The long non-coding RNA Dali is an epigenetic regulatorof neural differentiation

Conserved Dali genomic organisation and transcription

Full-length mouse Dali is approximately 500 nt (2.6 kb) longer thana previously identified AK034039 cDNA cloned from the telencephalon(Figure 1—figure supplement 1A).Its locus, downstream of the Pou3f3 gene, contains mammalianconserved sequence both just upstream of its transcriptional start site, whichpresumably contributes to this locus’ promoter, and within its transcribedsequence. A positionally equivalent and sequence-similar human DALI(∼3.7 kb) transcript was identified by RT-PCR and RACE in human foetal brain(Figure 1B; Figure 1—figure supplement 1B,C). Transcriptionalevidence also exists for the orthologous locus in rat embryonic, as well as heart andkidney, samples (data not shown).

Dali is a chromatin-associated transcript that is co-expressedwith Pou3f3 in neural cell lineages

ENCODE data indicate that both mouse and human Dali loci have theproperties of a weak (or poised) enhancer in both brain and kidney tissues (Figure 1—figure supplement 1D,E).Consistent with this, Dali was most highly expressed in the adultbrain and kidney, two of the three tissues displaying highest Pou3f3expression, when profiled across a panel of adult mouse organs (Figure 1—figure supplement 1F,G). In adult mouse (P56),Dali and Pou3f3 were expressed in all threeregions of adult neurogenesis, the sub-ventricular zone (SVZ), olfactory bulb (OB),and dentate gyrus (DG) (Figure 1—figuresupplement 1I) (Reviewed in Ming andSong, 2011). Dali was also co-expressed withPou3f3 temporally and spatially in the developing mouse embryonicbrain (Figure 1C,D). Both transcripts wereup-regulated with the first appearance of cortical neurons (E10.5), and increased inexpression further as the ratio between neurons and progenitors grew (Figure 1C,D). Furthermore, bothDali and Pou3f3 transcripts were undetectable inself-renewing mouse E14 embryonic stem (ES) cells, but after 3 days of retinoic acid(RA)-induced differentiation, a stage corresponding to the cell cycle exit ofneuronal progenitors and their differentiation into neurons, these transcripts wererapidly up-regulated, their levels subsequently peaking at days 7(Pou3f3) and 8 (Dali) (Figure 1E,F).

Mouse neuroblastoma N2A cells, which are frequently used as a neuronalprogenitor-like cell type and an in vitro model of neuronal differentiation (Tremblay et al., 2010), express bothDali (at a population-average level of 2 copies per cell (Figure 1—figure supplement 1K)) andPou3f3. When first detected in neuronal-progenitor-dominatedareas of the developing brain (E10.5), Dali is expressed at a levelat least two orders of magnitude higher than in N2A cells (Figure 1—figure supplement 1H). However, in N2A cellstreated with RA for 72 hr, Dali is up-regulated approximately4.5-fold, similar to the up-regulation observed in embryonic cortical plate (bothdorsal and lateral) between days E10.5 to E18.5 (Figure 1—figure supplement 1H). Therefore, despiteDali expression level differences in N2A cells and the in vivosystem, N2A cells represent an appropriate model system in which to studyDali function. Furthermore, Dali, but not acontrol mRNA (Gapdh), was highly enriched in the nucleus of N2Acells, most abundantly in the chromatin fraction (Figure 1G,H). Taken together, the data suggest that Dalimay be involved in regulating nuclear function during neuronal development,potentially in coordination with Pou3f3.

Dali regulates neural differentiation of N2A cells

We next investigated whether Dali regulates neural differentiationby generating three independent stable Dali knockdown N2A cell lineseach showing approximately 50–70% reduction of Dalitranscript levels and inducing neural differentiation using RA (Figure 2A). Compared to a stable non-targeting control line,fewer differentiated cells of Dali knockdown lines developedneurites. Those that did exhibited shorter neurites, often with multiple shortoutgrowths emanating from the same cell, compared to one or two long neuritesdeveloped by differentiated control cells (Figure2B,C) indicating that Dali is required for normaldifferentiation of N2A cells.

10.7554/eLife.04530.006Figure 2.

Dali plays a role in regulating genes in neuronalcells.

(A) qRT-PCR analysis validates reduced levels ofDali in three clonal Dali knockdowncell lines compared to a control line. Mean values ± s.e., n =3. (B) Reduced neurite outgrowth in RA-differentiatedDali knockdown cells. Cells were imaged using brightfield microscopy. Cells with ≥1 neurites of length greater thantwice the cell body diameter were scored as positive. Average values± s.e., n = 3. 500-600 cells were counted in each case acrossat least three non-overlapping fields. (C) Representativeimages of control and stable Dali knockdown cellsdifferentiated with RA for 72 hr. Scale bar = 200 μm.(D) N2A cells were transfected with either anon-targeting control (scrambled) or a Dali targetingshRNA expression vector (shDali) for 72 hr. Mean values ± s.e., n= 3. (E) Transient Dali knockdowninduces statistically significant changes in the expression of 270 genesin N2A cells (10% FDR) (Supplementary file 2). (F) GeneOntology (GO) categories significantly enriched amongDali regulated genes (5% FDR, hypergeometric test,Benjamini and Hochberg correction; Supplementary file2). (G) Decreased Pou3f3expression upon Dali knockdown. Normalised usingGapdh, shown relative to a non-targeting control (setat 1). Mean values ± s.e., n = 3, one tailed t-Test, unequalvariance. (H) Reduced Pou3f3 levels instable Dali knockdown cells (see panel A).qRT-PCR results were normalised using Gapdh andpresented relative to expression in control cells (set arbitrarily to 1).Mean values ± s.e., n = 3, one tailed t-Test, unequalvariance.

DOI:http://dx.doi.org/10.7554/eLife.04530.006

10.7554/eLife.04530.007Figure 2—figure supplement 1.

Non-coding transcripts in the Pou3f3 locus form a network ofregulatory interactions.

(A) Dali regulates expression of lncRNA AK011913. N2A cellswere transfected with either a non-targeting control (scrambled) or threeindependent Dali targeting shRNA expression vectors. Dali and AK011913levels were measured by RT-qPCR 72 hr post- transfection. Mean values± s.e., n = 3, one tailed t-Test, unequal variance.(B) AK011913 regulates expression of Dali and Pou3f3positively and linc-Brn1a negatively. N2A cells were transfected witheither a non-targeting control (scrambled) or a AK011913 targeting shRNAexpression vector (shRNA). AK011913, Dali, linc-Brn1a and Pou3f3 levelswere measured by RT-qPCR 72 hr post- transfection. Mean values ±s.e., n = 3, one tailed t-Test, unequal variance. (C)RT-qPCR validation of transient Dali knockdown microarray results. Goodagreement between microarrays and RT-qPCR and across independent shRNAconstructs is observed, indicating that the results are unlikely torepresent technical artefacts or off-target effects. Mean values ±s.e., n = 3.

DOI:http://dx.doi.org/10.7554/eLife.04530.007

Dali regulates neural gene expression

To investigate the molecular function of Dali, we performedmicroarray analysis to profile the transcriptome of N2A cells in whichDali transcript levels had been depleted by ∼70% usingtransient transfection of a specific Dali targeting shRNA expressionvector (Figure 2D; shRNA and RT-qPCR oligosequences and positions can be found in Supplementary file 1). Dali knockdownresulted in statistically significant changes in expression levels for 270 genes(False Discovery Rate [FDR] < 10%) compared to a non-targeting control (Supplementary file 2;Figure 2E). 14 of 15 of these genes werealso determined as being differentially expressed, with similar fold changes, usingRT-qPCR and two additional independent shRNA expression constructs targetingDali (Figure 2—figuresupplement 1C). Gene expression changes we observed using microarrays werethus unlikely to be dominated by off-target effects of the shRNA used. Gene Ontology(GO) analysis revealed that Dali-regulated genes were significantlyenriched in cell cycle, DNA repair, cellular response to stimulus, and cellprojection assembly functions (Figure 2F andSupplementary file2; Benjamini-Hochberg corrected p ≤ 0.05). Taken together, theseexpression and loss of function studies suggest that Dali acts as apro-differentiation factor in neural development.

Dali and Pou3f3 share transcriptionaltargets

To investigate whether Dali knockdown affects expression of theadjacent Pou3f3 gene, we reduced its levels by transienttransfection of three different shRNA constructs in N2A cells. After 72 hr, reductionof Dali levels by an average of 60–70% was found to reducePou3f3 transcript levels by approximately 40% (Figure 2G). Three independent stableDali knockdown clones in which Dali levels werereduced by 50–60% (Figure 2A) alsoshowed ∼15–40% lower Pou3f3 levels (Figure 2H). This suggests that theDali transcript positively regulates Pou3f3expression in an RNA-dependent manner. The genome-wide transcriptional response toDali knockdown thus could be explained, in part, by its effect onPou3f3.

Levels of another transcript, AK011913, expressed downstream ofPou3f3 (Figure 1A) werereduced by approximately 55% upon Dali knockdown (Figure 2—figure supplement 1A).Reduction of AK011913 levels by approximately 60% using shRNAsresulted in Dali and Pou3f3 levels decreasing by73% and 82%, respectively (Figure 2—figuresupplement 1B). Linc-Brn1a, a lncRNA upstream of andsharing a bi-directional promoter with Pou3f3, was up-regulated byapproximately 90% upon AK011913 depletion (Figure 2—figure supplement 1B). This is reminiscent ofthe down-regulation of Pou3f3 and up-regulation oflincBrn-1a following knockdown of another lncRNA downstream ofPou3f3, lincBrn-1b (Figure 1A) (Sauvageau et al.,2013). Together with previous reports, our data show the opposingregulatory influences of lncRNAs transcribed up- and downstream ofPou3f3 on its expression. Non-coding transcripts expressed fromthe extended Pou3f3 locus thus contribute to a complex network ofregulatory interactions.

Furthermore, Chromatin Conformation Capture (3C) showed that theDali promoter contacted three regions across thePou3f3 locus (Figure 1A)in ES derived neuronal precursors (Figure1—figure supplement 2) : 1) an enhancer element sequence lyingupstream of Pou3f3 within the linc-Brn1a locus, 2)a region overlapping the 3′ UTR of Pou3f3 and full-lengthAK53590 (which are both regulated by Dali), aswell as parts of TCONS_00000039 and linc-Brn1b,including a differentially methylated region reported to be important in regulatingPou3f3 expression (Mutai etal., 2009), and 3) a region lying within another non-coding locus(TCONS_00000040) (Ramos et al.,2013). Neither Dali nor Pou3f3 appears toplay a role in initiating these DNA looping interactions because these contacts werepresent in E14 ES cells where neither is expressed (Figure 1—figure supplement 2B). Nevertheless, theDali locus appears to contribute to an extended structurally andtranscriptionally complex region centred on the Pou3f3 gene.

To examine to what extent the transcriptional response to Daliknockdown can be explained by its effect on Pou3f3, we reduced thelevel of Pou3f3 transcript in N2A cells by 35% using transienttransfection of a Pou3f3 targeting shRNA vector (Figure 3A) and using microarrays observedstatistically significant expression changes in 1041 genes (FDR <10%; Figure 3B). Dali transcriptlevels do not change upon Pou3f3 depletion (Figure 3A). Genes differentially expressed afterPou3f3 knockdown were enriched in categories related to celldivision and cell cycle (Figure 3C). Theintersection between the sets of genes differentially expressed inDali or in Pou3f3 knockdown cells was 6.2-foldgreater than expected by chance (p-value < 2.2 × 10⁻¹⁶),and represented 31% of all genes differentially expressed in Daliknockdown cells (Figure 3D). Approximatelyequal numbers of genes shared between the two datasets were down- (43 genes) orup-regulated (41 genes) in both experiments (Supplementary file 3). A strong correlation was observedbetween the fold-change values of differentially expressed genes inDali and Pou3f3 knockdown experiments (R =0.74; Figure 3E). Genes that weresignificantly differentially expressed only when Dali was depletedwere enriched in chromatin assembly and MAPKKK signalling functions, whilst genesthat were differentially expressed only when Pou3f3 transcripts weredepleted were preferentially involved in dendrite development and axon guidance(Figure 3F). Cell cycle, DNA repair, andcellular response to stimulus genes were regulated by Dali in bothPou3f3-dependent and -independent manners. We conclude thatDali and Pou3f3 interact, either genetically ormolecularly, to regulate a subset of common targets involved in neuraldifferentiation, and that Dali also likely possessesPou3f3-independent transcriptional regulatory functions.

10.7554/eLife.04530.008Figure 3.

Dali regulates transcription in bothPou3f3-dependent and -independent manners.

(A) N2A cells were transfected with either a non-targetingcontrol (scrambled) or a Pou3f3 targeting shRNA expressionvector (knockdown). Pou3f3 and Dali levelswere measured by qRT-PCR 72 hr post- transfection. Mean values ± s.e.,n = 3. (B) Pou3f3 knockdown resulted instatistically significant changes in the expression of 1041 genes in N2Acells ((10% FDR, Supplementary file 3). (C) GO-analysis of genesdifferentially expressed upon Pou3f3 analysis (5% FDR,hypergeometric test, Benjamini and Hochberg correction; Supplementary file3). (D) Intersection of Pou3f3 andDali targets shows a significant (Fisher’s exacttest) overlap approximately 6.2 times as large as expected by chance alone.(E) Target genes common between Dali andPou3f3 show correlated expression, with the nearly allbeing positively or negatively regulated by both factors (R = 0.74;Supplementary file3). (F) Enrichments of Gene Ontology categories ofPou3f3-dependent or -independent Dalitargets.

DOI:http://dx.doi.org/10.7554/eLife.04530.008

Dali regulates gene expression programmes during neuraldifferentiation of N2A cells

To further investigate the role of Dali in neuronal differentiationwe profiled the transcriptomes of proliferating or RA differentiated control andDali stable knockdown N2A cell lines. In proliferating cells, 733genes were differentially expressed between Dali knockdown andcontrol cells (Figure 4A), including manygenes with functions related to neuronal differentiation, apoptosis, neuronalfunction (Figure 4B). RA-mediated neuronaldifferentiation induced expression changes in 958 genes in control cells and 1016genes in Dali knockdown cells (Figure 4—figure supplement 1A,B). Based on GO category annotations,differentiation of control or Dali knockdown cells was broadlysimilar, and was associated with altered expression of cell cycle, celldifferentiation, energy metabolism, and neuron projection (Figure 4—figure supplement 1C,D). However, 804 geneswere differentially expressed between terminally differentiated control andDali knockdown cells (Figure4C), of which 376 genes (46.8%) also differed in expression betweenDali knockdown and control cells prior to their differentiation(Figure 4E). The 428 genes that weresignificantly altered in expression only between stable Dali andcontrol differentiated cells were enriched in functional categories relating tosterol biosynthesis, energy metabolism, cell cycle, response to chemical stimulus,cell cycle, adhesion and small GTPase signalling (Figure 4D). All 11 (of 34 known) sterol biosynthesis genes weredown-regulated in Dali knockdown cells. This observation isconsistent with the impaired neurite outgrowth of stable Daliknockdown cells because neuritogenesis and neurite outgrowth critically rely onmembrane biosynthesis, and therefore, on expression of sterol biosynthesis genes(Paoletti et al., 2011).

10.7554/eLife.04530.009Figure 4.

Gene expression analysis of stable Dali knockdowncells.

(A) Stable Dali knockdown resulted instatistically significant changes in the expression of 747 genes in N2Acells (1.3-fold, 5% FDR, Supplementary file 4). 332 genes wereup-regulated, 415 down-regulated. (B) GO-analysis of genesdifferentially expressed upon stable Dali depletion (5%FDR, hypergeometric test, Benjamini and Hochberg correction).(C) Stable Dali knockdown and controlcells were differentiated with retinoic acid for 72 hr. 825 genes weredifferentially expressed between differentiated knockdown and controllines ((≥1.3-fold, 5% FDR, Supplementary file 4). 436 genes wereup-regulated, 389 down-regulated. (D) GO-analysis of genesdifferentially expressed only between differentiated stableDali knockdown and control cells (5% FDR,hypergeometric test, Benjamini and Hochberg correction). (E)Intersection of gene sets differentially expressed between stableDali knockdown and control cells prior to(undifferentiated) and after retinoic acid addition (differentiated).(F) GO-analysis of genes responding to retinoic acidtreatment differently between stable Dali knockdown andcontrol cells (5% FDR, hypergeometric test, Benjamini and Hochbergcorrection). ‘Differential responder’ genes were identifiedusing multifactorial analysis of the stable Daliknockdown arrays using limma (Smyth,2004).

DOI:http://dx.doi.org/10.7554/eLife.04530.009

10.7554/eLife.04530.010Figure 4—figure supplement 1.

Transcriptomics.

(A and C) Heatmap displaying expression changesin control (A) and stable Dali knockdown (C)cells treated with RA for 72 hr. (B and D)GO-analysis of genes differentially expressed (≥1.3-fold, 5% FDR)upon RA treatment of control (B) and stable Dali knockdown(D) cells (5% FDR, hypergeometric test, Benjamini andHochberg correction). GO categories significantly enriched among geneschanging ≥1.5-fold are marked with an asterisk (*).

DOI:http://dx.doi.org/10.7554/eLife.04530.010

In addition, several key neuronal differentiation genes, for example Nrcam,Dscam, Dlx1 and Pax3, were differentially expressedbetween Dali knockdown and control cells both prior to and afterdifferentiation. Furthermore, multifactorial analysis of RA-induced gene expressionchanges in control and stable Dali knockdown cells showed that 174genes responded to RA differently depending on the presence or knockdown ofDali (FDR 5%; Supplementary file 4). These were significantly enriched incategories relating to neuronal development (Figure4F), including pro-differentiation factors such as the inhibitor of Wntsignaling Dkk1 (Cajanek et al.,2009) and Wnt receptor Fzd5 (Kemp et al., 2007).

In summary, compared to control cells, stable Dali knockdown cellsexhibit contrasting alterations in gene expression programmes before and afterRA-induced differentiation. In both cases, these programmes are enriched infunctional categories related to neural differentiation and function, consistent withthe proposed role for Dali in neural development.

Dali preferentially binds to active promoters intrans

We next sought to identify and characterise genes that are both bound and regulatedby Dali. To do so, we determined the genome-wide binding profile ofDali in N2A cells using Capture Hybridisation Analysis of RNATargets (CHART)-Seq (Simon et al., 2011;Simon, 2013) (Figure 5—figure supplement 1A–C). We discovered1427 focal Dali-associated regions genome-wide (Figure 5A,B; Supplementary file 5), of which all nine selected loci werevalidated by CHART-qPCR in an independent experiment (Figure 5—figure supplement 1D).

10.7554/eLife.04530.011Figure 5.

CHART-Seq analysis of Dali genomic bindingsites.

(A) Peaks were called against control CHART-seq experimentsand input DNA. We consider only the 1427 peaks common to both comparisons(SupplementaryFile 5). (B) Sequencing of Dalibound DNA reveals focal peaks, including those at the promoter ofAche, E2f2, andHmgb2. (C and D)Dali peaks are broadly distributed across the mousegenome (C) but are particularly enriched in 5′ UTRsand gene promoters (D). Red arrowheads in (C)mark the Dali locus. (E) A third ofDali peaks are situated within 5 kb of a TSS.(F) Dali-bound loci are enriched inactive chromatin marks (H3K4me3, H3K27ac, PolII), DNase Ihypersensitivity regions, enhancers and CpG islands annotations (CGI),and CTCF-bound regions, while being depleted of gene body marks(H3K36me3) and repressive chromatin marks (H3K9me3 and H3K27me3).(G) Representative categories from GO analysis of genesassociated with Dali binding sites (within 1 Mb) includegene expression, cell cycle, signalling, synaptic transmission andcytoskeleton organization among others. Categories marked with anasterisk (*) are significantly enriched also among genes associatedwith peaks within 10 kb of a TSS, with two asterisks(**)—among genes with peaks within 100 kb (Supplementary File5). (H) The intersection of genes proximal(<1 Mb) to Dali peaks, regulated byDali and changing expression uponPou3f3 (10% FDR) knockdown identifies those bothbound and regulated by Dali, as well as genes regulatedby both Dali and Pou3f3 and directlybound by Dali.

DOI:http://dx.doi.org/10.7554/eLife.04530.011

10.7554/eLife.04530.012Figure 5—figure supplement 1.

CHART Analysis.

(A) CHART-seq was performed using cocktails of capture (C-)oligos oligonucleotides complementary to accessible (violet) and/orevolutionary conserved (blue) regions of Dali and a non-targetingcontrol. (B) Specific enrichment of Dali genomic locus (atposition 1250) using C-oligos compared to controls was assayed by qPCR.Mean values ± s.e., n = 3 (technical replicates).(C) Specific purification of Dali RNA using C-oligoscompared to controls was assayed by RT-qPCR. (D) CHART-seqresults were validated by performing an independent experiment with thesame three cocktails of oligonucleotides, a control sense oligocomplementary to the opposite strand of Dali locus and a non-targetinglacZ control oligo. Enrichment of genomic regions identified as peaks wasassayed by qPCR. (E) Dali binds to chromatin in a focalmanner, with most peaks being <1000 bp wide. (F)Computational analysis of CHART-seq peak set and Dali showed that DNAsequences under peaks are not more complementary to Dali sequence thancontrol flanking regions, as judged by either length of aligned regions(left) or alignment quality score (right). (G) DNA sequencesunder peaks are also not predicted to form RNA:DNA–DNA triplexeswith the Dali transcript than control flanking regions.

DOI:http://dx.doi.org/10.7554/eLife.04530.012

Dali binding sites were typically limited to less than 1 kb inlength (Figure 5—figure supplement1E) and were distributed across the genome with no apparent chromosomal biasesother than a depletion on the X chromosome which may reflect the inactivation of oneX chromosome copy in these female N2A cells (Figure5C). These sites were preferentially located at the 5′ end ofprotein coding genes (Figure 5D): 30.5% ofpeaks were within 5 kb of a transcriptional start site (TSS) (Figure 5E). Dali bound sequences weresignificantly enriched for H3K4me3, H3K4me1 and H3K27ac modified histones and PolIIoccupancy, and were depleted for repressive histone marks (Figure 5F). This suggests that Dalipreferentially associates with regions of active chromatin. GO category enrichmentanalysis showed that genes associated with Dali peaks contribute toprocesses related to neuronal differentiation (cell cycle), neuronal projectiondevelopment (cytoskeleton organization and small GTPase mediated signaltransduction), neuronal function (synaptic transmission), and more general cellularprocesses, such as gene expression, intracellular signalling, and cellularhomeostasis (Figure 5G). 150 genes (8.6% ofall Dali bound genes) regulated by Dali containedDali binding sites within their regulatory regions (Figure 5H) and presumably represent directtranscriptional targets.

Dali interacts with chromatin modifying proteins

To investigate the mechanisms of its genomic targeting, we next performedcomputational analysis of Dali bound sequences. We discovered thatDali binding sites do not exhibit significant sequencecomplementarity with the Dali transcript (Figure 5—figure supplement 1F, see Methods), and arenot likely to form RNA-DNA:DNA triplex structures (Figure 5—figure supplement 1G), suggesting thatDali does not bind DNA directly. We therefore speculated thatDali may be targeted to the genome indirectly thoroughRNA-protein interactions. To identify proteins that interact directly withDali, we performed a pull down assay in which in vitrotranscribed and 5′ end-biotinylated Dali was incubated withnuclear extract prepared from day 4 RA-differentiated ES cells. We identified, usingmass spectrometry, 50 proteins that associated with Dali, but notwith antisense Dali or a size-matched unrelated control transcript(Supplementary File7). Direct interactions between the endogenous Dalitranscript and four of these candidate binding proteins, the DNA methyltransferaseDNMT1, the BRG1 core component of the SWI/SNF family chromatin remodelling BAFcomplex, and the P66beta, and SIN3A transcriptional co-factors, were subsequentlyvalidated using UV-crosslinked RNA Immunoprecipitation (UV-RIP) in N2A cells (Figure 6A,B). Human DALI wasalso found, using UV-RIP, to interact with human DNMT1, yet not with BRG1, in humanneuroblastoma SH-SY5Y cells (Figure 6B).Consequently, in further experiments, we focused on the evolutionarily conservedDNMT1 interaction.

10.7554/eLife.04530.013Figure 6.

Dali associates with chromatin and transcriptionalregulatory proteins.

Dali interacts with BRG1, SIN3A, and P66beta in mouse N2Acells (A) and DNMT1 in mouse N2A and human SH-SY5Y cells(B). Nuclear extracts prepared from UV cross-linked cellswere immuno-precipitated using either anti-DNMT1 or control IgG antibodies.Associated RNAs were purified and the levels of Dali andcontrol Gapdh mRNA were quantified using qRT-PCR. Resultsare expressed as fold enrichment relative to an isotype IgG controlantibody. Mean value ± s.e., n = 3. (C) Denovo discovery of a near-perfect match to a CTCF motif in125/1427 (8.8%) Dali CHART-Seq peaks. (D)Dali co-occupies several locations shared with CTCF.Control regions are not predicted to be bound by CTCF and are not bound byDali. ChIP assays were performed in N2A cells usingeither an antibody against CTCF or an isotype specific control. Theindicated DNA fragments were amplified using qPCR. Fold enrichment wascalculated as 2-ΔΔCt (IP/IgG). Mean value ± s.e., n =3. (E) Dali does not directly interact withCTCF protein in mouse N2A cells. Nuclear extracts were prepared from UVcross-linked cells and immuno-precipitated using either anti-CTCF or controlIgG antibodies. Associated RNAs were purified and the levels ofDali and control U1 snoRNA weredetected in each UV-RIP using qRT-PCR. Results are expressed as foldenrichment relative to an isotype IgG control antibody. Results arepresented as mean value ± s.e. of three independent experiments.(F) De novo discovery of a motif for POUIII family transcription factors (which includes POU3F3) in 115/1427 (8.1%)Dali CHART-Seq peaks. (G) UV-RIP in N2Acells: FLAG-tagged POU3F3 protein directly interacts withDali. Mean value ± s.e., n = 3.(H) ChIP-qPCR in N2A cells: POU3F3 occupies a subset of locibound by Dali and regulated by both Pou3f3and Dali. Loci associated with known(Dali-independent) Pou3f3 targets wereused as positive control, while loci not regulated by eitherPou3f3 or Dali and not bound byDali were used as negative control. Mean value ±s.e., n = 3.

DOI:http://dx.doi.org/10.7554/eLife.04530.013

Interestingly, 9 of 58 human transcription factors reported by Hervouet et al. asinteracting with the DNMT1 protein (Hervouet etal., 2010), including CTCF, but also AP-2, C-ets-1, LRH1, PARP, PAX6,STAT1, YY1, and Sp1, were found to have binding site motifs that were significantlyenriched within our stringent Dali bound CHART-seq peaks (Supplementary File 6).Motifs for none of 42 transcription factors that do not interact with DNMT1 butinteract with DNMT3a and/or DNMT3b (Hervouet etal., 2010) were enriched in these peaks (Supplementary File 6). Inparticular, using a de novo motif discovery approach, we found a highly-enrichedCTCF-binding site-like motif in 125 out of 1427 Dali peaks (9%; MEMEE-value = 3.1 × 10⁻⁶²; Figure 6C) (Supplementary File 7). This result was concordant with the greater thanexpected overlap between Dali-associated regions and known CTCFbinding sites in neuronal tissues (Figure 5F)(Shen et al., 2012). Using ChromatinImmunoprecipitation and qPCR (ChIP-qPCR) in N2A cells, we confirmed theCTCF-enrichment of previously-known CTCF-binding sites within 7Dali-bound and regulated promoters, but not at four control regions(Figure 6D). However, despite CTCF andDali thus occupying a subset of shared genomic binding sites,UV-RIP provided no evidence of a direct physical interaction (Figure 6E). Consequently, Dali and CTCF may benon-interacting molecular subunits of a larger ribonucleoprotein complex, oralternatively they might independently bind adjacent sequence, or compete for bindingto the same region. Taken together, the data suggest that Dali isrecruited to chromatin via indirect interactions with several DNA-binding proteinsthrough its direct association with DNMT1.

Depletion of Dali leads to DNA methylation changes at bound andregulated promoters

Increasing numbers of lncRNAs have been shown to direct DNA methylation changes attheir sites of synthesis (Mohammad et al.,2010; Di Ruscio et al., 2013). Thedirect interaction of Dali with DNMT1, however, suggests that it maybe able to regulate DNMT1-mediated CpG methylation at CpG island-associated promotersof Dali-bound and -regulated genes in trans. Toinvestigate this, we performed Combined Bisulfite Restriction Analysis (COBRA) (Xiong and Laird, 1997) in parallel at 10different CpG islands. Selection of these regions was on the basis that they eachcontained several COBRA-compatible restriction enzyme sites and could be efficientlyamplified from bisulfite-converted template. COBRA demonstrated that five of theseregions (corresponding to four genes) exhibited altered restriction profilesindicative of altered DNA methylation status after Dali depletiondepletion (Figure 7—figure supplement1). The inability of COBRA to detect changes at all sites may indicate thatthe DNA methylation status of the remaining regions did not change uponDali depletion or that changes that occurred were undetected dueto technical limitations of the assay.

Bisulfite sequencing demonstrating that the Dlgap5,Hmgb2, and Nos1 promoters each display increasedCpG methylation in two independent stable Dali knockdown linescompared to control further confirmed these results (Figure 7A). Importantly, these data show that methylation changes occurwithin the core of these CpG islands and are not limited to their shores. Althoughother unidentified factors are also likely to play a role, our results are consistentwith Dali (or a Dali:POU3F3 complex) acting intrans, as part of a multi-subunit ribonucleoprotein complex, toreduce DNMT1-mediated CpG methylation at a subset of bound and regulated genepromoters away from its site of transcription.

10.7554/eLife.04530.014Figure 7.

Dali modulates DNA methylation at bound andregulated promoters.

(A) DNA methylation status of three CGI-associated promotersbound and regulated by Dali was assessed using bisulfitesequencing in control and two stable Dali knockdownlines. DNA methylation levels were found to be increased in knockdownlines. The degree of increase was correlated with the degree ofDali knockdown (see Figure 7—figure supplement 1). (B)Nos1 gene has two clusters of alternative TSSs (Exon1 and Exon 2). The upstream neuronal tissue-specific cluster (Exon 1) isassociated with a CpG island and is bound by Dali.(C) Down-regulation of Nos1 observed instable Dali knockdown lines can be explained by reducedinitiation from the Dali-bound TSS (Exon 1), as theratio between Exon1 and an internal Exon 3 is diminished, while the ratiobetween Exon 2 and Exon 3 is not changed. Mean values ± s.e, n= 3, one tailed t-Test, unequal variance. (D)Dali transcript regulates Pou3f3locally and E2f2 distally in ES mouse cells.Dali is expressed from its endogenous locus innon-expressing mouse E14 ES cells using custom-designed TALE-TF (left).De novo induction of the endogenous Dali locus issufficient to up-regulate the neighbouring Pou3f3 geneand down-regulate the distally located E2f2 gene(right). Mean value ± s.e., n = 3.

DOI:http://dx.doi.org/10.7554/eLife.04530.014

10.7554/eLife.04530.015Figure 7—figure supplement 1.

DNA Methylation analysis.

(A) DNA methylation status of three Dali-bound and regulatedCGI-associated promoters (Dlgap5, Nos1, and Hmgb2; see Figure 7) was assayed in a stablecontrol and two independently isolated Dali knockdown lines (mean value± s.e., n = 3). The degree of DNA methylation increasecorrelated with the degree of Dali depletion observed. (B,C, D, E) Combined bisulfiterestriction analysis (COBRA assay) results for Hmgb2 (B),Fbn1 (C), Dlgap5 (D), Nos1 (E).COBRA was performed by bisulfite-treating genomic DNA of control andstable Dali knockdown cells, proliferating and differentiated with RA(+RA), PCR-amplifying the CpG-island associated promoters of theindicated Dali-bound and regulated genes, and digesting the PCR productswith COBRA-compatible or control enzymes.

DOI:http://dx.doi.org/10.7554/eLife.04530.015

One of these genes, Nos1, has multiple alternative promoters fallinginto two distinct regions (for simplicity referred to here as Exon 1 and Exon 2)whose differentiated use is proposed to fine-tune its expression in response tovarious physiological and developmental stimuli (Bros et al., 2006). Only the 5′-most region contains a CpG islandand is bound by Dali (Figure7B). By measuring expression levels of the three 5′-mostNos1 exons in stable Dali knockdown and controllines we observed that the expression level of the 5′ mostDali-bound Exon 1 was reduced, relative to that for Exon 3, whenDali was depleted, whereas the expression ratio between Exons 2and 3 was unaffected (Figure 7C). Thepreferential use of the 5′ most CpG site could reflect a secondary effect ofDali knockdown. Nevertheless, the observation that this site isbound by Dali transcript suggests that Dali mayfunction by promoting the preferential use of a distantly located (and more rarelyused) alternative promoter potentially through its effect on promoter-associated CpGisland methylation.

Dali and POU3F3 protein form a trans-actingtranscriptional regulatory complex

A recognisable binding motif for POU III family transcription factors, such asPOU3F3, was present in 115 out of 1427 Dali CHART-Seq peaks (8.0%;E-value = 3.8 × 10⁻⁵; Figure 6F). This finding, together withDali and Pou3f3 regulating a set of common genes(Figure 3D) and Dalioccupying regulatory regions within 135 (13%) of Pou3f3 targets(Figure 5H), suggested thatDali and POU3F3 protein may interact physically. Indeed, weobserved direct RNA-protein interactions between over-expressed FLAG-tagged POU3F3and co-transfected Dali, using UV-RIP in N2A cells (Figure 6G). Using ChIP-qPCR, we then determinedthat at least five genes that were regulated by both Dali andPou3f3 contained regions that were bound both byDali and by POU3F3 protein (Figure 6H). These results provide further mechanistic insights intoDali's mode of action and indicate that Dali andPOU3F3 form a complex that binds to and regulates a subset of genes intrans in N2A cells.

Induction of the endogenous Dali transcript in mouse ES cellsregulates Pou3f3 locally and E2f2 distally

Finally, we tested whether de novo expressed Dali transcript can actas a transcriptional regulator in order to further substantiate the observation thatDali functions as a novel regulator of both local and distal geneexpression. To achieve this, we induced Dali expression from itsendogenous locus in E14 mouse ES cells, which do not express Dali orPou3f3 to detectable levels, using transient transfection of anartificial Transcription Activator-Like effector (TALE) transcription factor. After72 hr, up-regulation of Dali transcript was shown to significantlyincrease Pou3f3 expression (Figure7D). Dali expression from its own locus is thus sufficientto induce the expression of its genomically neighbouring Pou3f3 gene(Figure 7D). We next investigated theexpression levels of E2f2, a gene that we found to be negativelyregulated by Dali using shRNA mediated knockdown (Supplementary file 2), andfound that Dali up-regulation reduced E2f2transcript levels by approximately 40% (Figure7D). Taken together, these results indicate that Dali canregulate both local and distal target genes when its expression is induced from itsendogenous locus.

Plasmid construction

We used the Whitehead Institute siRNA selection program to design shRNAs that targetmultiple regions of Dali or Pou3f3. To minimise thepossibility of off-target effects, we compared candidate sequences against the NCBIRefSeq database and removed those with ≥15 bases in the anti-sense strand thatmatched a database entry. We then cloned the double stranded DNA oligonucleotidescontaining sense-loop-antisense targeting sequences downstream of the U6 promoter inpBS-U6-CMVeGFP (Sarker et al., 2005) bylinker ligation. The Dali expression plasmid was constructed by PCRamplifying the full length Dali sequence as anEcoRI-XhoI fragment from mouse N2A cell genomicDNA and inserting it into pcDNA3. The FLAG-tagged Pou3f3 expressionplasmid was constructed by excising the full length Pou3f3 ORF fromPou3f3 (NM_008900) mouse cDNA clone in pCMV Entry vector(Cambridge Biosciences, UK) and inserting it into the multiple cloning site (MCS) ofthe N-terminal pFLAG-CMV-6a vector (Sigma–Aldrich, UK) betweenEcoRI and EcoRV sites. The sequences of alloligonucleotides used for cloning are shown in Supplementary file 1.

Dali and Pou3f3 knockdown

Cells were plated at a density of approximately 2 × 10⁵ cells perwell in a six well plate. 16–24 hr later cells were transfected with 1.5μg shRNA expression construct using FuGENE 6 (Promega, UK) following themanufacturer's instructions. Total RNA was extracted from the cells 48–72 hrlater using TRIzol-chloroform extraction method. For stable transfections, N2A cellswere co-transfected with the pBSU6-shRNA expression vector and pTK-Hyg (Clontech,Mountain View, CA) at a 5:1 ratio. 72 hr post-transfection 200 μg/ml HygromycinB was added to the cells to select individual drug resistant clones that were laterisolated and expanded under selective conditions. Dali expression inindividual clones was measured by qRT-PCR.

qRT-PCR and RACE

Reverse transcription was performed using the QuantiTect Reverse Transcription Kit(Qiagen, Netherlands). SYBR Green quantitative PCR was performed using a Step OnePlus Real-Time PCR System (Applied Biosystems, UK). For RACE, GeneRacer Kit(Invitrogen, UK) was used according to the manufacturer's instructions. Human foetalbrain RNA was purchased from Promega. Primers are listed in Supplementary file 1.

Cell culture

Mouse N2A neuroblastoma and E14 ES cells were cultured as described in (Vance et al., 2014). The N2A cell line waschosen because it has been used extensively as a model to study neuraldifferentiation in vitro (Shea et al.,1985). Human neuroblastoma (SH-SY5Y) cells were grown in DMEM/F12 mediumsupplemented with 10% FBS, 1% penicillin-streptomycin, and 1% L-glutamine at37°C in a humidified atmosphere with 5% CO₂. Biochemicalfractionation, ChIP and UV-RIP experiments was performed exactly as described inVance et al. (2014). The followingantibodies were used: anti-DNMT1 (ab87656; Abcam, UK), anti-BRG1 (ab4081; Abcam),anti-P66beta (ab76924; Abcam), anti-SIN3A (Active Motif, Belgium, 39,865), anti-CTCF(Abcam, 70,303), anti-rabbit IgG control antibodies (Millipore, Billerica, MA) andmouse monoclonal anti-FLAG M2 beads (Sigma–Aldrich) for FLAG-tagged POU3F3experiments.

Animal work

All animal experiments were conducted in accordance to schedule one UK Home Officeguidelines (Scientific Procedures Act, 1986). C57BL/6J, postnatal day P56 male andpregnant females were killed by cervical dislocation; whole brains were dissected inice-cold phosphate-buffered saline (PBS) from adult (n = 2), and intrauterinestages E9 (n = 6), E10.5 (n = 6), E13.5 (n = 6), E15.5 (n = 6)and E18.5 (n = 6) mice. Brains were embedded in 5% agarose (low melting,Bioline) and sectioned using a vibrating microtome (Leica, VT1000S) into 200 μmcoronal sections using a chilled solution of 1:1 mixture of RNAlater (Ambion) andPBS. Regions of interest (adult: dentate gyrus, subventricular zone and olfactorybulb; embryos: preplate, proliferative compartmenst combining ventricular andsubventricular zones, and cortical plate from lateral and dorsal tiers) weredissected from individual sections using 27 gauge needles under visual guidance,using transillumination on a dissecting microscope (MZFLIII, Leica, Switzerland).Dissected samples were rinsed in RNAse free PBS/RNAlater 1:1, submerged in ice-coldRNAlater kept for 24 hr at 4°C and stored at −80°C in RNAlater untilprocessing.

Transcriptomic analysis

Total RNA was isolated using the Qiagen Mini RNeasy kit according to themanufacturers' instructions. RNA integrity was assessed on a BioAnalyzer (AgilentTechnologies, UK). 200 ng RNA was used to produce labelled sense single stranded DNA(ssDNA) for hybridization with the Ambion WT Expression Kit, the Affymetrix WTTerminal Labelling and Controls Kit and the Affymetrix Hybridization, Wash, and StainKit following the manufacturer’s instructions. Sense ssDNA was fragmented andthe distribution of fragment lengths was assessed on a BioAnalyzer. Next, fragmentedssDNA was labelled and hybridized to the Affymetrix GeneChip Mouse Gene 1.0 ST Array(Affymetrix, UK). Arrays were processed on an Affymetrix GeneChip Fluidics Station450 and Scanner 3000.

CEL files were analysed using the Limma, oligo, and genefilter R Bioconductorpackages (Smyth, 2004; Carvalho and Irizarry, 2010). Arrays were RMA backgroundcorrected and quantile normalised. Summary expression values were calculated at thegene level. Genes whose expression changed upon Dali andPou3f3 knockdown, as well as upon retinoic acid induceddifferentiation of control and stable Dali knockdown cells, werefiltered to remove genes showing little variation in expression (variance cut off of0.5) before the identification of significant changes. In every case, the LimmaEbayes algorithm was used to identify differential expression between three knockdownand three control samples (biological replicates). 1.3-fold change cutoff was appliedin every case. GOToolbox was used to perform Gene Ontology analyses ((Martin et al., 2004); http://genome.crg.es/GOToolBox/). Representative significantlyenriched categories were selected from a hypergeometric test with aBenjamini-Hochberg corrected p-value threshold of 0.05.

CHART

CHART Enrichment and RNase H Mapping experiments were performed as described in(Simon, 2013). We designed 10biotinylated DNA capture (C)-oligos: 5 oligos complementary to the most accessibleregions of Dali, as determined by RNase H mapping, and 5 oligostargeting the most evolutionarily conserved regions of the transcript (Figure 5A). These oligos were used as twococktails of 5 oligos, and as a pool of all 10. As controls, we used an oligodesigned to target the antisense Dali sequence (absent from the N2Atranscriptome). Additionally we require peaks to not overlap with those identified inan analogous CHART-sequence experiment using the E. coli lacZsequence (GSE52571) (Vance et al.,2014). Compared to controls, all three cocktails of Dalioligos showed significant enrichment of the Dali transcript (10-foldcompared to lacZ), but no enrichment of the abundant mRNAGapdh (Figure 5B). Withoutany prior information about Dali genomic binding, we considered itsendogenous site of synthesis to assess the enrichment of transcript-associated DNAloci. Specific enrichment of Dali at its locus was observed asexpected (Figure 5—figure supplement1).

CHART extract was prepared from approximately 3 × 10⁸ N2A cells perpull down and hybridized overnight with 810 pmol biotinylated oligonucleotidecocktail (Supplementary File1) at room temperature with rotation. 250 μl MyOneC1 streptavidinbeads (Invitrogen) were used to capture the complexes overnight at room temperaturewith rotation. After extensive washes, bound material was eluted using RNase H (NewEngland Biolabs (NEB), UK) for 30 min at room temperature. Samples were treated withProteinase K and cross-links were reversed. RNA was purified from 1/5 total samplevolume using the QIAGEN miRNeasy kit. DNA was prepared from the remaining sampleusing the phenol:chloroform:isoamyl alcohol extraction and ethanol precipitationmethod. DNA was further sheared to an average fragment size of 150–300 bpusing a Bioruptor (Diagenode, Belgium) and sequenced on an Illumina HiSeq (50 bppaired end).

Computational analysis of CHART-seq data

CHART-seq was performed with three independent pull down samples (using twoindependent cocktails of 5 C-oligos, and one cocktail containing all 10 C-oligos) andsequenced simultaneously with a matched input sample. 50 bp, paired-end reads weremapped to the mouse genome (mm9) using bowtie with the options ‘–m1–v2 –best–strata–a’. For eachDali sample, peaks were called against the matched N2A inputsample (4208 peaks) and CHART-seq peaks previously analogously identified in N2Acells using two lacZ controls (1928 peaks) (Vance et al., 2014). Peak calls were made using the MACS2algorithm ((Zhang et al., 2008); https://github.com/taoliu/MACS/blob/master/README) with the options‘–mfold 10 30 –gsize = 2.39e9 –qvalue =0.01’ using the CGAT pipeline ‘pipeline_mapping.py’ (https://github.com/CGATOxford/cgat). Peak calls were then filteredsuch that only peak calls with a −log10 q value >5 were retained (FDR0.001%).

We discovered 1427 Dali-associated regions genome-wide calledagainst both input and lacZ control samples (Figure 5A; Supplementary file 5).

Characterisation of Dali binding sites

The chromosomal distribution of Dali peaks was visualised using theR Bioconductor package ‘ggbio’ (Yinet al., 2012). Genome territory enrichments analysis was performed usingthe Genome Association Tester (GAT; (Heger et al.,2013)). 10,000 simulations were performed using a mappability filteredworkspace and an isochore file partitioning the genome into eight bins based onregional GC content. For the chromosomal enrichment analyses, chromosomal territorieswere proportionally assigned to a single virtual meta-chromosome before using GAT totest for GC and mappability corrected enrichments as above. Gene Ontology categoriesenriched for Dali binding were identified by intersecting regulatoryregions for known coding genes with Dali binding sites. Regulatoryregions for genes were defined following the GREAT definition (McLean et al., 2010) as a basal domain surrounding the TSS(from −5 kb to +1 kb) and extending domains upstream and downstream tothe nearest gene's basal domain or to a maximum distance of 1 Mb. Enrichments wereidentified using GOToolbox.

Dali peaks were characterised using DNase I hypersensitivity (HS)data generated by the Stamatoyannopoulos lab at the University of Washington andchromatin features identified by the Ren lab at the Ludwig Institute for CancerResearch ((Shen et al., 2012); ENCODE Project Consortium, 2012). Enrichmentsof DNase I HS and chromatin features overlapping Dali peaks were assessed using GATto control for mappability and regional GC content as above.

Complementarity between Dali sequence and binding locations wasassessed using the EMBOSS Water algorithm (Rice etal., 2000) which performs Smith-Waterman alignment with a range of gapopening and extension penalties. RNA-DNA:DNA triplex formation was assessed using theTriplexator search software suit (Buske et al.,2012). The MEME-ChIP (Machanick andBailey, 2011) algorithm was used to perform de novo motif discoveryanalysis by examining the unmasked DNA sequence of the central regions of peaklocations. MEME-ChIP was run with the options ‘-meme-mod zoops -meme-minw 5-meme-maxw 30–meme-nmotifs 50’ using a custom background file preparedfrom regions flanking the peak locations using the command ‘fasta-get-markov-m 2’. Enrichment of known vertebrate transcription factor binding sites fromthe TRANSFAC Professional database (Matys et al.,2006) was assessed using the AME algorithm (McLeay and Bailey, 2010) with the options‘–method fisher–length-correct’ using the sequence andbackground file prepared for MEME-ChIP analysis.

3C

E14 ES cells or day 4 ES-derived neuronal were cross-linked with 2% formaldehyde.Nuclei were prepared and permeabilized with 0.3% SDS in 1.2× restriction buffer(NEB3 for BglII) for 1 hr at 37°C. Then, SDS was sequestered byadding 1.8% Triton X-100. 1 × 10⁶ nuclei (∼15 μg ofchromatin) were digested with 400 units of BglII restriction enzymeovernight, and the enzyme was inactivated. Nuclei were diluted in 1.15× T4 DNAligation buffer (NEB), and SDS sequestered by adding 1% Triton X-100. The digestedchromatin was ligated using 100 Weiss units of T4 DNA ligase for 4 hr at 16°Cand treated with Proteinase K to reverse cross-links. Samples were further treatedwith RNase A, and DNA was phenol-chloroform extracted and ethanol precipitated.

A RP23-92N4 (CHORI; BACPAC) Bacterial Artificial Chromosome (BAC) clone covering thePou3f3-Dali locus was treated as above and used as a controltemplate for the 3C assay. Ligation products of 3C and BAC samples were quantified byqPCR. PCR reactions consisted of 300 ng 3C sample, 0.2 μM test primers and aprimer corresponding to Dali promoter and 1× SYBR Green PCRMastermix (Life Technologies, UK). All reactions were performed in triplicate. Themean threshold cycle (Ct) value was calculated and used to calculate relative amountsof PCR products. To normalise for different primer efficiencies, interactionfrequencies were calculated by dividing the amount of PCR product obtained from the3C sample by the amount of DNA obtained from control BAC DNA. Interaction frequencieswere also normalised to Gapdh internal controls prepared fromgenomic DNA in the same manner as the BAC clone sample. All primers used are listedin Supplementary file1.

COBRA

We used COBRA to study 9 out of 44 CpG island-containing promoters bound byDali and associated with genes differentially expressed betweenstable Dali knockdown and control cell lines prior to or subsequentto the RA-induced differentiation. 80–350 ng of genomic DNA wasbisulfite-treated using EZ DNA Methylation Gold kit according to the manufacturer'sinstruction and used for PCR amplification. Primers for amplifying bisulfiteconverted template DNA were designed using MethPrimer software accessible athttp://www.urogene.org/methprimer/ (Li and Dahiya, 2002). PCR products were on-column purified with QIAquickPCR Purification Kit. 250 ng to 1 μg of purified products were incubated withappropriate COBRA-compatible (BstUI (NEB), MspI(NEB), TaqI (Thermo Scientific), HpyCH4IV (NEB)) orcontrol (Hsp92II (Promega), BfaI (NEB)) restrictionenzymes overnight. Restriction products were analysed on 3% low melting point agarosegels.

TALE-mediated up-regulation

Target regions were selected and TAL effector constructs were designed usingsoftware, tools, and information found on the TAL Effector Nucleotide Targeter2.0 website accessible from https://tale-nt.cac.cornell.edu/. Construction of custom TALE-TFsdesigned to target promoter-proximal region of Dali to up-regulatetranscription from the locus was performed as described by Sanjana et al. (2012). The TALE-TF was designed to target thefollowing region lying upstream of the TSS of Dali: chr1 (mm9):42807019-42807038 ("TGTCCCTTGTCCACATATCT"). The TAL domain sequence used was asfollows: NH NG HD HD HD NG NG NH NG HD HD NI HD NI NG NI NG.

Data deposition

Microarray and CHART-Seq data have been deposited in the GEO database under accessionnumber GSE62035 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE62035).