Evolution Increases Primates Brain Complexity Extending RbFOX1 Splicing Activity to LSD1 Modulation

RbFOX1 promotes the alternative usage of a cryptic 3′ AG splice site in LSD1 gene generating additional splicing isoforms

Considering that RbFOX1 is a brain-enriched RNA-binding protein implicated in homeostatic splicing regulation (Lee et al., 2009; Gehman et al., 2011; Amir-Zilberstein et al., 2012), we hypothesized its possible role in controlling, alongside nSR100 and NOVA1 (Rusconi et al., 2015), alternative splicing of LSD1 exon E8a, and the consequent generation of neuroLSD1. RbFOX1 exerts its function as splicing factor binding the conserved consensus motif (U)GCAUG within introns. In the human LSD1 INT8a-9 (downstream of exon E8a), we identified two RbFOX1 consensus sequences very close to each other (Fig. 1A). To evaluate RbFOX1 contribution to E8a splicing, we exploited a minigene reporter assay (Baralle and Baralle, 2005; Rusconi et al., 2015). We cloned the inherent 2749 nucleotide-long human genomic region (chr1:23065790-23068538) in pBSplicing plasmid, spanning from ∼400 bp upstream of exon E8a to the whole downstream INT8a-9 (hMG2700).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

RbFOX1 regulates the inclusion of a novel LSD1 cryptic exon (E8b) in minigene reporter assays. In human tissues, this cryptic exon is present in two versions: brain-specific E8b and ubiquitously expressed E9-long. A, The “peaks and valley” graph, generated with Genome Vista browser, shows a high degree of conservation among mammalian species at specific genomic coordinates (human GRCh38/hg38) of LSD1 INT8-9. The enlarged schematic representation of exon E8a downstream intron shows the localization of E8a (purple), E8b (green), the two RbFOX1 consensus motifs (red), and E9 (purple). DNA sequence of exon E8b (highlighted in green) and its intronic flanking regions are reported. Embedded premature (TGA) STOP codon is in bold. Bold represents E8b splice sites. Red represents RbFOX1 consensus motifs. Purple represents first 110 nucleotide of exon E9. B, GeneMapper capillary gel electrophoresis analysis of minigene amplicons by rqfRT-PCR performed SH-SY5Y cells transfected with hMG2700 alone, or cotransfected with nSR100 and/or RbFOX1. The transfection of hMG2700 alone generates the minigene minimal transcript (MG, 246 bp, peak 1) and a second transcript including the 12 nt-long E8a (MG-E8a, 246 + 12 = 258 bp, peak 2) whose expression is increased by nSR100 transfection. After RbFOX1 transfection, unexpected splicing event entails exonization of a further 77-bp-long LSD1 intronic region (MG-E8b, 246 + 77 = 323 bp, peak 3), the exon E8b. On nSR100 cotransfection, exon E8b can also be included in E8a-containing minigene transcripts (MG-E8a+E8b, 246 + 12 + 77 = 335 bp, peak 4). B′, By loading a higher amount of minigene PCR products on the capillary electrophoresis, regardless the consequent saturation of the minigene-derived transcript without alternative exons (peak 1), E8a and E8b corresponding peaks (peak 2 and 3) appeared in the electropherogram trace, showing that E8a and E8b, respectively, are included in minigene-derived transcripts also when nSR100 and RbFOX1 are not overexpressed. Left, When nSR100 is overexpressed, E8a+E8b-including isoform appears (peak 4) in addition to MG-E8b (peak 3). C, GeneMapper capillary gel electrophoresis analysis of minigene amplicons by rqfRT-PCR performed in SH-SY5Y cells transfected with hMG2700 alone (CTR), or cotransfected with RbFOX1 or RbFOX2. The transfection of hMG2700 alone generates the minigene minimal transcript (MG, 246 bp) and a second transcript including the 12-nt-long E8a (MG-E8a, 246 + 12 = 258 bp). Upon transfection, both RbFOX1 and RbFOX2 induce the exonization of E8b (MG-E8b, 246 + 77 = 323 bp). Comparison of E8b inclusion in minigene-derived transcripts on RbFOX1 and RbFOX2 overexpression (F_(2,15) = 780.3, p < 0.0001). Data are mean ± SEM. *Compared with CTR. ^#Compared with RbFOX1. ****p < 0.0001, ^####p < 0.0001, one-way ANOVA, Tukey's post hoc test. D, panLSD1-isoforms RT-PCR (top, PCR1) and LSD1-E8b specific RT-PCR (middle, PCR2) performed in human tissues (β-actin is shown). PCR strategies are indicated. *An aspecific band. E, Schematic representation of PCR2 amplicons. In human tissues, the prevailing isoform is LSD1-E9-long; LSD1-E8b is only present in brain tissues.

We cotransfected the hMG2700 reporter plasmid with or without the E8a master regulator nSR100 (pCDNA3.1 Flag-nSR100) in SH-SY5Y cells and analyzed all minigene-derived transcripts using rqfRT-PCR with vector-specific primers (Baralle et al., 2003; Baralle and Baralle, 2005) (Fig. 1B). The minigene-derived transcript is identified by a PCR fragment of 246 bp (MG) (Fig. 1B). As previously shown, Flag-nSR100 cotransfection highlights a second peak of 258 bp relative to a second minigene-derived transcript that includes 12-bp-long exon E8a (MG-E8a) (Fig. 1B). Unexpectedly, when contransfecting the minigene with HA-RbFOX1 alone, an unknown transcript appeared, including additional 77 bp to the minigene-derived transcript (total of 323 bp corresponding to MG + 77bp). When HA-RbFOX1 is cotransfected with Flag-nSR100, the novel 77 bp additional fragment is included within MG and MG-E8a transcripts. Hence, a total of four peaks are detectable, which include MG (peak 1), MG-E8a (peak 2), MG + 77bp (peak 3), and another transcript MG-E8a + 77bp of 335 bp (peak 4) (Fig. 1B). To assess the nature of this 77-bp-long fragment, we sequenced the PCR product relative to the MG + 77bp and the MG-E8a + 77bp amplicons. This analysis unraveled the existence of a novel LSD1 alternative exon, whose splicing is promoted by RbFOX1 that we named exon E8b (Fig. 1A). MG-E8a (peak 2), MG-E8b (peak 3), and MG-E8a+E8b (peak 4) corresponding peaks can be detected also without nSR100 and RbFOX1 cotransfection when a higher amount of amplified DNA is loaded, which however leads to peak 1 and 2 saturation (Fig. 1B′). Exon E8b seems to represent a true cryptic exon with acceptor (AG) and donor (GU) splice sites that includes an in-frame stop codon potentially leading to the degradation of endogenous LSD1 transcripts via NMD (Fig. 1A). Notably, exon E8b has never been reported. This is presumably because of rapid physiological NMD, a cellular process that requires protein synthesis (Chang et al., 2007). We could easily identify minigene-derived transcripts including exon E8b because transcripts generated from this plasmid do not assemble ribosomal subunits, due to the designed absence of the AUG codon in pBSplicing vector (Baralle and Baralle, 2005); hence, they are not subjected to NMD (Rusconi and Battaglioli, 2018).

RbFOX1 is highly enriched in brain and muscles, while in all other tissues an RbFOX1 paralog, namely, RbFOX2, is prevalent (Kuroyanagi, 2009). Indeed, RbFOX1 and RbFOX2 RNA-binding domains share the same target sequence (UGCAUG) (Kuroyanagi, 2009). We therefore tested the ability of RbFOX2 to induce E8b exonization in hMG2700 when overexpressed, showing that also RbFOX2 is able to induce E8b inclusion in minigene transcripts. However, RbFOX2 promotes E8b exonization 4 times less efficiently compared with RbFOX1 (Fig. 1C, CTR vs RbFOX1, 0.74 ± 0.22 vs 47.18 ± 1.35, ****p < 0.0001; CTR vs RbFOX2, 0.74 ± 0.22 vs 11.83 ± 0.58, ****p < 0.0001; RbFOX1 vs RbFOX2, 47.18 ± 1.35 vs 11.83 ± 0.58, ^####p < 0.0001), suggesting that RbFOX1 and RbFOX2 are not fully redundant as LSD1 regulators.

In order to verify the existence of LSD1-E8b isoforms in vivo, we analyzed a panel of human tissues. LSD1 can be visualized with two RT-PCR bands corresponding to LSD1 with or without alternative exon E2a (E8a cannot be resolved being only 12 nt in length) (Fig. 1D, PCR1). PCR1 does not show any further amplicon, except the two above mentioned, including LSD1 and LSD1-E2a related sequences. For this reason, we decided to modify the PCR probes to increase isoform selectivity and detectability. Hypothesizing that transcripts including the further 77 bp could be expressed at very low levels because of a rapid degradation, we designed E8b-specific forward primer and LSD1 E10-annealing reverse primer (PCR2, Fig. 1D). With this approach, we successfully detected E8b-including amplicons in the human brain. Interestingly, with PCR2, we also detected in all tissues, a second slower migrating PCR band, which on sequencing resulted to include exon E8b along with its intronic downstream sequences. As such, E9-long novel exon is generated, which comprises E8b sequence together with exonized INT8a-9 and E9 (Fig. 1D, right). Thus, the exon E8b identified by minigene experiments (77 bp long) does exist in brain-derived LSD1 transcripts (Fig. 1D). However, in all other human tissues analyzed, the E8b GU donor splice site is generally neglected, becoming the dominant donor splice site of exon E9, which makes E9-long prevailingly spliced into LSD1 transcripts. Figure 1E shows a schematic representation of the two possible LSD1 alternative exons, respectively, E8b and E9-long highlighted by PCR2.

Taking into consideration that E8b and E9-long have in common the same in-frame stop codon, the two isoforms should be functionally equivalent and both candidates to be subjected to NMD. As both are present in human tissues, although E9-long is more widespread and E8b restricted to the brain, we will refer to E8b/E9-long as the alternative LSD1 exons whose functional significance is studied throughout this work.

Splicing of the cryptic LSD1 exons E8b/E9-long is a higher primates' acquisition

While the genomic region encompassing exons E8a and E8b is not conserved in other vertebrates, in mammals, this intronic sequence is particularly highly conserved (Fig. 1A). We proceeded by aligning the identified exon E8b/E9-long sequence among mammals (Fig. 2A). Interestingly, our analysis showed that exon E8b/E9-long acceptor splice site (AG dinucleotide) is only present in the genome of higher primates, suggesting that E8b/E9-long splicing is restricted to these species. To confirm this assumption, we performed exon E8b/E9-long specific PCR on cDNA obtained from a representative panel of nonprimates (rodents), lower primates (nonprimate monkeys, prosimians), and higher primates (great apes and humans) tissues (Fig. 2B). As E8b/E9-long inclusion in LSD1 transcripts is a ubiquitous process, we used cDNA from the mouse hippocampus, fibroblast cell lines from lemur (Eulemur fulvus 1 [EFU1], Lemur catta 1 [LCA1]), lymphoblast cell line of MMU, and human hippocampus. Two different RT-PCR strategies were performed: the PCR2, already described above, and a third PCR using a forward primer specifically annealing to LSD1-E7 and the reverse primer on exon E8b (PCR3) (Fig. 2B). As expected, from both analyses emerged that only those mammalian species holding the E8b acceptor splice site (i.e., higher primates) show exon E9-long inclusion in LSD1 transcripts (Fig. 2B). While in the human hippocampal samples PCR3 generates two amplicons (including E8b with or without E8a), in the MMU lymphoblast samples, only one PCR product was present. This is because the upper band includes neurospecific LSD1 exon E8a sequences that are not expressed in lymphoblast. As positive control, we assayed the presence of LSD1 transcripts with PCR1, as described in Figure 1D. These data indicate that the A to G substitution creates in all higher primates a new acceptor splice site for E8b/E9-long inclusion (Fig. 2A). To unambiguously demonstrate that such a genome modification is per se necessary to exonize E8b, we mutagenized the acceptor splice site AG in the hMG2700, back into the archaic AA to re-create the lower primate layout. Notably, this single mutation impedes E8b exonization (Fig. 2C) also when RbFOX1 is overexpressed. In order to investigate whether AG dinucleotide was also sufficient to induce E8b/E9-long splicing, we cloned lemur genomic sequences corresponding to the human MG2700 and mutagenized the 3′ acceptor splice site of E8b/E9-long (the AA dinucleotide) into AG (Fig. 2D). Regardless of RbFOX1 overexpression, we did not observe E8b inclusion in the mutagenized lemur minigene. This experiment indicates that AG dinucleotide is necessary but not sufficient to promote E8b/E9-long exonization. Other E8b-flanking genomic sequences among the few which differ between lemur and simiiformes, must be necessary but not sufficient for E8b inclusion.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Exon E8b/E9-long inclusion is a higher primate acquisition. A, Multiple alignment of the human LSD1 exon E8b/E9-long 5′ end in mammals. E8b/E9-long AG splice acceptor site (highlighted in yellow) is only present in higher primates. B, LSD1 isoform detections by RT-PCR (PCR1, PCR2, and PCR3) in mRNA extracted from mouse hippocampus, lemur (EFU1, LCA1) fibroblast cell lines, MMU lymphoblast cell line, and human hippocampus. PCR strategies are indicated. PCR3 reverse primer anneals on E8b, amplifying both E8b- and E9-long-including LSD1 isoforms (a schematic view of all LSD1 splicing isoforms is presented in Fig. 6). C, In the human minigene MG-2700, mutation of AG E8b acceptor splice site into the archaic AA dinucleotide prevents RbFOX1-mediated E8b exonization. GeneMapper capillary gel electrophoresis analysis of minigene amplicons by rqfRT-PCR performed in SH-SY5Y cells transfected with WT hMG2700 [AG], or mutant hMG2700 [AA] alone, or cotransfected with RbFOX1. D, GeneMapper capillary gel electrophoresis analysis of minigene amplicons by rqfRT-PCR performed in SH-SY5Y cells transfected with WT lemur MG2700 [AA], or mutant lemur MG2700 [AG] alone or cotransfected with RbFOX1.

Exon E8b/E9-long inclusion regulates LSD1 expression via NMD

Near the 5′ end of exon E8b/E9-long, a premature stop codon is present (Fig. 2A). Hence, we investigated the ability of its inclusion to mediate LSD1 transcript degradation through NMD. To inhibit NMD, we treated human and rodent cell lines with a compound capable of blocking protein synthesis, such as CHX. We applied CHX for 6 and 8 h, scoring LSD1 transcripts increase by qRT-PCR. Notably, NMD inhibition increased total LSD1 transcripts in the human neuroblastoma cell line SH-SY5Y (Fig. 3A, LSD1 CTR vs CHX 6 h, 0.94 ± 0.05 vs 1.46 ± 0.16, **p = 0.0073; CTR vs CHX 8 h, 0.94 ± 0.05 vs 1.36 ± 0.14, *p = 0.0383). As positive control, we measured in SH-SY5Y a transcript known to undergo NMD, namely, PRKCA (Cardamone et al., 2017) (Fig. 3A, PRKCA CTR vs CHX 6 h, 0.85 ± 0.17 vs 2.05 ± 0.20, *p = 0.0261; CTR vs CHX 8 h, 0.85 ± 0.17 vs 2.50 ± 0.32, **p = 0.0061). We also analyzed the specific modulation of E8b/E9-long-including LSD1 transcripts on CHX treatment and observed more than a fourfold increase of this isoform mRNA when NMD is inhibited (Fig. 3B, LSD1-E8b/E9-long CTR vs CHX 6 h, 1.02 ± 0.07 vs 4.53 ± 0.48, ***p = 0.0003; CTR vs CHX 8 h, 1.02 ± 0.07 vs 4.98 ± 0.61, ***p = 0.0001). Thus, on NMD inhibition, E8b/E9-long including transcripts appear to increase more than total LSD1 transcripts (compare Fig. 3B with Fig. 3A), which is consistent with E8b/E9-long splicing responsible for LSD1 transcript degradation. Indeed, among all LSD1 transcript isoforms, only the fraction including E8b/E9-long exon is prevented by CHX treatment from NMD-induced degradation, supplying a mechanistic explanation for overall increase of total LSD1 isoforms. We repeated these experiments with two other human cell lines (SK-N-BE and COV434), confirming E8b/E9-long-induced LSD1 degradation (not shown). To strengthen the link between exon E8b/E9-long and NMD-mediated decrease of LSD1, we performed the same CHX experiment in a murine cell line, in which E8b/E9-long splicing inclusion cannot occur because of the absence of 3′ AG acceptor splice site. In the murine neuroblastoma cell line N2a, NMD inhibition did not exert any effect on LSD1 transcripts, whereas it increased the levels of a known positive control, namely, HNRNPL (Fig. 3C, LSD1 CTR vs CHX 6 h, 0.69 ± 0.18 vs 0.50 ± 0.04, p = 0.36; HNRNPL CTR vs CHX 6 h, 1.36 ± 0.20 vs 28.91 ± 2.01, ***p = 0.0002).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Splicing inclusion of E8b/E9-long in LSD1 mature transcripts causes LSD1 transcript degradation via NMD. A, LSD1 transcript level (normalized over RPSA) on NMD inhibition (CHX, 100 µg/ml for 6/8 h) in human cell line SH-SY5Y assessed by qRT-PCR (F_(2,14) = 7.79, p = 0.0053). PRKCA1 was used as NMD-positive control (F_(2,6) = 13.18, p = 0.0064). B, Representative agarose gel visualization of upstream LSD1-E8b/E9-long specific RT-PCR (PCR3) on SH-SY5Y cell line after CHX treatment and corresponding quantification of LSD1-E8b transcripts, normalized over RPSA (F_(2,8) = 42.02, p < 0.0001). C, LSD1 transcript level (normalized over GJA1) on NMD inhibition in a mouse (N2a) cell line, assessed by qRT-PCR. HNRNPL was used as NMD-positive control. Data are mean ± SEM. Compared with CTR (untreated cells): *p < 0.05, **p < 0.01, ***p < 0.001, Student's t test for simple comparisons; one-way ANOVA, Tukey's post hoc test for multiple comparisons. D, Relative abundance of LSD1 canonical transcripts (LSD1 and LSD1-E2a) and of their variants with E8b/E9-long (LSD1-E8/E9-long and LSD1-E2a+E8/E9-long) in RNA-seq data of UPF1 KD and SMG6-7 kD performed in HeLa cells (Colombo et al., 2017) (n = 3 individual for each condition) KD, knock down; scr, scramble. For box-and-whisker plots, the central line, box, and whiskers represent the median, interquartile range from first to third quartiles, and 1.5 × interquartile range, respectively.

To provide a further proof that LSD1 levels are regulated by NMD via alternative inclusion of E8b/E9-long, we took advantage of available RNA-seq datasets obtained by specific knockdown of NMD-instrumental factors, such as UPF1 and SMG6-7 in HeLa cells (Colombo et al., 2017). With this approach, we observed a modest increase of E8b/E9-long-including LSD1 transcript isoforms (Fig. 3D). However, in this cell line, the overall level of E8b/E9-long-including LSD1 transcript is extremely low (not shown), and we could not detect any appreciable modulation of total LSD1 mRNA. This could because of the very low levels of RbFOX1 and RbFOX2 expression in HeLa cells.

Together, these results suggest a novel post-transcriptional mechanism of negative LSD1 regulation, which is restricted to higher primates and humans.

RbFOX1 negatively modulates LSD1 expression via exon E8b/E9-long splicing, ultimately regulating the IEGs

Minigene splicing assays suggested that E8b/E9-long inclusion into mature LSD1 transcripts requires RbFOX1, which seems to be necessary and sufficient to promote LSD1 exon E8b/E9-long alternative splicing in higher primates. To verify whether RbFOX1 regulates LSD1 expression via controlling E8b/E9-long splicing into LSD1 transcripts, we generated a RbFOX1 DOX-inducible cell line. We used SH-SY5Y TetOn cells, stably transfecting pTRE2hyg-HA-RbFOX1 (TetOn RbFOX1). As control, we used SH-SY5Y TetOn parental cell line. At this point, we examined HA-RbFOX1 effects on LSD1 expression at the protein and mRNA levels. Upon DOX treatment, which induces HA-RbFOX1 protein expression, we could observe significant LSD1 protein downregulation compared with its control (Fig. 4A, TetOn RbFOX1 CTR vs DOX, 0.93 ± 0.04 vs 0.73 ± 0.04, *p = 0.0108). On the contrary, LSD1 protein level did not change in SH-SY5Y TetOn cells with or without DOX (Fig. 4A, TetOn CTR vs DOX, 1.12 ± 0.08 vs 1.12 ± 0.08, p = 0.9989). Consistently with the proposed NMD-mediated mechanism of LSD1 regulation, also total LSD1 transcript levels, amplified by qRT-PCR using panLSD1 primers, were clearly reduced only in SH-SY5Y TetOn HA-RbFOX1 cells but not in TetOn parental cell line (Fig. 4B, TetOn RbFOX1 CTR vs DOX, 1.24 ± 0.15 vs 0.79 ± 0.06, **p = 0.0055; TetOn CTR vs DOX, 0.90 ± 0.06 vs 0.93 ± 0.13, p = 0.7907). Relevantly, all these results rely on RbFOX1-mediated exonization of LSD1 exon E8b/E9-long, whose inclusion is eventually increased by DOX as evidenced with RT-PCR performed with an exon E8b-specific primer (Fig. 4C, TetOn RbFOX1 CTR vs DOX, 0.96 ± 0.15 vs 1.67 ± 0.21, *p = 0.0165; TetOn CTR vs DOX, 1.22 ± 0.28 vs 1.07 ± 0.15, p = 0.6540). As positive control, DOX-induced RbFOX1 expression modulates alternative splicing of the known RbFOX1 target, NUMA1 (Zhang et al., 2008; Denichenko et al., 2019) (Fig. 4D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

RbFOX1 regulates LSD1 expression through NMD by promoting exon E8b/E9-long inclusion. A, LSD1 protein quantification using anti-LSD1 antibody (normalized over GAPDH) on DOX induction (1 µg/ml for 24 h) of HA-RbFOX1 in SH-SY5Y clonal line (SH-SY5Y TetOn HA-RbFOX1) compared with SH-SY5Y parental line (TetOn). HA antibody was used to visualize RbFOX1 overexpression (+, positive control for HA-RbFOX1 expression). B, LSD1 transcript, assessed by qRT-PCR (normalized over RPSA) in the two SH-SY5Y clones as in A. C, E8b/E9-long-specific RT-PCR quantification over RPSA in the two SH-SY5Y clones as in A. D, RbFOX1 overexpression regulates alternative splicing of NUMA1 exon E13 (known RbFOX1/2 target gene) in SH-SY5Y TetOn RbFOX1 but not in TetOn parental line. E, LSD1 overexpression in SH-SY5Y cells elicits IEG downregulation: qRT-PCR analysis of the IEGs Egr1, C-Fos, and Npas4 (normalized over RPSA) on LSD1 overexpression (LSD1 OEX) in SH-SY5Y cells. F, qRT-PCR analysis shows that RbFOX1 overexpression regulates LSD1 downstream IEGs targets Egr1, c-Fos, and Npas4 (normalized over RPSA) in TetOn HA-RbFOX1 but not in TetOn parental line. G, RbFOX1 does not regulate LSD1 in lower primates. Overexpression of RbFOX1 (OEX) in mouse-derived N2a cell line (24 h) does not affect LSD1 protein level and mRNA expression (VEC, vector). H, RT-PCR analysis of grin1 E5 alternative splicing (AS) on RbFOX1 overexpression (OEX). I, qRT-PCR analysis of the IEGs egr1, c-fos, and npas4 (normalized over RPSA) on RbFOX1 overexpression (OEX) in N2a. In RbFOX1 overexpression experiments, data are mean ± SEM. *Compared with CTR or VEC in SH-SH5Y or N2a cells, respectively. *p < 0.05, **p < 0.01, Student's t test. In LSD1 overexpression experiment, data are mean ± SEM. *Compared with VEC. *p < 0.05, **p < 0.01, Mann–Whitney test.

LSD1 is a transcriptional regulator of the IEGs (Toffolo et al., 2014; Wang et al., 2015; Rusconi et al., 2016); consistently, its overexpression in SH-SY5Y cell line leads to Egr1 and C-fos downregulation (Fig. 4E, VEC vs LSD1 OEX; Egr1, 1.08 ± 0.08 vs 0.77 ± 0.09, *p = 0.0291; C-Fos, 1.05 ± 0.08 vs 0.64 ± 0.06, **p = 0.0012; Npas4, 1.08 ± 0.12 vs 0.78 ± 0.09, p = 0.0676). In order to verify whether RbFOX1-mediated LSD1 downregulation also impacts the expression of LSD1 downstream targets, we measured IEG expression on RbFOX1 induction by DOX treatment. In Figure 4F, we show that Egr1, C-Fos, and Npas4 are significantly upregulated on RbFOX1 overexpression. This modulation did not occur in the TetOn parental SH-SY5Y clone (Fig. 4F, TetOn RbFOX1, Egr1 CTR vs DOX, 0.92 ± 0.09 vs 1.58 ± 0.15, **p = 0.0055; C-Fos CTR vs DOX, 0.75 ± 0.12 vs 1.30 ± 0.10, **p = 0.0084; Npas4 CTR vs DOX, 0.95 ± 0.04 vs 1.44 ± 0.21, *p = 0.0318; TetOn, Egr1 CTR vs DOX, 0.96 ± 0.09 vs 1.02 ± 0.19, p = 0.7811; C-Fos CTR vs DOX, 1.12 ± 0.10 vs 1.05 ± 0.13, p = 0.6822; Npas4 CTR vs DOX, 1.78 ± 0.57 vs 1.30 ± 0.31, p = 0.4760).

Once again, to fortify the link between exon E8b/E9-long and NMD-mediated decrease of LSD1, we overexpressed RbFOX1 in mouse neuroblastoma N2a cell line. This approach did not affect LSD1 level at either the mRNA or the protein level (Fig. 4G, LSD1 protein VEC vs RbFOX1 OEX, 0.97 ± 0.04 vs 0.84 ± 0.05, p = 0.0625; LSD1 mRNA VEC vs RbFOX1 OEX, 0.98 ± 0.04 vs 0.91 ± 0.07, p = 0.3862), although able to modify the splicing of the known RbFOX1 target grin1 (Lee et al., 2009) (Fig. 4H). This is again consistent with the absence of a functional E8b/E9-long dinucleotide acceptor splice site in lower mammals (Fig. 2A). Remarkably, LSD1 target genes were not modified in mouse N2a cells overexpressing RbFOX1 with the eminent exception of c-fos (Fig. 4I, egr1 VEC vs RbFOX1 OEX, 0.88 ± 0.05 vs 0.79 ± 0.09, p = 0.3722; c-fos VEC vs RbFOX1 OEX, 0.96 ± 0.06 vs 1.75 ± 0.26, **p = 0.0074; npas4 VEC vs RbFOX1 OEX, 1.03 ± 0.08 vs 0.95 ± 0.17, p = 0.6648), whose increased level suggests an alternative, LSD1-independent RbFOX1-mediated regulation.

Lsd1 exons E8a and E8b/E9-long are preferentially spliced together impacting LSD1/neuroLSD1 ratio

In mammals LSD1 intron INT8-9 shows many peculiarities. Indeed, this single intron contains two alternative exons of predicted functional importance, E8a and E8b/E9-long. As discussed until now, inclusion of exon E8b/E9-long can modulate overall LSD1 expression in a neuronal but in principle also in non-neuronal contexts. However, in the brain, exon E8b/E9-long inclusion could have an additional effect that is to fine-tune the LSD1/neuroLSD1 ratio. This could happen if E8b/E9-long were preferentially included in a specific LSD1 isoform (i.e., if E8a and E8b/E9-long were preferentially spliced together).

We quantitatively analyzed minigene splicing experiments performed in SH-SY5Y cells, in which minigene hMG2700 was cotransfected with Flag-nSR100 or with both Flag-nSR100 and HA-RbFOX1 (Fig. 5A, left). As previously shown in Figure 1B, on Flag-nSR100 overexpression, the MG transcript, including E8a (peak 2) increases and after Flag-nSR100/HA-RbFOX1 cotransfection, the two MG transcripts, including exon E8b alone (peak 3), or together with exon E8a (peak 4) rise. We observed that overall E8a inclusion, calculated as (peak2 + 4)/(peak1 + 2+3 + 4), is only dependent on nSR100 presence, being not modified by RbFOX1 (Fig. 5A, middle, nSR100 vs nSR100+RbFOX1, 31.63 ± 1.59 vs 26.85 ± 2.21, p = 0.0758). We then examined E8a inclusion frequency in the two different populations of minigene transcripts: those skipping E8b (MG and MG+E8a, peaks 1 and 2) and those including E8b (MG+E8b and MG+E8a/E8b, peaks 3 and 4). E8a frequency within E8b nonincluding and including transcripts was plotted and calculated as peak2/(peak1 + 2) and peak4/(peak3 + 4), respectively (Fig. 5A, right, E8a inclusion in nSR100+RbFOX1, MG vs MG-E8b, 16.85 ± 2.41 vs 31.55 ± 1.96, ****p < 0.0001). From this analysis, we clearly established that E8a is preferentially included together with E8b. Considering that minigene-derived transcripts do not undergo NMD, with this experiment, we foresee that preferential E8a/E8b inclusion should favor NMD-mediated decrease of E8a containing isoforms, namely, neuroLSD1 in the brain.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

LSD1 exons E8a and E8b/E9-long are preferentially spliced together impacting LSD1/neuroLSD1 relative ratio. A, Quantification of GeneMapper capillary gel electrophoresis of minigene products amplified by rqfRT-PCR performed SH-SY5Y cells transfected with hMG2700, together with nSR100 alone or with RbFOX1. Overall E8a splicing (left histogram) depends on nSR100 presence and is calculated as follows: (peak2 + 4)/(peak1 + 2+3 + 4). E8a inclusion percentage (right histogram) in E8b nonincluding peak2/(peak1 + 2) and including peak4/(peak3 + 4) transcripts (MG vs MG-E8b: F_(1,20) = 26.65, p < 0.0001; nSR100 vs nSR100+RbFOX1: F_(1,20) = 21.51, p = 0.0002). Data are mean ± SEM. Not significant, p > 0.05 (Student's t test). ****p < 0.0001 (two-way ANOVA Sidak's post hoc test). B, Sequence analysis of purified LSD1-related PCR band (PCR3) from human hippocampus shows LSD1 E8-E8b/E9long and E8a-E8b/E9-long splice junctions. C, Quantification of exon E8a (neuroLSD1) inclusion frequency by rqfRT-PCR in total LSD1 transcripts (PCR1, see Fig. 2B) and E8b/E9-long-containing LSD1 transcripts (PCR3, see Fig. 2B) in human hippocampus (n = 17) and in human cerebellum (n = 6); average E8a inclusion values are mechanistically evaluated with different quantitative PCRs that are anyway internally standardized (Zibetti et al., 2010). Human postmortem brain specimen description is reported in Table 1. D, Quantification of exon E8a (neuroLSD1) inclusion frequency by rqfRT-PCR in total LSD1 transcripts, in white (PCR1, see Fig. 2B), and E8b/E9-long-containing LSD1 transcripts, in green (PCR3, see Fig. 2B) in human hippocampus (n = 17) and in human cerebellum (n = 6) for each individual brain sample. Age (years) for each donor is reported. Data are mean ± SEM. **p < 0.01 (paired Student's t test). E, Nuclear (RbFOX1-FAPY) and cytoplasmic (RbFOX1-TALVP) RbFOX1 splicing isoforms (Lee et al., 2009; Gehman et al., 2011) are modulated during ASDS in the mouse hippocampus. Quantification of the relative percentage of RbFOX1 nuclear isoform in the mouse hippocampus comparing defeated mice (ASDS) to control ones (CTR) (n = 4 or 5 mice per condition) (F_(3,14) = 36.03, p < 0.0001). Agarose gel images are reported: nuclear and cytoplasmic RbFOX1 isoforms are indicated. Data are mean ± SEM. *Compared with CTR. ^#Compared with ASDS 2 h. **p < 0.01, ****p < 0.0001, ^###p < 0.001, one-way ANOVA Tukey's post hoc test.

To further investigate this hypothesis, we studied E8a and E8b/E9-long splicing in human cerebral tissues. We sequenced PCR3 amplicons obtained from human hippocampus to unambiguously confirm the existence of LSD1-E8b/E9-long including or skipping the exon E8a (Fig. 5B). Considering that exon E8b/E9-long-including isoforms are present at very low levels in vivo, as they are rapidly degraded by NMD, we exploited a strategy based on two specific internally standardized rqfRT-PCR (Zibetti et al., 2010), allowing the former (PCR1), a reliable evaluation of highly expressed LSD1 isoforms and the latter (PCR3) a measure of E8b/E9-long-containing transcripts (Fig. 5C). In principle, PCR1 should also amplify E8b/E9-long-containing amplicons, but this is not the case because they fall under the resolution threshold of the PCR. PCR3 (with E7- and E8b/E9-long-annealing forward and reverse primers, respectively) allows the identification of E8a inclusion frequency in E8b/E9-long-containing LSD1 transcripts. We reasoned that, if E8a were included in LSD1 transcripts independently on E8b/E9-long, then E8a inclusion frequency in PCR1 and PCR3 should be identical. Vice versa, if exon E8a is preferentially included together with E8b/E9-long, the E8a percentages should be different in the two PCRs. We conducted PCR1 and PCR3 on RNA extracted from human postmortem hippocampal and cerebellar samples (Table 1), comparing E8a inclusion frequency in total LSD1 transcripts (PCR1) and in E8b/E9-long-containing LSD1 isoforms (PCR3). We observed that neuroLSD1 relative percentage is higher in E8b/E9-long-including LSD1 transcripts compared with those who do not, both in human hippocampal and cerebellar samples. In the hippocampus, E8a relative percentage in E8b/E9-long-containing LSD1 transcripts is higher compared with E8a inclusion in total LSD1 transcripts (Fig. 5C, Total LSD1 vs LSD1-E8b/E9-long, 23.35 ± 1.58 vs 50.71 ± 6.56, **p = 0.0011). Similarly, in the cerebellum, relative E8a percentage inclusion in E8b-containing LSD1 transcripts is higher compared with that of total LSD1 transcripts (Fig. 5C, Total LSD1 vs LSD1-E8b/E9-long, 58.87 ± 4.21 vs 78.02 ± 4.61, p = 0.0547). Individual differences in E8a-E8b/E9-long co-splicing are shown in Figure 5D. Thus, also in endogenous human brain-derived transcripts, E8a and E8b/E9-long exons are preferentially spliced together, entailing increased degradation of neuroLSD1 (the LSD1 isoform including E8a) and affecting LSD1/neuroLSD1 ratio. All this said, we can conclude that, in addition to the above-described RbFOX1-mediated LSD1 general-level modulation, RbFOX1-operated preferential degradation of neuroLSD1 in the human brain represents a further molecular layer of LSD1/neuroLSD1 ratio regulation.

Considering the fundamental role played by LSD1/neuroLSD1 ratio modulation in allowing environmental adaptation to stress and its important readouts on stress-coping and emotional behaviors, these data indicate a new molecular pathway linking RbFOX1 function to stress responses. We here show that the social defeat stress, which significantly elicits LSD1/neuroLSD1 splicing modulation in the mouse hippocampus, also shifts the ratio between the nuclear and the cytosolic RbFOX1 isoforms in favor of the nuclear one, namely, RbFOX1-FAPY, involved in splicing regulation (Fig. 5E, RbFOX1 nuclear isoform percentage, CTR vs ASDS 2 h, 37.27 ± 0.91 vs 44.42 ± 1.44, **p = 0.0076; CTR vs ASDS 7 h, 37.27 ± 0.91 vs 55.79 ± 1.67, ****p < 0.0001; CTR vs ASDS 7 h + 24 h resting, 37.27 ± 0.91 vs 38.42 ± 1.47, p = 0.9323; ASDS 2 h vs ASDS 7 h, 44.42 ± 1.44 vs 55.79 ± 1.67, ^###p = 0.0002). Notably, in the mouse brain, the LSD1/neuroLSD1 ratio modulation cannot be contributed by increased nuclear RbFOX1 splicing isoform, indeed, as explained above, RbFOX1-LSD1 functional interaction is a higher primates' acquisition. Nonetheless, the evolutionary conserved (functionally independent) molecular stress-homeostatic pathways of RbFOX1 and LSD1 are functionally linked in higher primates thanks to LSD1 E8b/E9-long exons.