Kalirin-RAC controls nucleokinetic migration in ADRN-type neuroblastoma

ADRN NB cells migrate nucleokinetically via DCX- and LIS1-dependent mechanism

We reasoned that, like in neurons, DCX could be part of the migratory program in NB. In primary NB, DCX mRNA correlated positively with mRNAs for ADRN SE–associated TFs, GATA3, and PHOX2B, but not MES FOSL2, suggesting ADRN-specific expression (R_GATA3 = 0.53, R_PHOX2B = 0.51; R_FOSL2 = −0.09; Fig S1A). mRNA profiling in NB cell lines, which enabled detection of ADRN-type and MES-type cells (Fig S1B), confirmed this finding and attributed DCX expression to ADRN-type cells (Fig 1A). Chromatin immunoprecipitation sequencing (ChIP-seq) for activation and repression marks demonstrated that DCX was transcribed exclusively in ADRN cell lines, whereas the DCX locus was silenced in MES cell lines, SH-EP and GI-ME-N (Fig 1B), indicating that ADRN-type cells exclusively express DCX mRNA. An inspection of t-SNE–processed expression data from mouse sympathetic precursors (Furlan et al, 2017) showed that Dcx was highly expressed by sympathoblasts (Fig S1C). A lower amount of Dcx mRNA was present in chromaffin cells and the bridge population. In neuroepithelia (NE) and nascent neurons of the central nervous system (CNS), DCX specifically regulates nuclear translocations or NUC (Tsai & Gleeson, 2005). We reasoned that NUC could affect the ADRN type of NB cells. Live-cell imaging of MNA ADRN cell lines, IMR-32 and NB-S-124 revealed instances of nuclear mobility (Fig 1C). We pursued this observation further by comparing movies of migrating IMR-32 and SH-EP cells expressing histone H2B fused to a fluorescent protein (nuclear marker or “_NM”) and detected variability in nuclear positioning in migrating TP53wt (Carr et al, 2006) IMR-32 cells, compared with MES-type SH-EP (Fig 1D and Video 1). Next, we recorded positions of cellular (CC) and nuclear centroids (NCs) (CC; NC tracking). Inspection of these tracks revealed NUC events in IMR-32 cells with the nucleus surpassing the cell centroid, followed by cell contraction or the nucleus leapfrogging over the cellular centroid (Fig 1E). We resolved CC; NC tracks (Lan et al, 2016) by linking each NC from a time point n to the CC in a time point n + 1, generating ∠NCn/NCn+1/CCn+1 (NNC/NCC) angle distribution and NC-CC maps (Fig 1F, left). This analysis revealed an overrepresentation of ≥140°; ≤180° block (140–180°), which should reflect leading process (LP) formation, MES mode and posterior NUC, in SH-EP_NM, as well as the overrepresentation of >0°; ≤40° block (0–40°), which should reflect cell contractions and anterior NUC, in IMR-32_NM and SK-N-BE(2)c_NM (TP53mut; Carr et al, 2006) (Fig 1F, left). We decoded NUC events from CC; NC tracks (linkages of the frames: [140–180°; CC > NN], followed by [0–40°; CC < NN]; [0–40°; CC < NN], followed by [0–40°; CC > NN]) which revealed NUC prevalence in IMR-32_NM and SK-N-BE(2)c_NM, compared with SH-EP_NM (Fig 1F, right). Cell velocity in IMR-32_NM and SK-N-BE(2)c_NM cells showed correlation with NUC footprint, compared with SH-EP_NM cells (Fig 1G). To corroborate these findings, we inspected phenotypes of IMR-32, IMR-32_NM, NB-S-124, and three other ADRN cell lines growing on top of three-dimensional collagen (pseudo-3-D assay), and observed traits of neurons migrating through 3-D matrix–bead-like dilations within LPs as well as nuclear deformations (Schaar & McConnell, 2005; Nishimura et al, 2014) (Fig 1H and Video 2). Next, we checked the nuclear migration in IMR-32_NM after cytochalasin B–induced actin depolymerisation or colcemid-induced MT depolymerisation. Both compounds inhibited cell motility; however, cytochalasin B-treated cells retained mobile nuclei, whereas colcemid treatment prevented nuclear migration (Fig S1D and E and Video 3). These data highlighted the role of MT in NUC in ADRN-type cells. Yet, MT-dependent NUC migration is normally observed in postmitotic cells, which leaves open the question as to how this process blends with the cell cycle in NB. In postmitotic neurons, DCX acts jointly with a dynein regulator LIS1 (PAFAH1B1) (Caspi et al, 2000), whereas in proliferating neural progenitors, DCX and LIS1 function differentially during the process named interkinetic nuclear migration (IKNM) (Carabalona et al, 2016). Two correlative studies have already implicated DCX and LIS1 in NB migration (Messi et al, 2008; Evangelisti et al, 2009), which was in support of their concert action in NB. LIS1 mRNA had a fit with GATA3, PHOX2B, and DCX mRNAs in primary NB (R_GATA3 = 0.53, R_PHOX2B = 0.52; R_DCX = 0.42; Fig S1A). Unlike DCX mRNA, LIS1 mRNA was present in cancers of non-NE origin and MES-type cells (Fig S1F and G), which could reflect LIS1 function during spindle assembly (Moon et al, 2014). Lis1 mRNA did not have an affinity with a particular subgroup in t-SNE–resolved expression data from mouse sympathetic precursors (Furlan et al, 2017; Fig S1C). While IKNM was not amenable to examination in dissociated cultures because of the absence of adherens junctions present in vivo (LaMonica et al, 2013), differential knockdown (KD) of DCX, controlling G1-specific, kinesin-dependent NUC (Carabalona et al, 2016), and of LIS1, controlling G2-specific, dynein-dependent NUC, and spindle assembly during IKNM (Tsai et al, 2005; Yingling et al, 2008; Carabalona et al, 2016), could help to tell IKNM from NUC in NB. 2D exclusion assays revealed inhibition of migration after either DCX or LIS1 KD in ADRN cell lines (Fig 1I). Proliferation in NB cell lines was not affected in DCX-KD and LIS1-KD cells (Fig S1H), which ruled out DCX and LIS1 involvement in spindle assembly or cell cycle transition in NB. Our tracking of randomly walking cells in DCX-KD and LIS1-KD IMR-32_NM spheroids showed inhibition of motility (Fig 1J and Video 4; data not shown). We then tested cell cycle specificity of migration in ADRN-type cells by performing cell tracking in asynchronously growing IMR-32 and an IMR-32 derivate, IMR5-75 expressing a FUCCI cell cycle sensor (Ryl et al, 2017), which revealed a tendency for migration in G1 phase (Fig 1K, left). No significant difference in NNC/NCC angle distributions in G1 and S/G2 phases was observed in IMR-32_NM (Fig 1K, right). Expression of a construct encoding γ-tubulin fused to mCherry in IMR-32 showed that nuclei surpassed γ-tubulin signals during migration (Fig 1L and Video 5), which was in agreement with the nucleus-centrosome (N-C) inversion mechanism (Umeshima et al, 2007).

• Download figure
• Open in new tab
• Download powerpoint

Figure S1. Expression of NUC genes, DCX, and LIS1, in NB.

(A) Scatter plots showing correlation of expression of DCX or LIS1 and ADRN transcription factors GATA3, PHOX2B, or MES transcription factor FOSL2 (498 patients; GEO: gse49710). (B) ADRN and MES cell lines used in the work (Boeva et al, 2017; van Groningen et al, 2017) (top left) and representative phase-contrast images of N-type NB-S-124 (bottom left). ChIP-seq (right) showing H3K4me3 and H3K36me3 binding at GATA3 and PHOX2B loci in I-type cell lines KELLY and LAN-2 and MES-type SH-EP. (C) Dcx and Lis1 expression in t-SNE-resolved E12.5 and E13.5 sympathetic precursors: sympathoblasts, Schwann cell precursors, bridge population, and chromaffin cells (Furlan et al, 2017). t-SNE maps are annotated according to Furlan et al (2017). (D)_NM fluorescence (negative grayscale) for nuclear morphologies in IMR-32_NM cells after treatment with vehicle, colcemid or cytochalasin b. Scale bar 100 μm. (E) Random walk plots and accumulated migration distances in vehicle, colcemid or cytochalasin B-treated IMR-32_NM (13 h, 15-min intervals). For colcemid experiments, the cells were washed after treatment and left to recover for 10 h before imaging. Cells were tracked using an ImageJ plugin for manual tracking. Mean migration distances + SD are presented. (F) LIS1 mRNA expression in MES and ADRN (marked in red) cell lines. (G) Affymetrix LIS1 mRNA expression analysis (left) in two series of NB tumor samples (red) and series of various other tumor types and neural crest cell line (blue) (probe set 229349). Affymetrix LIS1 mRNA expression data (right) from various cancer cell lines (CCLE). NB cell lines are indicated in red. (H) Validation of DCX and LIS1-KD (left) by qRT-PCR 48 h after transfection in IMR-32 and cell viability of NB cell lines after DCX or LIS1-KD (right). Cell survival was assessed by Alamar Blue assay at 48 and 72 h after transfection. Values are reported as mean percent ± SD of vehicle-treated control.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. NB cells exhibit NUC during migration.

(A) DCX mRNA expression in MES and ADRN (marked in red) cell lines (the names of ADRN cell lines whose identity was assigned based on the phenotype and/or PHOX2b, GATA3, and FOSL2 transcription are represented in grey color [Fig S1A], the cell type identity in the other cell lines [black color] is described in the literature). (B) ChIP-seq showing H3K4me3, H3K36me3, and H3K27me3 binding at the DCX locus in three MES and ADRN (marked in red) cell lines. For H3K27me3 binding, the region surrounding DCX locus (white box) was visualized in IGV program. (C) Time-lapse images showing migrating NB-S-124 and IMR-32 cells. (D) Time-lapse images of IMR-32_NM and SH-EP_NM cells during migration. Nuclear (in blue) and cellular center (in orange) of migrating cells are indicated. Scale bar 20 μm. (E) CN trajectories and CN plots of representative IMR-32_NM and SH-EP_NM cells. (F) Schematic of nuclei positioning determined by NCn+1-NCn/|NCn+1-CCn+1 angle (NNC/NCC) (left) and NNC/NCC angle frequency distribution in IMR-32_NM, SK-N-BE(2)c_NM, and SH-EP_NM cells (right). NUC events decoded from CC; nuclear centroid tracks in concatenated tracks from IMR-32_NM, SK-N-BE(2)c_NM, and SH-EP_NM (top right). Mapping of NUC (exemplary track in red), positive and negative noise-corrected NC-CC distances, 0–40° and 130–180° signatures (two or more sequential frames within the same angle block) (exemplary multi-colored track) in concatenated tracks from IMR-32_NM, SK-N-BE(2)c_NM and SH-EP_NM (bottom right). (G) Correlation plots between cell velocity and NUC footprint (weighted mean NUC distance) in IMR-32_NM, SK-N-BE(2)c_NM and SH-EP_NM cells. (H) Live imaging of pseudo-3-D-assayed IMR-32_NM cells. Scale bar 10 μm. (I) 2D exclusion assay in NB cell lines after RNAi against DCX or LIS1 72 h post-transfection. Relative cell migration is quantified by cell density’s normalization to control siRNA-transfected control. Graphs represent the mean relative difference migration ± SD. (J) Random walk plots and accumulated migration distances in control IMR-32_NM cells and after RNAi against DCX or LIS1 72 h post-transfection (13 h, 15-min intervals). Mean migration distances + SD are presented. (K) Box plots showing cell migration distances in 216 and 1,528 (G1 phase), 167 and 985 (S/G2 phases) sequential timepoints (20 and 5 min per timepoint, respectively) from tracings of 13 cells from IMR5-75 and 22 cells from IMR-32 expressing the G1 cell cycle sensor (left); P-values: IMR5-75: 6.168 × 10⁻⁵, IMR-32: 2.2 × 10⁻¹⁶ (two-sample Kolmogorov-Smirnov test). NNC/NCC angle frequency distribution and noise-corrected nuclear centroid-CC distances (right) in 0–40° and 140–180° signatures in concatenated tracks from IMR-32_NM expressing the G1 cell cycle sensor. (L) Time-lapse images showing representative migrating IMR-32_NM expressing γ-tubulin. Scale bar 20 μm.

The plotting of nuclear positions (NCn-CCn or |NC-CC|) showed that nuclei were less present in the periphery of the cells after DCX- or LIS1-KD (Fig 2A). Staining for NUC-relevant neuronal βIII-tubulin (Xie et al, 2003) and γ-tubulin showed that cell phenotypes produced by RNAi of DCX and LIS1 were different and resembled neurons after Lis1 or Dcx-KD (Youn et al, 2009; Nishimura et al, 2014), that is, higher variability in the N-C distance and dose-dependent unipolar neurite outgrowth in LIS1-KD cells as well as defective LPs in DCX-KD cells (Fig 2B). Accordingly, suppression of 0–40° signature, NUC as well as no NNC/NCC angle overrepresentations were seen after DCX-KD and LIS1-KD, respectively (Fig 2C and D). Yet, cell velocity in DCX-KD and LIS1-KD IMR-32_NM cells showed substantial correlation with the NUC footprint (Fig 2E). Altogether, these analyses showed no evidence for MES migration mode in LIS1-KD and DCX-KD IMR-32_NM cells. Also, gene set enrichment analysis (GSEA) of DCX- and LIS1-KD IMR-32 RNA-seq-resolved expression profiles showed that there was no induction of the MES program (Fig S2A and Table S1), which supported our idea regarding the maintenance of ADRN program after DCX- and LIS1 RNAi. DCX-KD and LIS1-KD transcriptomic profiles showed remarkable overlap (Fig 2F). Particularly, MT-related gene signatures and gene signatures associated with cortical cytoarchitecture were depleted in DCX-KD and LIS1-KD transcriptomes (Table S1). Gene signatures related to non-sense–mediated decay (NMD) and mRNA transport, mitochondrial function, oxidative phosphorylation (oxphos) signature, as well as genes related to cell contraction and Ras signalling regulation were specifically depleted in DCX-KD and LIS1-KD cells, respectively (Figs 2G and S2B). The down-regulation of pathways related to mitochondrial function was the only expression trait shared with KD of a DCL family gene, dclk1 in mouse neuroblastoma (Verissimo et al, 2010) (Table S1). On the other hand, GSEA of the up-regulated genes showed less consistency even under the relaxed threshold (q-values ≤ 0.25) (Table S1). Pseudogene transcripts, signatures related to G-protein coupled receptor signalling and PTPRB neighbourhood were overrepresented by the up-regulated genes in DCX-KD cells (Fig S2C and Table S1). Next, we assessed DCX- and LIS1-KD expression profiles in the context of primary NB using three clinically relevant dichotomies (stage 4S versus stage 4, stages 1|2 versus stage 4; MYCN-nonamplified versus MNA tumors). This analysis showed that up-regulated, but not down-regulated genes in the DCX-KD cells had an affinity toward transcriptomic profiles of stage 4S, stages 1|2 and MYCN-nonamplified tumors (Figs 2H and S2D). We then extracted then the genes from DCX- and LIS1-KD profiles that did not recapitulate association with low-risk characteristics, which we named “mis-expressed” genes (DCX(LIS1)-KD^UP ∩ stage 4S [stages 1|2; MYCN-nonamplified]^DOWN; DCX(LIS1)-KD^DOWN ∩ stage 4S [stages 1|2; MYCN-nonamplified]^UP [P-values ≤ 0.05, no logFC threshold]). GSEA showed that mis–down-regulation engaged TP53 targets and mis–up-regulation engaged CHEK2 neighbourhood in LIS1-KD cells (Fig 2I and Table S1). Given the association of LIS1 expression with favourable prognosis in NB (Garcia et al, 2012), these results might reflect LIS1 involvement in the potentially tumor-suppressive process(es) in NB. In DCX-KD cells, little consistency was found in the list of mis–up-regulated genes, whereas mis–down-regulation engaged genes encoding extracellular matrix, MYC targets as well as genes marked bivalently in the CNS (Fig 2J). These observations showed that phenotypic NUC inhibition in ADRN-type NB could generate disparate transcriptomic alterations. In DCX-KD cells, the down-regulated genes showed positional enrichment for NB-relevant regions of loss of heterozygosity (Mora et al, 2001; White et al, 2001; Lasorsa et al, 2020), including 19p13, 19q13, and 1p36 (Table S1). We checked the literature on NB for similar observations and found that a link between MYCN down-regulation and oxphos inhibition was identified previously in MNA NB cells (Dzieran et al, 2018; Oliynyk et al, 2019). Also, the positional depletion of 19p13 genes was observed in disseminated NB tumor cells (NB DTC) from relapsed patients (Table S8 by Rifatbegovic et al, 2018). Although the 19p13 gene overlap in the DCX-KD and NB DTC profiles was negligible (Table S1), 19p13 shutdown in relapse DTCs and DCX-KD cells could imply similar regulatory mechanisms. We retrieved the RNA-seq data from NB DTCs and reanalysed with the following settings: no logFC cutoff; P-values ≤ 0.05; DTC^TOTAL versus Tumor (TU)^TOTAL; DTC^MNA versus TU^MNA; DTC^RELAPSE versus TU^RELAPSE to identify whether similarities between transcriptomic profiles of DCX-KD IMR-32 and NB DTC go beyond the 19p13 genes. DCX-KD^UP gene set showed splitting when mapped onto NB DTC transcriptome profiles, whereas DCX-KD^DOWN showed a match with NB_DTC^DOWN (Fig 2K). Cross-checking for immune cell-specific signatures showed that the presence of DCX-KD^UP ∩ NB_DTC^UP overlap could not be explained by bone marrow cell contamination (Fig S2E). Several gene signatures associated with stage 4S, stages 1|2 and MYCN-nonamplified status, as identified by the parametric analysis of gene set enrichment (PAGE analysis) using gse49710 signature (Table S1), showed similar splitting (Fig S2F), suggesting cellular heterogeneity of NB DTCs rather than a gene signature functional variegation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. NUC genes are involved in migration in ADRN-type cells.

(A) |NC-CC| plots for control IMR-32_NM cells and the cells after DCX- or LIS1-KD 72 h post-transfection. (B) βIII-tubulin immunolabeling (top left) in control IMR-32 cells and after RNAi against DCX and LIS1 72 h post-transfection. Neurite outgrowth after LIS1 RNAi (bottom left; grayscale negative field) is indicated by white arrows. Scale bar 100 μm. Violin plots showing the nucleus-to-centrosome distance distribution in DCX- and LIS1-KD IMR-32 cells (right). Means are indicated. P-values: DCX siRNA: 0.009898, LIS1 siRNA: n.s.; (Welch t test). (C) Mapping of NUC, positive and negative noise-corrected NC-CC distances, 0–40° and 140–180° signatures in concatenated tracks from DCX-KD (27 cells, 762 time points) and LIS1-KD (41 cells, 1,256 timepoints) IMR-32_NM. (D) NNC/NCC angle frequency distribution in concatenated tracks in DCX- and LIS1-KD IMR-32_NM cells. (E) Correlation plots between cell velocity and NUC footprint in DCX- and LIS1-KD IMR-32_NM cells. (F) Venn diagram showing overlaps between the differentially expressed genes (DEGs; |logFC| cutoff: 0.3) in DCX- and LIS1-KD IMR-32 versus control IMR-32. (G) Gene set enrichment analysis plots of the indicated gene sets in DCX-KD IMR-32. False discovery rate (FDR)-adjusted P-values (q-values) are listed. (H) Volcano plots showing log2FC expression (y-axis) and log10 FDR-adjusted P-value (−log10 FDR-P-value, x-axis) of DEGs in DCX- and LIS1-KD IMR-32 in stage 4S versus stage 4 tumors (GEO: gse49710). Each dot represents an individual spot. (I, J) Gene set enrichment analysis plots of the indicated gene sets in LIS1- and DCX-KD IMR-32 based on their mis-expression in stage 4S versus stage 4 tumors (P-values by two-way t test ≤ 0.05, no logFC cutoff). (K) Plots showing mapping of DEGs of DCX-KD IMR-32 onto NB disseminated tumor cell transcriptome profiles (Rifatbegovic et al, 2018).

• Download figure
• Open in new tab
• Download powerpoint

Figure S2. Analysis of DEGs in DCX- and LIS1-KD IMR-32.

(A) Gene set enrichment analysis (GSEA) plots showing gene set “epithelial–mesenchymal transition” in DCX- and LIS1-KD IMR-32 versus control. q-values are listed. (B) GSEA plots showing gene signatures related to cell contraction and Ras signalling regulation in LIS1-KD IMR-32 versus control IMR-32. q-values are listed. (C) Enrichment plots for pseudogene transcripts (Pei et al, 2012), gene signatures related to G-protein coupled receptor signalling and PTPRB neighbourhood in DCX-KD IMR-32 versus control IMR-32. q-values are listed. (D) Volcano plots showing expression of DEGs in DCX-KD and LIS1-KD cells in MYCN-nonamplified versus MNA and stages 1|2 versus stage 4 tumors. (E) GSEA plots for immune cell specific signatures of DCX-KD^UP ∩ NB_disseminated tumor cell list. (F) GSEA plots for the signatures associated with stage 4S, stages 1|2 and MYCN-nonamplified status, as identified by parametric analysis of gene set enrichment (PAGE analysis) using gse49710 signature (Table S1) in NB disseminated tumor cell transcriptome profiles (Rifatbegovic et al, 2018).

Altogether, the results show that migration in ADRN-type cells depends on the N-C inversion mode of NUC that requires both DCX and LIS1, which rules out the possibility of proliferative NUC, IKNM. The results also imply that DCX-KD–like situations appear in disseminated NB cells.

ADRN SE-associated TF, SOX11, is involved in NUC regulation in ADRN NB

The parallelism between migration modes in neurons and ADRN-type NB cells is likely to extend to the transcription control, which could involve neuron-specific TFs embedded in NB pathogenesis. We reasoned that DCX co-expression signature may shed information on NUC control in NB, as the expression of DCX mRNA was a specific trait of cancers of NE origin, including NB (Figs 3A, top and S1A, and S3A). DCX expression did not discriminate between low-stage and advanced-stage NB tumors (Fig S3B), which could emerge from differential TF control; therefore, we focused on the stage 4 subset (gse49710). Interrogation of DCX co-expression signature for the general TFs involved in neuronal migration (Kwan et al, 2012; Hoshiba et al, 2016) retrieved five prospective candidates (Fig S3C). One candidate, the high-mobility-group domain-containing TF encoding SOX11 was a strong hit (ninth most positively DCX mRNA-correlated TF in stage 4; R_stage4 = 0.54). SOX11 is mutated in Coffin-Siris syndrome (Tsurusaki et al, 2014; Nemani et al, 2014) and is a likely candidate for 2p25.2 deletion syndrome (Lo-Castro et al, 2009); these disorders are characterised by microcephaly with hindbrain abnormalities. In mice, Sox11 deficiency generates complex CNS defects that stem from proliferation deficits in NE and neuronal migration errors (Hoshiba et al, 2016). In peripheral nervous system (PNS), Sox11 regulates proliferation during the early development of sympathetic ganglia (Potzner et al, 2010). It was previously shown that Sox11 was indispensable for the proliferation of tyrosine hydroxylase-expressing precursors in developing sympathetic ganglia (Potzner et al, 2010). In line with this, Sox11 expression was observed in sympathoblasts and the bridge population (Furlan et al, 2017; Fig S3D). In NB, SOX11, identified previously as an SE-associated TF in ADRN-type cells (van Groningen et al, 2017), was highly expressed in MYCN-nonamplified stage 4 and MNA NB (localised [LOC] versus MYCN-nonamplified stage 4, P-value < 0.005; MYCN-nonamplified versus MNA, P-value < 0.001; one-way ANOVA test; Fig S3E). SOX11 expression profile across cancer cell lines and primary tumors showed concordance with DCX expression, except for human neural crest cells and pheochromocytoma, which could partially reflect earlier SOX11 induction during neural crest development and SOX11 lineage specificity (Potzner et al, 2010; Figs 3A, bottom and S3F). Consistently, SOX11 mRNA was prevalent in ADRN cell lines, as compared with MES cell lines (Fig 3B). SOX11 directly controls DCX expression in neurons (Mu et al, 2012), which could also hold true in NB and, hence, affect cell migration. We observed decreased migration from SOX11-KD IMR-32_NM spheroids and the closure of |NC-CC| distance along with DCX downmodulation (Fig 3C–E and Video 4). Real-time qRT-PCR in two other SOX11 high expressors, ADRN cell lines CLB-GA and NGP, also showed DCX mRNA depletion after SOX11 RNAi (Fig 3E). As RNA-seq data showed, SOX11 targets identified in mouse ES-derived nascent neurons (Bergsland et al, 2011) were significantly down-regulated (q-value < 0.05), whereas other SOX11 target datasets (Lachmann et al, 2010; Kuo et al, 2015) did not demonstrate depletion (Fig S3G). DCX- and SOX11-KD transcriptomic profiles showed little overlap, but formed a DCX-KD^DOWN ∩ SOX11-KD^DOWN semantic match that involved signatures related to MT function and mRNA transport as well as MYC targets (Fig 3F). No overlapping signatures were found in the up-regulated genes (Table S2). In contrast to the findings in neurons (Piens et al, 2010; Μu et al, 2012), DCX promoter was not bound by SOX11 in NB cells (Decaesteker et al, 2020 Preprint). Also, SOX11-KD IMR32_NM had different morphology, as compared with the DCX-KD cells (Fig 3G) and retained parental NNC/NCC angle distribution (Fig 3H, top). The impact of 0–40° signature and 0–40°-linked NUC was preserved in SOX11-KD cells (Fig 3H, bottom). Given the intact NNC/NCC angle distribution, it was worth checking nuclei localisation in the pre-mitotic KD cells, which revealed a higher percentage of asymmetrically localised metaphases in SOX11-KD IMR-32_NM, compared with DCX-KD and control cells (Fig 3I). Cell velocity in SOX11-KD IMR-32_NM cells showed no correlation with NUC footprint, compared with control or NUC-suppressed, DCX-KD and LIS1-KD cells (Figs 3J and 2E). This suggested that SOX11-KD IMR-32_NM acquired slower, NUC-independent migration mode. In line with this, we observed the overrepresentation of gene hallmarks “wound healing” and “EMT” by the up-regulated genes, which supported our idea about SOX11 RNAi-induced reprogramming (Fig 3K and Table S2). SOX11 RNAi resulted in down-regulation of several ADRN TFs, ISL1, KLF7, MYCN, and neuron-specific RNA binding protein-encoding ELAVL2 and ELAVL4, as well as induction of MES TFs, ETS1 and JUND (Boeva et al, 2017; van Groningen et al, 2017; Zeid et al, 2018) (Table S2) and the targets of ETS-1 and AP-1. The targets of ELAVL proteins identified previously in IMR-32 (Scheckel et al, 2016) were down-regulated in SOX11-KD IMR-32, which was in striking contrast to the DCX-KD profile (Fig S3H). Also, gene signatures “oxphos,” “NMD” and gene lists associated with migration/actin cytoskeleton regulation and cortical cytoarchitecture showed inverted pattern of regulation in SOX11-versus DCX-KD cells (SOX11-KD^UP ∩ DCX-KD^DOWN) (Fig 3L and Table S2). Further assessment of SOX11-KD transcriptomic profile in the context of the clinically relevant dichotomies revealed that mis–up-regulation engaged MYC targets, RAN neighbourhood, whereas mis–down-regulated involved bivalently marked genes (Figs 3M and S3I and Table S2). The SOX11-KD^DOWN gene set matched with NB DTC^UP set (Fig S3J). This indicated that SOX11^HIGH, rather than SOX11^LOW IMR-32 reflected a NB DTC sub-compartment. Next, we inspected whether SOX11 was capable of initiating the MES-to-NUC transition by forcibly expressing SOX11 in the inducible format in SH-EP. We observed a moderate velocity gain in the cells with enforced SOX11 expression, compared with the controls (Fig S3K). NNC/NCC angle distribution revealed acquisition of 0–40° block in SH-EP expressing SOX11, but the highest cell velocity was associated with 140–180° block in SH-EP expressing SOX11 (Fig S3L), suggesting that SOX11 forced expression interfered with nuclear positioning mechanisms in SH-EP without reprogramming migration type.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. SOX11 regulates NUC.

(A) Affymetrix DCX mRNA (probe set 229349, top) and SOX11 mRNA (probe set 204914, bottom) expression analysis from primary tumors. (B) SOX11 mRNA expression in MES and ADRN (marked in red) cell lines. (C) Random walk plots and accumulated migration distances in control IMR-32_NM and after RNAi against SOX11 72 h post-transfection (6 h, 15-min intervals). Mean migration distances + SD are presented. The efficiency of SOX11-KD was determined by WB. (D) Representative |NC-CC| plots in IMR-32_NM controls and after RNAi against SOX11. (E) Relative qRT-PCR for SOX11 and DCX in IMR-32 after SOX11 RNAi 72 h post-transfection and in NGP and CLB-GA after SOX11 RNAi 48 h post-transfection (left). Mean relative difference values + SE of control are presented. WB for DCX in IMR-32 after 48 h post-transfection (right). (F) Venn diagram showing the number of gene sets (extracted from MSigDB [22,596 gene sets]) in IMR-32 DCX-KD^DOWN ∩ SOX11-KD^DOWN semantic overlap (q-values ≤ 0.05). (G) Cell morphology in SOX11-and DCX-KD cells. Scale bar 20 μm. (H) NNC/NCC angle frequency distribution (top) and mapping of NUC, positive and negative noise-corrected NC-CC distances, 0–40° and 140–180° signatures (bottom) in SOX11-KD (57 cells, 1,314 timepoints). (I) The percentages of asymmetrically localised metaphases (left) in control, SOX11- and DCX-KD IMR-32_NM. Mean percentages + SD are presented. Representative images of dividing SOX11-KD cell (right, scale bar 20 μm). (J) Correlation plot between cell velocity and NUC footprint (weighted mean NUC distance) in SOX11-KD IMR-32_NM. (K) Gene set enrichment analysis plots showing gene sets “wound healing” and “epithelial–mesenchymal transition” in SOX11-KD versus control. (L) Venn diagram showing the number of gene sets (MSigDB) in IMR-32 SOX11-KD^UP ∩ DCX-KD^DOWN semantic overlap (q-values ≤ 0.05). (M) Volcano plots showing expression of DEGs in SOX11-KD IMR-32 in stage 4S versus stage 4 tumors.

• Download figure
• Open in new tab
• Download powerpoint

Figure S3. Modulation of SOX11 levels in ADRN and MES NB cell lines.

(A) Affymetrix DCX mRNA expression analysis in cancer cell lines. (B) Box plots demonstrating DCX expression in primary NB: MYCN-nonamplified (stages 1–3 [localised, LOC], stage 4, stage 4S) and MNA (gse49710) and P-values by one-way ANOVA analysis. (C) Scatter plots (left) showing correlation of expression of DCX mRNA and the transcription factors involved in neuronal migration (gse49710) and Top10 of transcription factors positively correlated with DCX mRNA (right) in stage 4 NB (gse49710). (D) Sox11 expression in t-SNE-resolved E12.5 and E13.5 sympathetic precursors: sympathoblasts, Schwann cell precursors, bridge population and chromaffin cells (Furlan et al, 2017). Population annotation is as described in Fig S1C. (E) Box plot demonstrating SOX11 mRNA expression in primary NB (gse49710): MYCN-nonamplified (LOC, stage 4, and stage 4S) and MNA. (F) Affymetrix SOX11 mRNA expression analysis in various cancer lines. (G) Gene set enrichment analysis plots showing humanized gene list of Sox11 bound genes (Bergsland et al, 2011) and SOX11 targets (Lachmann et al, 2010; Kuo et al, 2015) in SOX11-KD versus control. (H) Gene set enrichment analysis plots showing targets of ELAVL proteins (Scheckel et al, 2016) in SOX11-KD, DCX-KD IMR-32 versus control. (I) Volcano plots showing expression of DEGs in SOX11-KD IMR-32 in MYCN-nonamplified versus MNA and stages 1|2 versus stage 4 tumors. (J) Mapping plots of DEGs in SOX11-KD IMR-32 onto NB disseminated tumor cell transcriptome profiles (Rifatbegovic et al, 2018). (K) Random walk plots in SH-EP_SOX11-TAT (top left) after application of doxycycline (doxy) or vehicle (1 h of pretreatment + 6 h 30 min of tracking, 15-min intervals) and box plots demonstrating cell velocity (bottom left) in vehicle (n = 2,065) and doxy-induced (n = 2,112) SH-EP_SOX11-TAT. P-value: 2.2 × 10⁻¹⁶ (Kruskal-Wallis test). SOX11 overexpression was confirmed by fluorescence monitoring and WB (right, exemplary tracks from three experiments). (L) NNC/NCC angle frequency distribution in concatenated tracks from control (22 cells; 923 timepoints) and doxycycline-induced (106 cells; 1,904 timepoints) SH-EP_SOX11-TAT (left). Box plots showing cell velocities (right) across NNC/NCC angle frequency distribution in SH-EP_SOX11-TAT.

Altogether, these data show that NUC is controlled by SOX11 in NB. SOX11 RNAi fosters morphological asymmetric cell divisions and causes reprogramming of NUC migration in ADRN-type NB cells. The latter was in contrast to the observations carried out in nascent neurons previously.

Inhibition of ROCK or RAC1 blocks different steps of NUC migration

We reasoned that targeting the intrinsic regulators of the MT cytoskeleton to generate DCX RNAi-like NUC errors was a prospective direction for the development of migration blockers in the ADRN-type NB. In neurons, the formation of LPs is regulated by RAC1, whereas actomyosin contractions at the trailing end are regulated by RHOA-ROCK (Kawauchi et al, 2003; Martini & Valdeolmillos, 2010). The down-regulation of RAC1 neighbourhood, RHOA-related gene signatures, “BERENJENO_transformed_by_RHOA_DN,” and “BERENJENO_transformed_by_RHOA_UP,” after DCX RNAi indirectly supported the role of the RAC1 and RHOA pathways in DCX-KD phenotype in NB (Fig 4A). To assess these processes, we applied a ROCK inhibitor, Y27632 (5 μM), which reduced migration in all tested ADRN and MES cells, whereas cell viability was unaffected (Figs 4B, left and S4A and B). In contrast, an RAC1 inhibitor, NSC23766 (10 μM) (Bid et al, 2013), reduced migration only in ADRN cell lines without affecting cell viability (Figs 4B, left and S4B and C), in line with the observed effect upon RAC1 siRNA treatment (Figs 4B, right and S4D). In neurons, the RAC1 function is mainly associated with the modulation of cytoskeletal dynamics in the growth cones, dendritic spines, and lamellipodia, which are absent in IMR-32 (Fig S4E, left). In line with this, per equal amount of total protein, basal RAC1 activity was lower in ADRN IMR-32 and SK-N-BE2c, as compared with SH-EP (Fig S4E, right). The absence of lamellipodia was reflected by perinuclear and nuclear localisation of RAC1 identified with either immunostaining in IMR-32 or forced expression of fluorescent protein-tagged full-length RAC1 (Fig 4C). Visually, only ROCK inhibitor-exposed ADRN-type cells underwent a profound morphological transformation (Figs 4D and S4F). Tracking showed motility defects after either RAC1 or ROCK inhibition in IMR-32_NM, which was further confirmed by RNAi against ROCK1, ROCK2, and RAC1 (Figs 4E and F, and S4G and Video 5). Both RAC1- and ROCK inhibition–induced defects in IMR-32_NM fit the neuroblastic cell phenotype and did not demonstrate MES traits (Fig 4G and Videos 5–Videos 7). Cell morphology after RAC1 suppression revealed NUC defects, which was also confirmed by the results of |NC-CC| plotting (Fig 4H). In agreement with RAC1/ROCK antagonism (Petrie et al, 2009), RAC1 inhibitor-treated cells maintained cell contractions, whereas ROCK inhibitor–exposed cells retained nuclear migrations but failed to detach (Fig S4H and Videos 5–Videos 7). Staining for βIII-tubulin revealed that ROCK inhibition up-regulated NUC fork–like MT elements (Xie et al, 2003), whereas fewer MT structures were observed after RAC1 inhibition, which was an indication of RAC1 involvement in the regulation of NUC MTs (Fig 4I). RNA-seq analysis showed that there was an overlap between transcriptomic alterations induced by RAC1 inhibitor and DCX-KD (Fig 4J). Similar to DCX RNAi, genes up-regulated by RAC1 inhibition genes had an affinity towards stage 4S, stages 1|2 and MYCN–nonamplified status (Figs 4K and S4I). GSEA showed that the genes up-regulated by RAC1 inhibition were enriched for TCF21-dependent genes and bivalently marked genes (Table S3). Although we did not find GSEA-based evidence for MES signature in the transcriptomic profile of RAC1-inhibited cells, several hallmarks, including ETS fusion (EWS/ETS) targets, formed an overlap with the SOX11-KD^UP gene set. Also, RAC1 neighbourhood genes that showed inverted, DCX-KD^DOWN/SOX11-KD^UP, expression pattern, were not depleted in the RAC1 inhibitor-treated cells (Table S3). We thought that identification of NUC RAC1-GEF(s) could help to refine RAC1 suppression, and hence, inhibition of NUC migration in the context of NB. As previous studies showed, three GEFs, T-lymphoma and metastasis gene 1 TIAM1, triple functional domain protein TRIO and TRIO homologue kalirin, are inhibited by NSC23766 (Gao et al, 2004; Zeinieh et al, 2015). These GEFs, which were already implicated in NB biology (Molenaar et al, 2012; Pugh et al, 2013), were highly expressed in primary NB and NB cell lines (Fig S5A and B). To date, only Tiam1 was reported to activate Rac1 in the context of migrating mouse CNS neurons (Kawauchi et al, 2003). In primary NB, KALRN showed stronger co-expression with DCX and SOX11, compared with TRIO and TIAM1 (Fig S5C), indicating ADRN-type–specific expression of KALRN, which was also reflected by RNA profiles in NB cell lines (Fig 5A). We checked the expression of the three GEFs by WB using RAC1–GEF domain-relevant ABs, which showed a 160-kD band, identified as delta-kalirin-8 (Mains et al, 1999, in primary MNA NB as well as ADRN cell lines, and a 115-kD band in several MYCN-nonamplified primary NBs (Figs 5B and S5D). An AB against kalirin PDZ-binding motif (STYV) detected a 100-kD delta-kalirin-7 in several primary MYCN-nonamplified NB. TIAM1 isoforms were found in ADRN cell lines as well as several primary NBs. An anti-TRIO AB detected several annotated isoforms in one primary MYCN-amplified NB and a 50-kD band in the cell lines. As immunostaining showed, kalirin isoforms, like RAC1, had perinuclear and nucleolar localisation, whereas TRIO and TIAM1 were found only in nucleoli and in cell boundaries, respectively, which supported the likelihood of kalirin involvement in RAC1 activation and MT regulation (Figs 5C and S6A–C). TIAM1, TRIO, and KALRN are potentially regulatable by SOX11 (Table S2); therefore, we checked their expression by WB, which revealed a reduction of kalirin and TIAM1 levels in SOX11-KD IMR-32 cells (Fig S6D). Kalirin-SP and kalirin-STYV signals had uneven distribution at migration fronts of immunostained NB-S-124 (Fig S6E) and IMR-32 spheroids (data not shown), which was also confirmed by WB of sparsely seeded cells, suggesting that kalirin isoforms were functionally diversified (Fig 5D). We down-regulated the kalirin RAC1-GEF (GEF1) pharmacologically with the compounds that do not interfere with TIAM1–ITX3 (kalirin–GEF1 inhibitor#1, 10 μM) and NPPD (kalirin–GEF1 inhibitor#2, 5 μM) (Blangy et al, 2006; Ferraro et al, 2007; Bouquier et al, 2009; Yan et al, 2015). Active RAC1 levels were lower in IMR-32 and SK-N-BE(2)c treated with either RAC1 inhibitor or each of kalirin–GEF1 inhibitors. KALRN RNAi in IMR-32 (Fig S7A and B) also reduced RAC1 activity, thus confirming kalirin involvement in RAC1 activation (Figs 5E and S7C). This effect was not observed in the reprogrammed after SOX11 RNAi cells (Johnson et al, 2000).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. ROCK and RAC1 inhibition interferes with cell detachment and NUC.

(A) Gene set enrichment analysis plots showing RHOA-related and RAC1-related gene sets in DCX-KD IMR-32 versus control. (B) 2D exclusion assay in NB cell lines after treatment with the RAC1 inhibitor or ROCK inhibitor and after RAC1 RNAi (left). Relative cell migration (right) is quantified by normalization of cell density to vehicle- or control siRNA-treated cells. Mean relative difference values ± SD are presented. (C) RAC1 immunostaining and CyPet-RAC1 subcellular localisation in IMR-32 (L, laminin; CIV, collagen IV; F, fibronectin). Scale bar 20 μm. (D) Images of IMR-32 spheroids stained with Calcein AM after 72 h of treatment with vehicle or ROCK inhibitor in pseudo 3-D. (E) Random walk plots in IMR-32_NM treated with vehicle, RAC1 inhibitor or ROCK inhibitor for 72 h (13 h, 15-min intervals). Scale bar 200 μm. (F) Accumulated migration distances of control, ROCK inhibitor– and RAC1 inhibitor–treated IMR-32_NM cells and after ROCK1, ROCK2, or RAC1 RNAi. Mean migration distances + SD are presented. (G) Time-lapse images of IMR-32_NM after treatment with ROCK inhibitor (left) or RAC1 inhibitor (right). Nuclei and cell leading edges of migrating cells are indicated. Scale bar 20 μm. (H) |NC-CC| plots in control IMR-32_NM and after treatment with the RAC1 inhibitor or RNAi against RAC1. (I) βIII-tubulin immunolabeling in IMR-32 after treatment with vehicle, ROCK- or RAC1 inhibitor. Fork-like structures are showed in grayscale negative field. Scale bar 100 μm. (J) Venn diagrams showing the numbers of genes in DCX-KD ∩ RAC1 inhibitor transcriptomic overlap (P-values ≤ 0.09; |logFC| cutoff [DCX-KD]: 0.3, |logFC| cutoff [RAC1 inhibitor]: 0.2). (K) Volcano plots showing expression of DEGs in RAC1-inhibitor-treated IMR-32 in stage 4S versus 4 tumors.

• Download figure
• Open in new tab
• Download powerpoint

Figure S4. Cell analyses in NB cell lines after RAC1 and ROCK inhibition.

(A) Cell viability of NB cell lines treated with various concentrations of RAC1 inhibitor NSC23766 or ROCK inhibitor Y27632. Cell survival was assessed by Alamar Blue assay at 48 and 96 h after treatment. Mean percent values ± SD of vehicle-treated control are reported. (B) Cell viability of NB cell lines treated with ROCK inhibitor Y27632 (5 μM) or RAC1 inhibitor NSC23766 (10 μM) in 2D exclusion assay. Cell survival was assessed by Alamar Blue assay. Mean percent values + SD of vehicle-treated control are reported. (C) Random walk plots and accumulated migration distances of vehicle, RAC1 inhibitor- and ROCK inhibitor-treated NB-S-124 cells (2.5 h; 5-min intervals). (D) Cell viability of NB cell lines transfected with control siRNA or RAC1 siRNA in 2D exclusion assay. Cell survival was assessed by Alamar Blue assay. Mean percent values + SD of control are reported. (E) Representative images of IMR-32 showing filopodia and growth cone-free cell edges (left). Scale bar 20 μm. RAC1 activity in IMR-32, SK-N-BE(2)c and SH-EP (right). The cell lysates were incubated with PAK-PBD beads and the bound RAC1 was analysed by Western blotting. (F) Images of SK-N-BE(2)c spheroids stained with Calcein AM after 72 h of treatment with vehicle or ROCK inhibitor in pseudo 3-D. Cell cycle analysis by flow cytometry in SK-N-BE(2)c and SK-N-BE(2)c after the treatment with ROCK inhibitor. Cell cycle distribution of VioBlue Dye-labelled cells was analysed by flow cytometric analysis. Percentage of cells in each phase of the cell cycle and representative cell cycle images are shown. Mean values ± SD are presented. (G) qRT-PCR analysis of RAC1-, ROCK1- and ROCK2-KD in IMR-32 cells. (H) Tracings of IMR-32_NM nuclei after treatment with vehicle and ROCK inhibitor. (I) Volcano plots showing expression of DEGs in RAC1-inhibitor-treated IMR-32 in MYCN-nonamplified versus MNA and stages 1|2 versus stage 4 tumors.

• Download figure
• Open in new tab
• Download powerpoint

Figure S5. Expression of RAC1-GEF-encoding KALRN, TIAM1, and TRIO in NB.

(A) Affymetrix mRNA expression analysis (probe sets: KALRN, 230596, 206078, 227750, 232717; TIAM1, 213135; TRIO, 209011, 209012, 208178, and 209013) in two series of NB tumor samples (red) and series of various other tumor types (blue). (B) Expression of KALRN, TIAM1, and TRIO in NB cell lines (red) and series of different cancer cell lines (black). (C) Graphs showing correlation of expression of DCX (left) or SOX11 (right) and KALRN, TIAM1, or TRIO in primary NB (gse49710). (D) WB analysis of kalirin expression with kalirin-SP, TRIO, TIAM1, kalirin-STYV, and anti-kalirin-PH-2 in NB cell lines and primary tumors. Asterisks indicate previously annotated isoforms.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. Kalirin is a NSC23766-sensitive RAC1-GEF in ADRN-type cells.

(A) TIAM1, KALRN, and TRIO mRNA expression in MES and ADRN (marked in red color) cell lines. (B) WB analysis of TIAM1, TRIO and kalirin in a panel of primary NBs, NB cell lines. Location of antigens and RAC1-GEF domains in TIAM1, TRIO, and kalirin proteins (top) is marked by red lines and boxes, respectively. MNA status in primary tumors and cell lines is indicated. The names of ADRN cell lines are marked in red color. (C) IMR-32, SK-N-BE(2)c, and KELLY were stained with anti-STYV and anti–kalirin–SP antibodies and visualized with Cy3- or Alexa 488–conjugated secondary antibodies. Nuclei are indicated by the dashed lines. (D) WB analysis of kalirin–SP and kalirin–STYV in MNA cell lines harvested at 20% and 50% confluency. (E) RAC1 activity in IMR-32 and SK-N-BE(2)c cell treated with kalirin–GEF1 inhibitor#1 (10 μM), kalirin–GEF1 inhibitor#2 (5 μM), RHOA inhibitor (3 μM), RAC1 inhibitor (10 μM), or after SOX11 and KALRN RNAi. The cell lysates were incubated with RBD–Rhotekin or PAK-PBD beads and the bound activated RHOA and RAC1 was analysed by Western blotting. Flow through fraction was analysed for RHOA and RAC1 expression as loading control. (F) H3K4me3, H3K36me3, and H3K27me3 ChIP-seq showing binding events at KALRN promoter and gene body in NB cell lines. ADRN cell lines are marked in red color. Arrows indicate transcription start sites. (G) Expression of kalirin isoforms in primary NBs. (H) RT-PCR for 3′-most exons of kalirin-7, kalirin-12, Duet and kalirin-9 in IMR-32 and SK-N-BE(2)c. (I) KALRN gene profile obtained from RNA-seq data analysis of 27 primary NBs. 5′-most exons, 3′-most exons of the isoforms are indicated, exons encoding known and high score calpain cleavage sites (predicted for the kalirin-12 [uniprot: O60229-1]) are marked in purple color.

• Download figure
• Open in new tab
• Download powerpoint

Figure S6. Expression and localisation of kalirin, TRIO, and TIAM1 proteins in NB cells.

(A) Immunolabelling of TRIO and TIAM1 in IMR-32, SK-N-BE(2)c, and KELLY grown on (collagenIV/laminin/fibronectin)-coated plastic. Scale bar 20 μm. (B) Subcellular localisation of kalirin-STYV, Golgi apparatus and γ-tubulin (left) in IMR-32 cells. Scale bar 10 μm. For superimposed images (right), centrosome-specific (centrin-2), and Golgi apparatus signals were colored green. Scale bar 5 μm. (C) Heterogeneity of kalirin-SP and kalirin-STYV subcellular localisation in IMR-32 and SK-N-BE(2)c. Scale bar 20 μm. (D) WB analysis of TIAM1, TRIO, and kalirin in IMR-32 cells after 48 h of SOX11 RNAi. (E) Migration fronts of NB-S-124 stained by kalirin-STYV and kalirin-SP. Scale bar 10 μm.

• Download figure
• Open in new tab
• Download powerpoint

Figure S7. Modulation of KALRN levels in ADRN NB with RNAi and RA-induced differentiation.

(A) Immunolabelling of kalirin in IMR-32 after KALRN RNAi. Scale bar 20 μm. (B) Western blot analysis of kalirin-8 and kalirin-STYV levels in IMR-32 and SK-N-BE(2)c after KALRN RNAi and kalirin-8 levels in IMR32 after KALRN RNAi or additional negative control siRNAs: control siRNA #2, HNRNPK siRNA, and DISC1 siRNA. (C) RAC1 activity in IMR-32 and SK-N-BE(2)c cell treated with kalirin-GEF1 inhibitor#1 (10 μM), kalirin-GEF1 inhibitor#2 (5 μM), RHOA inhibitor (3 μM), RAC1 inhibitor (10 μM), or after SOX11 and KALRN RNAi. (D) Kalrn expression in t-SNE-resolved E12.5 and E13.5 sympathetic precursors: sympathoblasts, Schwann cell precursors, bridge population and chromaffin cells (Furlan et al, 2017). (E) Sequencing electrophoregrams showing 3′-UTR of kalirin-9 isoform (top) and the exon scheme based on the results of sequencing (bottom). Location of stop codon is marked with “-.” (F) KALRN, TRIO, and TIAM expression in SK-N-BE(2)c (left) after RA (10 μM) treatment. x-axis indicate timepoints in hours. Western Blot analysis of kalirin, TRIO, and βIII-tubulin in SK-N-BE2c after 72 h of RA-treatment (top right) and of TRIO and kalirin-8 in IMR-32 and SK-N-BE(2)c after 72 h of RA treatment (bottom right). (G) Cell viability of NB cell lines treated with vehicle, kalirin-GEF1 inhibitor#2, or kalirin-GEF1 inhibitor#1. Values are reported as mean percent ± SD of vehicle-treated control.

Based on co-expression with DCX mRNA and WB results, several ADRN type-specific KALRN transcripts were produced in primary NB, which was confirmed by ChIP-seq profiling for H3K36me3 in NB cell lines (Fig 5F). H3K4me3 peaks’ presence at exon-B and RAC1–GEF–unrelated Duet exon in all tested cell lines along with H3K27me3 loading at the KALRN in MES cell lines was indicative of bivalency, associated with the genes involved in neuronal specification (Liu et al, 2017). Kalirin is expressed in the murine heart, adrenal medulla and superior cervical ganglia (May et al, 2002). Kalrn mRNAs demonstrated no affinity with a particular population on the t-SNE–processed expression data from mouse sympathetic precursors (Furlan et al, 2017; Fig S7D). More specifically, kalirin isoforms −9 and −12 are expressed in mature sympathetic neurons (May et al, 2002), kalirin-9 protein is expressed in cardiac outflow tract (Wu et al, 2013), whereas kalirin-8 protein is found in rat and mouse neuroendocrine cells (Hansel et al, 2001; Ferraro et al, 2007). ADRN-type cells are likely to inherit KALRN expression from a sympathoadrenal precursor. In line with this, kalirin-9 and -12 expression was higher in stage 4S NB (4S versus 4: P-value = 4.4 × 10⁻⁶ [kalirin-9]; P-value = 9.1 × 10⁻³ [kalirin-12]) (Fig 5G). Given the low contribution of Duet 5′-exons into KALRN isoform repertoire in NB, this indicated that kalirin-9 was the most highly expressed KALRN isoform in stage 4S. 3′-UTR diversity was identified for kalirin-9 because a sub-isoform with a cryptic 3′-UTR exon carrying a stop codon was present in IMR-32 (Fig S7E). On the other hand, as WB in primary tumors showed, kalirin-8 expression was associated with MYCN amplification, but not with MYCN expression, indicating its sub-lineage specificity. RT-PCR for 3′-most exons of kalirin-9 and -12 showed positive results in untreated IMR-32 and SK-N-BE2c (Fig 5H), whereas the full-length kalirin-9 and -12 proteins were barely detectable and a kalirin–GEF2–relevant AB (Fig S5D) did not detect kalirin-8, which implied that other mechanisms (e.g., cleavage by calpains [Miller et al, 2017]) were involved in generating kalirin-8 in NB. Consistently, KALRN gene profiles from primary NB did not support kalirin-8 3′-UTR, which, along with the absence of 5′-most exons of A_23_P307563–detectable Duet and negative RT-PCR results for Duet, indicated the prevalence of kalirin-12 and kalirin-7 in ADRN NB (Fig 5H and I). Given the repertoire of kalirin isoforms in sympathetic neurons (May et al, 2002), we checked KALRN expression in a retinoic acid (RA)-induced neuronal differentiation model, SK-N-BE(2)c, and observed induction of kalirin-9 and kalirin-12 mRNAs and proteins as well as suppression of kalirin-8 after RA treatment. We also noticed TRIO transcripts and proteins’ strong induction, whereas TIAM1 mRNA levels remained unchanged (Fig S7F, left and top right). In IMR-32 cells, which are not amenable to differentiation, we observed neither kalirin-8 down-regulation nor induction of kalirin-9, -12, and TRIO (Fig S7F, bottom right).

Taken together, these results suggest that kalirin is involved in RAC1 activation in ADRN-type NB, particularly in its most aggressive, MNA sub-type.

Kalirin inhibition hinders NUC migration, evoking transcriptional signatures associated with low-risk characteristics in primary NB

Kalirin has not been identified as a regulator of NUC or MT function, as its functions are mainly attributed to the regulation of exocytosis and actin cytoskeleton in the post-migratory neurons. Moreover, kalirin is an extremely multifaceted molecule; therefore, we sought to carefully assess cell phenotype after kalirin suppression. Kalirin suppression did not affect cell viability (Fig S7G) and cell cycle distribution in NB cell lines (data not shown), which was in agreement with the pro-migratory role of RAC1. 2D exclusion assay revealed a decrease in migration after kalirin suppression in MNA ADRN cells and, to a lesser extent, in MYCN-nonamplified NB69, which was confirmed by cell tracking in IMR-32_NM and NB-S-124 (Figs 6A and B and S8A and B). As cell kymographs and |NC-CC| plotting demonstrated, kalirin-GEF1-suppressed IMR-32_NM displayed cell polarisation defects and NUC defects (Fig 6C and D and Video 7). Immunostaining showed that kalirin was not a centrosomal protein (Fig S6B), but rather colocalised with MTs and Golgi complex, which supported the idea of kalirin involvement in the regulation of the NUC function of MTs (Fig S8C). We noticed that, similar to DCX- or SOX11-KD, the proportion of the nuclei with distally located centrosomes as well as a variability in the N-C distance were lower in the cells after the suppression of kalirin or RAC1 inhibition (Figs 6E and S8D), indicating that the cells failed to translocate centrosomes, which was in line with the findings in neurons (Tanaka et al, 2004; Yang et al, 2012). Also, inhibition of either kalirin or RAC1 altered γ-tubulin distribution, suppressing extra-centrosomal γ-tubulin signals (Fig S8E). The heterogeneity of nuclear shapes, observed in fixed IMR-32 cells, was decreased in the cells treated with the RAC1 inhibitor or after kalirin suppression, which was an additional indication of a reduction in the cell fraction primed for NUC (Fig 6F and G). We found less distal βIII-tubulin staining in the cells after kalirin suppression (Fig 6H). These alterations closely resembled the defects of nuclear elongation and LP formation, observed after the block of MT polymerization in neurons (Nishimura et al, 2017). Thereafter, we stained IMR-32 for F-actin by fluorescently labelled phalloidin, which revealed a reduction in perinuclear staining in kalirin-GEF1 inhibitor#2-treated cells; therefore, actin regulation by kalirin could not be completely ruled out (Fig S8F). As a further test for kalirin involvement in NUC, we checked whether kalirin–GEF1 inhibition affected the expression of DCX, LIS1 and SOX11 and found down-regulation of DCX, but not SOX11, in the kalirin–GEF1–inhibited IMR-32 cells. These results suggested that kalirin suppression could reinforce migration inhibition via DCX downmodulation (Fig S8G). We checked whether DCX re-introduction (Tanaka et al, 2004) rescued migration in KALRN-KD cells. The substantial compensation of migration was observed in KALRN- and DCX-KD, but not in SOX11-KD IMR32_NM cells (Fig S8H). The most plausible explanation for this situation is the reprogramming taking place in SOX11-KD cells, which makes DCX irrelevant to cell migration. Kalirin-7 forced expression did not significantly affect motility nor induced differentiation in IMR-32, rescuing migration after application of 5 μM kalirin–GEF1 inhibitor#2 (Fig S9A; data not shown). We resolved migration defects by inspecting NNC/NCC angle distribution in IMR-32_NM after the suppression of RAC1 or kalirin (Fig 6I, left). Less NUC was observed in the RAC1- and kalirin–GEF1–inhibited cells, but NUC impact onto cell velocity was retained after RAC1- or kalirin-GEF1 inhibition (Fig 6I, right and Fig 6J). This, along with the results of |NC-CC| plotting, implied that NUC suppression by RAC1 or kalirin–GEF1 inhibitors did not lead to migration mode reprogramming, which was consistent with the results of RNA-seq in RAC1- and kalirin–GEF1–inhibited IMR-32 (Table S3). Next, we resolved RAC1- and kalirin–GEF1–inhibited IMR-32 cells in the pseudo-3D assay and observed suppressed single cell motility after treatment with either of the inhibitors (Fig 6K). As NC-CC mapping showed (Fig 6I), RAC1- and kalirin–GEF1–suppressed cells retained the impact of negative NC-CC within 0–40° block, which was an indicator of cell contractions. This prompted us to test combination treatments with the ROCK inhibitor and RAC1 or kalirin–GEF1 inhibitors. The treatment with either of kalirin–GEF1 inhibitors or RAC1 inhibitor aided with ROCK inhibitor reduced cell viability in SK-N-BE(2)c (Fig S9B). This effect was not present in other cell lines. In the pseudo 3-D assay, we noticed spheroids’ reduced dissociation after the addition of either of the kalirin–GEF1 inhibitors or RAC1 inhibitor to the ROCK-suppressed cells (Fig S9C) or addition of the ROCK inhibitor aided with either the RAC1 inhibitor or kalirin–GEF1 inhibitors (Fig S9D). Our data indicate that double treatments with ROCK- and kalirin–GEF1 inhibitors could reinforce inhibition of migration. The data provided mechanistic evidence for the NUC function of kalirin in ADRN-type NB cells.

• Download figure
• Open in new tab
• Download powerpoint

Figure S8 Analysis of NUC defects in kalirin-GEF1-inhibited cells.

. (A) Random walk plots and accumulated migration distances of vehicle- and kalirin–GEF1 inhibitor–treated NB-S-124 cells (13 h, 15-min intervals). Mean values + SD are presented. (B) Cell viability of NB cell lines treated with vehicle, kalirin–GEF1 inhibitor#2 (5 μM) or kalirin–GEF1 inhibitor#1 (10 μM) in 2D exclusion assay. Values are reported as mean percent + SD of vehicle-treated control. (C) Kalirin–STYV, βIII-tubulin, and γ-tubulin co-localisation in IMR-32. (D) Violin plots showing the nucleus-to-centrosome distance in IMR-32 cells treated with kalirin–GEF1 inhibitor#1, kalirin–GEF1 inhibitor#2, or RAC1 inhibitor and in KALRN-KD or SOX11-KD IMR-32 cells. P-values: RAC1 inh: 0.0001028, kalirin–GEF1 inhibitor#1: 1.933 × 10⁻⁶, kalirin-GEF1 inhibitor#2: 1.191 × 10⁻⁵, KALRN siRNA: 0.0001527, SOX11 siRNA; 2.388 × 10⁻⁵ (Welch t test). (E) γ-tubulin distribution in control, RAC1 inhibitor–treated and kalirin–GEF1 inhibitor#2–treated IMR-32 cells. Scale bar 20 μm. (F) Representative negative images of IMR-32 in control cells and after treatment with RAC1 inhibitor, kalirin–GEF1 inhibitor#2 or after KALRN-KD stained with phalloidin-Alexa Fluor-555. Scale bar 200 μm. (G) WB for DCX in RAC1-inhibited and kalirin-GEF1-suppressed IMR-32. (H) Cell migration (16 h, 15-min intervals) in expressing DCX-RFP KALRN-KD, SOX11-KD and DCX-KD (Ambion siRNA:145587) IMR-32 cells (72 h of RNAi). Values are reported as mean + SE of control. P-values: (DCX-RFP [−] versus DCX-RFP [+]): control siRNA: n.s., DCX siRNA: 9.906 × 10⁻⁷; SOX11 siRNA: 0.1; KALRN siRNA: 3.737 × 10⁻⁷ (Welch t test). The efficiency of DCX-KD was determined by WB, and the selected siRNA is marked in red.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6. Kalirin inhibition hinders migration in MNA cells and perturbs cell polarisation and MT structure.

(A) Relative migration in 2D exclusion assay after treatment with vehicle, kalirin–GEF1 inhibitor#1 (10 μM) and kalirin–GEF1 inhibitor#2 (5 μM) or after KALRN RNAi. Relative cell migration is quantified via normalization of cell density to vehicle- or siRNA-treated control. Graphs represent mean relative difference in migration + SD. (B) Random walk plots and accumulated migration distances in control IMR-32_NM cell after treatment with kalirin–GEF1 inhibitor#1 or kalirin–GEF1 inhibitor#2, and after 48 h of KALRN RNAi (13 and 10 h, 15-min intervals). Mean values + SD are presented. (C) Time-lapse images of IMR-32_NM after kalirin–GEF1 inhibitor#2 treatment. Nuclei and leading processes are indicated. Scale bar 20 μm. (D) |NC-CC| plots in control IMR-32_NM and after treatment with kalirin–GEF1 inhibitor#2, kalirin–GEF1 inhibitor#1 or KALRN siRNA. (E) The percentage of centrosomes located distally in control, kalirin, or RAC1-suppressed IMR-32 cells. Mean values + SD are presented. (F) DAPI staining showing changes in the nucleus shape in IMR-32 after kalirin–GEF1 inhibition. (G) Box plots demonstrating nuclear roundness in IMR-32 cells treated with RAC1- or kalirin–GEF1 inhibitor. Data represent three independent experiments (819, 1,008, 375, and 772 cells). (H) βIII-tubulin staining in control IMR-32 cells and after treatment with kalirin–GEF1 inhibitor#2 or KALRN RNAi. The representative fields were photographed. Scale bar 100 μm. (I) NNC/NCC angle frequency distribution (left) and NUC and noise-corrected NC-CC distances in 0–40° and 140–180° signatures in concatenated tracks from control (25 cells and 866 cells), RAC1-inhibited (30 cells, 563 timepoints), kalirin–GEF1–inhibited (25 cells and 606 timepoints) and KALRN siRNA treated (30 cells and 860 timepoints) IMR-32_NM (right). (J) Correlation plots between cell velocity and NUC footprint in control IMR-32_NM and after treatment with RAC1 inhibitor, kalirin–GEF1 inhibitor#1, kalirin–GEF1 inhibitor#2, or KALRN siRNA. (K) Phase contrast images, random walk plots, and accumulated migration distances of randomly migrating cells treated with vehicle, RAC1 inhibitor (10 μM), kalirin–GEF1 inhibitor#1 (10 μM) or kalirin–GEF1 inhibitor#2 (5 μM) for 48 h in pseudo-3-D (21 h, 90-min intervals). Mean values + SD are presented.

• Download figure
• Open in new tab
• Download powerpoint

Figure S9. Migration inhibition after double treatments with ROCK- and kalirin-GEF1 inhibitors.

(A) 2D exclusion assay in IMR-32-6/TR and IMR-32-6/TR-kalirin-7 (left) after treatment with kalirin–GEF1 inhibitor#2 (5 μM). Relative cell migration is quantified by normalization of cell density to vehicle-treated control (100%). Graphs represent the mean percent migration + SD from three independent experiments. Cell viability of IMR-32-6/TR and IMR-32-6/TR-kalirin-7 (right) treated with vehicle or kalirin–GEF1 inhibitor#2 (5 μM) in 2D exclusion assay. Values are reported as mean percent + SD of vehicle-treated control. (B) Cell viability of IMR-32, SK-N-BE(2)c, NB69, and KELLY treated with vehicle, kalirin–GEF1 inhibitor#1 (5 μM), kalirin–GEF1 inhibitor#2 or RAC1 inhibitor in combination with ROCK inhibitor (5 μM). Values are reported as mean percent + SD of vehicle-treated control. (C) IMR-32_NM spheroids treated with vehicle, ROCK inhibitor (5 μM), RAC1 inhibitor (5 μM), kalirin–GEF1 inhibitor#1 (10 μM), or kalirin–GEF1 inhibitor#2 (2.5 μM) for 48 h or treated with ROCK inhibitor (72 h) before addition of RAC1- or kalirin–GEF1 inhibitors (48 h) and stained with Calcein AM in pseudo-3-D. Scale bar 500 μm. (D) IMR-32 spheroids treated with ROCK inhibitor (5 μM) alone and in combination with RAC1 inhibitor (5 μM), kalirin–GEF1 inhibitor#1 (5 μM), or kalirin–GEF1 inhibitor#2 (2.5 μM) for 96 h in pseudo-3-D and stained with crystal violet.

Analysis of RNA-seq–resolved profiles of kalirin-GEF1–inhibited IMR-32 showed that the up-regulated genes had an affinity with transcriptomes of stage 4S, MYCN-nonamplified tumors and stages 1|2 (Figs 7A and S10A). There was a significant transcriptomic overlap between kalirin–GEF1 inhibition and DCX RNAi or RAC1 inhibition in IMR-32, compared with the SOX11 RNAi (Fig S10B and C; data not shown). Similar to DCX RNAi, gene signatures RAC1 neighbourhood, mRNA transport and NMD, MYC(N) and TP53 targets, gene sets pertaining to mitochondrial function and cortical cytoarchitecture were depleted in kalirin–GEF1–inhibited cells (Fig 7B and Table S3). The depletion of TP53 and MYC(N) targets was a trait present under several NUC inhibitory conditions, including DCX-KD, LIS1-KD, SOX11-KD, and kalirin–GEF1–inhibited IMR-32 (Fig S10D); and TP53 RNAi induced migration defects in IMR-32_NM (Fig S10E), although the TP53-KD cells remained adherent (Video 8). The up-regulation of pseudogenes and TCF21-dependent genes was also a trait that RAC1-inhibited and kalirin–GEF1–inhibited IMR-32 shared with the DCX-KD cells (Figs 7B and S10F, left). TCF21 was highly expressed in stage 4S (stage 4S versus stage 4, P-value < 0.001; stage 4S versus MNA, P-value <0.001; one-way ANOVA test; Fig S10F, right). Yet, we found neither a difference in TCF21 expression in the profiles of NUC-suppressed cells nor a possible proxy TF downstream of TCF21, which suggested other regulatory mechanisms behind the TCF21 signature in NB. The transcriptomic overlap within the group “RAC1 inhibitor and kalirin-GEF1 inhibitors” (P-value cutoff, 0.12; Kruskal–Wallis test) included up-regulation of ZNF702P, ETV1 (lowest P-value < 0.02), and SIAH1 (lowest P-value < 0.03), as well as down-regulation of KIF6, SCUBE1, EHMT1, and MPDZ (lowest P-value < 0.03) (Fig 7C and Table S3). The expression of ZNF702P and SIAH1 was associated with certain low-risk characteristics in NB (stage 1 versus stage 4, P-value < 0.05; MYCN-nonamplified versus MNA, P-value < 0.05; one-way ANOVA test), whereas KIF6 was highly expressed in stage 4 and MNA NB (stage 4 versus LOC, P-value < 0.001; MNA versus MYCN-nonamplified tumors, P-value < 0.001, one-way ANOVA test, Fig 7D). The genes correlated with ZNF702P, SIAH1, and KIF6 extracted from the gse49710 dataset showed overlaps with the profiles of kalirin–GEF1– and RAC1-inhibited cells (Fig S11A). As ZNF702P was down-regulated in the SOX11-KD cells (Table S2), we checked the lineage affinity of ZNF702P mRNA, which revealed ADRN type–specific ZNF702P expression (Fig S11B). Pseudogene function as miRNA decoys was documented before (Poliseno et al, 2010); therefore, we checked available miRNA–pseudogene interaction data (Li et al, 2014), which revealed the binding of several miRNA species to ZNF702P mRNA (Table S4).

• Download figure
• Open in new tab
• Download powerpoint

Figure S10. Analysis of DEGs in kalirin-GEF1-inhibited IMR-32.

(A) Volcano plots showing expression of DEGs in kalirin–GEF1–inhibited IMR-32 versus control IMR-32 in MYCN-nonamplified versus MNA and stages 1|2 versus 4 tumors. (B) Venn diagrams showing the numbers of genes in transcriptomic overlap between kalirin–GEF1 inhibition and DCX-KD or SOX11-KD in IMR-32. (C) Enrichment plots showing positions of DCX-KD and SOX11-KD transcriptomic overlaps in kalirin-GEF1-inhibited IMR32 versus control IMR-32. (D) Enrichment plots showing TP53 and MYC(N) targets in DCX-KD, RAC1-inhibited/kalirin-GEF1 LIS1-KD, SOX11-KD, and IMR-32 versus control. q-values are listed. (E) Random walk plots and accumulated migration distances in control IMR-32_NM cell after 48 h of TP53 RNAi (4 h, 5-min intervals). Mean values + SD are presented. P-value: 0.003 (Welch t test). (F) Enrichment plots showing “CUI_TCF21_TARGETS_DN“ gene set (left) in DCX-KD, RAC1-, and kalirin–GEF1–inhibited cells and box plot demonstrating TCF21 expression in primary NB tumors (right).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7. Inhibition of kalirin-GEF1 engages pathways of post-transcriptional gene regulation.

(A) Volcano plots showing expression of DEGs in kalirin–GEF1–inhibited cells versus control in stage 4S versus 4 tumors (gse49710). (B) Gene set enrichment analysis plots showing genes sets with similar pattern of regulation in kalirin–GEF1–inhibited and DCX-KD IMR-32. (C) Venn diagram showing the number of genes in transcriptomic overlap between DEGs in RAC1 inhibitor-, kalirin–GEF1 inhibitor#1– and kalirin–GEF1 inhibitor#2–treated IMR-32 versus control. (D) Box plots demonstrating ZNF702P, SIAH1, and KIF6 expression in primary NB tumors. (E) Color-coded scheme showing gene overlaps between 3′-UTR- and intronic-AU–rich element (ARE)-containing genes (Bakheet et al, 2018) in the ranked lists of DEGs in RAC1- and kalirin–GEF1–inhibited IMR-32 versus control IMR-32. (F) Enrichment plots for miR-573 targets (“miR-573_PITA”) based on PITA algorithm (Kertesz et al, 2007) in IMR-32–treated with kalirin–GEF1 inhibitor#1 versus control (top). Diagrams demonstrating miR-573 targets in the ranked gene subsets of 3′-UTR-ARE and intronic ARE containing genes from IMR-32–treated with kalirin–GEF1 inhibitor#1 versus control (bottom). (G) Box plot demonstrating MIR573 expression in primary NB tumors (gse62564) (left) and volcano plot showing expression of predicted miR-573 targets (right) in stage 4S versus 4 tumors. (H) Box plot showing distances between AREs and miR-573–binding sites in 3′-UTRs of DEGs in kalirin–GEF1–inhibited cells. P-value: 0.005 (two-sample Kolmogorov–Smirnov test). (I) Gene set enrichment analysis plots for the indicated gene sets (top) in the combined list of genes mis-regulated (based on their mis-expression in stage 4S versus stage 4 tumors [P-values by two-way t test ≤0.05; no logFC cutoff]) in RAC1- and kalirin–GEF1–inhibited IMR-32. Color-coded scheme showing gene overlaps (bottom) between mis-expressed genes extracted from profiles of kalirin–GEF1– and RAC1-inhibited cells and ranked according to their expression in stage 4S versus stage 4. (J) Diagrams showing positions of the genes up-regulated by dacarbazine in ADRN-type cell lines (Henrich et al, 2016) in the combined lists of genes mis-regulated in RAC1– and kalirin–GEF1–inhibited (top) and DCX-KD IMR-32 (bottom).

• Download figure
• Open in new tab
• Download powerpoint

Figure S11. Analysis of AU-rich element (ARE)–containing genes and miRNA targets in kalirin-GEF1– and RAC1-inhibited IMR-32.

(A) Diagrams showing positions of genes correlated positively or negatively with ZNF702P, SIAH1 and KIF6 expression in kalirin–GEF1–inhibited IMR-32 versus control IMR-32. (B) ZNF702P mRNA expression in MES and ADRN (marked in red) cell lines. (C) Enrichment plots for ARE-containing genes (Bakheet et al, 2018) in kalirin-GEF1- and RAC1-inhibited IMR-32 versus control IMR-32 (q-values are listed). (D) Enrichment plots for miR-573 targets (Kertesz et al, 2007) in the ranked gene subsets of 3′-UTR-ARE and intronic ARE containing genes from DCX-KD IMR-32 versus control IMR-32 (q-values are listed). (E) Venn diagrams showing the number of genes potentially targeted by miR-573 (Kertesz et al, 2007) in the subsets of genes correlated positively (“correlated”) and negatively (“anticorrelated”) with the expression of MIR573 in primary NB (gse62564). (F) Enrichment plots for the genes targeted by miR-873 (Li et al, 2014) in kalirin-GEF1-inhibitor#1-treated, DCX-KD and LIS1-KD IMR-32 versus control (q-values are listed) (top) and volcano plot showing expression of miR-873 targets (bottom) in stage 4S versus 4 tumors. (G) Box plots showing distances between AREs and miR-873 target sites in 3-UTRs of DEGs in kalirin–GEF1–inhibited cells. P-value: 0.04 (two-sample Kolmogorov–Smirnov test). (H) Mapping of DEGs in kalirin–GEF1–inhibitor#1- and kalirin–GEF1–inhibitor#2-treated IMR-32 (left) and miR-573 targets and ARE-containing genes (right) onto NB_disseminated tumor cell transcriptomic profiles. (I) Cell viability in RAC1-inhibited and kalirin–GEF1–inhibited IMR-32 treated with vincristine. Mean relative viabilities + SD are presented.

Given the regulation of RNA-binding protein-encoding mRNAs in kalirin–GEF1–inhibited cells, cross-checking of our datasets for genes containing AU-rich elements (AREs) (Bakheet et al, 2018) was reasonable, which revealed the prevalence of ARE-containing mRNAs in the up-regulated genes of kalirin–GEF1–inhibited cells (Figs 7E and S11C and Table S4). The enrichment involved mRNAs with 3′-UTR- and intronic AREs. Sub-setting for 3′-UTR-ARE- and intronic ARE-containing genes revealed that they were enriched for the targets of miR-181-5p, miR-153-5p, miR-335-3p, miR-493-5p, miR-12136, and miR-548-5p/3609-3p and the motif “AACTTT_UNKNOWN” (Table S4). Given the criteria for AACTTT selection (8 kb surrounding a transcription start site; Xie et al, 2005), we reasoned that “AACTTT” motif represented an intronic binding site for a miRNA. miRBase inspection showed that nucleotide string “AACUUU” is potentially targeted by miR-148a-5p, miR-548at-5p, miR-561-5p, and the juxtaposed to miR-12136 sequence “326_104” at 1p36.33, cloned previously from an MNA NB tumor (Afanasyeva et al, 2008). The miRNA precursors, MIR181A2 and MIR3609, were down-regulated in kalirin–GEF1–inhibited cells (Table S3), which could explain the up-regulation of miR-181 and miR-3609 targets. These two miRNA genes, together with MIR573 and MIR873, were present in profile overlaps after at least two of the NUC inhibitory treatments used in this study (DCX RNAi, LIS1 RNAi, SOX11 RNAi, and RAC1/kalirin–GEF1 inhibition), suggesting essential functions of these miRNAs in NUC regulation. Yet, ARE-positive genes potentially targeted by miR-573 were up-regulated in kalirin–GEF1–inhibited cells (Figs 7F and S11D). A similar effect was visible in DCX-KD, after subsetting the profile for ARE-positive genes (Fig S11D). The expression of MIR573 precursor was higher in stage 4S and LOC tumors (Fig 7G, left; P-values < 0.001; one-way ANOVA test). Also, the expression of predicted miR-573 targets showed affinity to stage 4S and stages 1|2. (Fig 7G, right, data not shown). Genes correlated positively with MIR573 expression in primary NB were enriched for miR-573 targets (Fig S11E). miR-873 targets showed a similar pattern of expression in kalirin–GEF1–inhibitor#1–treated IMR-32, but not in the DCX- and LIS1-KD cells (Fig S11F). Potential miR-573 binding sites and confirmed experimentally miR-873 binding sites (Li et al, 2014) were localised further from 3′-UTR AREs in the ARE-containing up-regulated genes compared with the ARE-containing down-regulated genes in kalirin–GEF1–inhibited cells (Figs 7H and S11G). Therefore, the up-regulation might require a function in cis of ARE-binding complexes and miRNA-induced silencing complexes (miRISCs). The enrichment for ARE-positive miRNA targets of miR-181-5p, miR-153-5p, miR-335-3p, miR-493-5p, miR-12136, and miR-548-5p/3609-3p was also apparent in the transcriptomes of stage 4S and stages 1|2, compared with stage 4 primary tumors (Table S4). Similar to DCX-KD, kalirin–GEF1 inhibitor^UP and NB_DTC^UP gene sets formed an overlap (Fig S11H, left), which was an indication that kalirin–GEF1 inhibition took place in NB in vivo. Also, affinity of miR-573 targets, 3′-UTR- and intronic-ARE–positive mRNAs with transcriptomes of DTCs was observed (Fig S11H, right).

Next, we checked whether and RAC1- and kalirin–GEF1 inhibition affects cell death induced by doxorubicin, paclitaxel, vincristine, and exherin in IMR-32 and SK-N-BE(2)c and observed only minor interaction with the MT drugs, vincristine and paclitaxel (Fig S11I, data not shown). Our data indicated that mis-expression mainly affected down-regulation in RAC1- and kalirin–GEF1–inhibited cells; therefore, possible vulnerabilities were to be found in the mechanisms of down-regulation. We combined mis-expressed genes extracted from the transcriptomic profiles of RAC1- and kalirin–GEF1–inhibited cells and subjected derived lists to GSEA, which revealed the presence of CHEK2 neighbourhood and DREAM targets in a small subset of mis–up-regulated genes (Fig 7I and Table S4). The subset of mis–down-regulated genes was enriched for bivalently marked genes (Fig 7I). This suggested that the activity of the polycomb repressive complex was responsible for the gene down-regulation in the NUC-suppressed cells. Several prospective epigenetic modifiers, including EZH2 inhibitors, were already tested in NB (Henrich et al, 2016; Chen et al, 2018). Yet, the genes up-regulated after treatment of ADRN-type cell line, IMR-5/75, with an EZH2 inhibitor (Henrich et al, 2016) showed little overlap with the profiles of kalirin–GEF1– and RAC1-inhibited IMR-32 (data not shown). On the other hand, the genes up-regulated after the treatment with the DNA-methyltransferase inhibitor (DNMT) dacarbazine showed an overlap with mis–down-regulated genes in kalirin–GEF1– and RAC1-inhibited IMR-32 as well as DCX-KD cells (Fig 7J), implying that dynamic DNA methylation might take place in the NUC-suppressed cells. Taken together, these data show that pharmacological inhibition of kalirin–GEF1 evokes transcriptomic traits of low-risk NB.

Our data provide evidence that gene up-regulation driven by kalirin–GEF1 inhibition operates via miRNA-dependent posttranscriptional mechanisms. Most importantly, our analyses suggest that suppressors of epigenetically driven gene down-regulation might interact with kalirin–GEF1 inhibitors synergistically, which should be checked in future experiments.