Introduction

Motor neurons (MNs) generate complex networks that propagate nerve impulses from the central nervous system (CNS) to peripheral tissues, where they are translated into muscle contraction [1, 2]. Special attention has recently been trained to MN development and function, since these cells are particularly vulnerable in severe degenerative diseases [3]. In mammals, upper MNs establish glutamatergic connections with lower cholinergic MNs, conveying the information to the target muscles. Among lower MNs, spinal MNs (spMNs) represent the ultimate link of the neuronal circuit to skeletal muscles.

MN lineage determination is dictated by signaling pathways coordinating the assembly of cell type-specific transcription factors (TFs), which trigger MN-specific gene programs, eventually suppressing alternative differentiation fates. The TFs OLIG2 and HB9 are paradigmatic for inducing spMN specification while inhibiting the interneuron (IN) identity [4, 5]. A class of regulatory noncoding RNAs that also plays a central role in MN differentiation are the long noncoding RNAs (lncRNAs), that share several features with messenger RNAs (mRNAs) but are not translated into proteins [6]. They are implicated in most cellular events for their function as transcriptional and post-transcriptional regulators of gene expression, conjugated with their capability to simultaneously interact with nucleic acids and proteins [7]. These distinctive traits rely on their ability to fold into multiple functional domains, achieving a modular organization through which they bind and coordinate the activity of different factors. This makes lncRNAs very versatile molecules endowed with high specificity of action.

Their pervasive expression in the CNS [8], where they are tightly regulated in space and time, strongly argues for their implication in neurodevelopmental processes and in human brain evolution [9, 10]. Moreover, the profound impact of lncRNA deregulation on MN pathophysiology is diagnostic of their involvement in orchestrating cell fate decision, cell identity and function of this neuronal subtype [6]. Despite this, very few lncRNAs have been associated with MN development and activity [11], with a main function as transcriptional regulators [6, 7].

We previously characterized the neuronal isoform of HOTAIRM1, which we referred to as nHOTAIRM1, as a lncRNA upregulated during in vitro neuronal differentiation of SH-SY5Y cells, considered as equivalent to neuronal precursors [12]. Physiologically, among the brain tissues, nHOTAIRM1 is exclusively expressed in the spinal cord (SC). In line with this observation, we found that it accumulates in MN-enriched ventral SC lineages, differentiated from human induced pluripotent stem cells (iPSCs) [12]. This evidence prompted us to explore whether nHOTAIRM1 might impact MN generation and function, to finally realize how its deregulation may affect MN life cycle. Here, we applied a genome editing-based loss-of-function approach to a model system that recapitulates spMN differentiation and identified key nHOTAIRM1 target genes implicated in MN development and function. We show that nHOTAIRM1 is critical in multiple aspects of MN physiology, from their formation, at the expenses of INs, to the achievement of their correct morphology and activity.

Results

nHOTAIRM1 is induced in a model of spMN differentiation

iPSCs were differentiated into functional spMNs, following a protocol based on lineage-specific inducible transcriptional programming modules (NIL) [13, 14]. The progression of differentiation induced by doxycycline (Dox) treatment (Fig. 1A) was monitored via qRT-PCR analysis of specific cellular markers (Fig. 1B). NANOG and OCT4 were used as iPSC pluripotency markers (D0), whereas the conversion of iPSCs into MN progenitors (MNPs, D5) was signed by the expression of the pan-neuronal marker TUJ1 and of the early marker of cholinergic MNs MNX1 (referred as HB9 throughout the text) [15]. At this stage, MNPs were dissociated and re-plated for further maturation. Three days later (D8), cells displayed consistent expression of ChAT, required for the synthesis of the neurotransmitter acetylcholine [16]. On day 12 (D12), a marked induction was observed for LHX3 and ISLET1, whose coordinated expression directs cells to become spMNs [17]. At this stage, the cell population is almost exclusively composed of post-mitotic spMNs displaying the properties of functional MNs [14]. In this differentiation set, we profiled nHOTAIRM1 expression (Fig. 1B). Almost undetectable in iPSCs, it was gradually induced in MNPs, reaching its peak of expression in spMNs (D12). This marked increase in expression suggests the involvement of nHOTAIRM1 in spMN programming, whereas its high expression levels in post-mitotic MNs also indicate a potential role in spMN activity.

Fig. 1: nHOTAIRM1 expression profile along spMN differentiation and its distribution in post-mitotic spMNs.
figure 1

A Schematic representation of a 12-day protocol for differentiating functional spMNs from iPSCs. Cells were induced for 5 days by doxycycline treatment (DOX induction, D0-D5) for the expression of programming TFs and then cultured for further 7 days (D12) for spMN maturation. B Expression profiling of differentiation markers and nHOTAIRM1 along iPSC conversion to spMNs. qRT-PCR analysis of markers for: pluripotency (NANOG, OCT4), MN progenitors (TUJ1, HB9), MN specification (ISLET1), spinal identity (LHX3), electrophysiological activity (ChAT). nHOTAIRM1 expression is indicated by black bars. Differentiation days are indicated on the x-axis. Data (means ± SEM) are expressed as arbitrary units, relative to ATP5O mRNA levels, as internal control. For each target, the peak of expression was set as 1. D0 was considered as the reference sample for statistical tests. C qRT-PCR analysis of nHOTAIRM1 nucleus/cytoplasm localization in spMNs (D12). Left panel: the efficiency of nucleus/cytoplasm fractionation in spMNs was assessed by evaluating pre-GAPDH and GAPDH RNA levels, respectively. Normalization was performed on total RNA. Data (means ± SEM) are expressed as percentage of total GAPDH or pre-GAPDH levels. Right panel: nHOTAIRM1 distribution in the nucleus (dark gray) and cytoplasm (light gray) of spMNs. Normalization was performed on total RNA. Data (means ± SEM) are expressed as percentage of total nHOTAIRM1 levels. D qRT-PCR analysis of nHOTAIRM1 distribution in the soma and the neurites of spMNs (D12). Left panel: efficiency of soma/neurite separation from spMNs was assessed by evaluating the levels of GNG3 and COL3A1 mRNAs, respectively. Normalization was performed on total RNA. Data (means ± SEM) are expressed as percentage of total GNG3 and COL3A1 levels. Right panel: nHOTAIRM1 distribution in the neuritic (dark gray) and somatic (light gray) compartments. Normalization was performed on total RNA. Data (means ± SEM) are expressed as percentage of total nHOTAIRM1 levels. In all the panels, statistics is as follows: N = 3 biological replicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, two-tailed Student’s t test.

Full size image

nHOTAIRM1 is mainly cytoplasmic and localized in the soma and the neurites of spMNs

To unravel the function of nHOTAIRM1 in spMNs, we analyzed its subcellular localization. Biochemical fractionation and expression analysis in spMNs (D12), using GAPDH and pre-GAPDH as controls (Fig. 1C, left panel), revealed that nHOTAIRM1 is predominantly cytoplasmic (about 83%, Fig. 1C, right panel). This suggests its potential role as a post-transcriptional regulator of gene expression. We next deepened the distribution of nHOTAIRM1 in spMN cell bodies (somata) and neurites. Upon soma/neurite separation, assessed by checking GNG3 and COL3A1 localizations, respectively [18, 19] (Fig. 1D left panel), the relative enrichment of the lncRNA was quantified (Fig. 1D, right panel), revealing its occurrence at comparable amounts in both districts, with the cell bodies being also contributed by the nuclear counterpart.

This distribution was confirmed by imaging assays combining RNA fluorescence in situ hybridization (RNA FISH) with immunofluorescence (IF). Figure 2 shows that nHOTAIRM1 (indicated by red spots) is enriched in the soma, but also accumulates in the neurites, as well as in distal axonal segments (yellow arrowheads). These findings imply potential roles for nHOTAIRM1 in diverse functions such as neurite outgrowth, axonal transport, local translation of axon-resident mRNAs or synaptic activity.

Fig. 2: Visualization of nHOTAIRM1 subcellular localization in post-mitotic spMNs.
figure 2

RNA FISH of nHOTARIM1 (red spots) combined with TUBB3 IF (gray) in iPSC-derived spMNs (D12). Nuclei are counterstained in cyan. Square inserts are digitally magnified in yellow-bordered panels. Yellow arrowheads point to nHOTAIRM1 spots in distal axonal segments. Scale bar units are indicated in each panel.

Full size image

Identification of nHOTAIRM1 target genes through transcriptome analysis

To decode nHOTAIRM1 function, its target genes were identified through transcriptomic analysis performed in wild-type (WT) and HOTAIRM1 knockout (KO) spMNs (D12). The KO iPSC lines were obtained using CRISPR/Cas9 genome editing by blocking HOTAIRM1 gene transcription through a premature poly-A signal [20], inserted at the 5’ region of its first exon, downstream to the annotated transcription start site peaks reported in Zenbu (https://fantom.gsc.riken.jp/zenbu/) (Fig. 3A). Two independent HOTAIRM1 KO homozygous iPSC clones (KO#1 and KO#2) were generated. Figure 3B shows that the induction of the lncRNA is drastically reduced along the entire differentiation process of both KO clones, achieving a reduction by about 87% and 96% in differentiated KO#1 and KO#2 spMNs (D12), respectively.

Fig. 3: nHOTAIRM1 functional KO.
figure 3

A Schematic representation of HOTAIRM1 gene editing experiment. Two sgRNAs (carried by the plasmid PX333) were designed to target exon 1 of HOTAIRM1 locus. The polyA signal, the selection cassette, and the homology arms for homology-directed repair through DNA recombination were carried by the HR110PA plasmid. Cell manipulation is schematized below. B qRT-PCR analysis of nHOTAIRM1 expression in wild type (WT) and nHOTAIRM1 KO (KO#1 and KO#2) spMNs (D12). Data (means ± SEM) are expressed in arbitrary units, relative to ATP5O mRNA levels, as internal control. N = 3 biological replicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, two-tailed Student’s t test, referred to D0 as the reference sample for statistical tests.

Full size image

Data mining of RNA-Seq analyses, performed on three independent biological replicates of WT and KO#2 spMNs, showed that both samples clustered homogeneously and produced read counts for many genes (Supplementary Fig. 1A, B). We detected 16,504 genes expressed in at least one sample (Dataset 1) and, among them, those having |logFC | > 1 and adjusted p-value < 0.05 were considered as significantly differentially expressed. Applying these filters, we found 1,887 upregulated and 654 downregulated genes (Dataset 1) in KO spMNs, considered as bona fide nHOTAIRM1 targets (Supplementary Fig. 1C–E). Gene ontology over-representation analysis was performed on the lists of differentially expressed genes using WebGestalt R tool (http://www.webgestalt.org/). From this test, only the gene sets with FDR < 0.05 were considered significantly enriched. Concerning the upregulated genes, the analysis of “Biological Process”, “Cellular Component” and “Molecular Function” domains revealed highly heterogeneous ontological annotations, not strictly related to MN biology (Supplementary Figs. 2–4, panels A). Similar indications derived from querying the KEGG and the REACTOME pathway databases (Supplementary Figs. 5,6, panels A). When applied to downregulated genes, all these computational inspections indicated, instead, that they were enriched for neuronal functions (Supplementary Figs. 2–6, panels B). Of note, the “Biological Process” ontology showed annotations clearly associated to MN (i) differentiation, (ii) morphology, and (iii) activity (Dataset 1 and Supplementary Fig. 2B), which guided the subsequent analyses.

To first validate the RNA-Seq analysis, subsets of ten upregulated and ten downregulated genes were analyzed by qRT-PCR in WT and KO spMNs. Candidates with an adjusted p-value < 0.05 and having different significance values ranging from logFC −2.72 to logFC 7.65 were selected. Notably, all of them followed the expression trend highlighted in the RNA-Seq analysis (Supplementary Fig. 7A, B).

Subsequent functional studies focused on three main groups of genes belonging to seven downregulated biological processes, implicated in fundamental aspects of MN biology. They included over-represented genes engaged in: (i) MN differentiation (“cell differentiation in spinal cord” and “CNS neuronal differentiation”); (ii) morphology (“neuron projection guidance”, “axonogenesis”, and “axon guidance”), and (iii) activity (“synapse organization”, and “modulation of chemical synaptic transmission”).

nHOTAIRM1 regulates binary fate decision between MNs and INs

To explore the involvement of nHOTAIRM1 in neuronal cell fate decision, the expression profile of target genes related to differentiation was compared between WT and KO iPSCs along their conversion into spMNs (Fig. 4). We confirmed in KO cells the altered expression pattern of genes implicated in neuronal development, such as the TF and cofactor SALL1 [ref. [21]] and LMO4 [ref. [22]], expressed in the developing cerebral cortex, the nuclear receptor RORa, which identifies excitatory neurons in the CNS [23] and the transmembrane receptor EPHB1, important for the regulation of contact-dependent cell-to-cell interactions [24] (Fig. 4A). We also validated the deregulation of genes linked to SC or MN development (Fig. 4B), such as DLL4, with a role in neuronal subtype specification in the SC [25], PROX1, a regulator of ventral SC patterning [5], LHX4, required for ventral MN differentiation [26] and DCC, associated with MN migration [27]. The expression profiles of these genes were altered along the formation of KO spMNs from iPSCs, with a range of decrease spanning from 54% to 90% at D12. This suggests that nHOTAIRM1 is required for (motor) neuron development.

Fig. 4: Expression analysis of genes related to neuronal differentiation along the conversion of WT and nHOTAIRM1 KO iPSCs into post-mitotic spMNs.
figure 4

A, B Expression profile, by qRT-PCR analysis, of genes (indicated above each panel) implicated in neuronal development (A) and linked to SC or MN development (B), along the differentiation (from D0 to D12, x-axis) of WT or KO for nHOTAIRM1 spMNs (KO#2 clone). C Expression profile, by qRT-PCR analysis of HB9 and OLIG2 genes along the differentiation (from D0 to D12, x-axis) of WT or KO for nHOTAIRM1 spMNs (KO#2 clone). Data (means ± SEM) are expressed as arbitrary units in all the panels, relative to ATP5O mRNA levels, as internal control. For each target and each cell line, the expression at D0 was set as 1. N = 3 biological replicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, two-tailed Student’s t-test, referred to D0 as the reference sample for statistical tests.

Full size image

A consistent downregulation along the differentiation of KO spMNs from iPSCs was also appreciated for HB9 and OLIG2 (Fig. 4C), two genes crucially implicated in acquiring alternative neuronal fates. The HB9 gene, a selective marker of MNs in the developing SC [4, 28], is critical for specifying MN identity and repressing the IN character. Similarly, its regulator OLIG2, primes MNPs to become MNs [29] and inhibits IN identity [5]. On this basis, we wondered whether nHOTAIRM1 could positively control the generation of MNs at the expense of INs. To answer this question, we analyzed the expression of the cholinergic neuron marker ChAT [30] and the major MN marker ISLET1 [ref. [4]] during the conversion of iPSCs into spMNs. We found their expression levels drastically reduced along the differentiation of KO iPSCs (Fig. 5A), highlighting the relevance of nHOTAIRM1 in spMN generation. This finding was corroborated by IF analysis (Fig. 5B and Supplementary Fig. 8), in which we observed a significant decrease by about 42% in the number of CHAT + /ISLET1 + MN cells upon nHOTAIRM1 depletion. In this scenario, where the lack of nHOTAIRM1 impairs spMN differentiation, we asked whether an alternative cell fate was undertaken. Given that nHOTAIRM1 regulates the expression of HB9, a molecular switch between MN and IN cell fate, we wondered whether the lack of the lncRNA could partially mimic the absence of HB9.

Fig. 5: Analysis of MN and HB9-dependent IN genes upon nHOTAIRM1 KO.
figure 5

A Expression profile, by qRT-PCR analysis, of ChAT and ISLET1 genes along the differentiation (from D0 to D12, x-axis) of WT or KO for nHOTAIRM1 spMNs (KO#2 clone). Data (means ± SEM) are expressed as arbitrary units, relative to ATP5O mRNA levels, as internal control. Expression levels at D0 were set as 1. N = 3 biological replicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, two-tailed Student’s t-test, referred to D0 as the reference sample for statistical tests. B Quantification by immunostaining of CHAT + /ISLET1+ cells, expressed as percentage of the total number of cells, in WT and nHOTAIRM1 KO (KO#2) spMNs (D12). N = 3 groups of 60 cells randomly selected from 9 different fields, both for WT and KO, were analyzed. ***P ≤ 0.001. C qRT-PCR validation, in WT and nHOTAIRM1 KO (KO#2 clone) spMNs (D12) of the expression of HB9-dependent IN genes (indicated above each panel) significantly upregulated in HB9 KO MNs and nHOTAIRM1 KO spMNs compared to WT spMNs. Data (means ± SEM) are expressed in arbitrary units, relative to ATP5O mRNA levels, as internal control. For each target, the expression in WT spMNs was set as 1. N = 3 biological replicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, two-tailed Student’s t test.

Full size image

To verify this, we first intersected the lists of significantly upregulated genes in nHOTAIRM1 KO spMNs and HB9-null mutant mouse-induced MNs [31]. We found a signature of 10 shared genes expressed in INs that we validated in KO spMNs (D12) via qRT-PCR (Fig. 5C). Collectively, these results indicate that nHOTAIRM1, through the control of HB9 expression, regulates the binary fate decision between INs and MNs, to the latter’s advantage. This evidence was definitively confirmed by the analysis of specific IN marker genes, namely CALB2, SST, ETV1, LHX1, LAMP5, PVALB, and VIP, that we found upregulated upon nHOTAIRM1 depletion (Supplementary Fig. 9A).

nHOTAIRM1 is necessary for the correct MN neurite outgrowth

Neurite outgrowth depends on the activity of several genes promoting the formation of functional neuronal connections [32]. Interestingly, comparative transcriptome analysis revealed a list of genes critically involved in neurite outgrowth which were downregulated in KO spMNs. We focused on six of them, namely NrCAM, UNC5A, ROBO1, DCC, SEMA3E and SEMA6D, found enriched in three ontological gene categories, i.e. “neuron projection guidance”, “axon guidance”, and “axonogenesis”. In particular, the expression of NrCAM, crucially implicated in neurite outgrowth activity in vitro [33], and of UNC5A, involved in axonal pathfinding [34], declined by about 53% and 95%, respectively (Fig. 6A). The expression of the receptor ROBO1, engaged in axon guidance through the interaction with SLIT ligands [35, 36], decreased by about 66% in KO spMNs, whereas the expression of DCC, encoding for the receptor of NETRIN-1 with outgrowth-promoting activity [37], was reduced by about 80% (Fig. 6A). Moreover, the expression of SEMA3E and SEMA6D, belonging to the family of Semaphorins that instruct axon guidance in the developing nervous system [38], was reduced in KO spMNs by about 66% and 46%, respectively. These data imply that nHOTAIRM1 regulates the activity of several genes crucially implicated in the assembly of neurite networks.

Fig. 6: Analysis of the neurite network in WT and nHOTAIRM1 KO post-mitotic spMNs.
figure 6

A qRT-PCR analysis of genes involved in neurite outgrowth and branching (indicated above each panel) in WT and nHOTAIRM1 KO (KO#2 clone) spMNs (D12). Data (means ± SEM) are expressed in arbitrary units, relative to ATP5O mRNA levels, as internal control. For each target, the expression in WT spMNs was set as 1. N = 3 biological replicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, two-tailed Student’s t test. B Representative acquisition and skeletonization of neurite networks in WT and nHOTAIRM1 KO (KO#1 and KO#2 clones) spMNs (D12). Scale bar units are indicated in each panel. C–G Number of somata (C), number of branches (D), number of junctions (E), total branch length (F), and average branch length (G) per acquisition in WT and nHOTAIRM1 KO (KO#1 and KO#2 clones) spMNs (D12). Horizontal lines in boxplots indicate median values; boxes extend from the 25th to the 75th percentile of each group’s distribution of values; black dots represent outliers. Boxplots represent the value distribution from N = 3 independent biological replicates. In all, 10–20 acquisitions with a size of 212.3 μm × 212.3 μm have been analyzed for each replicate. *P ≤ 0.05, ****P ≤ 0.0001 correspond to one-way ANOVA multiple comparison test.

Full size image

Next, we compared the morphology of both the KO#1 and KO#2 spMNs with that of WT spMNs to verify whether a defective branching phenotype mirrored these molecular analyses. Images of neurite networks composed of WT or KO spMNs (Fig. 6B) were analyzed as described in ref. [39] (Supplementary Fig. 10). Starting from an equal number of cells among all the samples (Fig. 6C), we counted the total number of branches (Fig. 6D) and junctions (Fig. 6E) and measured both the total branch length (Fig. 6F) and the average branch length (Fig. 6G). A statistically significant reduction in the counts of both branch and junction number was found in the two KO neurite networks compared to WT. A lower total branch length was observed in the KO samples in line with these observations. No significant difference in these parameters was estimated between the two KO samples as control. On the contrary, the average branch length was higher in the KO samples compared to WT, as the KO neurite network’s lower density resulted in fewer intersections and, therefore, in longer slab segments. Collectively, these results reveal that nHOTAIRM1 is necessary for proper neurite outgrowth in spMNs.

nHOTAIRM1 controls genes involved in synaptic transmission

Other genes affected by nHOTAIRM1 depletion in spMNs were those involved in “synapse organization” and “modulation of chemical synaptic transmission”. They include CACNA1D, CNR1, GRIA1, GRIK3, and SHANK2 genes, all of which were significantly downregulated in KO spMNs (Fig. 7A).

Fig. 7: Analysis of synaptic genes in WT and nHOTAIRM1 KO post-mitotic spMNs.
figure 7

A qRT-PCR analysis of synaptic genes (indicated above each panel) in WT and nHOTAIRM1 KO (KO#2 clone) spMNs (D12). Data (means ± SEM) are expressed in arbitrary units, relative to ATP5O mRNA levels, as internal control. For each target, the expression in WT spMNs was set as 1. N = 3 biological replicates; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, two-tailed Student’s t test. B qRT-PCR analysis of ChAT in WT and nHOTAIRM1 KO (KO#2 clone) spMNs (D12). Data (means ± SEM) are expressed in arbitrary units, relative to ATP5O mRNA levels, used as internal control. The expression in WT spMNs was set as 1. N = 3 biological replicates; ***P ≤ 0.001, two-tailed Student’s t test. C, D Quantification of the intensity signal of CHAT protein upon immunostaining in WT and nHOTAIRM1 KO (KO#2 clone) spMNs (D12). For each sample, the analysis was carried out on 260 cells randomly selected in two different biological replicates. ****P ≤ 0.0001. Unpaired two-tailed Student’s t test.

Full size image

The expression of CACNA1D gene encoding for Cav1.3, a member of voltage‐gated L‐type Ca2+ channels essential for synapse maturation and pruning, declined by about 60% in KO spMNs [40,41,42], whereas the expression of CNR1 gene, regulating synaptic transmission [43], dropped down by about 56%. GRIA1, responsible for MN dendritic architecture [44], and GRIK3, relevant in synaptic plasticity and potentiation [45], decreased by about 66% and 47%, respectively, in KO spMNs. Finally, the expression of SHANK2, which affects the synaptic connectivity [46], was reduced by about 64%. Besides the aforementioned genes, we also focused on ChAT, a gene encoding for choline acetyltransferase, the rate-limiting enzyme in the synthesis of acetylcholine [47], essential for neurotransmission in MNs [48]. We evaluated its mRNA levels via qRT-PCR (Fig. 7B) and protein levels via IF (Fig. 7C, D) in D12 KO vs WT spMNs. We found that their levels were significantly reduced when nHOTAIRM1 was depleted, pointing to a role for the lncRNA as a regulator of synaptic activity in spMNs.

nHOTAIRM1 establishes RNA-RNA interactions with a subset of target mRNAs

Being predominantly cytoplasmic, we asked how nHOTAIRM1 could exert post-transcriptional control of its target genes. A typical role for the cytoplasmic lncRNAs is the competing endogenous RNA (ceRNA) involved in microRNA sponge activity [49]. No data on microRNA-nHOTAIRM1 interactions in neuronal systems were reported by the LncBase database (https://diana.e-ce.uth.gr/lncbasev3) Therefore, we investigated the interaction between the lncRNA and Argonaute 2 (AGO2), a major component of the miRNA-induced silencing complex (miRISC) [50]. In silico analyses through the catRAPID algorithm (http://s.tartaglialab.com/page/catrapid_group), that predicts protein partners of any given transcript, did not reveal binding propensity of nHOTAIRM1 for AGO2 protein. Accordingly, AGO2 was not retrieved from nHOTAIRM1 RAP-Mass Spectrometry experiments, previously performed to identify its protein interactors in spMNs [12]. A cross-linking immunoprecipitation (CLIP) assay, carried out from cytoplasmic extracts of differentiating neurons, further excluded the occurrence of direct interactions between nHOTAIRM1 and AGO2 (Supplementary Fig. 11). Together, all these data indicate that nHOTAIRM1 does have the pre-requisites to perform as ceRNA in the tested neuronal context.

Another mechanism establishing the functional crosstalk between coding RNAs and ncRNAs is mediated by RNA-RNA interactions [49]. RNA pull-down assays were performed in cytoplasmic extracts of WT D12 spMNs, under native conditions. The pull-down fractions enriched for nHOTAIRM1 were analyzed by qRT-PCR for the presence of mRNAs downregulated in KO spMNs. Among the tested candidates, the transcripts PROX1, ROBO1, SEMA6D, ChAT, GRIK3, CNR1, UNC5A, and SHANK2 were significantly enriched in the nHOTAIRM1 pull-down fractions (Fig. 8A). To detect direct lncRNA-mRNA pairings occurring in living cells, 4′-aminomethyl-4,5′,8-trimethylpsoralen (AMT)-crosslinked RNA pull-down assays were performed in spMNs. Two out of the eight nHOTAIRM1 mRNA targets identified in the native RNA pull-down assays, namely ROBO1 and SHANK2, were identified as direct interactors of the lncRNA (Fig. 8B). Through the IntaRNA algorithm (http://rna.informatik.uni-freiburg.de/) we inspected the regions of interaction between nHOTAIRM1 and the two mRNAs (Supplementary Fig. 12). Interestingly, the region of nHOTAIRM1 establishing RNA-RNA interactions encompasses the 5′end of exon 3 (nucleotides 644–689) that includes a G-rich tract, predicted as a putative G-quadruplex forming sequence (Supplementary Fig. 13) according to QGRS mapper tool (https://bioinformatics.ramapo.edu/QGRS/). Computational predictions of ROBO1 or SHANK2 regions of interaction with nHOTAIRM1 performed through IntaRNA showed that they map in the 5’ untranslated regions (5′UTRs) of the two mRNAs, according to Ensembl (https://www.ensembl.org/) (Supplementary Fig. 14). This observation supports the idea that nHOTAIRM1 might play a role in the control of mRNA stability and/or translatability [51].

Fig. 8: Identification of nHOTAIRM1 RNA interactors in post-mitotic spMNs.
figure 8

A qRT-PCR analysis of nHOTAIRM1 RNA pull-down experiments performed in native conditions in WT spMN (D12) cytoplasmic extracts. B qRT-PCR analysis of nHOTAIRM1 AMT-crosslinked RNA pull-down experiments performed in WT spMN (D12) total extracts. In both panels, data (means ± SEM) are expressed as percentage of RNA enrichment relative to Input. Boxplots represent the mean value of N = 3 independent biological replicates. Symbols (triangle, square and circle) represent values for each biological replicate.

Full size image

Overall, these results emphasize the influence of lncRNA-mRNA crosstalk in the multi-step process leading to the generation of functional spMNs.

Discussion

This study demonstrates that the lncRNA nHOTAIRM1 is a novel player in the regulatory network governing spMN production and function. Being specifically expressed in the SC, we explored its function in human iPSC-derived spMNs, where it localizes both in the soma and the neurites (Fig. 1C, D). Applying a reverse genetics approach combined with transcriptomics analyses, RNA pull-down experiments, and RNA FISH/IF assays, we uncovered the broad-spectrum regulatory action of nHOTAIRM1.

The identification in functional spMNs of nHOTAIRM1 target genes, whose expression is significantly altered when the lncRNA is depleted, proved that nHOTAIRM1 participates in (i) spMN differentiation at the expense of IN fate, (ii) neurite outgrowth, and (iii) modulation of synaptic transmission.

Notably, the depletion of nHOTAIRM1 causes a strong downregulation of several differentiation genes, among which HB9 and its regulator OLIG2, which are key to triggering MN specification and repressing IN commitment. Accordingly, the absence of the lncRNA during the conversion of iPSCs into spMNs causes a dramatic reduction of specific MN markers, such as ChAT [30] and ISLET1 [ref. [4]] (Fig. 5A) and a concomitant decrease in the number of spMNs, marked by the co-expression of CHAT and ISLET1 proteins (Fig. 5B).

As a corollary to this and similarly to what happens when HB9 is not functional in mouse-induced MNs [31], the nHOTAIRM1-dependent reduction of MN markers is accompanied by the induction of specific IN markers (Fig. 5C and Supplementary Fig. 9), pointing to the lncRNA as a pro-MN molecule, hierarchically positioned upstream of the OLIG2/HB9 module, at the crossroad between MN and IN cell fate (Supplementary Fig. 9B).

To include nHOTAIRM1 more comprehensively in the complex regulatory network underlying neuronal differentiation, its role in controlling the transition between neuronal precursors and differentiating neurons should also be considered. Indeed, we demonstrated that the lncRNA is required to suppress the expression of the master proneural TF Neurogenin 2 (NGN2) and its downstream neurogenic pathway [12], allowing the differentiation to proceed and the acquired neuronal cell identity to be maintained. This regulative axis fits with the role played by OLIG2 and NGN2 in SC development. In MNPs, OLIG2 expression preserves their undifferentiated state by repressing post-mitotic MN genes, such as HB9. On the other side, OLIG2 triggers the expression of NGN2 in a subset of progenitor cells, priming them to become MNs. Therefore, the relative levels of OLIG2 and NGN2 govern the cell fate decision which converts MN progenitors to post-mitotic MNs [29]. Acting on both OLIG2 and NGN2, nHOTAIRM1 may contribute to establish their proper ratio. The resulting overall picture is consistent with the speculation that nHOTAIRM1 stands at the apex of a sophisticated molecular circuit composed of OLIG2, NGN2 and their targets, which serves as a gate for timing the correct (motor) neuronal gene expression in the differentiating SC.

Regarding the involvement of nHOTAIRM1 in establishing proper neuronal connections, it was assessed at both the molecular and phenotypic level. The depletion of nHOTAIRM1 led to a significant downregulation of genes that influence neuronal wiring and targeting (NrCAM), axon guidance (SEMA3E and SEMA6D), and motor axon trajectories out of the SC (DCC and ROBO1) (Fig. 6A). Phenotypically, the absence of nHOTAIRM1 resulted in a defective MN morphology (Fig. 6B), with a significant reduction of the number of branches and junctions and of the total branch length (Fig. 6D–F).

Closely related to this aspect is the control exerted by nHOTAIRM1 on target genes encoding modulators of synaptic activity (Fig. 7A, B), whose dysfunction has been associated with CNS disorders. This consideration suggests a pathophysiological role for nHOTAIRM1 in spMNs. For example, diminished activity of ChAT, observed in the SC of patients with Amyotrophic lateral sclerosis (ALS), was supposed to be implicated in MN loss-of-function [52] (Fig. 7B, C). In line with this hypothesis, we have demonstrated that nHOTAIRM1 physically interacts with the RNA binding protein FUS, and that nHOTAIRM1 expression depends on the levels of FUS [12], whose loss-of-function contributes to neuronal dysfunction and ALS neurodegeneration [53]. Based on this evidence, we plan to better unravel the interplay between nHOTAIRM1 and ALS mutant FUS in the future.

Mechanistically, nHOTAIRM1 establishes RNA-RNA interactions with a number of its target genes. PROX1, SEMA6D, ChAT, GRIK3, CNR1, and UNC5A mRNAs establish indirect interactions with nHOTAIRM1 (Fig. 8A), whereas ROBO1 and SHANK2 mRNAs directly interact with the lncRNA (Fig. 8B). It is tempting to speculate that the dynamic association of direct and indirect targets with the lncRNA may play a role along spMN differentiation. Since these interactions were detected in the cytoplasm of post-mitotic MNs, we propose nHOTAIRM1 as a post-transcriptional regulator of spMN gene expression in this context. To the best of our knowledge, this represents the first evidence of a lncRNA-mediated post-transcriptional control in human spMNs, one only example being described earlier in mouse MNs [54]. The prediction that the direct mRNA interactors of nHOTAIRM1, ROBO1 and SHANK2, are bound at their 5’UTRs calls for further mechanistic investigations to elucidate whether the lncRNA regulates its target mRNAs at the level of their stability or translation. Intriguingly, RNA-RNA interaction predictions also indicate that nHOTAIRM1 binding may occur through a putative a rG4 structure. rG4s are known to be involved in the modulation of gene expression, lncRNA function and disease progression [55].

In conclusion, our findings reveal that nHOTAIRM1, as a pleiotropic regulator of multiple molecular pathways involved in MN differentiation, arborization and activity, is essential for ensuring the production of functional spMNs. Since its activity impinges on SC cell fate decisions, this study also provides a rationale for future investigations to clarify how nHOTAIRM1 contributes to the correct balance between neuronal subtypes in the motor circuits responsible for the overall coordination of locomotion.

Materials and methods

Cell culture and MN differentiation

Human induced pluripotent stem cells (iPSCs) used in this study (iPSC-NIL) were derived, maintained and induced to differentiate into spMNs following the methods described in [14].

Generation of nHOTAIRM1 KO iPSC-NIL lines

sgRNAs (Dataset 2) were designed using Benchling design tool (https://www.benchling.com/) targeting the regions of HOTAIRM1 locus. The Cas9 double-strand break target regions were selected, considering FANTOM5 transcription start site data in Zenbu (https://fantom.gsc.riken.jp/zenbu/). PX333 plasmid, encoding the WT Cas9 protein, was purchased from Addgene and sgRNAs were ordered as single-stranded DNA probes (Bio-Fab Research – Rome, IT) and cloned as recommended by the Zhang Lab Protocol (https://media.addgene.org/data/plasmids/62/62987/62987attachment_KiOWQSPn2egB.pdf) resulting in a vector identified as PX333-sgRNAs. HR110PA (System Biosciences – Palo Alto, CA, USA) was used as a backbone to create the donor vector (DONOR). A Poly-adenylation sequence (PAS) was cloned into the DONOR followed by a Neomycin resistance cassette using In-Fusion® HD Cloning Plus Kit (Cat. #638910). Two homology arms (HA) HA1 and HA2, with a length of 800 nt were amplified by PCR using iPSC-NIL gDNA (Kapa HiFi, Takara Bio - Saint-Germain-en-Laye, FR). HA1 was cloned upstream of the PAS, and HA2 was cloned downstream of the PAS. iPSCs were transfected on matrigel-coated dishes through the Neon Transfection System (Life Technologies – Carlsbad, CA, USA), using 100 μl tips in R buffer with the following settings: 1200 V, 30 ms, one pulse. The selection was performed in 800 μg/ml G418 for five days. Single HOTAIRM1 KO clones were amplified and genotyped.

Cellular fractionation

iPSC-derived spMNs were subject to subcellular fractionation using the Ambion PARIS Kit (AM1921, Life Technologies). After RNA extraction, equal cytoplasmic and nuclear RNA volumes were retro-transcribed and analyzed by qRT-PCR. Normalizations were based on the total amount of RNA.

Soma/neurite separation

Soma/neurite separation from iPSC-derived spMNs was performed using modified Boyden chambers and assessed by immunostaining of TUBB3 protein as shown in [19]. RNA samples collected from spMN soma and neurites were then analyzed by qRT-PCR. Enrichment of the neuronal projection marker COL3A1 in the neurite compartment and of GNG3 in the soma served as control of the proper separation of the two compartments.

RNA-Seq and gene ontology analyses

TruSeq Stranded mRNA Library Prep Kit (Illumina - San Diego, CA, USA) was used to obtain sequencing libraries from polyA+ RNA extracted from iPSC-derived WT and nHOTAIRM1 KO spMNs (3 independent biological replicates).

The sequencing reaction produced 100 nucleotide-long paired-end reads and was performed on a Novaseq 6000 sequencing system (Illumina) with a depth of more than 20 M. To remove adapter sequences and low-quality end bases Trim Galore188 (version 0.6.4_dev) software was used; the minimum read length after trimming was set to 20. Alignment to human GRCh38 genome primary assembly was performed using STAR (version 2.7.9a189).

About 85% of the total reads were successfully mapped to the human genome and most of them were aligned to unique locations. The quantMode TranscriptomeSAM option was used to generate alignments translated into transcript coordinates. The RSEM method was used to quantify the expression levels of the transcripts.

Differential expression analysis was performed using the Bioconductor package DESeq2 that fits a negative binomial generalized linear model on each gene. Only genes displaying in at least three samples a minimum absolute expression value of 10 - as suggested by DESeq2 package developers - were retained. Shrinked log fold change values were obtained by the function lfcShrink which considers the largest fold changes that are not due to low counts and uses these to originate a prior distribution. Genes with padj < 0.05 and absolute logFC > 1 were considered differentially expressed and used for further analysis.

GO over-representation analysis (ORA) was performed on significantly upregulated and downregulated genes having padj < 0.05 and absolute logFC > 1 using the WebGestalt R tool on the GO database (http://geneontology.org/).

RNA–RNA interaction prediction

IntaRNA version 3.3.2178 was used to map the binding regions between nHOTAIRM1 and the mRNA interactors found in the AMT-crosslinked RNA pull-down experiments. To obtain in silico predictions of RNA-RNA interactions, we filtered out the RNA transcript isoforms that were not expressed in our system from the RNA-Seq results, by looking at the RSEM output isoforms.

Among the expressed splicing isoforms, we considered the TPM expression values in the WT and KO samples in order to select the protein-coding splicing isoforms that were highly expressed in our system. We subsequently retrieved the FASTA files of these transcripts and used IntaRNA with default parameters to predict mRNA targets sites for nHOTAIRM1.

RNA extraction and analysis

Total RNA was extracted by Direct-zol RNA MiniPrep (R2052, Zymo Research – Irvine, CA, USA). For qRT-PCR assays, cDNA was synthetized by Takara PrimeScript RT Reagent Kit (RR037A, Takara Bio). qRT-PCR detection was performed using SensiFAST SYBR Lo-ROX Kit (BIO-94020, Bioline – Memphis, TN, USA) on a 7500 Fast Real-Time PCR (Applied Biosystem - Carlsbad, CA, USA). ATP5O mRNA was used as a reference target. For RNA pull-down assays, total RNA was extracted using miRNeasy Micro Kit (Qiagen - Germantown, MD, USA) and cDNA was synthetized by SuperScript VILO cDNA Synthesis Kit (ThermoFisher Scientific - Carlsbad, CA, USA).

RNA FISH coupled with IF

iPSC-derived MNs were plated on 12 mm diameter coverslips coated with matrigel (hESC-qualified Matrix Corning 354277) and fixed in 4% paraformaldehyde (Electron Microscopy Sciences - Hatfield, PA, USA) for 10 min at RT. A dehydration step was performed with ice-cold ethanol series (50%, 70%, 100%) in order to store cells at −20 °C until use.

nHOTAIRM1 was detected via RNA FISH with a mix of 18 biotinylated probes (Dataset 2) as described in [56, 57]. MNs were rehydrated by descendent ice-cold ethanol series (100%, 70%, 50%) and permeabilized in a solution of 0,05% Triton X-100 and 2 mM VRC (R3380, Sigma-Aldrich – St. Louis, MO, USA) in DPBS for 5 min. Cells were then washed three times in DPBS before replacing with 2X SSC buffer (3 M NaCl; 0,3 M sodium citrate in nuclease-free water for a 20X stock solution). Five min incubation in SSC was followed by incubation with pre-hybridization buffer (10% deionized formamide, 47671, Sigma-Aldrich,; 2X SSC in nuclease-free water) for 15 min at 37 °C. MNs were then incubated ON at 37 °C in a slide hybridizer machine (ACD HybEZ™ II Hybridization System) with hybridization buffer (10% deionized formamide; 2X SSC; 10% w/v Dextran sulfate, 2 mM VRC in nuclease-free water) complemented with the biotinylated probes at a final concentration of 50 nM each. The following day cells were washed twice with 2X SSC for 5 min first at 37 °C and then at RT. SSC buffer was then discarded, and coverslips were incubated with TN buffer (Tris HCl pH 7.5 1 M; NaCl 5 M in nuclease-free water) at RT for 10 min. Finally, biotinylated oligos were stained incubating with 1:200 diluted 568-conjugated streptavidin (S11226, Invitrogen – Waltham, MO, USA,) in 4% w/v BSA/TN buffer for 1–2 h at RT in a humid box.

When FISH staining was combined with IF, cells were washed twice with TN buffer and once with DPBS for 5 min at RT, and then were incubated for 1 hour at RT with primary antibodies (anti-βTUBIII, T2200, Sigma Aldrich; anti-Map2, 17490-1-AP, Proteintech, Manchester, UK) diluted in 1% w/v BSA/DPBS. Subsequently, samples were washed three times with DPBS for 5 min at RT and incubated for 45 min at RT with secondary antibodies (Goat anti-Mouse 488, A-11029, Invitrogen; Goat anti-rabbit 488, A-11008, Invitrogen; Donkey anti-Rabbit 594, IS-20152-1, Immunological Sciences - Rome, IT) diluted 1:300 in 1% w/v BSA/DPBS. Lastly, cells were washed three times with DPBS for 5 min at RT, nuclei were counterstained with DAPI solution (1ug/ml/PBS, D9542, Sigma Aldrich) for 5 min at RT and coverslips were mounted with Prolong Diamond Mounting Media (P-36961, ThermoFisher Scientific).

IF

MNs at D12 were cultured on 0.01% poly-L-ornithine/Murine Laminin 20 μg/ml (Sigma Aldrich), fixed in 4% paraformaldehyde (Electron Microscopy Sciences) for 20 min at 4 °C and washed with PBS. Fixed cells were then permeabilized and blocked with 0.2% Triton X-100/3% BSA/PBS for 15 min at RT. Subsequently, cells were incubated ON at 4 °C with anti-CHAT (ab223346 - Abcam, Cambridge, UK) and anti-ISLET1/2 primary antibodies (mouse clone 39.4D5 - DSHB, Iowa City, IA, USA) diluted 1/200 and 1/50 in 3% BSA/PBS solution, respectively. After washing with PBS, cells were incubated for 45 min at RT in a solution containing 1% donkey serum,1% goat serum in PBS, donkey anti-mouse 647 1:300 (A32787, Invitrogen) and goat anti-rabbit 555 1:300 (A32732, Invitrogen). Nuclei were counterstained with DAPI solution (1 µg/ml/PBS, D9542, Sigma Aldrich).

Images were acquired with inverted confocal Olympus IX73 microscope equipped with a Crestoptics X-LIGHT V3 spinning disk system and a Prime BSI Express Scientific CMOS camera, UPlanSApo 60X (NA 1.42) oil objective, and were collected with the MetaMorph software (Molecular Devices - San Jose, CA, USA). The Z-stack confocal microscopy images were taken automatically (200 nm Z-spacing).

Signal intensity analysis for CHAT was performed on Maximum Intensity Projection with ImageJ software, and it was measured for each single cell taken in account. A total of 260 cells randomly selected in two different biological replicates were considered for WT and KO. Cell count of CHAT + /ISLET1+ cells was performed with “Analyze Particle” tool on binary images obtained with automatic threshold on ImageJ.

Neurite network analysis

Neurite network analysis was performed in iPSC-derived WT, nHOTAIRM1 KO#1 and KO#2 spMNs as described in [39] with minor modifications. A Fiji-imageJ macro for semi-automated analysis of MN neurite networks was created adapting the workflow described in [39]. Briefly, Z-projections of confocal images of WT and nHOTAIRM1 KO spMNs stained with the neurite marker Map2 [ref. [58,59,60]] were generated. To segment somas starting from the DAPI channel, a nuclei mask was generated with the Huang thresholding algorithm and outliers with a radius lower than 20 pxs were removed. A selection was generated starting from the nuclei mask that was then expanded by 5 pxs (1 px = 0.207 mm) using the “Enlarge selection” function of ImageJ, to account for the small cytoplasmic portion typical of mature MNs.

Subsequently, the Map2 channel was duplicated and two different masks were generated. From the first copy, a high-intensity mask was created thresholding with Moments algorithm upon contrast enhancement (saturated pxs = 0.1%) and Gaussian Blur (σ = 2 μm) filtering. This mask accounts for high-intensity parts of the image, including somas, axon hillock and neurite edges. The second copy was used to generate a low-intensity mask, the LoG mask, to account for thinner parts of neurites. To create the LoG mask, the contrast was enhanced (saturated pxs = 0.1%), the LoG filter from the Process > Math ImageJ menu was used and a threshold was applied to the image using Moments algorithm.

The “Image Calculator” function of ImageJ was then used to combine the high intensity mask with the LoG mask and to subtract the nuclei mask to obtain a final neurite mask.

Finally, the neurite mask was skeletonized and the “Analyze skeleton” function was exploited to determine number of branches, junctions, and end-points per image, as well as total branch length and average branch length.

Number of cells per image was manually counted starting from the nuclei masks, taking advantage of the multipoint tool. Samples were imaged on a Nikon Instrument A1 Confocal Laser Microscope equipped with a 1.49 NA ×100 objective (Apo TIRF 100x Oil, Nikon, Tokyo, Japan) and with a 60x objective. Confocal images were collected with NIS-Elements AR software (Nikon, Tokyo, JP): ND Acquisition module was used for multipoint acquisition of Z-stack images (150–175 nm Z-spacing) of 4 mm thickness.

RNA pull-down

Native RNA pull-down experiments were performed on cytoplasmic extracts of iPSC-derived spMNs (D12) as described in [61]. Biotinylated DNA probes used to pull down nHOTAIRM1 are listed in Dataset 2.

AMT-crosslinked RNA pull-down experiments were performed as described in [62]. Biotinylated DNA probes used to pull down nHOTAIRM1 are listed in Dataset 2.

Cross-linking immunoprecipitation assay

CLIP assay was performed in cytoplasmic extracts of RA-treated SH-SY5Y neuroblastoma cells as described in this work [12].

Statistical analyses

Histograms show the mean ± SEM of at least three independent biological replicates. N is indicated in Figure Legends. Errors were calculated from relative quantities and then opportunely propagated; statistical significance was determined by two-tailed paired or unpaired Student’s t test, as indicated in Figure Legends. A p-value < 0.05 was considered statistically significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.