRes. 1462, 3–15.

Potthast, T., 2009. Paradigm shifts versus fashion shifts? Systems and synthetic biologyas new epistemic entities in understanding and making ‘life’. EMBO Rep. 10 (Suppl.1), S42–S45.

Prudencio, M., Jansen-West, K.R., Lee, W.C., Gendron, T.F., Zhang, Y.-J., Xu, Y.-F., Gass, J.,Stuani, C., Stetler, C., Rademakers, R., Dickson, D.W., Buratti, E., Petrucelli, L., 2012.Misregulation of human sortilin splicing leads to the generation of a non-functionalprogranulin receptor. PNAS 109, 21510–21515.

Rabin, S.J., Kim, J.M., Baughn, M., Libby, R.T., Kim, Y.J., Fan, Y., Libby, R.T., La Spada,A., Stone, B., Ravits, J., 2010. Sporadic ALS has compartment-specific aberrantexon splicing and altered cell-matrix adhesion biology. Hum. Mol. Genet. 19,313–328.

Rademakers, R., Neumann, M., Mackenzie, I.R., 2012. Advances in understanding themolecular basis of frontotemporal dementia. Nat. Rev. Neurol. 8, 423–434.

Renoux, A.J., Todd, P.K., 2012. Neurodegeneration the RNAway. Prog. Neurobiol. 97, 173–189.Sanchez-Pla, A., Reverter, F., Ruiz de Villa, M.C., Comabella, M., 2012. Transcriptomics:

mRNA and alternative splicing. J. Neuroimmunol. 248, 23–31.Sanford, J.R., Coutinho, P., Hackett, J.A., Wang, X., Ranahan, W., Caceres, J.F., 2008. Identifica-

tion of nuclear and cytoplasmic mRNA targets for the shuttling protein SF2/ASF. PLoSOne 3, e3369.

Schena, M., Shalon, D., Davis, R.W., Brown, P.O., 1995. Quantitative monitoring of geneexpression patterns with a complementary DNA microarray. Science 270, 467–470.

Schlitt, T., Kemmeren, P., 2004. From microarray data to results. Workshop on genomicapproaches to microarray data analysis. EMBO Rep. 5, 459–463.

Sephton, C.F., Good, S.K., Atkin, S., Dewey, C.M.,Mayer III, P., Herz, J., Yu, G., 2010. TDP-43 isa developmentally regulated protein essential for early embryonic development.J. Biol. Chem. 285, 6826–6834.

Sephton, C.F., Cenik, C., Kucukural, A., Dammer, E.B., Cenik, B., Han, Y., Dewey, C.M.,Roth, F.P., Herz, J., Peng, J., Moore, M.J., Yu, G., 2011. Identification of neuronalRNA targets of TDP-43-containing ribonucleoprotein complexes. J. Biol. Chem.286, 1204–1215.

Sephton, C.F., Cenik, B., Cenik, B.K., Herz, J., Yu, G., 2012. TDP-43 in central nervoussystem development and function: clues to TDP-43-associated neurodegeneration.Biol. Chem. 393, 589–594.

Sieben, A., Van Langenhove, T., Engelborghs, S., Martin, J.J., Boon, P., Cras, P., De Deyn,P.P., Santens, P., Van Broeckhoven, C., Cruts, M., 2012. The genetics and neuropa-thology of frontotemporal lobar degeneration. Acta Neuropathol. 124, 353–372.

Tarazona, S., Garcia-Alcalde, F., Dopazo, J., Ferrer, A., Conesa, A., 2011. Differentialexpression in RNA-seq: a matter of depth. Genome Res. 21, 2213–2223.

Tollervey, J.R., Curk, T., Rogelj, B., Briese,M., Cereda,M., Kayikci, M., Konig, J., Hortobagyi, T.,Nishimura, A.L., Zupunski, V., Patani, R., Chandran, S., Rot, G., Zupan, B., Shaw, C.E., Ule,J., 2011a. Characterizing the RNA targets and position-dependent splicing regulationby TDP-43. Nat. Neurosci. 14, 452–458.

Tollervey, J.R., Wang, Z., Hortobagyi, T., Witten, J.T., Zarnack, K., Kayikci, M., Clark, T.A.,Schweitzer, A.C., Rot, G., Curk, T., Zupan, B., Rogelj, B., Shaw, C.E., Ule, J., 2011b.Analysis of alternative splicing associated with aging and neurodegeneration inthe human brain. Genome Res. 21, 1572–1582.

Ugras, S.E., Shorter, J., 2012. RNA-binding proteins in amyotrophic lateral sclerosis andneurodegeneration. Neurol. Res. Int. 2012, 432780.

Ule, J., Jensen, K.B., Ruggiu, M., Mele, A., Ule, A., Darnell, R.B., 2003. CLIP identifies Nova-regulated RNA networks in the brain. Science 302, 1212–1215.

Page 10: TDP-43 high throughput screening analyses in neurodegeneration: Advantages and pitfalls

474 E. Buratti et al. / Molecular and Cellular Neuroscience 56 (2013) 465–474

Ule, J., Ule, A., Spencer, J., Williams, A., Hu, J.S., Cline, M., Wang, H., Clark, T., Fraser, C.,Ruggiu, M., Zeeberg, B.R., Kane, D., Weinstein, J.N., Blume, J., Darnell, R.B., 2005.Nova regulates brain-specific splicing to shape the synapse. Nat. Genet. 37, 844–852.

van Blitterswijk, M., van Es, M.A., Hennekam, E.A., Dooijes, D., van Rheenen, W., Medic,J., Bourque, P.R., Schelhaas, H.J., van der Kooi, A.J., de Visser, M., de Bakker, P.I.,Veldink, J.H., van den Berg, L.H., 2012. Evidence for an oligogenic basis ofamyotrophic lateral sclerosis. Hum. Mol. Genet. 21, 3776–3784.

Volkening, K., Leystra-Lantz, C., Yang, W., Jaffee, H., Strong, M.J., 2009. Tar DNA bindingprotein of 43 kDa (TDP-43), 14-3-3 proteins and copper/zinc superoxide dismutase(SOD1) interact to modulate NFL mRNA stability. Implications for altered RNAprocessing in amyotrophic lateral sclerosis (ALS). Brain Res. 1305, 168–182.

Witten, J.T., Ule, J., 2011. Understanding splicing regulation through RNA splicing maps.Trends Genet. 27, 89–97.

Wu, L.S., Cheng, W.C., Hou, S.C., Yan, Y.T., Jiang, S.T., Shen, C.K., 2010. TDP-43, a neuro-pathosignature factor, is essential for early mouse embryogenesis. Genesis 48, 56–62.

Xiao, S., Sanelli, T., Dib, S., Sheps, D., Findlater, J., Bilbao, J., Keith, J., Zinman, L., Rogaeva,E., Robertson, J., 2011. RNA targets of TDP-43 identified by UV-CLIP are deregulatedin ALS. Mol. Cell. Neurosci. 47, 167–180.

Xu, Z.S., 2012. Does a loss of TDP-43 function cause neurodegeneration?Mol. Neurodegener.7, 27.

Youmans, K.L., Wolozin, B., 2012. TDP-43: a new player on the AD field? Exp. Neurol.237, 90–95.

Zhang, C., Darnell, R.B., 2011. Mapping in vivo protein–RNA interactions at single-nucleotide resolution from HITS-CLIP data. Nat. Biotechnol. 29, 607–614.