Genome-wide analysis of familial dysautonomia and kinetin target genes with patient olfactory ecto-mesenchymal stem cells

Authors


  • Communicated by Mireille Claustres

Abstract

Familial dysautonomia (FD) is a rare inherited neurodegenerative disorder. The most common mutation is a c.2204+6T>C transition in the 5′ splice site (5′ss) of IKBKAP intron 20, which causes a tissue-specific skipping of exon 20, resulting in lower synthesis of IKAP/hELP1 protein. To better understand the specificity of neuron loss in FD, we modeled the molecular mechanisms of IKBKAP mRNA splicing by studying human olfactory ecto-mesenchymal stem cells (hOE-MSCs) derived from FD patient nasal biopsies. We explored how the modulation of IKBKAP mRNA alternative splicing impacts the transcriptome at the genome-wide level. We found that the FD transcriptional signature was highly associated with biological functions related to the development of the nervous system. In addition, we identified target genes of kinetin, a plant cytokinin that corrects IKBKAP mRNA splicing and increases the expression of IKAP/hELP1. We identified this compound as a putative regulator of splicing factors and added new evidence for a sequence-specific correction of splicing. In conclusion, hOE-MSCs isolated from FD patients represent a promising avenue for modeling the altered genetic expression of FD, demonstrating a methodology that can be applied to a host of other genetic disorders to test the therapeutic potential of candidate molecules. Hum Mutat 33:530–540, 2012. © 2011 Wiley Periodicals, Inc.

Introduction

Familial dysautonomia (FD, Riley-Day syndrome, hereditary sensory and autonomic neuropathy type III, MIM# 223900) is a rare neurodegenerative disease with autosomal recessive inheritance and a carrier frequency of 1 in 31 in the Ashkenazi Jewish population [Scott et al., 2010]. The disease is characterized by anatomical selective depletion of sensory and autonomic neurons [Axelrod et al., 1981; Pearson and Pytel, 1978; Pearson et al., 1978] resulting in variable symptoms including: decreased sensitivity to pain, lack of overflow tearing, inappropriate blood pressure control manifested as orthostatic hypotension and episodic hypertension, poor oral coordination resulting in poor feeding and swallowing, and gastrointestinal dysmotility [Axelrod, 2004]. FD is a disease for which no cure is currently available, and treatment is aimed at controlling symptoms and prevention of complications.

FD is caused by mutations in the IKBKAP gene (MIM# 603722), which encodes a protein termed IKAP/hELP1 [Anderson et al., 2001; Slaugenhaupt et al., 2001]. The most prevalent mutation, is the T-to-C transition in position six of the 5′ splice site (5′ss) of intron 20 (c.2204+6T>C), occurring in >99.5% of cases of FD [Anderson et al., 2001; Dong et al., 2002; Scott et al., 2010; Slaugenhaupt et al., 2001]. This mutation leads to a tissue-specific skipping of exon 20 of IKBKAP mRNA (MU isoforms). The defective splicing leads to low levels of transcripts including exon 20 (WT isoforms), reduced synthesis of IKAP/hELP1 protein, and appears to be more severe in sensory and autonomic nervous systems than others tissues [Cuajungco et al., 2003].

IKAP/hELP1 was identified as the scaffold protein required to assemble a well-conserved six-protein complex (ELP1-6), also called the holo-Elongator complex [Hawkes et al., 2002], which is recruited to the transcribed regions of some human genes essentially involved in actin cytoskeleton regulation and cell motility migration. Subsequently, IKAP/hElongator was also shown to have functions in cell migration [Close et al., 2006; Creppe et al., 2009], acetylation of microtubules, and neuronal development [Solinger et al., 2010]. It was also proposed to play a role in exocytosis [Rahl et al., 2005], and zygotic paternal genome demethylation [Okada et al., 2010], but most likely as a result of tRNA modifications [Chen et al., 2009a; Esberg et al., 2006; Huang et al., 2005; Li et al., 2009].

Several studies aimed at investigating transcriptional alterations revealed distinct patterns of gene expression in FD. Indeed, a subgroup of genes associated with cell migration and actin cytoskeleton was shown to be downregulated in IKAP/hElp1 deficient HeLa and FD fibroblast cells [Close et al., 2006]. Others identified genes known to be involved in oligodendrocyte development, myelin formation, and disorganization of microtubules from cerebrum of FD patients [Cheishvili et al., 2007, 2011]. Lee and colleagues determined that the neuron-specific splicing factor NOVA1 was underexpressed in FD versus control-induced pluripotent stem cell (iPSC) derived neural crest precursors [Lee et al., 2009]. Finally, a recent study showed that FD affects genes important for early developmental stages of the nervous system using neuroblastoma cell lines [Cohen-Kupiec et al., 2011]. Nevertheless, the specific means by which aberrant IKBKAP mRNA splicing causes the disease producing developmental and degenerative neuronal changes in FD neurons is still unclear. However, the plant cytokinin kinetin has been found to be a powerful agent that corrects IKBKAP mRNA splicing defects [Boone et al., 2010; Hims et al., 2007; Keren et al., 2010; Lee et al., 2009; Slaugenhaupt et al., 2004] and was effective when administered in transgenic mouse model [Shetty et al., 2011] and FD patients [Axelrod et al., 2011], which would make it a potential therapeutic agent for the treatment of FD and other disorders involving missplicing of mRNAs.

To better understand the cascade of events mediated by the c.2204+6T>C mutation, we used human olfactory ecto-mesenchymal stem cells (hOE-MSCs) from FD patients or from control individuals as an experimental model. This allowed us to modulate the rate of IKBKAP exon 20 skipping in vitro by varying culture conditions to produce spheres (with epidermal growth factor (EGF), and basic fibroblast growth factor (bFGF)) or to stimulate neuroglial differentiation (with a “rafnshh” cocktail including all-trans retinoic acid, forskolin, and sonic hedgehog) [Boone et al., 2010]. In this study, we performed the comparative transcriptome analysis between spheres and rafnshh-treated hOE-MSCs and also investigated the effect of kinetin at the genome-wide level.

Materials & Methods

Purification of hOE-MSCs

Human nasal mucosae were obtained by biopsying five FD patients (four females and one male aged 10–16 years) at the Dysautonomia Treatment and Evaluation Center, New York. Biopsies from five healthy controls (four females and one male, aged 10–34 years) were collected by the ENT Department in Marseille (University Hôpital Nord, France). Samples were obtained under a protocol approved by the local ethical committees in New York and Marseille. Biopsies were harvested as previously described [Boone et al., 2010] to obtain an olfactory cell culture of hOE-MSCs. Cells are routinely cultivated with DMEM/HAM'S F12 containing 10% FBS at 37°C in the presence of 5% CO2. Kinetin solution (1 mg/ml, Sigma-Aldrich, St. Louis, MO) was diluted in DMEM/HAM'S F12 at 100 µM concentration for dose-effect experiments, and at 80 µM in experiments of consecutive addition and washout of kinetin. For transcriptome analysis, four of the five control and FD hOE-MSCs have been used.

Generation of Spheres and Induction of Cell Differentiation

Multipotent spheres were obtained after 1 week of culture with EGF and bFGF as previously described [Boone et al., 2010]. For cell differentiation, hOE-MSCs were treated with the rafnshh cocktail consisting in 1% insulin-transferrin-selenium (ITS), 1 µM all-trans retinoic acid (Sigma-Aldrich), 5 µM Forskolin (R&D Systems, Minneapolis, MN), 15 nM Sonic hedgehog (R&D Systems), 1% B27 supplement (a serum substitute), and 0.5% N2 supplement (enhancing the growth and survival of neuronal cells) for 7 days without changing the medium.

RNA Isolation

Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Hilden, Germany) with DNAse treatment on the column following the manufacturer's guidelines. RNA concentration was determined using a nanodrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE). RNA integrity was assessed on an Agilent 2100 Bioanalyzer (Palo Alto, CA). All samples exhibited RIN>9.

End-Point Reverse Transcription-Polymerase Chain Reaction Analysis

Total RNA was subjected to reverse transcription (RT) using the High-Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA). End-point polymerase chain reaction (PCR) analysis was performed using the Go-Taq® DNA polymerase system (Promega, Madison, WI) and IKBKAP-specific primers (hIKBKAP 17-18F and hIKBKAP 22R; see Boone et al. [2010]). PCR products were separated on a 1.7% agarose gel by electrophoresis in 1X TBE buffer (Tris 0.89 M, boric acid 0.89 M, and EDTA 0.02 M). DNA was visualized under UV light after ethidium bromide incorporation and documented using BioVision Camera.

Real-Time PCR Assay

The PCR reactions were performed in duplicate in a final volume of 25 µl, including 300-nM primers, 200-nM TaqMan® probe, 12.5 µl of TaqMan® universal PCR master mix (Applied Biosystems) and 25–50 ng of cDNA in a AB Prism 7900 HT thermocycler with 50 cycles and the protocol recommended by the manufacturer. Primers hELP1 ex19F, hELP1 ex20-21R, and probe P-WTELP1 ex20R were used for detection of IKBKAP transcripts containing exon 20, while primers hELP1 ex19-21F, hELP1 ex21-22R, and probe P-MUELP1 ex21F were used for detection of IKBKAP transcripts skipping exon 20 [Boone et al., 2010]. To determine the level of expression of candidate genes dysregulated genes in FD, the following primer/TaqMan probe assays were obtained from Applied Biosystems: Hs_00176719m1 (LYN), Hs_01103338m1 (SNCA), Hs_01374916m1 (MAP1LC3C), Hs_00359592m1 (NOVA1), Hs_01120488m1 (SPON1), Hs_00216077m1 (LUC7L), Hs_00214302m1 (ZNF280D), and Hs00296608_m1 (WDR59) was used as a reference gene to normalize the data. Results were calculated using the 2(−ΔΔCT) method [Livak and Schmittgen, 2001].

Preparation of Samples and Microarray Assay

Sample amplification, labeling, and hybridization essentially followed the one-color microarray-based gene expression analysis (low input quick amp labeling) protocol (version 6.5, May 2010) recommended by Agilent Technologies. In brief, 500 ng of each total RNA sample was reverse transcribed into cDNA using oligo dT-T7 promoter primer. Labeled cRNA was synthesized from the cDNA. The reaction was performed in a solution containing dNTP mix, cyanine 3-dCTP, and T7 RNA Polymerase, and incubated at 40°C for 2 hr. Hybridization was performed into whole human genome microarray slides (4 × 44K G4112F, Agilent Technologies, Santa Clara, CA) containing 45,220 oligonucleotide probes at 65°C for 17 hr. Hybridized microarray slides were then washed according to the manufacturer's instructions and scanned using an Agilent DNA Microarray Scanner, using the Agilent Feature Extraction Software (Agilent Technologies). The microarray data are available from the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) under the series accession number GSE27915.

Microarray Data Analysis

Quantification files derived from the Agilent Feature Extraction Software were analyzed using the AgiND package (http://tagc.univ-mrs.fr/AgiND). We also used the AgiND R package for quality control and normalization. Quantile methods and a background correction were used for data normalization.

Statistical Analysis

For each comparison (“spheres” vs. “rafnshh”, “controls” vs. “FD”, “control spheres” vs. “FD spheres”, and “FD rafnshh” vs. “FD rafnshh treated with kinetin”), measurement of differential gene expression was obtained using the Multiexperiment viewer (MEV) program. Significant Analysis of Microarray (SAM version 1.13; Standford University) and Student's t-test were applied to determine fold changes (FC) and P-values (P), respectively. The data were analyzed using a two-class unpaired response type, which compared control versus FD samples as well as untreated versus treated FD samples with kinetin. To construct dendrograms, average linkage approximate hierarchical clustering of genes was performed using Pearson correlation (using Cluster [Eisen et al., 1998]) and visualized under Treeview software (http://jtreeview.sourceforge.net/). For each comparison of samples, the statistically relevant signaling pathways, corresponding to the differentially expressed genes, were identified using DAVID (Database for Annotation, Visualization, and Integrated Discovery; http://david.abcc.ncifcrf.gov/) [Huang et al., 2009] with high classification stringency, P < 0.05 and FDR < 20%.

Results

IKBKAP Splice Variants Ratio is Affected by Culture Conditions and Kinetin in FD hOE-MSCs

To observe the variation in IKBKAP mRNA alternative splicing, four control and four FD hOE-MSC cultures were either induced to form spheres or treated with the rafnshh cocktail. FD rafnshh-treated hOE-MSCs were also incubated with 100 µM kinetin for 48 hr. A semi-quantitative RT-PCR analysis confirmed that control hOE-MSCs expressed exclusively the WT IKBKAP mRNA transcript while FD hOE-MSCs expressed both the WT and the MU transcripts (Fig. 1A). In contrast, RT-qPCR analysis on the FD samples revealed a reduced WT/MU transcript expression ratio in rafnshh compared to sphere conditions, which was reversed with kinetin treatment (Fig. 1B). These results are consistent with the increased WT IKBKAP transcripts observed in spheres compared to adherent hOE-MSCs from our previous study [Boone et al., 2010].

Figure 1.

Expression profile of IKBKAP exon 20 alternative splicing in control and FD hOE-MSCs under defined culture conditions. A: Agarose gel electrophoresis of semi quantitative RT-PCR products obtained from four control and four FD hOE-MSCs cultivated in sphere and kinetin treated/untreated differentiation (rafnshh) conditions. IKBKAP transcripts are identified as WT for the correct transcript and MU for the exon 20-skipped transcript. B: Relative RT-qPCR was performed using cDNAs from the same samples of the three conditions (P < 0.01, *P < 0.05). Ct mean values for all samples from each condition were used and normalized with Ct mean values of WDR59.

Microarray Analysis Revealed Differential Transcriptional Expression of IKBKAP and Genes Implicated in Nervous System Function

The 16 RNA samples obtained after treating four control and four FD hOE-MSCs with either EGF and bFGF, or the rafnshh cocktail, were used to characterize the FD transcriptional signature. To confirm the strong impact of culture conditions on gene expression, we compared the control and FD sphere samples to the rafnshh-treated samples (without kinetin). After conducting a significant analysis of microarray (SAM) analysis, we visualized as a heatmap that more than 3,000 transcripts are differentially expressed (false discovery rate, FDR = 0) between spheres and neuroglial progenitors (Supp. Fig. S1). Of these genes, we analyzed only those with a more than 10-fold change (FC) superior and grouped them under five types of biological processes: nervous system development, cell adhesion, WNT/Shh signaling pathway, proteolysis, and retinoic acid activity (Supp. Table S1). All of the processes appear to be related to the factors added in culture media. Indeed, genes that show the greatest fold changes are involved in retinoic acid activity (RARRES1, DHRS3, RARRES2, RARB) and the WNT/Shh signaling pathway (SFRP4, CP, WNT11). In general, genes related to the nervous system are more highly expressed in spheres in comparison to the differentiated samples. In addition, many genes involved in proteolysis were upregulated in spheres samples (MMP1, MMP10, ADAMST14, MME, PRSS35, and ADAMST8).

Using the SAM analysis, we next compared the FD signature between control and FD samples. We assumed a FDR of 10% and characterized 35 differentially expressed genes with a FC>2 (Fig. 2). Although most of the genes were downregulated in FD, IKBKAP appears the second most discriminant marker between control and FD hOE-MSCs. Importantly, 10 differentially expressed genes encode proteins playing important role in neural cells: CD40, FXYD1, GPR37, LYN, NRG1, PACSIN1, RUNX3, SCN2B, SFRP2, SNCA [Aubert et al., 2002; Burré et al., 2010; Deng et al., 2007; Gibb et al., 2011; Hossain et al., 2010; Kramer et al., 2006; Lopez-Santiago et al., 2006; Marazziti et al., 2007; Newbern and Birchmeier, 2010; Perez-Otano et al., 2006; Tan et al., 2002]. When analyzing the gene ontology (GO) of the dysregulated genes in FD (P < 0.01 and FC>2, Supp. Table S2), the pathways with the most significant differential expression correspond to regulation of nervous system development and synaptic vesicle transport (Table 1).

Figure 2.

Heatmap of gene expression changes in control versus FD hOE-MSCs. Heatmap representation of overexpressed (red) and underexpressed (green) genes in four controls and four FD OE-MSCs in different culture conditions named as “SPHERES” and “RAFHSHH”. Normalized signal intensities were treated with the SAM software to highlight the most differentially expressed genes, with a FDR set at 10%. The color scale bar indicates Log2 ratio of intensities. Genes related to nervous system development are indicated in blue.

Table 1. Top Biological Process Gene Ontology (GO) Terms Overrepresented by Dysregulated Genes
Control versus FD cells:
IDTermCountP ValueFDR
FUNCTIONAL GROUP 1 ENRICHMENT SCORE: 2.35
ID:0051960Regulation of nervous system development51.0E-31.6
ID:0050767Regulation of neurogenesis46.9E-310
ID:0060284Regulation of cell development41.2E-217
FUNCTIONAL GROUP 2 ENRICHMENT SCORE: 1.46
ID:0048489Synaptic vesicle transport32.7E-34.1
Control versus FD sphere cells:
IDTermCountP ValueFDR
FUNCTIONAL GROUP 1 ENRICHMENT SCORE: 1.32
ID:0051960Regulation of nervous system development51.3E-218
FUNCTIONAL GROUP 2 ENRICHMENT SCORE: 1.31
ID:0007268Synaptic transmission61.3E-219

NOVA1 is Differentially Expressed in FD Versus Control Sphere-Derived hOE-MSCs

As previously shown, cells that have been induced to form spheres express a higher amount of IKBKAP WT transcript. Therefore, we were interested to identify genes that may be associated with this alternative splicing profile. We were surprised to find that spheres upregulated a significant number of genes related to nervous system development and synaptic transmission (Table 1). Detailed information about gene expression in spheres is supported in Supp. Table S3. Among nervous system-related genes, we identified genes such as SNCA that exhibited a 10-fold downregulation in FD. In addition to finding gene expression alterations for nervous system development in spheres, we also identified NOVA1 (neuro-oncological ventral antigen 1), encoding a neuron-specific RNA-binding protein [Jelen et al., 2007], as an upregulated gene in FD sphere hOE-MSCs. These results suggest that sphere-forming cells provide an FD-relevant signature even at an early undifferentiated state. Moreover, these results suggest that NOVA1 activity may be involved in the improvement of IKBKAP exon 20 inclusion in FD spheres.

Comparative Transcriptome Analysis Identify Convergent Pathways Affected in FD

Four previous studies from other laboratories have generated a wealth of data on the transcriptome variations in either FD or IKBKAP knockdown samples [Cheishvili et al., 2007; Close et al., 2006; Cohen-Kupiec et al., 2011; Lee et al., 2009]. Therefore, we procured the raw data from all studies and reanalyzed the data in search for the common candidates that may be involved in FD physiopathology. For each study, we identified genes that are differentially expressed between control and FD/IKBKAP knockdown samples with a FC>1.5 and a P-value <0.05, and cross-compared the lists of candidate genes for each study (Fig. 3). We did not find genes that were consistently dysregulated in all studies. Among the 3,228 candidate genes differentially expressed in at least one of the five studies, including our own, we found 10 genes shared by three different studies with the same kind of dysregulation. Seven genes were underexpressed in FD (CXCR7, PFKFB3, IKBKAP, SEMA5A, SEPT3, SNAI2, and TNC), and three genes were overexpressed (ARCHGAP28, MAN1A and XK) (Supp. Table S4). We also analyzed the GO of the 175 genes shared by at least two studies (Supp. Table S4). Nine processes emerged as significantly affected in FD: regulation of cell motion, guanyl ribonucleotide binding, contractile fiber part, neuron differentiation, regulation of protein kinase activity, regulation of apoptosis, cadmium ion binding, muscle tissue development, and osteoblast differentiation (Supp. Table S5).

Figure 3.

Common genes differentially expressed in FD. Intersection between the current study and the lists of four previous studies for the genes differentially expressed between control and FD/IKBKAP knockdown samples (FC>1.5, P < 0.05). The genes dysregulated in three different studies are listed and preceded by either a “↘” for underexpression or a “↗” for overexpression in FD samples. Capital letters define each study considered with the following order: A: Cheishvili et al. 2007; B: Current study; C: Lee et al. 2009; D: Close et al. 2006; E: Cohen-Kupiec et al. 2011.

Kinetin Modulated the Expression of Genes Involved in mRNA Splicing

Our microarray data were next examined for evidence of genes targeted by kinetin. Indeed, this plant cytokinin reproducibly induces rapid increase of IKBKAP transcripts with exon 20 inclusion through unknown mechanisms. To further understand the mechanism of kinetin in IKBKAP mRNA alternative splicing, we compared FD rafnshh-untreated hOE-MSCs versus FD rafnshh-treated hOE-MSCs with 100 µM of kinetin for 48 hr. Supp. Table S6 displays the list of genes affected by kinetin action in FD rafnshh hOE-MSCs. Interestingly, a majority of candidate genes were downregulated in response to kinetin. In addition to confirming an increased expression of IKBKAP in FD hOE-MSCs, we observed cellular responses that are consistent with predicted mechanisms of kinetin action. Indeed, our analysis detected differences in expression of genes involved in mRNA splicing: LUC7L, SNRPA, WDR70 (Supp. Table S6). Of particular interest, SNRPA and LUC7L are both related to the U1 snRNP splicing complex required for 5′ss selection. SNRPA, downregulated by 1.7-fold in response to kinetin in FD rafnshh-treated hOE-MSCs, encodes the U1 snRNP core protein U1A [Nelissen et al., 1991], LUC7L, upregulated by 2-fold, encodes a putative RNA-binding protein similar to the yeast Luc7p subunit of the U1 snRNP [Fortes et al., 1999; Tufarelli et al., 2001].

RT-qPCR Analysis of Candidate Genes Validates Microarray Data

To further confirm gene expression data from microarray analysis, we used relative qPCR to verify the differential expression of a subset of the identified genes based on statistical significance, as well as the biological relevance for each comparison. WDR59 was selected as the reference gene since it exhibited relatively stable expression in our microarray data. Using IKBKAP expression as a positive control for each experiment, we confirmed the differential expression of LYN and SNCA between control and FD cells (Fig. 4A), MAP1LC3C, NOVA1, SNCA, SPON1 between control and FD sphere-derived cells (Fig. 4B), and LUC7L between FD cells with or without kinetin treatment (Fig. 4C).

Figure 4.

Validation of microarray candidates by RT-qPCR. RT-qPCR using total RNAs extracted from four controls and four FD hOE-MSCs. Histograms represent the mean value of (A) IKBKAP, LYN, SNCA, (B) MAP1LC3C, NOVA1, SPON1, and (C) LUC7L transcript expression level, relative to WDR59 as a reference gene in control (gray) and FD samples (black). For dysregulated genes between control and FD hOE-MSCs, we pooled values of spheres and differentiated cells for each group. Error bars denote standard errors. (*P < 0.05; **P < 0.01, ***P < 0.001 using two-tailed Student's test).

ZNF280D is a Potential Sequence-Specific Target of Kinetin in FD hOE-MSCs

Among the list of genes whose expression is downregulated after kinetin treatment in FD OE-MSCs, we noted the presence of ZNF280D (Supp. Table S6). ZNF280D belongs to a unique group of 12 genes in the entire genome that contains an alternative 5′ss in one of its exons (exon 16) that is identical to the FD 5′ss (CAAguaagc) [Ibrahim et al., 2007]. Therefore, we hypothesized that kinetin may favor the splicing of introns flanked by the CAAguaagc 5′ss motif, resulting in a modification in the ratio of alternative 5′ss choice for ZNF280D exon 16 (Supp. Fig. S2). Since the use of the 5′ss identical to the FD IKBKAP intron 20 5′ss is also expected to induce a premature stop codon in ZNF280D exon 17 and make it a target for nonsense–mediated mRNA decay (NMD), this may explain why the total amount of ZNF280D transcripts is reduced in FD hOE-MSCs after kinetin treatment.

IKBKAP, LUC7L, and ZNF280D are Sensitive to Kinetin Treatment in hOE-MSCs

To corroborate the expression levels of LUC7L and ZNF280D detected in our microarray hybridization after kinetin treatment, we exposed adherent hOE-MSCs to increasing concentrations of kinetin (25 to 200 µM) over a 48-hr time course and determined the expression changes of these genes relative qPCR (Fig. 5A). Although a dose-dependent action of kinetin on increasing IKBKAP WT transcripts was only observed in FD samples (Fig. 5A, higher panel), LUC7L expression increased in both control and FD samples (Fig. 5A, middle panel). We also observed that increasing kinetin concentration leads to a dose-dependent inhibition of ZNF280D mRNA expression, supporting our hypothesis of sequence-specific targeting by kinetin (Fig. 5A, lower panel). To validate the action of kinetin on the expression of LUC7L and ZNF280D, we exposed hOE-MSCs to two consecutive rounds of 24-hr treatment with 80-µM kinetin followed by a 24-hr wash-out. At each 24-hr time point with kinetin treatment, we analyzed gene expression by RT-qPCR and observed that WT IKBKAP transcripts and LUC7L expression increased while MU IKBKAP transcripts and ZNF280D expression decreased (Fig. 5B). This variation in expression returned to basal levels during washout period. As expected, in control cells, kinetin treatment modulated expression of LUC7L and ZNF280D without acting on IKBKAP WT isoforms. These results strongly suggest that kinetin may increase the efficiency of 5′ss recognition in the FD context through the recruitment of U1 snRNP.

Figure 5.

Changes in gene expression after different exposures of hOE-MSCs to kinetin. A: Control and FD-adherent hOE-MSCs were incubated for 48 hr with different concentration of kinetin (25, 50, 100, and 200 µM) for dose effect experiment. B: Cells were exposed to 80 µM kinetin for 24 hr (K80), followed by the removal of the drug for another 24 hr (i.e., “W” for washout). Two rounds of drug addition/removal were performed and RNA was extracted each time after 24 hr for each condition. Total RNAs were reverse transcribed and levels of expression of IKBKAP alternative transcripts as well as LUC7L and ZNF280D expressions were analyzed by RT-qPCR. Each gene was normalized using WDR59 as a reference gene.

Genes Involved in mRNA Splicing Display an IKBKAP-Like Pattern of Expression

When analyzing gene expression data, it is informative to include a clustering algorithm to find groups of genes that behave similarly over a number of experiments [Eisen et al., 1998; Slonim, 2002]. To better understand the FD physiopathology and since IKBKAP represents the best biomarker to discriminate between control and FD samples, as well as samples with or without kinetin treatment, we wanted to identify genes with expression pattern similar to that of IKBKAP. We used hierarchical clustering to create dendrograms that capture the degree of similarity for each gene. An illustrative set of selected genes is shown in Supp. Figure S3A. Next, we looked for the cluster of genes that include IKBKAP (Supp. Fig. S3B). Significantly, among the few genes in the same cluster as IKBKAP, we identified DDX42, which encodes SF3b125, an RNA helicase involved in spliceosome assembly [Will et al., 2002], and NHP2L1 (nonhistone chromosome protein 2-like 1), which binds the 5′-stem-loop of U4 snRNA and may play a role in late stage spliceosome assembly [Nottrott et al., 1999].

Discussion

Genome-wide expression studies have been widely used in an effort to identify signatures that can define pathologies. In this study, we proposed to use properties of hOE-MSCs to perform a transcriptome analysis of FD. These cells have been used as a nervous system replacement cells in mice [Nivet et al., 2011] and demonstrate a potential to differentiate into nervous cell types [Delorme et al., 2010; Murrell et al., 2005]. Importantly, this novel patient-derived cellular model has allowed us to modulate IKBKAP alternative splicing by exposing cells to different culture conditions [Boone et al., 2010]. In this study, we discuss the opportunity to use hOE-MSCs derived from FD patients to analyze the transcriptional differences due to the alteration or improvement of IKBKAP mRNA alternative splicing. We focused on identifying gene expression differences in FD using two different cellular models to reproduce neuronal cells in early development (spheres), and neuroglial progenitors in later developmental stages using the “rafnshh” treatment. Retinoic acid (RA) and Sonic hedgehog (Shh) are known to regulate neuronal specification and differentiation during development [Probst et al., 2011]. Both RA and Shh induced expression of a set of genes and proteins that define peripheral nervous system sensory neurons in murine mesenchymal stem cells [Kondo et al., 2005]. These factors were also shown to stimulate the expression of motoneuronal transcription factors in parallel to neurite formation on hOE-MSCs [Zhang et al., 2006].

Previous microarray studies of FD were unable to discriminate IKBKAP expression between FD and control cells [Boone et al., 2010; Cheishvili et al., 2007; Keren et al., 2010; Lee et al., 2009]. However, in our analysis, we detected an IKBKAP signal above background level in both control and FD patient samples. In addition, we found that IKBKAP was the best marker for FD since this gene was initially underexpressed in FD cells but then showed even higher expression after kinetin treatment. These results increased confidence in interpreting our microarray data.

In accordance with previous microarray studies [Boone et al., 2010; Cheishvili et al., 2007; Lee et al., 2009], the FD transcriptional signature is characterized by a general decrease in transcriptional expression that might reflect a defect in transcription elongation due to impaired Elongator activity [Close et al., 2006]. Moreover, gene expression profiling studies have shown that most of gene expression differences between control and FD samples are involved in nervous system development, which correlates with FD physiopathology and findings from other cellular systems [Chen et al., 2009b; Cohen-Kupiec et al., 2011; Lee et al., 2009].

When we explored the transcriptome of spheres, we hypothesized that such cell populations maintained at a higher undifferentiated state would likely reveal discriminating markers of the “stem” state. Interestingly, rather than displaying a profile that is more consistent with stem cells, we identified nervous system-related genes in spheres. In fact, spheres contain a heterogeneous mixture of cells and progenitors whose identity and proportion still need to be characterized. However, this discrepancy with our hypothesis suggests that spheres can be a relevant model for predicting FD alteration, as also proposed for other diseases such as schizophrenia and Parkinson's disease [Cook et al., 2011; Matigian et al., 2010].

As in studies for all rare diseases, the sample size is unavoidably small, which may lead to moderate differences in gene expression variations. In addition, previous investigations, at the genome-wide level, aiming to identify transcriptional defects associated to FD used different cell types. Some investigators treated HeLa or neuroblastoma cells with siRNAs, while others generated FD iPSCs, hOE-MSCs, or analyzed FD brains [Boone et al., 2010; Cheishvili et al., 2007; Close et al., 2006; Cohen-Kupiec et al., 2011; Lee et al., 2009]. It was thus expected from such heterogeneity in cell types, genetic background, and methodologies that important discrepancies would characterize those studies and ours. Despite such limitations, we were able to identify a common set of genes in our microarray data and data from four previous studies (Supp. Table S4), that could contribute to the FD disease process [Cheishvili et al., 2007; Close et al., 2006; Cohen-Kupiec et al., 2011; Lee et al., 2009]. Among the dysregulated genes shared by at least two studies, several are related to nervous system development and characterize common alterations of neuronal cells. Notable downregulated genes include: SEMA5A and SEMA3C, which encode members of the semaphorin family, involved in axonal guidance during neural development [Hernandez-Montiel et al., 2008; Hilario et al., 2009]; NRCAM, which encodes an adhesion molecule acting as a co-receptor for SEMA3B and 3F [Falk et al., 2005]; ALCAM, involved in axonal guidance [Buhusi et al., 2009]; RELN, which regulates the migration of neuroblasts [Frotscher, 2010]; FEZ1, which promotes neurite elongation [Maturana et al., 2010]; and DLX5, which encodes a homeobox transcriptional factor promoting neuronal differentiation [Perera et al., 2004]. Therefore, we can speculate that in FD, the dysregulation of these candidate genes in FD will disrupt the precisely defined waves of migration, differentiation, and navigation of axonal growth cone for synapse formation, which are all essential for the formation of the peripheral nervous system. LYN is one of the genes that was found to be downregulated in our microarray data, as validated by RT-qPCR and IKBKAP knockdown in HeLa cells. LYN encodes a Src family tyrosine kinase that have many roles in the process of oligodendrocyte differentiation [Colognato et al., 2004; Hossain et al., 2010], and dopamine release in the mesolimbic system [Gibb et al., 2011]. Importantly, we highlighted 10 genes, including IKBKAP, whose dysregulation is shared by three independent genome-wide transcriptional studies (Fig. 3). Notably, four of them, CXCR7, SEMA5A, SNAI2, and TNC, are closely related to cell migration [Katafiasz et al., 2011; Nishio et al., 2005; Sadanandam et al., 2010; Sanchez-Alcaniz et al., 2011]. Since several studies previously suggested a contribution of altered migration pathways in the physiopathology of FD [Close et al., 2006; Cohen-Kupiec et al., 2011; Creppe et al., 2009; Johansen et al., 2008; Lee et al., 2009; Naumanen et al., 2008], future experiments will aim to investigate the role of those four genes in functional migration assays using hOE-MSCs.

Understanding the mechanisms underlying regulation of tissue-specific gene expression remains a challenging problem. So far, the only candidate gene that may explain increased aberrant splicing of IKBKAP mRNA in the nervous system is NOVA1, identified by Lee et al. as a downregulated gene in FD iPSC-derived neural crest precursors [Lee et al., 2009]. NOVA1 is a tissue-specific factor regulating alternative splicing in the brain of a large number of genes that function primarily at synapses [Ule et al., 2005]. Thus, it has been suggested that this splicing factor may participate in the balance of neuronal excitation and inhibition, and is necessary for proper synaptic development and function [Ruggiu et al., 2009]. In addition, one of the roles of NOVA proteins may be to enable neurons to adapt their synaptic inhibition in response to neuronal activity [Jelen et al., 2010]. In our system, we confirmed a NOVA1 dysregulation in FD hOE-MSCs-derived spheres supporting this gene's potentially critical role in modulating IKBKAP mRNA alternative splicing. Therefore, we can speculate that NOVA1 may not only act as a master candidate to regulate IKBKAP pre-mRNA splicing in FD, but also the regulation of many other targets involved in progression of this neurodegenerative disease. To understand the precise role of NOVA1 in mRNA splicing, further experiments modulating its expression in human control and FD cells will be necessary. In addition, it is clear from the initial analysis of postmortem tissues that most constitutive IKBKAP exon 20 skipping occurs in tissues representing a mixture of cell types, and not just neurons [Cuajungco et al., 2003]. Thus, the ability to derive pure cultures of neurons or glial cells from hOE-MSCs will be of great benefit to determine the cell type predominantly affected during FD development.

We report for the first time a genome-wide gene expression analysis of IKBKAP mRNA splicing in response to kinetin, a plant cytokinin. Surprisingly, although kinetin helps to increase WT IKBKAP transcript level, the compound does not seem to influence the expression of a large proportion of genes. This specificity in IKBKAP mRNA splicing is an encouraging result in light of its potential clinical use [Axelrod et al., 2011]. Although the mechanism by which kinetin improves exon inclusion is still unknown, a previous study has suggested that kinetin may target specific sequences within the 5′ss [Hims et al., 2007]. In this context, our finding that genes encoding a core component and a putative subunit of U1 snRNP, SNRPA and LUC7L, are regulated by kinetin, supports the hypothesis that this compound can induce the recruitment of splicing factors to reinforce 5′ss recognition. In addition, we demonstrated a consistent decrease of ZNF280D expression, which shares with IKBKAP an identical 5′ss motif that potentiates the presence of a premature stop codon most likely targeted by the NMD machinery. Therefore, we propose kinetin as a new sequence-specific agent that can affect U1 snRNP-mediated 5′ss recognition. Further experiments considering the 11 other alternatively spliced mRNAs sharing a 5′ss identical to the one bordering IKBKAP exon 20 will also be of interest to understand the mechanism underlying kinetin activity on mRNA splicing.

In conclusion, this study provides important clues to the physiopathology of FD. We identified several genes involved in nervous system development and differentiation that could represent the molecular-altered signature unique to the abnormal FD neuronal function. Knowledge of the commonly expressed genes from different cell types should facilitate their further characterization and functional studies. Our results also identified kinetin as a compound that affects genes involved in mRNA maturation and shed new light on its mechanism of action and its potential for therapeutic use.

Acknowledgements

We wish to thank the patients and their families for their contribution to this study. We also thank Jeanne Hsu for critical reading of the manuscript.

Ancillary