Kalirin-RAC controls nucleokinetic migration in ADRN-type neuroblastoma

Video and gene expression analyses coupled with the RNAi technique reveal microtubule-dependent, neuronal-like nucleokinetic migration in a noradrenergic type of neuroblastoma, providing a glimpse into the mechanism by which noradrenergic neuroblastoma cells may spread.

2) In line with my previous comment, the molecular relationship between Sox11 and kalirin-Rac1 is unclear, both of which act as an upstream regulator of Dcx. Do Sox11 and Rac1 independently regulate Dcx expression? Or is Sox11 is an upstream (or downstream) regulator of Rac1? The authors should examine whether suppression of Sox11 (or kalirin/Rac1) affects the activity of Rac1 (or Sox11).
3) As described above, underlying mechanisms of the migration of postmitotic neurons and the interkinetic nuclear migration (IKNM) are somewhat different. While the nuclear movement in postmitotic neurons essentially requires dynein motor activity, the IKNM during G1 phase of neural progenitors depends on kinesin and myosin (and increased nuclear density at the apical region of the ventricular zone also provides the force for IKNM during G1 phase) (Genes Cells (2013) Vol.18, 176-194). This may not match the authors' conclusion, although I think there is no need that all migratory behaviors of the ADRN-type NBs and developing postmitotic neurons are similar. In addition, previous reports indicate that suppression of Rac1 activity decreases the distance between the centrosomes and nuclei (Cell Rep (2012) Vol.2, 640-651), whereas Lis1 heterozygous deficiency and Dcx knockdown increase the distance (J Cell Biol (2004) Vol. 165, 709-721. J Neurosci (2008. It is also inconsistent with the case of the ADRN-type NBs, where Rac1 positively controls Dcx. How do you explain these incompatible results? 4) Regarding Fig. 6E, the analyzing method may not be a standard in the field of developmental neuroscience. The authors should measure the distance between the centrosomes and nuclei.
5) The immunoblot data in Fig. 5E are not good. Quantitative data are required. In addition, as I mentioned above, the nuclear elongation should also be quantified. In my eyes, the nuclei in the NBs treated with a ROCK inhibitor show abnormally elongated morphologies.

RAC1 pulldown replicates
Kalrn expression in t-SNE-resolved E12.5 and E13.5 sympathetic precursors (Furlan et al., 2017).   Table S1 RNA-seq data from DCX-KD and LIS1-KD Table S2 RNA-seq data from SOX11-KD IMR-32 Table S3 RNA-seq data from RAC1/kalirin-GEF1-suppressed IMR-32 Reviewer #1 (Comments to the Authors (Required)): Here the authors provide evidence for nucleokinetic migration in ADRN-type neuroblastoma cells. This is somehow expected since neuroblast migration require nucleokinetic migration during development. The authors beautiful demonstrate how this process require ROCK and RAC1 and provide evidence that NUC is controlled by SOX11 and DCX.
minor comments: the authors state in the introduction: "understanding the mechanisms implicated in migration of DCX-positive NB can shed light on initial steps of the metastatic process in NB" It is not clear from this paper however, if the metastatic nature of neuroblastoma originate from MES type or ADRN type, and thus if nucleokinetic migration is part of the metastatic program in high stage NB. Long distance migration of neural crest and Schwann cell precursors (SCP) during embryonic development do not depend on NUC and DCX. It would be helpful to comment on this in the discussion.  (Furlan et al., 2017).
Given the presence of Dcx in sympathoblasts and chromaffin cells, we can speculate that NUC is active in these cells and might be inherited by ADRN NB. We can approach this issue by revisiting the lissencephaly and tubulinopathy cases for defects in the peripheral nervous system. Only a few cases have been reported where patients with lissencephaly also had a neurocristopathy, or Hirschsprung disease (Hikita et al., 2014;Mittal et al., 2014). One case related to a mutation in an aristaless-related gene (ARX) was reported with regard to adrenal gland hypoplasia (Bonneau et al., 2014). There is evidence that Lis1 haploinsufficiency adversely affects the migration of (para)sympathetic preganglionic neurons, but it is not clear whether this defect also pertains to neural crest-derived neurons. Hereditary dysautonomias constitute a special type of neurocristopathy that can be mechanistically linked to NB migration. Familial dysautonomia (FD), which in most cases is caused by mutations in IKBKAP, was initially thought to be the result of defective neuronal migration (Naumanen et al., 2008). However, it was later suggested that neuron loss in FD was not a consequence of neuron migration failure (Jackson et al., 2014). All in all, it is unlikely that NUC is active during the development of sympathetic ganglia. Evolutionary younger than other neural crest derivatives postganglionic sympathetic elements emerge in gnathostoma through phox2, ascl1 and hand coalescence into an expression module (Häming et al., 2011). Therefore, the correspondence between NUC migration in NB and CNS might manifest during the development of the neuronal subtypes expressing NB-like transcription factor code, i.e., hindbrain (nor)adrenergic neuronal formations (Zeisel et al., 2018). Particularly, facial branchiomotor neurons demonstrate N-C inversions when migrating tangentially (Distel et al., 2010). Also, motor neurons migrate tangentially in a reelin-negative region (Rossel et al., 2005), which is an interesting notion when put in the context of RELN negativeness of advanced NB (Becker et al., 2012). Probably, migrating ADRN NB resembles motor neurons, but "gets stuck" at the stage of tangential migration-like centrosomal repolarisation.
Remarkably, murine orthologue of up-regulated in kalirin-GEF1-inhibited cells ETV1 is expressed in the nuclei of mouse cranial nerves (Zeisel et al., 2018). More precisely, Etv1 expression appears in facial motor neurons during the final post-tangential stage of their migration, the sub-nuclear segregation, and is indispensable for finalising neuronal differentiation (Zhu, Guthrie, 2013;Tenney et al., 2019). Kalirin-GEF1 inhibition upregulates low-risk-specific transcriptomic traits in ADRN NB (Table S3, the manuscript), which can be interpreted as a sign of a differentiation-like process. Yet, it is the kalirin paralog, Trio, that is involved in facial motor sub-nuclear segregation in mice (Backer et al., 2007). On the other hand, in xenopus, kalrn expression is present in the cranial nerves, while trio is expressed in (n = 493) and ITSC_P4 (n = 520) and the expression of the marker of migrating neural crest cells, SOX10, was determined by WB in nuclear protein lysates from JoMa1.3 and total protein lysates from ITSCs. F or positive control, total protein lysates from melanoma cell line, MeWo (Alver et al., 2016), were used.
The manuscript by Afanasyeva et al describes morphological and molecular similarities between neuroblastoma cell lines with a sympathetic noradrenergic identity (ADRN-type NBs) and migrating neurons in the developing brain. In normal developing brains, migrating neurons extend the leading and trailing processes. During migration, the centrosome moves forward and cytoplasmic dilation is formed at the proximal region of the leading process. Subsequently, the nucleus shows elongated morphology and moves into the cytoplasmic dilation. Thus, the migrating neurons exhibit "saltatory movement". In the submitted manuscript, the authors showed that the ADRN-type NBs exhibited saltatory movement ( Fig.  1F) with the formation of leading process-and cytoplasmic dilation-like structures (Fig. 1I). The authors also observed the nuclear elongation in the ADRN-type NBs, although they did not provide quantitative data. In addition, the migration of the ADRN-type NBs was shown to require Dcx, Lis1 and Rac1, both of which were previously reported to regulate neuronal migration in the developing brains. Inhibition of Rac1 and its activator, kalirin, disturbed centrosomal positioning (Fig. 6E).
These findings are interesting and informative to researchers in not only cancer biology but also developmental neuroscience. However, this reviewer finds several weak points in this manuscript. This study identified many molecules involved in the migration of ADRN-type NBs, but the epistasis of these molecules is unclear. Both the kalirin-Rac1 pathway and Sox11 upregulate the expression of Dcx, but suppression of Rac1, Sox11 and Dcx exhibits different phenotypes. The relationship between Rac1, Dcx and Sox11 in the migration of the ADRNtype NBs should be clarified. Second, several molecules, including Dcx, have previously been reported to be associated with neuroblastoma, which reduces the novelty of this manuscript. Third, the authors may confuse nucleokinesis in postmitotic neurons with interkinetic nuclear migration in neural progenitors, as described below.
Overall, this reviewer finds this manuscript potentially interesting, but many additional experiments are required to resolve the above-mentioned problems.
Thank you very much for your thoughtful and thorough review of our manuscript. We have read the comments and concerns very carefully. The idea that DCX is an NB-relevant gene has indeed been highlighted in several correlative studies. We believe that our work is the first attempt to explore whether there is a functional link between the presence of DCX and the migration mode in NB. Apart from technical issues, it appears to us that the reviewer's concerns focus on two important issues: 1) the link between SOX11, DCX and kalirin; and 2) the transitions between MES and ADRN states in KD and drug-treated cells. Below, we respond in detail to the reviewer's comments and describe our additional experiments.
Changes in the initial version of the manuscript are either highlighted for added sentences or striked through for deleted sentences in the revised version.
[Major points] 1) Knockdown of Sox11 dramatically reduced Dcx mRNA in IMR-32 cells, indicating that Sox11 is a major upstream regulator of Dcx. (*) However, the morphological abnormalities of Dcxknockdown cells seem to be more severe than those of Sox11-knockdown in Fig. 3H. (**) I wonder if some of the phenotypes of the Dcx-knockdown cells might result from off-targeting effects. Does re-expression of Dcx in the Dcx-knockdown cells restore all phenotypes? (***) In addition, the authors should also examine whether expression of Dcx could rescue the phenotypes of the Sox11-or kalirin-knockdown, which would resolve the above-mentioned first issue.
These are two excellent points, particularly the one concerning the DCX rescue.
(*) The phenotype in DCX-KD IMR-32 is not "more severe" but distinctly different in nature.
As we wrote in the previous version of our manuscript, GO term and motif analysis of SOX11-KD IMR-32 up-regulome revealed enrichment for AP-1 and ETS1 targets as well as EMT hallmark (Table S2, the manuscript and Figure R1), suggesting NUC(ADRN)-to-MES transition.
This transition was also observed in SOX11-KD IMR-32 using an independent siRNA (Decaesteker et al., publication in preparation). In line with this observation, SOX11-KD facilitated the inhibition of migration; however, evidence was found concerning slow, NUCindependent migration mode ( Figures 3K and 3L) that was accompanied with the depletion of tubulin-related expression signatures and upregulation of actin-related signatures (Table S2).  (Table   S1). Consistently, no evidence for MES program was found in the up-regulomes of DCX-(and LIS1-KD) cells (Table S1, Figures S2A (the manuscript) (Table S1). The overrepresentation of MYC targets also indicated possible functional overlap between DCX and SOX11. This point has been mentioned in the revised version of our manuscript. Yet, we are puzzled as to how DCX regulates transcription in NB. It is known that Dcx regulates motordriven neuronal transport mediated by Kif1a (Liu et al., 2012), thus having the potential to control neurite-to-nucleus signaling and hence transcription.
We would like to mention that statistical analysis shows that DCX expressions defined by RNA-seq in stage 4S and stage 4 belong to different distributions according to KS test ( Figure   R3). The microarray data show a similar trend, but p-values are higher than 0.05 ( Figure R3; (**) we took into consideration that the off-target effects of DCX RNAi have been noted previously (albeit in the experiments with a shRNA and not siRNA; Baek et al., 2014).
Following the reviewer's suggestion, we complemented our video analysis of DCX-KD by inserting two additional siRNAs § to DCX ( Figure R4), which revealed a migration-defective phenotype ( Figure R5). We would like to keep these data within this letter instead of the main text.  Table S2). We checked this finding using WB, which revealed kalirin and TIAM1 downregulation in IMR32 upon SOX11-KD ( Figure S6D). We did not observe the induction of previously annotated TRIO isoforms upon SOX11-KD. We also checked SOX11 expression in RAC1-/kalirin-GEF1-inhibited and KALRN-KD IMR32 cells and did not observe any significant changes in the SOX11 level. Next, we searched for SOX11 targets in our RNA-seq data from RAC1/kalirin-GEF1-inhibited cells (Table   S3). The first attempt to characterize NB-relevant SOX11 targets has been already undertaken (Decaesteker et al., under revision). We discuss SOX11 targets and SOX11-interacting proteins in the context of previously published works (Kuo et al., 2015;Heim Birgit, 2014 etc).
We also checked whether forced expression of the DCX construct corrected migration defects induced by KALRN-KD, which revealed the compensation of migration upon DCX-RFP transfection ( Figure S8H).
Likewise, we are intrigued by the idea of IKNM activity in NB. As we state in our manuscript ("IMR5-75 expressing a FUCCI cell cycle sensor [Ryl et al., 2017] and growing asynchronously, which revealed tendency for migration in the G1 phase [ Figure 2K; the manuscript]."), the fact that NB cells are less migratory in the S/G2 phase along with DCX involvement is an indirect evidence of neuron-like migration, rather than IKNM. Indeed, this does not necessarily mean that nuclear migrations do not take place during S and G2 phases. It is known that interkinetic nuclear migrations are minimal during S-phase (Hayes, Nowakowski;2000). To resolve this question, we used IMR-32 expressing G1 marker ( Figure R8) and mapped the nuclear and cytoplasmic centers. We did not observe a drastic switch in NNC/NCC mapping after the cells passed through G1 ( Figure R9), which suggested similar migration mechanisms ( Figure 2K).  Dcx controls basally directed nuclear movement in rat brain progenitor cells (Carabalona et al., 2016). While IKNM is not amenable to examination in dissociated cultures due to 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), helped to tell IKNM from NUC in NB.
To consolidate all the data, we checked the effect of CDK5 RNAi on the migration in IMR-32.
Thus, two programs, the nucleokinesis and the cell cycle, merge in ADRN NB. We would like to keep these data in the revision letter. Yet, these experiments do not provide an answer to the question whether G2 phase-specific mechanisms of nuclear movement are active in NB (particularly, a nuclear pore-mediated mechanism that involves RANBP2-BICD2 (Baffet et al., 2015)). We found that RANBP2 and BICD2 were co-expressed in NB (R = 0.648; Figure R13). A number of missense mutations and the presence of CpG methylation (as documented in the CCLE database) in RANBP2 gene in NB indicated that CDK1-BICD2-RANBP2 functionality might not be intact in NB. The second issue is the centrosome motility observed in NB cells, which might provide a clue to the whole problem. During the G1 phase of IKNM, centrosome is tethered at the apical surface via a primary cilium, which persists during the cell cycle until late G2 (Spear and Erickson, 2012). On the other hand, in migrating post-mitotic cortical neurons, primary cilia are highly dynamic (Higginbotham et al., 2012). This is a very interesting topic that should be pursued in further studies. In this respect, it is well worth mentioning that a ciliary dysfunction disorder, Bardet-Biedl syndrome, manifests in neural crest migration defects and is associated with Hirschsprung's disease (Tobin et al., 2008). Moreover, one of the primary regulators of cilia disassembly, Aurora A (Korobeynikov et al., 2017), is deregulated in NB (Faisal et al., 2011).
Overall, the question is also related to NUC origin in NB. A detailed response to this question is provided in the revised discussion. - In addition, previous reports indicate that suppression of Rac1 activity decreases the distance between the centrosomes and nuclei (Cell Rep (2012)  During the manuscript preparation, we checked the low-resolution images of the LIS1-KD cells and did not find extreme centrosome "overshoots" (Tanaka et al., 2004), which already seemed a discrepancy with regard to the previous publications. Following the reviewer's suggestion, we measured the nucleus-to-centrosome distance in DCX-, LIS1-, SOX11-KD and RAC1-/kalirin-suppressed IMR-32 cells at a higher magnification ( Figure 2B and S8D). This revealed that RAC1-or kalirin-suppression indeed reduced the nucleus-to-centrosome distance ( Figures R14 and R15; Figure S8D). Also, we observed a higher variability in the nucleus-to-centrosome distance in LIS1-KD, but not in DCX-KD IMR-32 cells. This conforms with the results of video analysis which demonstrated cell rounding and faint projections in DCX-KD IMR-32. We also noticed that fewer centrosomes were present distally in the cells with elongated nuclei after DCX-KD (see figure below). This is somewhat common (although coupled also with nuclear rounding) in RAC1-and kalirin-suppressed as well as SOX11-KD IMR-32 cells ( Figure 6E), which led us to conclude that both DCX and RAC1/kalirin might regulate the centrosome translocation in NB cells.  We have added the data from centrosome-to-nucleus measurements ( Figure 2B and S8D) but would like to retain Figure 6E. This is because we think that this panel provides useful information, and thus we respectfully ask the reviewer to re-consider his/her point.  Fig. 5E are not good. Quantitative data are required. In addition, as I mentioned above, the nuclear elongation should also be quantified. In my eyes, the nuclei in the NBs treated with a ROCK inhibitor show abnormally elongated morphologies.
Thank you very much for the opportunity to complement our RAC1 pulldown data with additional replicates and the results from SOX11-KD ( Figure 5E; Figure S7C). Given the upregulation of RAC1 neighborhood in SOX11-KD IMR-32 (Table S2), we weren't sure about the outcome of pulldown experiments in SOX11-KD. We think the results captured the redistribution of RAC1 activity in the SOX11-KD cells. The data on nuclear elongation are provided as the distribution of nuclear roundness across the cell population (several hundred cells) in Figure 6G. This panel shows that RAC1/kalirin-GEF1 inhibition increases nuclear roundness.
Yes, we also noticed that ROCK inhibition affects nuclear shape, albeit in a different way ( Figure R16). We noticed that nuclei look groove-less in the cells after ROCK inhibition, while the grooves are still visible in the nuclei of RAC1-/kalirin-GEF1-suppressed cells ( Figure R17).
Also, ROCK-treated nuclei may look elongated on the kymographs ( Figure 4G)). This is due to image scaling along the x-axis.
As Figure R18 demonstrates there is no nuclear elongation in ROCK inhibitor-treated cells. As ROCK is not the major focus of our paper, we have included these data in the response letter but not the main text. 1) First of all, we highly encourage our referees to go through our analysis of NUC-deficient cells and our conclusions about epigenetic downregulation in NUC-deficient cells. We invest a lot of hopes to this direction.
2) Next, according to PCA plotting (Decaesteker et al., unpublished observation, Figure R18) the clone we chose for the previous manuscript ("B6") version does not cluster with the other SH-EP_SOX11-TAT clones. We are still not sure whether this was a cloning artefact, or a result of SH-EP heterogeneity that might have something to do with the NB lineages. Finally, we decided to replace the results from clone "B6" with results from three other clones (previously Figure 3K, new Figure S3K-L). All three clones demonstrate slight but statistically significant migration increase upon induction of SOX11 expression.  3) At the moment, we are not sure whether SDHA is a competent reference gene since our RNA-seq analyses in DCX-KD IMR-32 revealed that this gene was co-regulated with DCX. We re-analysed our RT-qPCR data from the DCX-KD cells using HPRT1. Our western blots show that there was no co-regulation of DCX protein in LIS1-KD IMR-32 ( Figure R20). Figure R20. WB for DCX and LIS1 in IMR-32 after transfection (WB for DCX and LIS1 in IMR-32 (Trizol-isolated protein; RNA-seq replicate). LIS1 AB was from Cell Signaling.