J. Neurochem (2012) 120, 699–709.
L1 cell adhesion molecule (L1CAM), an adhesion/signaling protein encoded by a gene target of the transcription repressor RE-1-Silencing Transcription factor (REST), is expressed in two alternatively spliced isoforms. The full-length isoform, typical of low-REST neural cells, plays key roles in survival/migration, outgrowth/fasciculation/regeneration of axons, synaptic plasticity; the isoform missing two mini-exons, abundant in a few high-REST non-neural cells, maintains some effect on migration and proliferation. To investigate whether and how L1CAM alternative splicing depends on REST we used neural cell models expressing low or high levels of REST (PC12, SH-SY5Y, differentiated NT2/D1 and primary neurons transduced or not with REST). The short isoform was found to rise when the low-REST levels of neural cells were experimentally increased, while the full-length isoform increased in high-REST cells when the repressor tone was attenuated. These results were due to Nova2, a neural cell-specific splicing factor shown here to be repressed by REST. REST control of L1CAM occurs therefore by two mechanisms, transcription and alternative splicing. The splicing mechanism, affecting not only L1CAM but all Nova2 targets (∼7% of brain-specific splicing, including the mRNAs of other adhesion and synaptic proteins) is expected to be critical during development and important also for the structure and function of mature neural cells.
β-catenin construct truncated of its N-terminal 90 aminoacids
ChIP followed by ultra-high-throughput sequencing
fluorescence-activated cell sorter
full length isoforms of L1CAM
L1 cell adhesion molecule
- mAb and pAb
monoclonal and polyclonal antibody
nuclear factor I-A
RE-1-Silencing Transcription factor
L1 cell adhesion molecule (L1CAM), composed of six extracellular immunoglobulin-like domains followed by five fibronectin type III repeats, a single-pass transmembrane sequence and a short cytoplasmic domain, is a multifunctional surface protein encoded by a gene located in the X chromosome. L1CAM was discovered in neurons and in various neurosecretory cell lines, including PC12 (Salton et al. 1983; Rathjen and Schachner 1984). Its binding to adhesion proteins and receptors, with ensuing activation of multiple signalling cascades (Panicker et al. 2003; Islam et al. 2004; Kulahin et al. 2008), plays pivotal roles in a variety of processes such as survival and migration of neurons; outgrowth, fasciculation and regeneration of axons; synaptic plasticity (recent reviews: Gerrow and El-Husseini 2006; Maness and Schachner 2007; Schmid and Maness 2008). The functional importance of L1CAM in the brain is documented also by its mutations, known as the cause of an intensively investigated brain genetic disease, the L1CAM syndrome or CRASH (reviews: Kamiguchi et al. 1998; De Angelis et al. 2002; Schäfer and Altevogt 2010).
L1CAM expression in neural cells depends on RE-1-Silencing Transcription factor (REST, also known as NRSF) (Kallunki et al. 1997), the master factor governing neural cell differentiation. High levels of REST, typical of stem cells, repress the expression of hundreds of neural-specific target genes. A rapid down-regulation, taking place during neural progenitor maturation, relieves the expression of these genes and orchestrates the acquisition of the mature phenotype (Ballas and Mandel 2005; Ooi and Wood 2007). Even in mature neural cells, however, the levels of REST can increase, although moderately, both in physiological and pathological conditions (Garriga-Canut et al. 2006; Spencer et al. 2006). Recently, another transcription repressor, nuclear factor I-A (NFI-A), was found to participate in the regulation of L1CAM expression in neural cells (Schneegans et al. 2009). Whether the role of NFI-A depends on REST, or whether it works independently, is unknown.
In many non-neural cells, where levels of REST are high, expression of L1CAM remains low. In others, however, such as Schwann cells, oligodendrocytes, epithelial cells and cells of the immune system, L1CAM is high, relevant for multiple, ubiquitous or specialized functions including myelination, lymph node architecture, T cell activation and cell migration (see, among others, Di Sciullo et al. 1998; Itoh et al. 2000; Balaian et al. 2000; Maddaluno et al. 2009). In addition, L1CAM is now recognized as a surface protein essential for stem cells self-renewal (Son et al. 2011) and as a marker of invasiveness and metastases of various types of human cancers (reviews: Raveh et al. 2009; Siesser and Maness 2009; Schäfer and Altevogt 2010). In at least some tumours expression of L1CAM is activated by the β-catenin/TCF transcription system (Gavert et al. 2005), operative in parallel to REST (Pfeiffer et al. 2010). Finally, in pancreatic adenocarcinoma expression of L1CAM was shown to depend on an additional transcription factor, SLUG (Pfeiffer et al. 2010; Geismann et al. 2011).
The L1CAM protein exists in two alternatively spliced isoforms. The full-length isoform (fl-L1), composed by 28 exons, is typical of neural cells; the short (sh-L1) or non-neural isoform, lacking two mini-exons (exon 2, extracellular, and exon 27, intracellular) is predominant in the other cell types. By their participation in L1CAM binding to extracellular (De Angelis et al. 2001) and cytoplasmic (Herron et al. 2009) proteins, respectively, the two spliced exons appear functionally important in neural cells and also in some tumours where fl-L1, and not sh-L1, promotes liver and lung metastases (Hauser et al. 2011). In a human breast cancer line, however, even over-expression of sh-L1 induced disruption of adherens junctions, increased β-catenin transcription, increased motility and cell scattering (Shtutman et al. 2006).
In spite of its recognized relevance in physiology and pathology, L1CAM splicing and its regulation had never been investigated in mechanistic terms. The present work was carried out by employing several cell models: two clones of the rat pheochromocytoma line PC12 (D’Alessandro et al. 2008), one (wtPC12), expressing the typical phenotype, with low REST and high fl-L1 levels, the other (PC12–27) expressing in contrast high REST and low sh-L1 levels; the high-REST neuroblastoma SH-SY5Y line; rat primary neurons and the human NT2/D1 cells differentiated with retinoic acid (RA), analyzed before and after REST transduction. Our results demonstrate that the low REST of neural cells governs not only the expression of L1CAM but also its specific splicing, induced through the expression of Nova2, a brain-specific splicing factor. The REST-dependence of splicing is likely a widely important process, not limited to L1CAM but extended to the whole panel of Nova2 targets [∼7% of the brain-specific mRNAs spliced in the neocortex (Ule et al. 2005)].
Materials and methods
Goat polyclonal antibodies (pAbs) against the C-terminal (sc-1508) and one N-terminal domains (sc-31032) of L1CAM were from Santa Cruz, a mouse monoclonal Ab (mAb) against the N-terminal (CD171) was from BD Biosci. Pharmingen (San Diego, CA, USA). The anti-REST rabbit pAb was from Millipore Corporation (07–579; Bedford, MA, USA); the anti-Nova2 and anti-NFI-A rabbit pAbs from MBL International (RN044PW; Woburn, MA, USA) and Abcam (Ab41851; Cambridge, UK), respectively. Horseradish peroxidase-conjugated pAbs anti-mouse, anti-rabbit and anti-goat immunoglobulins were from Bio-Rad Laboratories (Hercules, CA, USA); tetramethylrhodamine isothiocyanate-conjugated rabbit anti-goat pAb from SouthernBiotech (Birmingham, AL, USA); FITC-conjugated annexin V from Boehringer Mannheim Italia (Monza, Italy). The Fungene HD transfection factor was from Promega; lipofectamine 2000 from Invitrogen (Carlsbad, CA, USA); the anti-β-tubulin mAb, DAPI, RA and the histone deacetylase inhibitor trichostatin A (TSA) from Sigma-Aldrich (St Louis, MO, USA); BSA from Roche Molecular Biochemicals (Indianapolis, IN, USA). The EZ-chromatin immunoprecipitation (ChIP)-A kit including the control rabbit IgGs was from Millipore; the BCA Protein Assay kit and the NE-PER® kit for nuclear protein extraction from ThermoScientific (Waltham, MA, USA).
Cells and cell clones
Rat pheochromocytoma wtPC12 and PC12–27 clones and the Myc-REST subclone stably transfected with REST were as in D’Alessandro et al. (2008); the wtPC12/Δ90βcat subclone and SH-SY5Y cells as in Tomasoni et al. (2011) and Racchetti et al. (2010), respectively. Rat P2 primary cortical neurons, grown in Dulbecco’s modified Eagle’s medium (Lonza, Basel, Switzerland) with 10% horse serum (Euroclone, Pero-Milano, Italy), 5% fetal clone III serum (Hyclone, Bonn, Germany), 2 mM ultraglutamine and 100 U/mL penicillin/streptomycin (Lonza) (Racchetti et al. 2010), were used after 6–10 days in vitro. NT2/D1 cells, described in Nikcevićet al. (2008), were grown in Dulbecco’s modified Eagle’s medium with 10% fetal clone III serum, 2 mM ultraglutamine and 5 U/mL penicillin/streptomycin.
Transient and stable transfections and transductions. Scratch assay
Transient and stable transfections of PC12 clones, using lipofectamine 2000, were as in D’Alessandro et al. (2008). Subclones were grown in complete medium supplemented with G418 (450 μg/mL). For the scratch assay, PC12–27 cells were transiently co-transfected (Hauser et al. 2011) with GFP together with either fl-L1 or sh-L1 constructs (gifts of Ugo Cavallaro and Achim Krueger, respectively), grown for 1 day and then sorted. Fractions expressing similar levels of GFP were incubated to confluence (1 day) in 24-well dishes. Monolayers, wounded with a p200 pipette tip, were washed carefully and photographed in a phase-constrast microscope every 24 h for 3 days (Liang et al. 2007).
Three rat Nova2-targeting shRNAs were generated and inserted in the pGFP-V-RS vector (OriGene Technologies, Rockville, MD, USA). The HT120832D (ACGCAAGCCGCTCAATACCTCATCAGCCA) shRNA, showing the best activity and a non-effective scramble cassette in the same vector were used for Nova2-knock-down and as negative control. Fractions exhibiting high GFP fluorescence, sorted by fluorescence-activated cell sorter (FACS) 72 h after transfection with Fungene HD, were used for the assays. For transduction of REST in neuronal primary cultures and NT2/D1 cells, #945.PCCL.sin.cPPT.SV40ployA.eGFP.minCMV.hPGK.deltaLNGFR.Wpre lentiviral vectors including GFP together with (or not, controls) the full length REST were employed. After 24 h incubation, the cells were washed, cultured for further 6 days and then sorted by FACS. Highly GFP-fluorescent fractions were used for the assays. For details, see Tomasoni et al. (2011).
Real-time polymerase chain reaction (RT-PCR)
Total RNAs, extracted using RNeasy mini columns (Qiagen, Valencia, CA, USA) and assayed by spectrophotometry, were used (1–2 μg) to generate cDNA templates for RT-PCR, using random primers, Rnase inhibitor and retrotranscriptase M-MLV (Applied Biosystems, Foster City, CA, USA). Assays of triplicate samples were carried out according to the SYBR Green procedure, using 50 ng template cDNA in the ABI 7900HT Sequence Detection System (Applied Biosystem). Results were normalized to the invariant mRNAs of calmodulin or GAPDH. PCR products in TBE buffer (45 mM Tris-borate; 1 mM EDTA, pH 7) were analyzed on 2% agarose gels for GAPDH and calmodulin, on 4% gels for L1CAM isoforms. Primers used in rat cells were: for calmodulin, fw: 5′-GAAGAAGAAATTAGAGAAGCGTTCC-3′; rev: 5′-GTAGTTTACCTGACCATCCCCAT-3′; for GAPDH, fw: 5′-CAAGGTCATCCATGACAACTTTG-3′; rev: 5′-GTCCACCACCCTGTTGCTGTAG-3′; for L1CAM-F2: 5′-ATGGTCATGATGCTGTGGTACGT-3′; for -R2: 5′-GGAACTCCACTTGGGGTCTGC-3′; for L1CAM-F27: 5′-AGGAGGACACTCAGGTAGATTCCG-3′; for -R27: 5′-GGTTTGATGTCTCCATTGAGAGATG-3′; for L1CAM-F2.1, 5′-CCTGATGAATATAAAGGACACC-3′; for L1CAM-F27.1: 5′-CGAGTACAGGTCCCTGGAGAGTG-3′; for L1CAM-Ex2, fw: 5′-CCCACCCCCGTCAGCAGGTA-3′; rev: 5′-ACCCTCAGCCACGAGCTGGA-3′; for L1CAM-Ex27, fw: 5′-CCGAGGGCTGGTTCATCGCC-3′; rev: 5′-CAGGTGTCTTGGCCCCTGGC-3′; for REST, fw: 5′-CTACATGGCACACCTGAAGCACCAC-3′; rev: 5′-GCGTAGTCACACACGGGGCAGTTGAAC-3′; for NFI-A, fw: 5′-AGCCCACATGCGACGCCATC-3′; rev: 5′-GGCATGGGCAGAGGCACAGG-3′; for Nova2, fw: 5′-CTCAATCATCGGCAAAGGTG-3′; rev: 5′-GGCTCTGTCGGGGTTCATC-3′. In human cells, primers used were: for hL1CAM-hF2: 5′-GGTCGTGGCGCTGCGGTACGTGTG-3′; -hR2: 5′-GCGGAACTGCACTTCGGGCTTGCC-3′; for hL1CAM-hF27: 5′-ACTCAGTGAAGGATAAGGAG-3′; -hR27: 5′-TTGAGCGATGGCTGGCTGCT-3′; hREST, fw: 5′-GAGGAGGAGGGCTGTTTACC-3′; rev: 5′-TCACAGCAGCTGCCATTTAC-3′; hNova2, fw: 5′-CATTGGCAAGGGCGGGCAGA-3′; rev: 5′-CACGCGCTCCTGCAGGTTGA-3′. SLUG expression was analyzed by a Taqman assay performed by the RT-PCR and DNA sequencing service of IFOM-IEO, Milan (assay IDs: Rn00709370_m1 and Hs00161904_m1).
Cell extracts in TD buffer containing 1% Triton X-100, 50 mM Tris–HCl (pH 7.5), 250 mM NaCl, 5 mM EDTA, 50 mM NaF and protease inhibitors, were rocked (15 min) and then centrifuged (13 000 g, 4°C, 15 min). Protein samples, separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred to nitrocellulose filters, were processed as in Cocucci et al. (2004). Photographic development was by chemiluminescence using ECL Western Blotting Detection reagent (Amersham Bioscience, Little Chalfont, UK) or Femto Signal (Pierce, Rockford, IL, USA). Signals were acquired by the Densitometer SI and Image Quant (Amersham Bioscience).
For surface immunolabeling of L1CAM, PC12 and NT2/D1 cells on coverslips were fixed with 95% ethanol at −2°C (10 min) or 4% formaldehyde at 4°C (10 min), then washed and immunolabeled for 1 h with anti-L1CAM goat pAbs against the extracellular or the intracellular domains. The bound Ab was revealed by tetramethylrhodamine isothiocyanate-conjugated rabbit anti-goat pAb. Nuclei were stained with DAPI. Intracellular immunolabeling was revealed similarly, however upon Triton X100 permeabilization of the fixed cells. The surface appearance of phosphatidylserine was checked by labeling formaldehyde-fixed, non-permeabilized cells with FITC-annexin V (0.5 μg/mL, 15 min, 25 °C, in the presence of divalent cations). Samples were studied in a Perkin-Elmer Ultraview ERS confocal microscope. Image deconvolution was carried out in a wide field microscope of the Delta Vision system.
Chromatin of PC12–27 cells was extracted and immunoprecipitated using the EZ-ChIP-A kit, using the anti-REST pAb and control rabbit IgGs from Millipore. To reveal the REST binding sites, the rat Nova2 gene nucleotide sequence was analyzed by the Genomatix Genome Analyzer. Immunoprecipitation of REST-associated DNA fragments was verified by RT-PCR using the following primers: for intron 2–3, fw: 5′-CCTGAGCAAGGGTTGTCCGGGA-3′; rev: 5′-CCACCTAGCCTGTCCCCATGCA-3′; for intron 4–5, fw: 5′-AGGAACTAGTAAGCTGTCATGGCG-3′; rev: 5′-TCCTTTAAGGGTCGGTCAGCAGGC-3′. ChIP-enriched DNAs were quantified by the qPCR-SYBR green method, using the ABI 7900HT Sequence Detection System (Applied Biosystem). ChIP samples were analyzed in triplicate by PCR. Relative enrichment = the ratio between the gene in the immunoprecipitates with specific and control Abs.
The significance of the data was assessed using the two-tailed unpaired t-test. Data shown are means ±SD. Statistical significance is indicated by asterisks: *p < 0.05; **p < 0.01; ***p < 0.001.
REST governs not only the expression but also the splicing of L1CAM
Expression of L1CAM was investigated in the wtPC12 and PC12–27 clones characterized by low and ∼50-fold higher levels of REST mRNA and protein (Fig. 1; D’Alessandro et al. 2008). The mRNA levels of L1CAM were the opposite of those of REST, over 20-fold higher in the wtPC12 than in the PC12-27 clone (Fig. 1a and b). The differences of the L1CAM protein levels were analogous (over 15-fold higher in wtPC12, Fig. 1c). Immunofluorescence revealed a preferential distribution of L1CAM to the surface, accompanied by some labeling of cytoplasmic puncta evident in wtPC12 (Fig. 1d). The opposite levels of the L1CAM and REST mRNA and protein were confirmed also in other cells: the high-REST human neuroblastoma SH-SY5Y line (Racchetti et al. 2010) and a subclone of wtPC12 transfected with a full-length REST construct (Myc-REST: total level of REST ∼4-fold higher than in wtPC12 cells, much lower than in PC12–27 cells, D’Alessandro et al. 2008) (Fig. 1a–c). In the high-REST PC12–27 and SH-SY5Y, the L1CAM mRNA and protein were increased when the REST tone was attenuated by TSA, a blocker of histone deacetylases, the main effectors of REST gene repression (D’Alessandro et al. 2008), employed at a concentration that fails to induce any surface increase of FITC-annexin V labeling, a marker of apoptosis (not shown; see D’Alessandro et al. 2008). In contrast, in the low-REST wtPC12 cells TSA induced no significant change (Figure S1A,B).
In addition to the difference in expression, the L1CAM protein of wtPC12 and PC12–27 clones differed also in apparent Mr (Fig. 1c). In western blots carried out in parallel with two pAbs and one mAb, over 90% of the wtPC12 protein migrated in the ∼220 kDa band, close to a minor, faster migrating band which was the only band appreciable in PC12–27 and SH-SY5Y cells (Fig. 1c). In terms of mRNA, a duality of L1CAM was due to the alternatively spliced fl-L1 and sh-L1 isoforms, as revealed by RT-PCR results obtained with the primers F2/R2 and F27/R27 (see Fig. 2a). The differences between wtPC12 and Myc-REST versus PC12–27 and SH-SY5Y (Fig. 2b) were localized in areas close to the 5′ and 3′ ends including, respectively, the 2 and 27 exons. A more refined analysis, focused on the 15 bases of exon 2 and the 12 bases of exon 27 (F2.1 and F27.1, Fig. 2a) demonstrated their presence in the mRNAs of wtPC12 and Myc-REST, and their lack in those of PC12–27 and SH-SY5Y (Fig. 2b, right), as expected for the fl-L1 and sh-L1 isoforms, respectively. This identification was confirmed by the sequencing of the two L1CAM mRNA transcripts in wtPC12 and PC12–27 cells (not shown).
fl-L1 and sh-L1 isoforms affect differently PC12 cell migration
To extend the characterization of the two spliced L1CAM isoforms we carried out a functional test, taking advantage of the very low expression of the endogenous protein in PC12–27 cells (see Fig. 1a and c). The analysis of FACS-sorted, transiently transfected populations of PC12–27 could therefore be employed to investigate the separate effect of either isoform on migration. Figure 2c shows the results of a scratch assay. Compared with control PC12–27 cells, transfected with GFP only (not shown), the cells transfected with sh-L1 did not exhibit any increased migration. In contrast, the cells transfected with fl-L1 migrated distinctly faster, approaching the reestablishment of the contact across the scratch within 3 days, when the sh-L1-transfected cells were still far away. These results confirm the profound functional difference between the two L1CAM isoforms previously anticipated by others based on different experimental approaches (see Schäfer and Altevogt 2010; Hauser et al. 2011).
Role of NFI-A, SLUG and β-catenin in L1CAM expression/splicing in PC12
Three transcription factors, the repressor NFI-A and the stimulators SLUG and β-catenin, had been reported to play roles in the expression of the L1CAM gene in various tumours (Gavert et al. 2005; Schneegans et al. 2009; Pfeiffer et al. 2010; Schäfer and Altevogt 2010). Working with wtPC12, PC12–27 and Myc-REST clones we found the levels of the NFI-A mRNA and protein to vary analogously, but less extensively, to those of REST (compare Figure S2A to Fig. 1b, c). Interestingly, these NFI-A levels were not changed significantly by cell treatment with TSA (not shown), excluding the expression of NFI-A to be directly controlled by REST. Concerning SLUG, traces of its expression by RT-PCR were found neither in the rat wtPC12 and PC12–27 clones nor in the human SH-SY5Y cells (not shown).
To investigate the role of β-catenin we employed Δ90-βcat, encoding a truncated, slow-turnover protein active in the transcription of the β-catenin/TCF target genes (Chenn and Walsh 2002). No significant increase of L1CAM mRNA was observed in wtPC12 cells transfected with the construct (not shown). However in these cells the sh-L1 mRNA, with skipped exons 2 and 27, was increased to levels over 50% of the fl-L1 isoform (Figure S2C,D). This was most likely an indirect consequence of the increased transcription of the REST gene induced by the higher β-catenin tone (see Tomasoni et al. 2011).
Splicing of L1CAM in PC12 cells: involvement of the Nova2 factor
Alternative splicing of L1CAM in our low- and high-REST PC12 cells could be due to a splicing factor dependent on REST for its expression. We wondered whether any recognized splicing factor gene had been reported to include in its promoter the specific REST binding sequence, RE-1. Large-scale chromatin immunoprecipitation (ChIP-Seq) studies (Johnson et al. 2007) had predicted the sequence in the promoter of the gene encoding Nova2, a neural cell-specific factor known to govern the splicing of various channels and synaptic proteins (Ule et al. 2005; Yano et al. 2010; Allen et al. 2010). So far, however, the role of Nova2 had not been validated.
In wtPC12 and PC12–27 clones, the levels of Nova2 mRNA and protein were found to differ opposite to those of REST (compare Fig. 3a and b to Fig. 1b and c). This confirms the possibility of a negative control of Nova2 expression by the repressor. Moreover, treatment of PC12–27 with TSA induced a 15-fold increase of the Nova2 mRNA (Fig. 3c, left), similar to the increase in L1CAM mRNA revealed by the amplification with F2.1/R2 and F.27/R27. In contrast, in wtPC12 the TSA treatment was ineffective (Fig. 3c). To investigate whether the differential expression of the Nova2 gene was due to the direct binding of REST or to an indirect mechanism, we carried out experiments of chromatin immunoprecipitation (ChIP). Fig. 3d shows that the Nova2 gene of PC12–27 cells immunoprecipitated by the anti-REST Ab was enriched 3.5- to 4.5-fold with respect to a control Ab. This result confirms that expression of the Nova2 gene is repressed by the direct binding of REST.
The role of Nova2 in wtPC12 was investigated also upon knock-down of the splicing factor to ∼1/4th level of the control, induced by transient transfection of a specific shRNA plasmid (Fig. 3e). Analysis of the F2/R2 and F27/R27 mRNA fragments failed to reveal significant differences in L1CAM expression between knocked-down and control cells even if the ratio between upper and lower band was markedly decreased (Fig. 3f). However, the analysis of the two spliced mini-exons (F2.1/R2 and F27.1/R27 in Fig. 3a) revealed large (−70/75%) and significant decreases of their expression (Fig. 3f right). The results strongly suggest the two L1CAM isoforms, fl-L1 and sh-L1, to depend on the differential expression of the REST-dependent Nova2 splicing factor. Specifically, Nova2 promotes inclusion of the two mini-exons, 2 and 27, in the mRNA of L1CAM.
REST governs Nova2 expression and L1CAM splicing also in NT2/D1 cells and in primary neurons
These experiments were carried out in two additional cell models transduced with REST-encoding or scrambled retroviral vectors: human neuron-like cells differentiated from NT2/D1 cells by a 4 week treatment with RA and primary cultures of rat cortical neurons. The results of Figs 4 and 5 recapitulate in the two models the results obtained in the low- and high-REST PC12 clones. Untreated NT2/D1 cells, which express high REST levels (Fig. 4b; Tomasoni et al. 2011), exhibited very low and inappreciable levels of L1CAM mRNA and protein, respectively (Fig. 4a and b). During the RA treatment NT2/D1 cells, while developing a neural-type phenotype (Andrews 1984), exhibited a progressively decreasing REST expression, reaching very low levels in the fully differentiated, neuron-like cells (NT2-N, Fig. 4b). Concomitantly, expression of L1CAM increased (Fig. 4a and b), with distribution of the protein to the cell body and neurites (Fig. 4c). The Nova2 protein became clearly appreciable only in the NT2-N cells (Fig. 4b). Trunsduction of the low-REST NT2-N cells with the REST lentiviral vector, with ensuing large increase of the REST protein level (Fig. 4e), was followed by the drop of both L1CAM and Nova2 mRNA and protein, reaching levels 25% or lower of controls, and by the conversion of the predominant fl-L1 mRNA and protein into the sh-L1 isoforms (Fig. 4d and e).
In the primary cultures of neurons, which express very low levels of REST (Fig. 5a and c), Nova2 and L1CAM were high both at the mRNA and the protein level (Fig. 5b and c). Upon transduction with REST the levels of the transcription factor increased considerably (over 35- and 7-fold at the mRNA and protein level respectively, Fig. 5a and c). Concomitantly, Nova2 and L1CAM were significantly decreased (−70% or more) and L1CAM was switched from the fl-L1 to sh-L1 isoform (Fig. 5b and c), with disappearance of the mini-exons 2 and 27 (Fig. 5b).
In conclusion, both the differentiated, neuron-like NT2-N cells and the primary neurons, when transduced or not with REST, resembled PC12–27 and wtPC12 cells, respectively (compare Figs 4 and 5 to Figs 1 and 2). The REST-dependent, Nova2-mediated mechanism governing the generation of the two L1CAM isoforms, identified in the PC12 clones, appears therefore of wider, possibly general importance in neural cells.
Expression of the L1CAM gene in two alternatively spliced isoforms, fl-L1 (neural) and sh-L1 (non-neural), had been described two decades ago (Reid and Hemperly 1992). So far, however, the mechanism of the two isoform generation had never been investigated. Our work, aimed at identifying this mechanism, was carried out first in clones of the rat pheochromocytoma PC12 line differing in the levels of REST, the master transcription factor of their differentiation. REST is low in wtPC12, as in the other neural cells, and 50- to 60-fold higher in PC12–27 cells (D’Alessandro et al. 2008).
In addition, the two PC12 clones exhibited differences also in L1CAM, abundant and prevalent of the fl-L1 isoform in wtPC12, scarce and represented only by sh-L1 in PC12–27 (and in the other high-REST line, SH-SY5Y). These properties were attenuated when the low-REST level of wtPC12 was increased, however of only 4-fold, by transfection of a full length construct, and when the high-REST tone of PC12–27 was weakened by TSA. Likewise, when NT2/D1 cells, differentiated to a neuronal-like phenotype by long-term treatment with RA, and primary cultures of rat neurons were transduced with REST, they decreased their L1CAM expression and switched large part of their L1CAM from the fl-L1 to the sh-L1 isoform. In the investigated neural cells, therefore, not only expression, but also splicing of L1CAM depends on REST. The different expression of the two L1CAM isoforms is important also functionally, with faster migration of PC12–27 cells transfected with fl-L1 with respect to those transfected with sh-L1, demonstrated by a scratch assay.
Additional factors previously reported to play a role in the transcription of the L1CAM gene do not seem as relevant as REST in neural cells. The repressor NFI-A appears to operate in parallel to, and possibly synergistically with REST, as previously hypothesized (Schneegans et al. 2009). β-Catenin, known as the stimulator of L1CAM gene transcription in various non-neural cancer cell types (Gavert et al. 2005; Pfeiffer et al. 2010), was investigated by transfection of its truncated construct, Δ90-βcat, known to stimulate transcription of target genes upon accumulation in the nucleus (Chenn and Walsh 2002). In spite of these properties, Δ90-βcat induced no significant change of the total L1CAM. A decrease of the fl-L1/sh-L1 ratio, observed in the Δ90-βcat-transfected cells, was most likely sustained by an increased expression of the REST gene (Nishihara et al. 2003) with ensuing 3- to 4-fold increase of the REST levels (Tomasoni et al. 2011). These results suggest that, in a low-REST neural cell such as wtPC12, the stimulatory role of β-catenin on L1CAM transcription is marginal, if any.
The most important result of this study was however the identification of Nova2, dependent on REST for its expression, as the splicing factor that governs the generation of the fl-L1 and sh-L1 isoforms of L1CAM. Together with Nova1, Nova2 was the first discovered mammalian splicing factor specific of neural cells. Its key relevance the brain was documented by a role in ∼7% of the specific splicing of the neocortex, with preference for mRNAs encoding proteins active at synapses, including 7 adhesion proteins (Ule et al. 2005). L1CAM, however, had not been identified as a Nova2 target. The REST dependence of Nova2 expression, envisaged by Johnson et al. (2007), was demonstrated here by both the ChIP of the Nova2 gene by anti-REST Abs and the large switch of the fl-L1 to the sh-L1 isoform, observed in the low-REST wtPC12 cells knocked-down of Nova2. These results, which explain why the inclusion in the L1CAM mRNA of exons 2 and 27 by alternative splicing takes place in the low-REST cells, where Nova2 is expressed, and not in the high-REST cells, where Nova2 is low to inappreciable, were duplicated in the neuron-like NT2-N cells and in primary neuronal cultures of the rat brain cortex. The REST-dependence of the L1CAM alternative splicing appears therefore as a property of many, possibly all types of neural cells. A model illustrating the REST control of Nova2 expression and the ensuing consequences on L1CAM splicing is shown in Fig. 6.
In conclusion, our studies of neural cells: PC12, SH-SY5Y, NT2/D1 differentiated by RA and primary neurons, have shown that not only the expression but also the alternative splicing of L1CAM are governed by REST acting in two ways: directly by repressing the transcription of the gene, and indirectly by repressing the transcription of the Nova2 gene. The first mechanism was known since long (Kallunki et al. 1997), the second is new, and its relevance appears already considerable. In fact, numerous important functions of L1CAM, including migration of neurons, outgrowth/fasciculation/regeneration of axons, and synaptic plasticity, depend specifically on fl-L1, the isoform spliced by Nova2. In addition, splicing by Nova2 extends to many other targets playing critical roles in brain-specific processes such as synapse assembly and axon guidance (Ule et al. 2005; Allen et al. 2010; Yano et al. 2010). The REST-dependence of the neural-specific alternative splicing, demonstrated here for Nova2 and by others for nSR100 in a paper (Raj et al. 2011) appeared during the revision of this paper, expand the role of the transcription factor to a new area. The gene targets affected by REST via this new way could be identified not by the occurrence in their promoter of RE-1, the classical REST-binding sequence, but by binding sites for REST-dependent splicing factors at critical sites within exons and/or introns of their encoded mRNA precursors. Changes of REST levels occur in neural cell during development (Ballas and Mandel 2005; Ooi and Wood 2007) and also upon maturation, when the transcription factor can moderately increase and fluctuate (Garriga-Canut et al. 2006; Spencer et al. 2006). The REST-dependent alternative splicing can therefore contribute to the array of effects governed by the transcription factor during the whole life of neural cells.
The authors, who have no conflicts of interest to declare, are grateful to Michela Dell’Orco and Stefania Fiordaliso who participated in some experiments; and to Gabriella Racchetti for general experimental help. We thank Anjen Chenn, Christopher Walsh and Luca Muzio for the truncated β-catenin construct; U. Cavallaro and A. Krueger for the fl-L1 and sh-L1 constructs; Davide Pozzi and Pietro Baldelli for the lentiviral vectors containing the REST and the scrambled constructs. Supported by Telethon (grant GGGP09066, to JM); the European Union and the Italian Ministry of Health (FP7-People-2009-IRG 249249 and RF-ICH-2009-1304134, to JMik); the Ministry of Education and Science, Republic of Serbia (173051, 143028 and a grant for training of young postdoctoral researchers, to AK).